0% found this document useful (0 votes)
13 views21 pages

Detailed Land Use Classification in A Rare Earth Mining Area Using Hyperspectral Remote Sensing Data For Sustainable Agricultural Development

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
13 views21 pages

Detailed Land Use Classification in A Rare Earth Mining Area Using Hyperspectral Remote Sensing Data For Sustainable Agricultural Development

Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 21

sustainability

Article
Detailed Land Use Classification in a Rare Earth Mining Area
Using Hyperspectral Remote Sensing Data for Sustainable
Agricultural Development
Chige Li 1 , Hengkai Li 1, * , Yanbing Zhou 2, * and Xiuli Wang 3

1 Jiangxi Province Education Department, School of Civil and Surveying & Mapping Engineering,
Jiangxi University of Science and Technology, Ganzhou 341000, China; [email protected]
2 Information Technology Research Center, Beijing Academy of Agriculture and Forestry Sciences,
Beijing 100097, China
3 Jiangxi Province Education Department, School of Economics and Management, Jiangxi University of Science
and Technology, Ganzhou 341000, China; [email protected]
* Correspondence: [email protected] (H.L.); [email protected] (Y.Z.)

Abstract: In China, ion-adsorbing rare earth minerals are mainly located in the southern hilly
areas and are important strategic resources. Extensive long-term mining has severely damaged
the land cover in mining areas, caused soil pollution and terrain fragmentation, disrupted the
balance between mining and agriculture, severely restricted agricultural development, and affected
ecological development. Precise and detailed classification of land use within mining areas is crucial
for monitoring the sustainable development of agricultural ecology in these areas. In this study,
we leverage the high spatial and high spectral resolution characteristics of the Zhuhai-1 (OHS)
hyperspectral image datasets. We create four types of datasets based on spectral, vegetation, red
edge, and texture characteristics. These datasets are optimized for multifaceted features, considering
the complex land use scenario in rare earth mining areas. Additionally, we design seven optimal
combination schemes for features. This is performed to examine the impact of different schemes on
land use classification in rare earth mining areas and the accuracy of identifying agricultural land
classes from broken blocks. The results show that (1) the inclusion of texture features has the most
Citation: Li, C.; Li, H.; Zhou, Y.; obvious effect on the overall classification accuracy; (2) the red edge feature has the worst effect on
Wang, X. Detailed Land Use improving the overall accuracy of the surface classification; however, it has a prominent effect on
Classification in a Rare Earth Mining
the identification of agricultural lands such as farmland, orchards, and reclaimed vegetation; and
Area Using Hyperspectral Remote
(3), following the combination of various optimization features, the land use classification yielded
Sensing Data for Sustainable
the highest overall accuracy, at 88.16%. Furthermore, the comprehensive identification of various
Agricultural Development.
agricultural land classes, including farmland, orchards, and greenhouse vegetables, yielded the most
Sustainability 2024, 16, 3582. https://
doi.org/10.3390/su16093582
desirable outcomes. The research results not only highlight the advantages of hyperspectral images
for complex terrain classification and recognition but also address the previous limitations in the
Academic Editor: Yingkuan Wang
application of hyperspectral datasets over wide mining areas. Additionally, the results underscore
Received: 31 January 2024 the reliability of feature selection methods in reducing information redundancy and improving
Revised: 29 March 2024 classification accuracy. The proposed feature selection combination, based on OHS hyperspectral
Accepted: 19 April 2024 datasets, offers technical support and guidance for the detailed classification of complex land use in
Published: 24 April 2024 mining areas and the accurate monitoring of agroecological environments.

Keywords: rare earth mining area; agricultural development monitoring; land use classification;
hyperspectral; feature optimization
Copyright: © 2024 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
1. Introduction
Attribution (CC BY) license (https:// The ecological environment in mining areas is closely related to the agricultural
creativecommons.org/licenses/by/ development in the area. The sustainable development of agriculture in mining areas can
4.0/). result in sustainable ecological development in the area. Southern China hosts significant

Sustainability 2024, 16, 3582. https://doi.org/10.3390/su16093582 https://www.mdpi.com/journal/sustainability


Sustainability 2024, 16, 3582 2 of 21

ion-adsorbing rare earth mineral resources in its hilly and mountainous regions. However,
early pool leaching, heap leaching, and in situ leaching processes can lead to soil erosion,
acidification, desertification, and other environmental problems [1,2]. Moreover, chaotic
mining processes lead to important changes in the physical and chemical properties of
soil in mining areas. Consequently, both vegetation and crops not only suffer significant
damage but also obstacles in their natural growth [3].
Agricultural development, whether in farmland, orchards, or other entities like re-
claimed vegetation, is affected by varying degrees of land fragmentation and poor growth
under the current status quo. In addition, there are many other land classes in mines, such
as tailings land, bare land, and sedimentation tanks, among others, which contribute to
the complex land cover classes. Therefore, it is crucial to have a precise understanding
of regional land use information. This is pivotal for correctly distinguishing between
agricultural and industrial land classes in mining areas, monitoring and comprehending
the agroecological environment conditions within these mining zones, and assessing the
overall ecological environment conditions [4,5].
There have only been a few studies on the detailed land use classification of complex
surfaces in mining areas and most of them focused on using existing land use datasets or
multispectral imagery to classify mining areas into simple land classes. For instance, Zhang
Zemin et al. [6] used 100 m spatial resolution land use datasets to study the change in land
use classes in a typical mining area and Jiaxing Xu et al. [7] combined 30 m spatial resolution
OLI satellite images with 30 m spatial resolution to classify a mining area into simple land
classes such as cropland, forest land, industrial and mining land, and water and proposed
a random forest classification method based on a multi-feature combination classification
scheme for remotely sensed images. However, these studies failed to achieve a detailed
classification of complex land classes within mining areas. Hyperspectral remote sensing
datasets, with their high spectral resolution, extensive information content, and multiple
narrow spectral bands, are highly sensitive to minute spectral details. This sensitivity
enables the detailed classification of land use, leveraging its potential to capture intricate
spectral features and differences in land classes [5,8]. Advances in hyperspectral technology
have endowed satellite hyperspectral images, such as those from Zhuhai-1 (OHS), with a
high spatial and spectral resolution, as well as enabling them to provide wide coverage
and large quantities. This capability enables the detailed classification of land use data in
rare earth mining.
However, there are still some problems in recognizing and detecting different kinds
of features using hyperspectral datasets. During the differentiation of complex landforms
or similar species, different landforms or species may present the same spectral features
or a mixed spectral phenomenon in a certain spectral segment [9], which can impact the
classification accuracy. To overcome this situation, many studies have adopted different
methods to extract and classify land classes or species for different study areas. Fei Xing
et al. [10] used a multi-end-element spectral mixing analysis method based on OHS hy-
perspectral imagery to extract and analyze alpine grass cover. Peng Qin et al. [11] used a
U-Net convolutional neural network to extract and use intersection as a basis for classifying
water in OHS hyperspectral imagery extraction and used intersection as the evaluation
index for model training. Guoli Zhou et al. [12] used spectral, topographic, texture, vegeta-
tion index, and other feature variables to construct an XGBoost model and combined this
with OHS hyperspectral datasets for bamboo species identification, achieving better user
accuracy and producer accuracy. Binge Cui et al. [13] proposed a small scene embedding
network (TSE-Net) based on scene representation and an attention mechanism for coastal
wetland small scene classification based on OHS hyperspectral images. Canran Tu et al. [11]
proposed a collaborative classification method fusing OHS hyperspectral imagery and
Gaofen-3 fully polarized synthetic aperture radar (SAR) imagery for wetland mapping and
achieved robust classification results. Previous studies have shown that whether it is the
extraction of individual landforms or the classification of multiple landforms/species, bet-
ter results can be achieved by using OHS hyperspectral datasets, which fully demonstrates
Sustainability 2024, 16, 3582 3 of 21

the advantages of OHS hyperspectral datasets in landform extraction and classification.


Furthermore, researchers have used different methods for different scenarios, revealing the
differences between each research method in different scenarios.
However, in the current study, the use of hyperspectral datasets for landform extraction
and classification focuses more on the classification of small scenes and the identification
of individual landforms. The study area is more concentrated in coastal wetlands, alpine
grasslands, and other common areas. Rare earth mining areas are located in the hilly
and mountainous areas of southern China, receiving less attention and, because of the
complexity and fragmentation of features caused by long-term mining, the identification of
agricultural-related features such as farmland and reclaimed vegetation is more difficult,
with fewer studies addressing these challenges. Using OHS hyperspectral datasets to
classify the land use in these mining areas serves a dual purpose. It not only addresses
the underutilization of hyperspectral datasets on a large scale in mining areas but also
enables accurate identification and monitoring of agricultural land classes linked to the
ecological environment. This, in turn, provides valuable insights for promoting sustainable
agriculture and ecological stability in mining areas.
In addition, as mentioned above, hyperspectral datasets are characterized by high
spatial and spectral resolution, rich spectral information, and high sensitivity to subtle
features leading to highly complex data; ordinary algorithms find it difficult to effectively
deal with its high-dimensional characteristics; nevertheless, the random forest algorithm
(RF) stands out due to its exceptional performance in handling this complex data, as
evidenced by its parallel training, resistance to overfitting, and adaptability to a large
number of input features. In current hyperspectral classification research, RF is widely used.
You Mo et al. [14] used an RF classifier combined with multiple feature indicators for land
use classification in the Pearl River Delta based on OHS hyperspectral datasets, achieving
robust results. Zhihua Wang et al. [15] proposed a method for the detailed classification
of crops using a combination of feature transformation and RF based on unmanned aerial
vehicle (UAV) hyperspectral imagery. Somnath Paramanik et al. [16] applied RF modeling
to classify the distribution of three species and two species combinations of mangrove
forests with the desired results. Haiming Qin et al. [17] used an RF classifier to classify
18 species of subtropical broadleaf forests based on different features extracted from UAV
hyperspectral images, radar images, and ultra-high resolution RGB images. It is evident that
the random forest algorithm (RF) has certain advantages and prospects for the classification
and identification of remote sensing images, including hyperspectral images.
In summary, addressing the current gaps in the large-scale application of hyperspectral
datasets in mining areas is crucial. Identifying agricultural land classes, including farmland,
orchards, and reclaimed vegetation, in these regions presents challenges due to fragmented
land surfaces and soil contamination resulting from mining. Given the close connection
between agriculture and the ecological environment, it is imperative to conduct an extensive
and detailed classification of land use in mining areas to accurately identify agricultural
land classes and understand the status of agricultural development.
This study focuses on analyzing the Lingbei rare earth mining area using OHS hy-
perspectral datasets. The analysis involves creating spectral, vegetation, red edge, and
texture feature datasets. These datasets are used to comprehensively interpret the indices
contained in the above different categories of feature variables from the separability of
land classes and feature relevance and rank the importance of features for vegetation
features, red edge features, and texture features. Combinations of these feature variables
are selected as the basis for identifying agricultural land classes and understanding the
agricultural development in the area. Selected feature variable combinations are utilized as
input parameters to assess the impact of different combinations on land use classification
and agricultural land identification in the mining area using the RF algorithm. The goal
is to identify the optimal feature selection combinations suitable for effective land use
classification and agricultural land identification in mining areas.
Sustainability 2024, 16, x FOR PEER REVIEW 4 of 22

Sustainability 2024, 16, 3582


to identify the optimal feature selection combinations suitable for effective land use clas-
4 of 21
sification and agricultural land identification in mining areas.

2. Data and Methodology


2. Data and Methodology
2.1.Overview
2.1. Overviewofofthe
theStudy
StudyArea
Area
TheLingbei
The LingbeiRare RareEarth
EarthMining
MiningArea Areaisislocated
locatedatatlatitude
latitude2424°51′24″–25°02′56″
◦ 51′ 24′′ –25◦ 02′ 56′′ NNand
and
longitude 114°58′04″–115°10′56″
◦ ′ ′′ ◦ ′ E;
′′ it is about 2 km north of Dingnan
longitude 114 58 04 –115 10 56 E; it is about 2 km north of Dingnan County, Jiangxi County, Jiangxi Prov-
ince, China,
Province, has an
China, has area
an of about
area 213 km
of about 213(Figure
2 1), and1),
km2 (Figure hasand
beenhas mined
beenfor over for
mined 30 years.
over
It has been affected by early unregulated mining processes and thus
30 years. It has been affected by early unregulated mining processes and thus has under- has undergone severe
internal
gone surface
severe fragmentation,
internal along with along
surface fragmentation, soil acidification and soil sanding
with soil acidification due
and soil to pro-
sanding
longed mining. These phenomena occur intermittently, impeding
due to prolonged mining. These phenomena occur intermittently, impeding the growth of the growth of vegeta-
tion and crops.
vegetation The introduction
and crops. The introduction of theofJiangxi Province
the Jiangxi Mineral
Province Resources
Mineral Master
Resources Plan
Master
(2008–2015) [18] has brought about some improvement in the
Plan (2008–2015) [18] has brought about some improvement in the ecological environmentecological environment of
the mining area. This improvement is evident in the alleviation
of the mining area. This improvement is evident in the alleviation of soil erosion and theof soil erosion and the
restorationofofthe
restoration theecological
ecologicalenvironment,
environment,achieved achievedthrough
throughactivities
activitieslike
likecultivating
cultivatingvari-
var-
iousagricultural
ous agricultural landland classes
classes suchsuch as orchards
as orchards andand reclaimed
reclaimed vegetation.
vegetation. Nevertheless,
Nevertheless, the
the presence
presence of tailing
of tailing land,land, bare land,
bare land, sedimentation
sedimentation ponds, ponds, and other
and other land classes
land classes arisingarising
from
from mining
mining activities
activities has complicated
has complicated the surface
the surface of the mining
of the mining area. Thisarea. This complexity
complexity poses
poses challenges
challenges for effectively
for effectively monitoring
monitoring the development
the development of agriculture
of agriculture and theand the eco-
ecological
logical environment
environment within the within the area.
mining mining area.

Figure1.1. Geographical
Figure location and
Geographical location and surface
surfacecondition
conditionofofthe
theLingbei
Lingbeimining
miningarea.
area.((a)
((a) Location
Location in-
information of China on the world map; (b) Location information of Jiangxi Province
formation of China on the world map; (b) Location information of Jiangxi Province in China; in China;
(c)
Location
(c) Locationinformation ofof
information the study
the area
study areainin
Jiangxi Province.).
Jiangxi Province.).

2.2. Data Sources and Pre-Processing


The ‘Zhuhai-1 hyperspectral satellite’ (OHS) is a micro-nano satellite for remote
sensing. It is financed, constructed, and operated by China’s Zhuhai Orbit Aerospace
Science and Technology Company Limited. Currently, eight hyperspectral satellites orbit
Earth. The OHS hyperspectral satellite carries a payload of Complementary metal-oxide-
2.2.2.2.
Data Sources
Data andand
Sources Pre-Processing
Pre-Processing
TheThe ‘Zhuhai-1
‘Zhuhai-1 hyperspectral
hyperspectral satellite’ (OHS)
satellite’ (OHS) is aismicro-nano
a micro-nano satellite forfor
satellite remote
remote sens-
sens-
ing.ing.
It isIt financed,
is financed, constructed,
constructed, andandoperated
operated by byChina’s
China’s Zhuhai
Zhuhai Orbit
OrbitAerospace
Aerospace Science
Science
and Technology
and Technology Company
Company Limited.
Limited. Currently,
Currently, eight
eighthyperspectral
hyperspectral satellites
satellitesorbit Earth.
orbit Earth.
Sustainability 2024, 16, 3582 TheTheOHS hyperspectral satellite carries a payload of of
Complementary metal-oxide-semi- 5 of 21
OHS hyperspectral satellite carries a payload Complementary metal-oxide-semi-
conductor
conductor (CMOS)
(CMOS) detectors
detectors andand is capable
is capable of acquiring
of acquiring data
datawith
witha spatial resolution
a spatial resolution of of
10 10mm and and a spectral
a spectral resolution
resolution of of
2.52.5
nmnm [19], providing
[19], providing data covering
data covering 32 32
bands
bands in in
thethe
semiconductor
visible
visibleand and (CMOS)
near-infrared
near-infrared detectors
withwith and is capable
a wavelength
a wavelength rangerangeofof400–1000
acquiring
of 400–1000 data
nm nm with
[19]. aThis
This
[19]. spatial resolution
information
information
of
captures 10 m
the and a spectral
visible- to resolution
near-infrared of 2.5 nm
wavelength [19], providing
band, crucial
captures the visible- to near-infrared wavelength band, crucial for accurately identifyingdata
for covering
accurately 32 bands
identifying in the
visible
vegetation
vegetation and
andandnear-infrared
other features.
other with
features. a wavelength range of 400–1000 nm [19]. This information
captures
This Thisstudytheused
study visible-
usedthetheto near-infrared
OHS OHShyperspectral wavelength
hyperspectral datasets
datasets band, crucial
consisting
consisting offor
of clear accurately
and
clear cloudless
and identifying
cloudless L1BL1B
levelvegetation
data
level data and
acquired
acquiredother
on on features.
7 December
7 December 2021 from
2021 from thetheOHS-2A
OHS-2A satellite. These
satellite. Thesedata
datainclude
include thethe
RPC RPC This
file,file, study
spatial
spatial usedfile,
extent
extent the OHS
and
file, hyperspectral
metadata
and metadata datasets
file.file.
Processing
Processingconsisting
steps such
steps of as
clear
such and cloudless
radiometric
as radiometric cali- L1B
cali-
levelatmospheric
bration,
bration, data acquired
atmospheric on 7 December
correction,
correction, andand 2021 from the
orthometric
orthometric OHS-2Aneed
correction
correction satellite.
need to These
to be applied
be data
applied include
individu-
individu- the
RPC
allyally
to the file,
to the spatial
data.
data. extent
TheThestudy
studyfile,
area and metadata
is the
area Rare
is the Rare file.
Earth Processing
Earth Mining
Mining steps
Area
Area such as
in Lingbei, radiometric
in Lingbei, China,
China,with calibra-
a a
with
tion,extent
spatial
spatial atmospheric
extentof 1916
of 1916×correction,
1788
× 1788 and orthometric correction need to be applied individually to
pixels.
pixels.
the data. The study area is the Rare Earth Mining Area in Lingbei, China, with a spatial
extent
2.3.2.3.
Classification
Classification ×System
of 1916System 1788and pixels.
Sample
and Sample Point Selection
Point Selection
TheThe
2.3. Bigmap
Bigmap
Classification GISGIS office
Systemoffice sky map
sky
and map
Sample imageimage
Point(Bigmap
(Bigmap
Selection v25.5.0.1
v25.5.0.1 ) provides
) provides access
access to atovariety
a variety
of map
of map information,
The Bigmap GIS office sky map image (Bigmap v25.5.0.1) provides access toaares-
information, including
including raster
rastermaps,
maps,satellite
satelliteimages,
images, and and aerial
aerialimages.
images. With With a res-
variety
olution
olution
of mapof up to 0.5
ofinformation,
up to 0.5m, m, it serves
it serves
including as asvaluable
valuable
raster support
maps, support forfor
satellite identifying
identifying
images, detailed
and aerialland
detailed landclasses
images. classes
With
[20]. Combining
[20]. Combining Bigmap
Bigmap and andOHSOHS hyperspectral
hyperspectral images,
images,
a resolution of up to 0.5 m, it serves as valuable support for identifying detailed the surface
the surface of the
of study
the study areaareawas was
land
divided
divided
classesinto artificial
into
[20]. artificial
Combining surfaces
surfaces and
Bigmap natural
and natural
and OHS surfaces
surfaces and
hyperspectral andthen then thethe
images, artificial surfaces
artificial
the surface surfaces were
of the were
study
divided
divided into
area was thethe
into
dividedhuman–work
human–work
into artificial object
object class
surfaces andand
class
and thethe
natural artificial
artificial
surfaces building
building
and then class to
class
the generate
to generate
artificial a a
surfaces
classification
classification system
were dividedsystem comprising
into the comprising
human–work12 land
12 land classes
classes
object (Table
class(Table 1).
and the The image
1).artificial
The image was acquired
was acquired
building in win-
in win-
class to generate
terter
whenwhen farmers
farmers
a classification use greenhouses
use
system greenhouses
comprising covered
covered
12 landwith withplastic
classesplastic films
(Table films to cultivate
1). to cultivate
The imagecertainwascrops,
certain crops,
acquired
extending
extending the period
the period of vegetable
of vegetable production.
production. Therefore,
Therefore,
in winter when farmers use greenhouses covered with plastic films to cultivate certain the classification
the classification system
system encom-
encom-
passes
passesboth
crops, bothgreenhouse
greenhouse
extending the vegetables
vegetables
period and farmland.
and
of vegetable farmland. Given
production. Given the noticeable
the
Therefore, thedifferences
noticeable differences
classification in system
the
in the
growth
growth of reclaimed
of reclaimed vegetation
vegetation in different
in different areas,
areas,it is
it observed
is
encompasses both greenhouse vegetables and farmland. Given the noticeable differences observed that some
that somereclaimed
reclaimed veg- veg-
etation
etationis systematically
is systematically
in the growth arranged,
of reclaimed arranged, predominantly
vegetation predominantly rectangular
in different areas, rectangular in shape,
in shape,
it is observed with
thatwith consistent
someconsistent
reclaimed
andandrobust
robust
vegetation growth.
growth. In contrast,
In contrast,
is systematically otherother
arranged, areas of reclaimed
areas of reclaimed
predominantly vegetation
vegetation
rectangular lack
inlackregularitywith in
regularity
shape, dis-
in dis-
consistent
tribution,
tribution,
and robustshowing
showing uneven
growth. In growth
uneven growth
contrast, within
within
other the same
the
areas same
of reclaimed
reclaimed
reclaimed area.
area.
vegetationAdditionally,
Additionally,
lack regularity thesethesein
areas
areasoften
often
distribution,exhibit
exhibit bare
showing bareground
ground
uneven near near
growththethe
vegetation.
vegetation.
within the sameTherefore,
Therefore,
reclaimed thethereclaimed
reclaimed
area. vegetation
vegetation
Additionally, these
is categorized
isareas
categorized into two
into groups:
two groups: thosethose with better
with bettergrowth
growth and and
often exhibit bare ground near the vegetation. Therefore, the reclaimed vegetation is thosethosewith poorer
with poorer growth.
growth.
TheThe quality
categorized quality of the
into oftwo samples
the samples
groups: isthose
directly
withrelated
is directly related
better to the
to the
growth accuracy
andaccuracy
those of with
the
of thelandform
landform
poorer recog-
growth. recog-
nition in The
nition mining
in mining areas
quality of[9].
areas the Insamples
[9]. this study,
In this study,Bigmap
Bigmap
is directly is used
related tofor
is used visual
for
the visual
accuracy interpretation
interpretation
of the landformbasedbased on on
recog-
a classification
anition
classification system.
in mining system. Sample
areas Sample points
[9]. In points
this are chosen,
are
study, chosen,
Bigmap resulting
resulting
is used infor
2400
in 2400 sample
visual sample pointspoints
interpretation of fea-
ofbased
fea-
tures—specifically
on a classification
tures—specifically 200200 forfor
system. eacheachland
Sample
land class—after
points areanomalous
class—after chosen,
anomalous image
resultingimage element
in element points
2400 sample points are
points
are
of features—specifically 200 for each land class—after anomalous image element points
eliminated.
eliminated.
are eliminated.
Table
Table1. Land useuse
1. Land classification scheme
classification andand
scheme image example
image of Lingbei
example rarerare
of Lingbei earth mining
earth area
mining in in
area
Table
Jiangxi 1. Land use classification scheme and image example of Lingbei rare earth mining area in
Province.
Jiangxi Province.
Jiangxi Province.
Examples of of
Examples
Examples
Examplesof of
Examples of OHS
OHS Hyper-
Hyper-
Examples of OHS
Land Class
Land Class Explicit Explanation
Explicit Explanation Bigmap
BigmapGISGIS
Land Class Explicit Explanation Bigmap GIS Hyperspectral
spectral Im-Im-
spectral
Office
Office
Office Imagery
agery
agery
Sustainability
Sustainability 2024,
Sustainability
2024,
Sustainability 16,
2024, x FOR
16,
16, 16,
2024, xOrchard
FOR PEER
xxOrchard
FOR
PEER
FOR REVIEW
PEER REVIEW
REVIEW
PEER REVIEW 66 of 22
66 of
of of 22
22 22
Artificial
Artificialsur-sur- TheThe
Orchard orchard is in
orchard is the shape
in the of aofladder
shape with
a ladder a distinct
with a distinct
The orchard is in the shape of a ladder
face
face (Ora)
(Ora) (Ora) with
slope.a distinct slope.
slope.

Pool
Pool Pool Ponds usually have edges that are easier
Artificial surface Ponds usually
Ponds have
usually edges
have edgesthat areare
easier to to
identify.
(Poo)
(Poo) (Poo) tothat easier
identify. identify.

Road
Road Road Roads Roads long
arelong
generally long and
areare
Roads generally
generally and continuous.
and continuous.
(Roa)
(Roa) (Roa) continuous.

Building
Building Buildings areare
Buildings usually rectangular
usually and
rectangular neatly
and ar-ar-
neatly
(Bui)
(Bui) ranged.
ranged.

Farmland
Farmland It is
It composed
is composed of of
a large number
a large numberof of
irregularly
irregularly
(Far)
(Far) shaped
shaped plots closely
plots connected,
closely connected,with obvious
with obviousfeatures.
features.
Sustainability
Sustainability 2024,
2024, 16, 16,
x x FOR
FOR PEER
PEER REVIEW
REVIEW 6 of66 22
of 22
Sustainability
Sustainability
Sustainability 2024,
2024,
2024, 16,
16, 16, x FOR
xx FOR PEER
PEER REVIEW
REVIEW 66 of of
22 22
Sustainability
Sustainability
Sustainability 2024,
16,16,
2024,
2024, x FOR
16,FOR PEER
xx FOR
FOR REVIEW
PEER
PEER
PEER REVIEW
REVIEW
REVIEW 6 of
of 22
66 of
of 22
22 22
Sustainability
Sustainability
Sustainability 2024,
2024, 16,
2024, 16,
xx FOR
16, x FOR PEER
PEER
xx FOR REVIEW
REVIEW
PEER REVIEW 66 of 6 of
66 22
of 22
22
Sustainability 2024,
Sustainability
Sustainability 16,
2024,
2024, 16,
16, FOR
x PEER
FOR
FOR REVIEW
PEER
PEER
Sustainability 2024, 16, x FOR PEER REVIEWREVIEW
REVIEW 6 of6
of 2222
of
of 22
22
Sustainability 2024, 16, 3582 6 of 21
Pool Pool
Pool
Pool Pool
Pool Ponds
Ponds usually
usually havehave edges edges that thatareare easier easier to identify.
to identify.
Pool
(Poo) Pool
(Poo) Ponds
Ponds
Ponds
Ponds usually
usually
usually
usually have
have have
have edges
edges edges
edges that
that that
that are
are are
are easier
easiereasier
easier to
to to identify.
identify.
identify.
Pool
(Poo)
(Poo)
Pool
Pool
(Poo)
Pool
(Poo) Ponds
Ponds
Table 1. usually
usually
Cont. have have edges edges that thatareare easiereasier to to to identify.
identify.
identify.
(Poo)
Pool(Poo)
Pool Ponds
Ponds
Ponds usually
usually
usually have
havehave edges
edges edges that
that thatare
are are easier
easiereasier to
to to identify.
identify.
identify.
(Poo) (Poo)
(Poo) Ponds
Ponds usually
usually have have edges edges that that are are easiereasier to to identify.
identify.
(Poo)
(Poo)(Poo)
Road Road Examples of Examples of OHS
Road
Road Road Land Class
Road Roads Roads are are generally
generally long
Explicit long and and continuous.
continuous.
Explanation Bigmap GIS Hyperspectral
Road
(Roa) Road
(Roa) Roads
RoadsRoads
Roads are
are are
are generally
generally
generally
generally long
long long
long and
and and
and continuous.
continuous.
continuous.
continuous.
(Roa)
Road
(Roa)Road
(Roa)
Road
(Roa) RoadsRoads are are generally
generally long long and and continuous.
continuous. Office Imagery
Road
(Roa)
Road Road
(Roa) RoadsRoads arearegenerally
generally long long and and continuous.
continuous.
(Roa) (Roa) Roads
RoadsRoads are
are aregenerally
generally
generally long
long long and
and and continuous.
continuous.
continuous.
(Roa)
(Roa)(Roa)
Building Building
Building Buildings
Buildings areare usually
usually
Buildings rectangular
rectangular
are and
usually rectangular andneatly neatly ar-ar-
and
Building
Building
Building
Building Buildings
Buildings
Buildings
Buildings are
are are usually
usually
usually
are usually rectangular
rectangular
rectangular
rectangular and
and and
and neatly
neatlyneatly
neatly ar-
ar- ar-
Building
Building
(Bui)(Bui) (Bui) Buildings
Buildings are are usually
usually ranged. rectangular
rectangular
ranged.
neatly arranged. and and neatlyneatly ar-ar-
ar-
Building
(Bui)
Building (Bui)
Building
(Bui)
(Bui) Buildings
Buildings
Buildings are are
are usually
usually
usually rectangular
ranged.
ranged.
rectangular
ranged. rectangular
ranged. and and
andneatly neatly
neatly ar- ar-
ar-
Building
Building
(Bui)
Building (Bui) Buildings
Buildings are
Buildings are usually usually
are usually rectangular
rectangular
ranged.
ranged.
rectangular and and neatly
and neatly ar- neatly ar- ar-
(Bui)(Bui) ranged.
ranged.
(Bui)(Bui)
(Bui) ranged.
ranged.
ranged.
Farmland
Farmland It isIt is composed of a large number of irregularly
FarmlandFarmland
Farmland
Farmland
Farmland It
It Itcomposed
is
is
It is composed
composed
composed
composed
is
Itofisa
of
of of large
composed
aa large
of a large
large
a large number
number
number
of a large
number
number
of irregularly
of
of ofnumber
irregularly
irregularly
irregularly
of irregularly
of
Farmland
Farmland
(Far)(Far) It
shaped
shaped It
is is
plots composed
composed
plots of
closely
closely of
a
irregularly a
large large
connected,
connected, numbernumber
shaped
with with of
plots of irregularly
irregularly
closely
obvious
obvious features.
features.
Farmland (Far)
Farmland
(Far)
(Far) shaped
(Far)
shaped
shaped It is plots
composed
It plots
is plots
composed closely
closely of of connected,
a largea large
connected, numbernumber
withwith of obvious
irregularly
of
obvious irregularly features.
features.
Farmland (Far)
(Far)(Far)
Farmland
Farmland shaped
shaped
shaped It
It It plots
is
is composed
is plots
plots closely
composed
composed closely
of
closely
closely of connected,
of connected,
aa large
connected, connected,
connected,a large
large number
with
number with
with
numberwith
obvious
with obvious
of
of obvious
irregularly
features.
obvious
obvious
of features.
irregularly
irregularly features.
features.
features.
(Far) (Far)
(Far) shaped
shaped
shaped plotsplots
plots closely
closely
closely connected,
connected,
connected, with with
with obvious
obvious
obvious features.
features.
features.
(Far)(Far)
(Far) shaped
shaped
shaped plots
plots
plots closely
closely
closely connected,
connected,
connected, with
withwith obvious
obvious
obvious features.
features.
Sedimentation
Sedimentation
Sedimentation They
They
They are are generally
generally
are generally round round
round and and
and
rectangular
rectangular
rectangular infeatures.
in shape
shape
in shape
Sedimentation
Sedimentation
Sedimentation They
TheyThey are
are aregenerally
generally
generally round
They
round roundareand
and and rectangular
generally
rectangular round and
rectangular in
in shape
shape
in shape
Sedimentation
Sedimentation
tank tank and They
andThey
are are
are are generally
generally
densely
densely round
distributed
distributed round and and
withinwithin rectangular
rectangular
a a
given given in
area in
shape
area of shape
of the
the
Sedimentation
tank
Sedimentation tank
Sedimentation
tank tank
Sedimentation
and
They
and andThey
andare
They
are tank
are
are are
densely generally
densely
generally
are
densely generally
densely distributed
rectangular
distributed
round
distributed
distributed round
round in
and and
within
shape
within
and
within rectangular
and
rectangular
within a given
aa given
rectangular
given
a given
are densely
area
in
area inof
area
shape
in
area ofshape
of
the
shape
thethe
Sedimentation
Sedimentation
tank
(Sed) tank
(Sed)
Sedimentation andThey
They
(Sed) They
andareare aregenerally
generally
densely
densely
generally round
distributed
distributed
round
distributed round
mine. and
mine.
and and
within
rectangular
within
within rectangular
rectangular
a given a given
a given area
in
area
inof in
area of of
shape
shape
the thethe
shape
of the
(Sed)
tank
(Sed)(Sed)
tank
(Sed) and andare are densely
densely distributed
distributed mine.
mine.mine.
within
mine. within a a
given given area area of of
thethe
tank
(Sed)
tank (Sed)
tank and
and and are
are aredensely
densely
densely distributed
distributed
distributed mine. within
mine. mine.a
within
within a given
a given
given area
area area of
of the
of
thethe
(Sed)
Greenhouse
Greenhouse (Sed)
(Sed) veg-
veg- mine. mine.
mine.
(Sed)
Greenhouse
(Sed)
Greenhouse (Sed) veg-
veg- It usually occurs in mine.
mine.
springmine. and winter and is more
Greenhouse
Greenhouse veg-
veg- It usually
It usually
usually occurs
occurs in spring
in spring and and winterwinter and and is more
is more
more
Greenhouse
Greenhouse etables
etables veg-veg-ItIt usually
usually
It occurs
occurs in
in
It usually
occurs spring
spring
in spring and
and
occurs and winter
winter
in spring
winter and
and
and and is
is more
more
winter
is
Greenhouse
Greenhouse
Greenhouseetables
etables
etables veg-
veg-
veg-
Greenhouse
etables It It usually
usually
neatly
vegetables
neatly occurs
occurs
arranged
arranged inby in
spring
by spring
addingaddingand and
topswinter
tops winter
to to
farm and
farmand is
fields. is more
more
fields.
Greenhouse
Greenhouse
Greenhouse
(Gre)etables
etables veg-
veg-
(Gre)veg- (Gre) It It usually
neatly
neatly
usually
It neatly
usually occurs
arranged
arranged
occurs andinis
occurs by in
in by
more
spring spring
addingadding
spring neatly
and and
tops
and tops
winterwinter
to
arranged
winter to
farm farm
and and
by
andfields.is
adding
is more more
fields.
is more
etables
(Gre)(Gre)
etables It
It neatly
usually
It neatly
neatly
usually
usually arranged
arranged
occurs
arranged
arranged
occurs
occurs in
in by
by in by
spring
by
springadding
adding
adding and
adding
spring and tops
tops
and tops
winter
tops
winter to
to
winter farm
to
to
farm farm
and
farm
and andfields.
is
is fields.
more
fields.
fields.is
more more
(Gre)
etables
(Gre)
etables(Gre)
(Gre)
etables neatly
neatly arranged
arranged by by tops
addingadding to farm
tops tops to fields.
to
farm farm fields.
fields.
(Gre)(Gre)
(Gre) neatly
neatly
neatly arranged
arranged
arranged by
by byadding
addingadding tops
tops tops to farm
to
tosoils
farm farm fields.
fields.
fields.
(Gre)
(Gre)(Gre) Mostly
Mostly abandoned
abandoned agricultural
agricultural land,land, soilsthat thatareare notnot
Unused
Unused landland Mostly
Mostly
Mostly
Mostly abandoned
abandoned
abandoned
abandoned agricultural
agricultural
agricultural
agricultural land,
land,land,
land, soils
soilssoils
soils that
that that
that are
are are
not
not
are not
Unused
Unused
Unused
Unused land
landland
land Mostly
Mostly
planted
planted withabandoned
abandoned
with vegetation
vegetation
Mostly agricultural
agricultural butbut show
abandoned show land,
land,
signs soils
signs
agricultural soils
of that
of thatareare notnot
cultivation,
cultivation,
land, not
Unused
Unused
(Unu)(Unu) landland planted
Mostly
Mostly
planted
planted
planted with
withwith
abandoned
abandoned
with vegetation
vegetation
vegetation agricultural
vegetation agricultural
but
but but
butshow
show show
land,
show land,
signs
signs signs
soils
signs soils
of
of of
that cultivation,
that are
cultivation,
cultivation,
of not
are
cultivation, not
Unused
(Unu)
Unused (Unu)landland Mostly
planted
planted
Mostly
Mostly
Unused land abandoned
with with
abandoned
abandoned agricultural
vegetation
vegetation but
agricultural
agricultural
soils that
etc. butshow
etc.
are land,
show signs
land,
land,
not soils
signs
soils
planted of
soilsthat
of
that
withthat are
are not
cultivation,
cultivation, are
notnot
(Unu)
Unused
Unused
(Unu)
Unused (Unu)land
(Unu)
landlandplantedplanted with with vegetation
vegetation etc.
but but
etc.
show show signs signs of of cultivation,
cultivation,
cultivation,
(Unu) planted
planted
(Unu)
planted withwith
with vegetation
vegetation
vegetation
vegetation etc.
butbutbut
etc.
butshow
showshow
show signs
signs signs
signs ofof of
cultivation,
cultivation,
of cultivation,
(Unu)(Unu) planted with vegetation etc.
but etc.
show signs of cultivation,
(Unu)
Original
Original (Unu) veveta-
veveta- etc.
etc. etc. etc.
Original
Original veveta-Undeforested
veveta- Undeforested primary etc. etc. exist in large continu-
forests
Original
Original veveta-
veveta- Undeforested primary
primary forests
forests existexist in large
in large continu-
continu-
Original
Original tion tion veveta-
veveta- Undeforested
Undeforested
Undeforested primary
primary
primary forests
forests
forests exist
existexist in
in large
large
in large continu-
continu-
continu-
Original
Original tion
Original
tion tion veveta-
veveta-
tionveveta- Undeforested
Undeforested primary
primary ous ous forests
forests
areas.areas. exist
exist in in
largelarge continu-
continu-
Original
Original tion
Original
(Ori) veveta-
tion
(Ori) veveta-
veveta- Undeforested
Undeforested
OriginalUndeforested
vevetation primary
primary ous
Undeforested ous
forests areas.
forests
areas. exist
primary exist in large
in
forests large continu-
exist continu-
in
(Ori)
tion (Ori)
tion Undeforested
Undeforested primary
primary
primary ous
ous ous
ous areas.
forests areas.
areas.
forests
forests exist
areas.
existexist in
in large
in
largelarge continu-
continu-
continu-
(Ori)
(Ori)
tion
(Ori)
tion tion
(Ori) (Ori) ous ous
large areas.
continuous
areas. areas.
(Ori) ous ous
ous areas.
areas.
areas.
Reclaimed
Reclaimed(Ori)(Ori) vege-
vege- The The saplings
saplings areare regularly ous
regularly areas. arranged,
arranged, with with good good aver-
aver-
Natural
Natural sur-
sur- Reclaimed
Reclaimed
Reclaimed
Reclaimed (Ori)
(Ori) vege-
vege-
vege-
vege- The
The The
The saplings
saplings
saplings
saplings are
are are regularly
regularly
regularly
are regularly arranged,
arranged,
arranged,
arranged, with
with with
with good
good good
good aver-aver-
aver-
aver-
Natural
Natural
Natural
Natural sur-
sur-
sur-
sur- Reclaimed
Reclaimed
tation_good
tation_good vege-
vege- The
age The
age saplings
saplings
growth,
growth, andare are
andare regularly
regularly
are
planted planted inarranged,
arranged, in
an an
overall with
overallwith good good
rectangular
rectangular aver-
aver-
Natural
Natural
face sur-
face sur- Reclaimed
tation_good
tation_good
Reclaimed
Reclaimed
tation_good
tation_good vege-
vege-
vege-age
The
age The
age
The saplings
growth,
growth,
saplings
saplings
growth,
age growth, and
are
and are
and
and The
are regularly
are
regularly
are
are regularly
areplantedplanted
saplings
plantedplanted arranged,
are
in
arranged,
in in
an
arranged,
an
in an
regularly
overall
overall
an with with
overall arranged,
with
overall good good
rectangular
rectangular
good
rectangular aver-
rectangular aver-
aver-
Natural
Natural
faceface
Natural
face sur- Reclaimed
sur- Reclaimed
tation_good
tation_good
sur-Reclaimed (Rec_g)
(Rec_g) vege-
vege- age The
Reclaimed
vege- The The
age saplings
saplings
growth,
growth,
saplings are
and and
are
shape regularly
are
are
shape regularly
are
planted
regularly
in in
the planted
the arranged,
arranged,
in in
arranged,
reclaimed an an
reclaimed overallwith with
overall
area.with good
good good aver-
rectangular
area.rectangular aver-
aver-
face
Natural
Natural
Natural sur-
face
face sur-
sur- (Rec_g)
tation_good
(Rec_g)
tation_good age age growth,
growth, and
shape
and with
shape
are are
in good
in
the
planted the
planted average
reclaimed
reclaimed
in in
an angrowth, area.
overall
area.
overall and are
rectangular
rectangular
Natural
face surface(Rec_g)
(Rec_g)
tation_good vegetation_good
age growth, shape
and shape
are in the
in
planted thereclaimed
reclaimed
in an area.
overall area. rectangular
face
face
face (Rec_g)
(Rec_g)
tation_good
tation_good
face Reclaimed age age growth,
growth, shape
and shape
and are
plantedin
are in
the
planted inthe
planted
an reclaimed
reclaimed
in
overallin
an an area.
overall area.
overall
rectangular rectangular
rectangular
shape
face (Rec_g)
(Rec_g)
(Rec_g)
Reclaimed (Rec_g)
vege-
vege- There
There is nois no shape
clear shape
shape
clearpattern in the
inpattern
the
in the
of reclaimed
reclaimed
reclaimed
of saplings
saplings area. area.
inarea.
in the
the reclaimed
reclaimed
(Rec_g)
(Rec_g)
Reclaimed
(Rec_g)
Reclaimed vege-
vege- There
There is is
no noshape
shape
clear
shape
clear in
in
pattern the
in
pattern
the in
thereclaimed
the
of of
reclaimedreclaimed
reclaimed
saplings
saplings area.
area.
in area.
area.
in
the the reclaimed
reclaimed
Reclaimed
Reclaimed
Reclaimed
Reclaimed
tation_bad vege-
vege-
vege-
vege- There
There
There
There
area is
is
andno
is
is
no clear
no
no clear
clear
their clearpattern
growthpattern
pattern
pattern isof
of saplings
of
of
clearly saplings
saplings
saplings in
uneven,in the
in
in
the thereclaimed
the
with reclaimed
reclaimed
reclaimed the sur-
tation_bad
Reclaimed
tation_bad vege-area areaand
There andtheir
is no
no
theirgrowth
cleargrowth is
pattern clearly
is uneven,
of saplings
clearly saplingsuneven, with
inwiththe
with the sur-
reclaimed
the sur-
tation_bad
Reclaimed
Reclaimed
tation_bad
tation_bad vege-
vege-area
There
areaareaand
There
andisandtheir
no
is clear
their
theirgrowth
clear pattern
There
growth
growth isis
pattern is clearly
noof saplings
of
clear
clearly
is clearly uneven,
pattern
uneven, in
uneven, the
in
of the
saplings
with with the
reclaimed
reclaimed
the sur-
in
sur-
Reclaimed
Reclaimed
Reclaimedtation_bad
tation_bad
(Rec_b)
(Rec_b) vege-
vege-
vege- There
area There
area
There
Reclaimed and is
andno
is
rounding
is
rounding no clear
no
their
their clear
clear pattern
growth
growth
ground
ground pattern
pattern moreismore of
is
of saplings
of
clearly
clearly saplings
similar
saplings
similar in
uneven,
uneven,
to the
in
to the
in
bare the
with
bare with thethe
reclaimed
reclaimed
ground.
reclaimed
ground. the sur-
sur-
sur-
(Rec_b)
tation_bad
tation_bad
(Rec_b)
(Rec_b)
(Rec_b) area area rounding
and
rounding
roundingandtheir
rounding their ground
growth
ground
ground growth
the
ground is
reclaimed
more
more more
clearly
is
more similar
clearly
area
similar
similar uneven,
similar uneven,
and
to
to to
their
bare
bare
to bare
with
bare with
growthground.
ground.
ground.the
ground. thesur-
is sur-
tation_bad
(Rec_b)
(Rec_b)
tation_bad
tation_bad area
vegetation_bad
areaareaand
and andtheir
rounding
rounding their
their growth
ground
ground growth
growth is
moreis clearly
more
is clearly
clearly uneven,
similar
similar to
uneven,
uneven, to
bare with
bare
with with the
ground.
ground. the sur-
the sur-
sur-
(Rec_b) (Rec_b)
(Rec_b) rounding
rounding groundclearly
ground more uneven,
more similar withtothe
similar tosurrounding
bare bare ground.
ground.
(Rec_b)
(Rec_b)
(Rec_b) rounding
rounding
rounding ground
ground
ground more more
more similar
similar
similar to to
bare bare ground.
ground.
Bare(Rec_b)
Bare ground Land
ground rounding
Land without
without ground
any ground
any cover more
cover more
or similar
or similar
treatment
treatment tototo bare
on bare
bare on
the ground.
ground.
ground.
the surface,
surface,
Bare
Bare
Bare
Bare ground
ground
ground
ground Land
LandLand
Land without
without
without
without any
any any
any cover
cover cover
cover or
or or treatment
treatment
treatment
or treatment on
on on
the
the
on the
the surface,
surface,
surface,
surface,
BareBare ground
(Bar) ground
(Bar) LandLand without
without
generally
generally any any
located
located cover
in cover
in
the or
the or treatment
treatment
vicinity
vicinity of of on
mining onthe
mining the
areas.surface,
surface,
areas.
Bare
Bare (Bar)
Bare
(Bar) ground Land
(Bar)
ground
ground
(Bar) Land
generally
Land
generally without
generally
without located
without
generally any
located any
located
any
Land
located in
cover
in cover
in
the
cover
the
without
in the
or
the or
vicinity
or
vicinity
any treatment
vicinity
treatment
treatment
cover
vicinity of
of of
mining
on
mining
ofor on
mining
the
on the
the
treatment
mining areas.surface,
areas.
surface,
areas.surface,
on
areas.
Bare
Bare
Bare(Bar)ground
ground
(Bar) BareLand
ground LandLand
ground without
without
generally
generally
without any any
located
located
any cover cover
in in
cover theoror
the treatment
or treatment
vicinity
vicinity
treatment of of on the
on
onmining
mining thesurface,
areas.
the surface, surface,
areas.
(Bar)(Bar) generally
generally located
located
the in thethe
in
surface, vicinity
vicinity
generally of of mining
mining
located inareas.areas.
the
(Bar)
(Bar)(Bar) Note: Note:generally
(Bar) human-work
human-work
generally
generally located object
object
located
located in
inclas: the
in clas:
the vicinity
Orc
the Orcandandand
vicinity
vicinity of
Far; Far;
ofFar; miningartificial
artificial
of mining
mining areas. building
building
areas.areas. class:
class: Roa,
Roa, Bui,Bui,
Poo,Poo, Gre,
Gre, andand Sed.
Sed.
Note:
Note:Note: human-work
human-work
human-work object object
object
clas: Orc andclas: clas:
Orc
vicinity
Far; artificialOrcand Far;
of mining
building class: Roa, artificial
artificial
areas.
Bui, Poo, Gre, and building
building
Sed. class:
class: Roa,
Roa, Bui,
Bui, Poo,Poo, Gre,
Gre, andand Sed.
Sed.
Note:
Note:
Note: human-work
human-work
human-work object
object
object clas:
clas:
clas: OrcOrc
Orc
andand
and Far;
Far;
Far; artificial
artificial
artificial building
building
building class:
class:
class: Roa,
Roa,
Roa, Bui,
Bui,
Bui, Poo,
Poo,
Poo, Gre,
Gre,
Gre, andand
and Sed.
Sed.
Sed.
Note: human-work
Note: human-workobject clas:
object OrcOrc
clas: andand
Far; artificial
Far; building
artificial class:
building Roa,
class: Bui,
Roa, Poo,
Bui, Gre,
Poo, andand
Gre, Sed.
Sed.
2.4.
Note:
2.4.
Note: Research
human-work
Research
Note:
human-workMethod
object
Method
human-workobject clas:
object
clas: Orc
clas: and
OrcOrc
and Far;
and artificial
Far;
Far; building
artificial
artificial class:
building
building Roa,
class:
class: Bui,
Roa,
Roa, Poo,
Bui,
Bui, Gre,
Poo,
Poo, and
Gre,
Gre, and Sed.
and Sed.
Sed.
2.4.
2.4.2.4. Research
Research
Note: Method
Method
human-work
Research Method object clas: Orc and Far; artificial building class: Roa, Bui, Poo, Gre, and Sed.
2.4.2.4.
2.4. Research
Research
Research
Optimizing
Optimizing
Method
Method
Method the the constructed
constructed multiple
multiple feature
feature categories
categories andand their
their associated
associated feature
feature
Research
2.4.2.4.OptimizingMethod
Research
OptimizingMethod
the the constructed
constructed multiple
multiple feature
feature categories
categories andand their
their associated
associated feature
feature
2.4. Research
2.4.Optimizing
2.4.indices Method
Optimizing
Research
Researchinvolves
Optimizing
Optimizing the
Method
Method the constructed
the constructed
analysis
the constructed
constructed and multiple
multiple
screening
multiple
multiple feature
of categories
feature
various
feature
feature categories
features
categories
categories and
andandtheir
from
and associated
their associated
different
their
their feature
feature
perspectives.
associated
associated feature
feature
indices2.4.
indices involves
Research
involves
Optimizing analysis
Method
the analysis and
constructed screening
and screening
multiple of various
of various
feature features
features
categories from
andfrom different
their different perspectives.
associated perspectives.
feature
indices
indices
indices involves
Optimizing
involves
involves
Optimizing analysis
the the
analysis
analysis and
constructed
and
constructed andscreening
multiple
screening
screening
multiple of
of various
feature
various
of various
feature features
categories
features
features
categories from
and
from
andfrom different
their
different
their different perspectives.
associated
perspectives.
associated feature
perspectives.
feature
ThisThis
indices
indices process
Optimizing
process
Optimizingis is
involves
involves aimed
Optimizing aimed
the analysis
analysis
theat at obtaining
and and
constructed
obtaining
constructed
the constructed feature
screening
screening
multiple
feature
multiple of
multiple combinations
of various
various
feature
combinations
feature that
features
features
categories
that
categories
feature are
categories are
andfrom
from
and
moremore
their
their
and suitable
different
differentassociated
suitable
associated
their for
associatedfor
the the
perspectives.
perspectives. fine
feature
fine
featurefeature
ThisThis
indices
indices
Thisindices
This process
involves
process
process is
involves
is
process isanalysis
involves
aimed
aimed
is aimed
analysis
at
analysis
aimed at at
at obtaining
and and
obtaining
and
obtaining
obtaining feature
screening
screening
feature
screening
feature
feature of
ofcombinations
various
ofcombinations
various
various
combinations
combinations that
features
features
that
features
that are
that are
from
from
are more
from
more
are more
more suitable
different
different
suitable
different
suitable
suitable for
for for
the
the
for the
perspectives.
perspectives.thefine
perspectives.
finefine
fine
indices
ThisThis
indices involves
classification
process
process
classification
indices is
involves
of
involves of
is
aimedanalysis
large-scale
aimed at
analysis
large-scale
analysis at and and
land
and screening
land
obtaining
obtaining use
screening
use
screeningin in
feature
feature of various
mining
of
mining
of variousfeatures
areas,
combinations
combinations
areas,
various that
features
enhancing
features from
enhancing
that
areare
frommore
from different
accuracy
more
accuracy
different perspectives.
suitable
suitable
differentin in
for classifying
for
the
classifying
perspectives.the
perspectives. fine
fine
This indices
This
Thisprocess involves
classification
process
classification is
process ofis of
isof aimedanalysis
large-scale
large-scale
aimed aimed at at land
obtainingand
land
at obtaining
obtaining screening
useusein
feature in
feature
mining
feature of
mining various
areas,
combinations
areas,
combinations
combinations features
enhancing
that
enhancing
that are
that arefrom different
accuracy
more
accuracy
more
are more suitable
suitablein
suitable in perspectives.
classifying
for the
classifying
for the
for thefinefine
fine
classification
This
landclassification
This
landprocess
use use is
process
and
classification
classification
andandof
of large-scale
aimed
is large-scale
aimed
precision
of
precision at
large-scale
large-scale in land
obtaining
at land
obtaininguseuse
recognizing
land
land use in
feature
usein mining
in
feature mining areas,
combinations
agricultural
in mining
mining areas,
combinations enhancing
land
areas,
areas, enhancing
that are
that more
are
classes.
enhancing
enhancing A accuracy
accuracy
more in
suitable
suitable
technological
accuracy
accuracy in classifying
in
for
in classifying
the
for fine
the
flowchart
classifying
classifying fine
This
land process
This
land useuse is
process
classification
and aimed
is aimed
precision
of
precision atin
large-scale
in
recognizing
obtaining
at
in obtaining
recognizing
land
recognizing use agricultural
featurefeature
agricultural
in mining
agricultural
land
combinations
combinations
land
areas,
land
classes.
that areA
that
classes.
enhancing
classes. A
technological
more
areA suitable
more
technological
accuracy
technological inflowchart
for
suitable the
for fine
the fine
flowchart
classifying
flowchart
classification
landclassification
land use and
use
classification and
classification of
of large-scale
of
precision large-scale
precision
large-scale
of in
large-scalein landland
recognizinguseuse
recognizing
landlanduseusein
in mining
in mining
agricultural areas,
agricultural
mining
in mining areas,
land
areas, enhancing
land
areas, enhancing
classes. A
classes.
enhancing
enhancing Aaccuracy
accuracy
technological in
technological
accuracy
accuracy in classifying
in classifying
flowchart
flowchart
classifying
in classifying
is is
land illustrated
landuseuse
illustrated
classificationandand
inof
classification in
Figure
ofFigure
precision
precision 2.
large-scale in2.
in
large-scale recognizing
recognizing
landlanduseusein inagricultural
agricultural
mining
mining areas,land
land
areas, classes.
classes.
enhancing
enhancing A A technological
technological
accuracy
accuracy in
in flowchart
flowchart
classifying
classifying land
is
land
is is illustrated
landillustrated
illustrated
use use
and in
and in
Figure
precision Figure
precision 2.
in 2.
in recognizing
recognizing agricultural
agricultural landland classes.
classes. A A technological
technological flowchart
flowchart
is illustrated
land
landis
land use
isuse
use and
illustrated
illustrated
use
and
in
in
and Figure
in
in Figure
precision
FigureFigure
precision
precision
and precision
2.
inin
2.
in 2.
recognizing
2.
in recognizing
recognizing agricultural
agricultural
agricultural
recognizing agricultural land
landlandclasses. A
classes.
classes. A technological
A technological
technological flowchart
flowchart
flowchart
classes. A technological flowchart is
is illustrated
is illustratedin inFigure
Figure 2. 2.
is
is illustrated
isillustrated in
illustrated
illustrated Figure
ininin
FigureFigure
Figure 2.
2.2.2.
bility 2024,Sustainability
16, x FOR PEER REVIEW
2024, 16, 3582 7 of 22 7 of 21

Figure 2. Technology roadmap.


Figure 2. Technology roadmap.

2.4.1. Feature Declaration


2.4.1. Feature Declaration
Due to theof
Due to the fragmentation fragmentation
the surface and of the
thecomplexity
surface andofthe the complexity
land classesof in the
ion-land classes in
adsorbing rare earth mining areas, the use of only traditional spectral features is not effec- features is not
ion-adsorbing rare earth mining areas, the use of only traditional spectral
effective
tive for the detailed for the detailed
classification classification
of features and land of class
features and land class
identification. identification.
Wanwan Yu et Wanwan Yu
et al.that
al. [21,22] showed [21,22] showed of
the addition that the addition
red-edge featuresof red-edge
can improve features can improve
the separability be-the separability
tween differentbetween different
land classes, whichland classes, which
is promising is promising
for detailed for detailed
land class land class
identification and identification
and mapping.
mapping. Therefore, Therefore,
32 raw spectral 32 rawfrom
reflectances spectral reflectances
hyperspectral from were
images hyperspectral
selected images were
selected as input parameters to construct 12 vegetation
as input parameters to construct 12 vegetation indices and 6 red-edge indices. Pei Huan indices and 6 red-edge indices.
et al. [23] proposed that the participation of texture information from remote sensing im- from remote
Pei Huan et al. [23] proposed that the participation of texture information
sensingcan
ages for classification images
yield for classification
a better can effect
classification yield than
a better
onlyclassification
using spectral effect
infor- than only using
spectral information and, therefore, in this study, eight
mation and, therefore, in this study, eight kinds of texture indices based on the grayscale kinds of texture indices based on
the were
covariance matrix grayscale covariance
constructed. matrix were
In summary, fourconstructed.
feature classesIn summary,
were selectedfourin feature
this classes were
selected in this study: spectral, vegetation, red
study: spectral, vegetation, red edge, and texture features, as shown in Table 2. edge, and texture features, as shown in
Table 2.
The results of the four feature categories above were combined sequentially to form
The resultsofofexperimental
seven different combinations the four feature categories
schemes (Tableabove wereschemes
3). These combined weresequentially
de- to form
seven different combinations of experimental schemes
signed to investigate how the accuracy of land use classification in rare earth mining areas (Table 3). These schemes were
designed to investigate how the accuracy of land use
changes when incorporating vegetation features, red-edge features, and texture features classification in rare earth mining
areas changes
using the RF classification when incorporating
algorithm. From them, we vegetation
can determine features, red-edge
the single features,
feature cat- and texture
features using the RF classification algorithm. From them, we can determine the single
egory with the most significant effect on improving the accuracy of land-use fine classifi-
feature category with the most significant effect on improving the accuracy of land-use
cation on a large scale in mining areas and identify agricultural land classes, including
fine classification on a large scale in mining areas and identify agricultural land classes,
farmland. In addition, these schemes allow us to intuitively explore whether the proposed
including farmland. In addition, these schemes allow us to intuitively explore whether
feature selection method effectively enhances the accuracy of land-use fine classification
the proposed feature selection method effectively enhances the accuracy of land-use fine
and agricultural land identification. The objective was to compare the schemes to identify
classification and agricultural land identification. The objective was to compare the schemes
the optimal feature combinations suitable for the detailed classification of large-scale land
to identify the optimal feature combinations suitable for the detailed classification of large-
use and the identification of agricultural land classes in rare earth mining areas.
scale land use and the identification of agricultural land classes in rare earth mining areas.
Table 2. Description of various characteristics.
Table 2. Description of various characteristics.
re Varia- Index Abbrevi-
Feature Variable OHS Image
Index Calculation Formula
Abbreviation Exponential Description
OHS Image Calculation Formula Exponential Description
ble ation
Spectral feature Band B1, B2, . . ., B32
tral fea-
Band B1, B2, …, B32 Suitable for areas with high density
ure GCVI b27/b7 − 1
vegetation cover.
Suitable
√ for areas with high density vegetation
GCVIfeature
Vegetation b27
RDVI b7 − 1 (b27 − b13)/ b27 + b13
It can be used for high and low
vegetation coverage.
etation cover.
h  i
b16 It is very sensitive to changes in
ature TCARI 3 × (b16 − b14) It can× be
(b16used
− b6)for
× high and low vegetation cover-
(b27EVI
− b13)
− 0.2
RDVI b 27 + b13   b14 chlorophyll content.
2.5 × b27−b13
b27+6×b13−7.5×b1+1
age. More sensitive to high vegetation coverage.
Sustainability 2024, 16, 3582 8 of 21

Table 2. Cont.

Feature Variable Index Abbreviation OHS Image Calculation Formula Exponential Description
Characterize vegetation coverage and growth
NDVI (b27 − b13)/(b27 + b13)
and health status.
Affected by chlorophyll and leaf tissue
TVI 0.5 × [120 × (b27 − b7) − 200 × (b13 − b7)] abundance, the difference between vegetation
was obvious.
b27−b13 It contains soil regulation coefficient and is
SAVI b27+b13+0.5 × (1 + 0.5)
more suitable for low vegetation cover area.
  q The index is based on an assessment of several
b27
MSR
b13 − 1 / b27
b13 + 1
vegetation indices derived from a combination
of two spectral bands.
It is used to estimate and measure vegetation
RVI b27/b13 biomass and is sensitive to high
vegetation coverage.
There was significant correlation with
gNDVI (b27 − b7)/(b27 + b7)
chlorophyll content and leaf area index.
It was responsive to chlorophyll concentration
MACRI [(b27 − b13) − 0.2 × (b27 − b7)] × (b27/b13)
and background reflectance of leaves.
It is sensitive to the change in soil background,
DVI b27 − b13 and the sensitivity to vegetation decreases
when the vegetation coverage is high.
It can be used to estimate leaf area index and
NDRE1 (b19 − b16)/(b19 + b16)
chlorophyll content of plants.
It can be used in fine agriculture, vegetation
NDRE2 (b22 − b17)/(b22 + b17)
stress detection, and so on.
√ Replace the near infrared band in MSR with a
MSRred (b27/b16) − 1/ b27/b16 + 1
valley with a red edge.
Red edge feature It is sensitive to chlorophyll content in
MTCI b20 − b17/b17 − b15
plant leaves.
It is more sensitive to the chlorophyll content
2.5×(b23−b14)−1.3×(b23−b6)
MCARI_red 1.5 × q √ in plants and the higher the value, the higher
(2×b23+1)2 −(6×b23−5× b14)−0.5 the chlorophyll content.
It is correlated with chlorophyll content and
leaf area index of plant canopy and can
IRECI (b22 − b14)/(b16/b19)
quantitatively characterize chlorophyll content
of plant.
Mean
Variance(Var)
Homogeneity(Hom) Calculated based on the first four principal
Contrast(Con) component bands after the original spectral
Texture feature Dissimilarity(Dis) principal component analysis, using window
Entropy(Ent) size: 5 × 5.
Second Moment(Sec)
Correlation(Cor)

Table 3. Combination scheme information.

Combination Scheme Specific Feature Information of the Scheme


Scheme 1 Spectral feature(Spe)(RF)
Scheme 2 Spectral feature+Vegetation feature(Spe+Veg)(RF)
Scheme 3 Spectral feature+Red edge feature(Spe+Red)(RF)
Scheme 4 Spectral feature+Texture feature(Spe+Tex)(RF)
Scheme 5 Spectral feature+Vegetation feature+Red edge feature+Texture feature(Spe+Veg+Red+Tex)(RF)
Scheme 6 Spectral feature+Feature importance ranking combination(Spe+Fea)(RF)
Scheme 7 Spectral feature+Vegetation feature+Red edge feature+Texture feature(Spe+Veg+Red+Tex)(SVM)

2.4.2. Feature Optimization Methods


Although the expansion and combination of feature variables can provide better
classification results, using all the features for the classification may lead to information
redundancy, resulting in a “curse of dimensionality”, which could impact the classification
performance [24]. Therefore, using the same feature class as a benchmark and performing
feature optimization on it can yield feature variables that are more favorable for fine-
grained feature classification in mining areas with various types of features. In order to
Sustainability 2024, 16, 3582 9 of 21

select feature indices with a higher separability of features and at the same time avoid
covariance among indices, we utilized the J-M (Jeffreis-Matusita) distance and principal
component analysis for feature variable optimization in the different feature categories.
The J-M distance provides superior results among the many methods for separability
determination [25,26]. It can determine the variability between categories based on the
degree of separation, with values ranging from 0 to 2—the larger the value, the better the
separation. For the training sample set E (i, j = 1, 2,. . ., N, i ̸= j) in two to-be-separated land
classes wi and wj , assuming that mi and mj are the feature means of wi and wj and τ i and
τ j are the feature standard deviations of wi and wj , the mathematical expression of the J-M
distance is as follows:  
J = 2 1 − e− B (1)
2 2
1 2 2 1 τi + τj
B= mi − m j + ln (2)
8 τi2 + τj2 2 2τi τj

where J denotes the J-M distance and B denotes the Bachmann distance.
To address variations in differentiation among land classes, we further process the J-M
distance by introducing weighting. These weights are determined based on the criterion
that the more indistinguishable the combination of land classes, the larger the weights,
and vice versa. In this study, we initially categorize land use in mining areas into two
major groups: highly distinguishable artificial and natural land surfaces (Table 1). These
two major categories are then assigned weight values equal to 1. In addition, the artificial
land surface category is divided into two subcategories: artificial objects and artificial
construction. These subcategories are assigned a weight value of 2, particularly for clearly
distinguishable land classes such as farmland and buildings. Finally, weights between land
classes within the same type of land use are assigned values of 3 or 4, determined by the
degree of similarity between features. The weights for specific landform combinations
are presented in Table 4; those not listed carry a weight of 1. We used the weighted J-M
distance to analyze vegetation, red edge, and texture features. Feature variables exhibiting
a strong differentiation of landforms are identified as the preferred characteristics.

Table 4. Weight allocation of land class combination.

Land Land Land


Weights Weights Weights
Assemblage Assemblage Assemblage
Rec_b, Far 2 Rec_g, Far 3 Gre, Roa 2
Rec_b, Rec_g 4 Far, Bar 2 Gre, Sed 2
Rec_b, Bar 4 Far, Ora 3 Bar, Ora 2
Rec_b, Ora 2 Far, Unu 4 Bar, Roa 4
Rec_b, Unu 3 Far, Ori 2 Bar, Unu 3
Rec_b, Ori 2 Rec_g, Bar 2 Ora, Unu 2
Bui, Gre 4 Rec_g, Ora 3 Ora, Ori 2
Bui, Roa 2 Rec_g, Unu 2 Poo, Sed 3
Bui, Sed 2 Rec_g, Ori 4 Roa, Sed 2
Unu, Ori 2

Principal component analysis (PCA) is a powerful technique for dimensionality reduction


and feature selection. With PCA, a low-dimensional feature space composed of projection
directions can better reflect the spatial structure information of the original high-dimensional
data [27,28] and facilitate the interdependence and correlation analysis of features [29]. As
many as 32 original spectral bands are provided by OHS hyperspectral datasets. Moreover,
using PCA for dimensionality reduction effectively addresses the “dimensionality catastrophe”
problem while substantially preserving the original band information. Covariance between
feature variables can lead to the inclusion of features that do not enhance classification accuracy
because of the diverse features used in this study. Thus, in the current study, correlation tests for
vegetation, red edge, and texture features were separately conducted using principal component
analysis. Feature variables with minimal covariance were prioritized for the classification of
land use to mitigate the adverse effects of covariance.
Sustainability 2024, 16, 3582 10 of 21

2.4.3. Importance Ranking of Features


The ranking of features by importance can help to filter out features with high con-
tributions to classification, regression, and other problems from multiple features in a
short time, improve calculation speed, and reduce the dimensions of data input [30]. This
study ranked the importance of all feature indices, encompassing vegetation, red edge, and
texture characteristics. This ranking provides an overall assessment of the contribution
of each feature type to classification. It allows for an intuitive identification of feature
categories and indices suitable for detailed feature classification in mining areas. The
Relief-F algorithm, a classic multivariate filtering feature selection method used in various
classification problems, assigns weights to features based on their relevance to the landform
class. Features weighing less than a specified threshold are rejected [31]. The algorithm is
computed as follows: given class labels, there are l classes C = [C1 , C2 ,. . ., Cl ], a sample Ri is
randomly selected from the training set, then k approximate samples of Ri are identified
in the same class, denoted as Hj (j = 1,2,. . ., k), and k approximate samples of Ri are also
identified from different categories, denoted as Mj (C) (j = 1,2,. . ., k), which repeats the
above steps on each feature dimension and obtains each feature weight as follows:
k di f f ( A, Ri , Hj ) k di f f ( A, R , M (C ))
P(C ) i j
W ( A) = W ( A) − ∑ m∗k
+ ∑ [ × ∑
1 − P(class( Ri )) j=1 m∗k
(3)
j =1 C ̸=class( R )
i

where A denotes the feature, W(A) denotes the weight of feature A, m is the number of
iterations, and P(C) denotes the probability that the category is C. diff (A, Ri , Hj ) and diff (A,
Ri ,Mj (C)) denote the difference between the samples Ri and Hj , Mj (C), respectively, with
respect to A, which is defined as follows:

R1 [ A]− R2 [ A]

 max( A)−min( A)
, i f A is continuous
di f f ( A, R1 , R2 ) = 0, i f ( A is discrete) and ( R1 [ A] = R2 [ A]) (4)

1, i f ( A is discrete) and ( R1 [ A] ̸= R2 [ A])

where R2 stands for Hj , Mj (C).

2.4.4. Land Use Classification


The RF algorithm facilitates the integration of high-dimensional data from multiple
sources, rendering it suitable for complex datasets [32]. It exhibits high adaptability to
overfitting in training samples and accomplishes classification tasks within a brief duration.
The results of a study on the classification of Mediterranean shrub vegetation types in
the Languedoc region of southern France showed that the RF algorithm outperforms
traditional classifiers in the identification of hyperspectral vegetation types for automatic
classification and is especially advantageous for distinguishing vegetation types with small
spectral differences [33]. The input classification data of the study included spectral features,
vegetation features, red edge features, and texture features, which involve more complex
data with higher dimensions. The study area constitutes various land classes with smaller
spectral differences. Therefore, the RF algorithm is effective and offers certain advantages
for the detailed classification of land use in mining areas. RF adopts Bootstrap resampling
to build a decision tree model for each of the k samples selected from the training sample
set. After each tree is classified, the final classification result is determined by voting [34].
k
H ( x ) = argmax∑i=1 I (hi ( x ) = Y ) (5)
y

where H(x) denotes the combination of classification models, hi denotes the decision tree
classification model, Y denotes the output variable, and I(◦ ) is the indicator function.
2.4.5. Accuracy Verification
In order to compare the effects of incorporating different features on the accuracy
of feature classification, this study evaluates the overall accuracy by using the confusion
Sustainability 2024, 16, 3582 11 of 21

matrix. The confusion matrix can help evaluate the accuracy by comparing the degree of
confusion between the classification results and the actual measurements and includes the
overall accuracy (OA), Kappa coefficient, producer’s accuracy (PA), and user’s accuracy
(UA). Among them, the overall accuracy and Kappa coefficient are used to evaluate the
overall performance of the classifier and the producer’s accuracy and user’s accuracy are
used to evaluate the misclassification and omission indicators of a particular class.

∑iK=1 Nii
OA = (6)
N

N ∑iK=1 Nii − ∑iK=1 Ni+ N+i


Kappa = (7)
N 2 − ∑iK=1 Ni+ N+i
Nii
PAi = (8)
N+i
Nii
U Ai = (9)
Ni+
where N represents the total number of samples; K represents the total number of categories;
Nii represents the number of samples assigned to the correct category; and N+i and Ni+
represent the true number of samples in category i and the number of samples predicted to
be in category i, respectively.
Sustainability 2024, 16, x FOR PEER REVIEW 12 of 22

3. Results
3.3.1. Feature Optimization Results
Results
3.1.1.
3.1. Determination
Feature of Spectral Features
Optimization Results
3.1.1. Determination of Spectral
Spectral features wereFeatures
calculated and downscaled using principal component analysis. The
Spectral features
contribution of eachwere calculatedcomponent
principal and downscaled wasusing
usedprincipal componentto
as a benchmark analy-
rank them cumulatively,
sis. The contribution of each principal component was used as a benchmark to rank them
as shown in Figure 3, where the horizontal coordinates indicate the new spectral bands generated
cumulatively, as shown in Figure 3, where the horizontal coordinates indicate the new
by thebands
spectral principal component
generated analysis
by the principal of the original
component analysisspectral features
of the original andfea-
spectral the vertical coordinates
tures and the vertical coordinates indicate the cumulative contribution of the new spectralinformation. Among
indicate the cumulative contribution of the new spectral bands to the image
bands
them, to the
theimage information.
cumulative Among them,
contribution ratetheofcumulative contribution
the first principal rate of the first
component to the fourth principal
principal component to the fourth principal component increases more in relative order
component increases more in relative order and, from the fourth principal component onwards,
and, from the fourth principal component onwards, the growth rate of the contribution
thedecreases
rate growthand ratetheof the contribution
contribution rate valuerate decreases
stabilizes and Therefore,
above 0.997. the contribution
this studyrate value stabilizes
above
finally 0.997.
selects theTherefore, this study
first four principal finally selects
components the first
as the result of thefour
finalprincipal components as the result
spectral feature
of the final spectral feature preference.
preference.

Figure 3. Cumulative
Figure contribution
3. Cumulative rate of principal
contribution components.
rate of principal components.
3.1.2. Determination of Vegetation Features
3.1.2. Determination of Vegetation Features
The 12 vegetation indices were subjected to weighted J-M distance calculations and
The
principal 12 vegetation
component indices
analysis and were
the results aresubjected to weighted
shown in Figures J-M 4distance
4 and 5. Figure shows calculations and
principal
that component
most of the vegetation analysis and
indices have thedivisibility
good results are shown and
of features in Figures
some of 4 and 5. Figure 4 shows
them
have
thatoutstanding
most of thedivisibility of specific
vegetation features
indices havebut good
that the divisibility ability
divisibility of TVI, and some of them
of features
DVI, RDVI, and TCARI for the land classes is not ideal, which is in line with the fact that
although they are sensitive to the content of chlorophyll and so on, most of the land classes
are construction land, mining land, etc. [35]. Figure 5 shows that the correlation coefficient
between NDVI and SAVI is close to 1, which is a strong correlation between the two, and
therefore, NDVI was preferred considering that it has a better ability to characterize the
surface reflectance of the features than SAVI [36]. MCARI was not considered because it
had a strong correlation with most of the indices. Therefore, the vegetation characteristics
Sustainability 2024, 16, 3582 12 of 21

have outstanding divisibility of specific features but that the divisibility ability of TVI,
DVI, RDVI, and TCARI for the land classes is not ideal, which is in line with the fact that
although they are sensitive to the content of chlorophyll and so on, most of the land classes
are construction land, mining land, etc. [35]. Figure 5 shows that the correlation coefficient
between NDVI and SAVI is close to 1, which is a strong correlation between the two, and
therefore, NDVI was preferred considering that it has a better ability to characterize the
surface reflectance of the features than SAVI [36]. MCARI was not considered because it
had a strong correlation with most of the indices. Therefore, the vegetation characteristics
were finally filtered to GCVI, EVI, NDVI, MSR, RVI, and gNDVI.

3.1.3. Determination of Red Edge Features


The red edge features were also feature optimized using weighted J-M distance and
principal component analysis, as shown in Figures 4 and 5. The figures show that both
NDRE1 and NDRE2 have low separability for most of the land classes and that the degree
of correlation with the other indices is high. Therefore, the final selected red-edge features
Sustainability 2024, 16, x FOR PEER REVIEW 13 of 22
are MSRred, MTCI, MCARI_Red, and IRECI.

3.1.4. Determination of Texture Features


3.1.4. Determination of Texture Features
The method of feature preference for texture features is the same as that for vegetation
The method of feature preference for texture features is the same as that for vegeta-
tion featuresand
features rededge
and red edge features,
features, as shownas shown
in Figuresin4 and
Figures 4 and
5. Figure 5. Figure
4 shows that the 4 shows that the
separability
separability of coroffor
cor forclass
each each class is obviously
is obviously extremely lowextremely low and of
and the separability the
homseparability of hom
for each
for class,
each although
class, higher than
although cor, than
higher is still cor,
downstream compared to thecompared
is still downstream other indices.
to the other indices.
Figure
Figure5 shows
5 showsthat sec,
thatent, andent,
sec, homandhavehoma high correlation,
have a highcon and dis havecon
correlation, a high
and dis have a high
correlation, and con and sec are more divisible when compared to each other and, there-
correlation, and con and sec are more divisible when compared to each other and, therefore,
fore, the preferred texture features were finally determined to be mean, var, con, and sec.
the preferred texture features were finally determined to be mean, var, con, and sec.

Figure 4. JM
Figure 4. distance weightsweights
JM distance and detailand
curves of each
detail feature
curves ofoneach
different ground
feature on classes. (“a”ground classes. (“a”
different
shows the index curve with significantly worse JM weighted value in the vegetation index; “b” is
shows
the the index
exponential curve
curve with with significantly
significantly worse JM
worse JM weighting weighted
value in the redvalue in the“c”vegetation
edge index; shows index; “b” is the
the exponential curve
exponential curve with significantly
with poor JM
significantly weighting
worse in the texturevalue
JM weighting index.).
in the red edge index; “c” shows the
exponential curve with significantly poor JM weighting in the texture index.).
Sustainability 2024, 16, x FOR PEER REVIEW 14 of
Sustainability 2024, 16, 3582 13 of 21

Figure 5. The correlation between the features of heat maps and high values of specific parts.
Figure 5. The correlation between the features of heat maps and high values of specific parts.

3.2. Importance3.2.
Ranking of Characteristic
Importance Variables
Ranking of Characteristic Variables
The Relief-F algorithm wasalgorithm
The Relief-F used to evaluate
was usedand sort the importance
to evaluate and sort the of all the feature
importance of all the featu
variables. Thevariables.
results are
The results are shown in Figure 6: texture index mean highest
shown in Figure 6: texture index mean had the score score a
had the highest
and was muchwas higher
muchthan the other
higher than indexes,
the otherreaching
indexes,0.1200. The0.1200.
reaching vegetation index SAVI,index SA
The vegetation
gNDVI, and NDVI decreased successively. NDRE1, the red index, secured
gNDVI, and NDVI decreased successively. NDRE1, the red index, securedthe fifth position.
the fifth po
The vegetationtion.
indexThe vegetation index in the upper and middle reaches of the scoreother
in the upper and middle reaches of the score was higher and the was higher a
texture indexesthe
except
otherthe meanindexes
texture were mostly
except in
thethe middle
mean werereaches
mostlyofinthe
thescore.
middle The indexof the sco
reaches
of the downstream
The index of the downstream region is the red edge index and vegetationred
region is the red edge index and vegetation index and half of the index and h
edge index is in this region. After comprehensive consideration, the importance of feature
tively small portion and, therefore, the red edge features failed to add constructively to
the distinction between land classes. Additionally, the mining area encompasses various
features such as farmland, greenhouse vegetables, sedimentation tanks, and bare ground,
where texture information is more pronounced. Particularly, the texture features not only
Sustainability 2024, 16, 3582 demonstrate a certain advantage in the detailed identification of plastic greenhouses 14 of 21 [38]
but also improve the classification and identification accuracy of texture information-rich
land classes [24], providing a distinct advantage in the identification of land classes within
variables was ranked
mining areas. This isinconsistent
descendingwith
order as texture
the follows:index
texture feature, vegetation
“MEAN” feature,
being ranked first and
and red edge feature.
being significantly higher than the other indexes.

Figure 6.6. Rank of the


Figure the importance
importanceof ofdifferent
differentfeatures
features(the
(thebar
barindicates
indicates
thethe magnitude
magnitude of of
thethe im-
portance score
importance scorefor
foreach
each feature index).
feature index).

Among the
3.3. Accuracy three types
Evaluation of feature
of Land variables selected,
Use Classification in Miningthe Area
feature importance of the
red-edge feature ranked last, which is related to the actual land class categories in the study
This study adopts overall accuracy, Kappa coefficient, producer’s accuracy, and
area: the most different spectral characteristics of green vegetation and other features is the
user’s accuracy
significant increaseasinthe evaluation
reflectance indexes
in the red-edgeof band
the land
[37].use classification
However, results of mining
the image-capturing
areaswas
time andDecember,
compares whenthe classification
some crops results of the seven
were processed intoschemes;
greenhouse the vegetables
classification
to accu-
ensure their continued growth and the land classes in the complex parcels in mining areas8, and
racy and classification results of the schemes are shown in Table 5, Figures 7 and
the confusion
mostly consistedmatrix of the scheme
of buildings, with the
bare ground, best classification
sedimentation results
tanks, and otherisconstruction
shown in Figure
9. TheThe
land. overall classification
vegetation land classaccuracy is the land
in the overall lowest when
class only using
accounted the spectral
for a relatively features
small
portion and, therefore,
after principal componentthe red edge features
analysis of the failed
original to add constructively
spectrum to the distinction
for the extraction of rare earth
between land classes;
mining area classes. Additionally, the mining
the overall accuracy andarea encompasses
Kappa coefficient various features
are only suchand
83.59% as 0.51,
farmland, greenhouse vegetables, sedimentation tanks, and bare ground,
respectively. Schemes 2, 3, and 4, respectively, include the vegetation features, red edgewhere texture
information is more pronounced. Particularly, the texture features not only demonstrate a
features, and texture features after feature optimization on the basis of the spectral
certain advantage in the detailed identification of plastic greenhouses [38] but also improve
the classification and identification accuracy of texture information-rich land classes [24],
providing a distinct advantage in the identification of land classes within mining areas.
This is consistent with the texture index “MEAN” being ranked first and being significantly
higher than the other indexes.

3.3. Accuracy Evaluation of Land Use Classification in Mining Area


This study adopts overall accuracy, Kappa coefficient, producer’s accuracy, and user’s
accuracy as the evaluation indexes of the land use classification results of mining areas and
compares the classification results of the seven schemes; the classification accuracy and
classification results of the schemes are shown in Table 5, Figures 7 and 8, and the confusion
matrix of the scheme with the best classification results is shown in Figure 9. The overall
classification accuracy is the lowest when only using the spectral features after principal
component analysis of the original spectrum for the extraction of rare earth mining area
classes; the overall accuracy and Kappa coefficient are only 83.59% and 0.51, respectively.
Schemes 2, 3, and 4, respectively, include the vegetation features, red edge features, and
features and the classification accuracy increases compared with that of Scheme 1, which
is specifically represented by the overall accuracy increase of 1.97%, 0.93%, and 3.09% re-
Sustainability 2024, 16, 3582 spectively, indicating that the inclusion of vegetation features, red edge features, and tex-
15 of 21
ture features can effectively improve the classification accuracy of land use in rare earth
mining areas; here, the texture features have the greatest effect on the improvement in
accuracy. Scheme 5 integrates the spectral features and feature indices after feature selec-
texture features after feature optimization on the basis of the spectral features and the
tion from each
classification feature increases
accuracy category and the overall
compared withaccuracy and Kappa
that of Scheme coefficient
1, which are again
is specifically
improved significantly. The overall accuracy of Scheme
represented by the overall accuracy increase of 1.97%, 0.93%, and 3.09% respectively, 5 is the highest among all
schemes, being improved by 4.84%, and the Kappa coefficient is improved
indicating that the inclusion of vegetation features, red edge features, and texture features by 0.11. Scheme
7 uses
can SVM toimprove
effectively classify thethesame combination
classification of features
accuracy of landanduseitsinaccuracy
rare earthismining
slightlyareas;
lower
here, the texture features have the greatest effect on the improvement in accuracy. Scheme as
than that of Scheme 5, which indicates that RF works better than other classifiers such 5
SVM for the detailed classification of large-scale land use in mining
integrates the spectral features and feature indices after feature selection from each feature areas; after adopting
the importance
category and therankingoverall of featureand
accuracy variables
Kappa for the classification
coefficient are again ofimproved
surface cover in mining
significantly.
The overall accuracy of Scheme 5 is the highest among all schemes, being improvedcoef-
areas (Scheme 6), the total feature preference results in overall accuracy and Kappa by
ficientand
4.84%, values the that
Kappa arecoefficient
essentiallyisequivalent
improvedto bythose
0.11. of Scheme
Scheme 5. The
7 uses SVM user
toaccuracies
classify theof
farmland
same and greenhouse
combination vegetables
of features and its are greatlyisimproved
accuracy in Scheme
slightly lower than 4, which
that correlates
of Scheme 5,
with the
which strong that
indicates texture information
RF works better of these
than twoclassifiers
other types of agricultural
such as SVMfeatures. The user
for the detailed
accuracies ofoforchards
classification large-scaleandland
reclaimed vegetation
use in mining areafter
areas; better in Scheme
adopting the 3, indicatingranking
importance that the
red-edge features are able to differentiate and recognize these
of feature variables for the classification of surface cover in mining areas (Scheme 6), two types of agricultural
land
the classes
total feature from the complex
preference surface
results environment.
in overall accuracy A and
comprehensive analysis
Kappa coefficient of thethat
values user
accuracy of the above four agricultural land classes in different
are essentially equivalent to those of Scheme 5. The user accuracies of farmland and programs shows that both
Program 5 vegetables
greenhouse and Program 6, which
are greatly are based
improved in on the RF
Scheme algorithm,
4, which are able
correlates withtothe
recognize
strong
them relatively
texture information effectively;
of thesethe twouser accuracy
types of Program
of agricultural 5 is higher
features. Thethan
userthat of Program
accuracies of
6 in general.
orchards and reclaimed vegetation are better in Scheme 3, indicating that the red-edge
Overall,
features are able thetofeature preference
differentiate method used
and recognize thesebytwothetypes
Institute not only effectively
of agricultural land classesim-
proves
from thethe accuracy
complex of the
surface detailed classification
environment. of land analysis
A comprehensive use in mining
of the areas but also of
user accuracy ac-
the aboveextracts
curately four agricultural
agriculturalland classes
land classesin different programs orchards,
such as farmland, shows thatreclaimed
both Programvegeta- 5
and
tion,Program 6, which are
and greenhouse based onfrom
vegetables the RF algorithm,
complex landare able to
classes, recognize
thus themthe
supporting relatively
under-
effectively;
standing and the monitoring
user accuracy of Program 5development
of agricultural is higher than that of
within Program
mining 6 in general.
areas.

Figure7.7.Results
Figure Resultsofofland
landuse
useclassification
classificationunder
underdifferent
differentschemes.
schemes.
Sustainability2024,
Sustainability 2024,16,
16,3582
x FOR PEER REVIEW 17 of 21
16 of 22
Sustainability 2024, 16, x FOR PEER REVIEW 17 of 22

Figure8.8.Comparison
Figure Comparisonofoffeature-specific
feature-specificclassifications
classificationsofofland
landuse
useclassifications
classificationsunder
underScheme
Scheme5.5.
Figure 8. Comparison of feature-specific classifications of land use classifications under Scheme 5.

Figure 9. Scheme 5 confusion matrix for land use classification. (The red numbers represent the
Figure 9.ofScheme
number 5 confusion
pixels 5correctly matrix for
classified land useuseclassification. (The red numbers representand
the
Figure 9. Scheme confusion matrixforforeach
landland classification
use classification. in red
(The Scheme 5. Commission
numbers represent the
number
Omission of pixels correctly classified for each land use classification in Scheme 5. Commission and
number of represent
Omission
the misclassification
pixels correctly andland
classified for each missed
useclassification
classificationerrors for each
in Scheme land-use category
5. Commission and
in Scheme represent the misclassification
5, respectively, and indicating
with lighter colors missed classification errors
smaller errors.). for each land-use category
Omission
in Schemerepresent the misclassification
5, respectively, and missed
with lighter colors classification
indicating errors for each land-use category
smaller errors.).
in Scheme 5, respectively, with lighter colors indicating smaller errors.).
Table 5. Precision statistics of the classification results of each scheme.
Table 5. Precision statistics of the classification results of each scheme.
Classifications Overall, the feature
Scheme preference
1 Scheme methodScheme
2 Scheme3 used by4 the Institute
Scheme not only
5 Scheme 6 effectively
Scheme 7
Classifications improves theScheme 1 of
accuracy Scheme
the 2 Scheme3
detailed Scheme
classification of 4 Scheme
land use in 5 Scheme
mining 6 but
areas Scheme 7
also ac-
Overall Accuracy 83.59% 85.56% 84.52% 86.68% 88.43% 88.26% 88.16%
Overall Accuracy curately extracts
83.59% 85.56%
agricultural 84.52%
land classes 86.68%
such as farmland,88.43% 88.26% vegetation,
orchards, reclaimed 88.16%
Kappa coefficient 0.51 0.55 0.53 0.58 0.62 0.59 0.61
Kappa coefficient and greenhouse vegetables0.55
0.51 from complex0.53 land classes,
0.58 thus supporting
0.62 the
0.59understanding
0.61
and PA%
monitoring 82.06
of 81.29 development
agricultural 83.59 82.36mining85.74
within areas. 80.21 84.97
Buildings PA% 82.06 81.29 83.59 82.36 85.74 80.21 84.97
Buildings UA% 57.47 63.70 52.40 63.25 58.84 66.88 56.07
UA% 57.47 63.70 52.40 63.25 58.84 66.88 56.07
PA% 25.48 26.44 25.00 35.58 38.94 17.79 31.73
Sedimentation tank PA% 25.48 26.44 25.00 35.58 38.94 17.79 31.73
Sedimentation tank UA% 12.83 15.71 15.25 16.37 24.92 14.23 21.36
UA% 12.83 15.71 15.25 16.37 24.92 14.23 21.36
Pool PA% 82.82 86.50 68.71 74.85 61.96 68.71 67.48
Pool PA% 82.82 86.50 68.71 74.85 61.96 68.71 67.48
Sustainability 2024, 16, 3582 17 of 21

Table 5. Precision statistics of the classification results of each scheme.

Classifications Scheme 1 Scheme 2 Scheme 3 Scheme 4 Scheme 5 Scheme 6 Scheme 7


Overall Accuracy 83.59% 85.56% 84.52% 86.68% 88.43% 88.26% 88.16%
Kappa coefficient 0.51 0.55 0.53 0.58 0.62 0.59 0.61
PA% 82.06 81.29 83.59 82.36 85.74 80.21 84.97
Buildings
UA% 57.47 63.70 52.40 63.25 58.84 66.88 56.07
Sedimentation PA% 25.48 26.44 25.00 35.58 38.94 17.79 31.73
tank UA% 12.83 15.71 15.25 16.37 24.92 14.23 21.36
PA% 82.82 86.50 68.71 74.85 61.96 68.71 67.48
Pool
UA% 13.53 15.06 13.49 17.40 21.40 14.18 22.40
PA% 41.29 41.94 40.65 56.13 50.97 34.19 49.68
Bare ground
UA% 4.26 4.93 3.96 7.95 7.47 12.77 7.59
PA% 52.04 51.68 58.54 65.51 66.98 54.52 64.86
Farmland
UA% 83.27 83.33 87.24 89.22 91.01 78.96 91.12
PA% 100.00 100.00 100.00 100.00 100.00 91.67 100.00
Road
UA% 3.46 3.54 4.08 3.66 3.58 1.96 3.79
PA% 28.38 25.73 27.32 25.20 28.91 4.24 28.65
Orchard
UA% 12.74 13.31 19.54 11.63 17.17 3.02 15.91
Original PA% 89.33 91.41 89.93 91.82 93.71 95.20 93.57
vevetation UA% 99.24 99.17 99.24 99.21 99.22 99.29 99.19
Reclaimed PA% 37.19 32.46 38.99 37.03 36.22 15.50 36.54
vegetation_good UA% 19.81 24.81 17.92 28.13 29.40 20.79 28.75
Reclaimed PA% 27.45 27.12 33.99 29.41 32.03 24.35 30.72
vegetation_bad UA% 24.21 25.74 33.66 30.20 36.57 22.11 36.22
Greenhouse PA% 50.90 61.44 52.33 60.68 61.16 60.11 60.78
vegetables UA% 70.81 73.69 70.73 66.91 74.62 72.01 70.25
PA% 50.00 55.71 55.71 67.14 62.86 85.71 67.14
Unused land
UA% 12.46 10.66 13.88 16.97 14.47 12.55 15.31
Note: PA: producer accuracy; UA: user accuracy.

4. Discussion
Hyperspectral datasets have multiple narrow bands and contain a large amount of
information, enabling more accurate and detailed feature variables to be obtained. Based
on the OHS hyperspectral datasets, four kinds of feature datasets were constructed: spec-
tral, vegetation, red edge, and texture features. J-M distance calculation and principal
component analysis were carried out for the indices of the different feature categories.
This process aimed to uncover the separability of feature indices to land classes and assess
the covariance between feature indices within the same land classes. This approach was
employed to achieve a more rational and detailed feature preference. The Relief-F algorithm
was used to rank the importance of features, excluding spectral features, and a more precise
and detailed feature preference was obtained from the overall features. This approach
provided feature preference results from an overall perspective, enabling an analysis to
identify the most effective feature categories for the detailed classification of land use in
mining areas. Based on the above analysis method, seven different combinations of feature
variables were designed and the effects of different indices on classification accuracy in
different schemes were analyzed using an RF algorithm to explore the feature preference
combinations applicable to land use classification in mining areas and for agricultural
land class identification. The study results indicate that after feature optimization, all
features of each category attain varying degrees of improvement in classification accuracy
(Table 5). (1) Texture features exert the most pronounced effect on overall accuracy across all
categories, consistent with their high significance in feature importance analysis. Notably,
Sustainability 2024, 16, 3582 18 of 21

incorporating texture features leads to significant enhancements in classification accuracy


for farmland, greenhouse vegetables, and other texture-rich land classes. Conversely, red-
edge features exert the least influence on overall; however, they more effectively recognize
agricultural land classes, such as farmland and reclaimed vegetation within complex land
parcels, compared with other indices. This observation aligns with the recognized issue
of severe fragmentation in farmland and reclaimed vegetation arising from uncontrolled
mining in mining areas. Integrating red-edge features effectively reduces landscape frag-
mentation and improves recognition accuracy [39]. Both types of indices contribute to
the accurate identification of agricultural land classes, facilitating precise monitoring of
agricultural development. (2) Among all schemes, combining the index results of each
category features yields the highest classification accuracy. It can also comprehensively
meet the requirements for improving the identification accuracy of different types of agri-
cultural land. These findings suggest that the feature selection method used in the current
study can provide more appropriate feature optimization results. The approach not only
enables fine-scale land use classification in mining areas on a large scale but also ensures
accurate recognition of different agricultural land classes within mining areas. This method
provides technical support for monitoring agricultural development in mining areas and
offers effective recommendations for further agricultural and ecological sustainability.
In summary, (1) hyperspectral datasets hold promise for the detailed land use classifi-
cation of large-scale complex landforms and (2) the feature selection method is effective in
reducing feature dimensions and improving classification accuracy. Hyperspectral images
can contain hundreds of subdivided spectral bands at consecutive wavelengths, yielding
rich spectral information and enabling more detailed detection and identification of target
features [40]. Many studies have been conducted to utilize hyperspectral datasets for classi-
fication, such as Zhao Peng et al. [41] who fused texture features and spectral features and
combined these with hyperspectral images to identify timber species; their classification
accuracy was higher than that of traditional methods. Fei Xing et al. [8] used a hybrid
analysis method of multiple end-element spectral analysis and extracted alpine grassland
cover using hyperspectral images. Binge Cui et al. [10] classified small scenes of coastal
wetlands using hyperspectral images and achieved robust results. Although previous
studies using hyperspectral datasets have achieved better extraction or classification results,
the research objects were mostly single species or small scenes. There is limited research
on the detailed classification of large-scale complex surface areas. To address this gap,
this study employs in-orbit hyperspectral images for detailed land use classification in
mining areas. Notably, it demonstrates the robustness of the classification accuracy remains,
regardless of the inclusion of features beyond spectral features. In order to further verify
the advantages of hyperspectral imagery for complex surface classification, we compare the
results of this study with those conducted for other similar mining areas. Zhang Chengye
et al. [42] used HF2 imagery to classify the surface of mining areas, dividing it into eight
categories, of which the overall accuracy was only 80.10% even though the classification
accuracy of the restoration and management areas, open-pit quarries, and water bodies
were as high as 95.00%, 85.10%, and 72.5%, respectively. Although the spatial resolution of
the OHS hyperspectral datasets used in this study is only 10 m, which is not better than that
of the Gaofen-2 image with a spatial resolution of better than 1 m, the spectral resolution
is as high as 2.5 nn and, therefore, it contains more detailed spectral information of the
landforms. Furthermore, the degree of spectral differentiation of the complex landforms is
stronger, so the classification precision is obviously higher and the overall classification
precision obtained using spectral indices alone is already 83.59%. The overall accuracy of
classification using only the spectral index is 83.59%. The use of hyperspectral datasets for
land use classification on a large scale in mining areas can improve accuracy, facilitating the
extraction of necessary information from complex surfaces to meet research requirements.
Due to the large number of bands contained in hyperspectral images and the various
feature variables that can be obtained from them, it is highly likely that the direct use of all
these features for classification will result in dimensional redundancy, which will have an
Sustainability 2024, 16, 3582 19 of 21

impact on classification accuracy [43]. To address this situation, many researchers have used
different feature preference methods for dimensionality reduction, such as Mahdianpari
et al. [44] who used the J-M distance to quantitatively analyze the separability of different
types of wetlands under different features; they then classified the land classes after feature
selection based on the separability results, which resulted in robust classification accuracy.
Md Rashedul et al. [45] combined MMF and mRMR methods for the dimensionality reduc-
tion in feature variables. Hao Yufeng et al. [46] and Fushuyu et al. [47] performed variable
optimization for wetland classification through the Relief-F algorithm and RFE algorithm,
respectively. Xiang Songyang et al. [30] selected features involved in hyperspectral image
classification by constructing the ReliefF-RFE feature selection algorithm. Although all of
the above studies used feature selection methods to derive effective preferred features, most
of them only considered land class separability or feature relevance individually. In this
study, we comprehensively considered the separability of land classes and the covariance
between features, not only analyzing each class of features individually for feature selection
but also mining the feature indices of different feature classes that are applicable to the
detailed classification of mining areas. We then ranked the importance of feature class
indices other than the spectral features to analyze the most important feature classes in
terms of the detailed classification of a wide range of mining areas. By designing a multi-
program format and using RF algorithms to explore the effects of the optimal combination
of feature indices after the above multi-faceted analysis, we effectively explored the optimal
feature classes and feature combination programs for detailed and large-scale land use
classification in mining areas and the identification of agricultural land classes such as
farmland. The study results indicate that the proposed feature preference method enhances
both the fine classification of land use and the identification of agricultural land classes.
This technique aids in monitoring agricultural development in mining areas and advancing
sustainable agriculture and ecological practices.

5. Conclusions
The primary aim of this research is to leverage hyperspectral imagery in constructing
diverse feature sets and employ a range of feature selection methods to pinpoint the
most suitable combination of features for rare earth mining areas. This endeavor seeks
to facilitate the refined classification of land use within expansive and intricate mining
territories, enabling the accurate identification of agricultural land types like farmland,
orchards, and reclaimed vegetation from fragmented data. Such efforts are instrumental in
advancing agricultural development monitoring, guiding sustainable agricultural practices,
and fostering ecological sustainability. The results show that
(1) The feature preference method combining J-M distance and principal component
analysis provides a better feature combination scheme from the perspective of land class
separability and feature correlation. It improves the accuracy of land use classification in
mining areas and agricultural land class identification significantly;
(2) Ranking the importance of the other three categories of features, excluding spectral
features, shows that the degree of importance of features in the land use classification of
rare earth mining areas is as follows: texture index, vegetation index, and red edge index;
(3) Combining the indices of the four features after feature optimization yielded the
most significant improvement in land use classification accuracy for mining areas, reaching
88.43%, with a Kappa coefficient of 0.61. This performance not only surpasses the overall
classification accuracy achieved using a single feature but also simultaneously results in an
effective increase in classification accuracy for specific farmland classes such as farmland,
orchards, reclaimed vegetation, and greenhouse vegetables.
Using hyperspectral datasets and feature selection methods, this study has led to
enhanced accuracy in fine land surface coverage classification in mining areas and the
identification of agricultural land classes. It lends technical support for monitoring agri-
cultural conditions in mining areas and promoting sustainable development. However,
the extent to which these results can be applied to other mining areas remains uncertain.
Sustainability 2024, 16, 3582 20 of 21

This uncertainty arises from the limitations imposed by the study area’s environment, the
classification of land classes, the time period, and other factors. Therefore, in the future,
more study areas can be added to further explore the universality of the research methods
and results.

Author Contributions: Conceptualization, methodology, and review and editing, H.L.; Experiment
construction, method implementation, software, and writing—original draft, C.L.; result calibration,
Y.Z.; Investigation and data curation, X.W. All authors have read and agreed to the published version
of the manuscript.
Funding: This research was funded by the National Natural Science Foundation of China, grant
number 42161057”, and the Jiangxi Province Graduate Innovation Special Fund Project, grant number
YC2023-S615.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: All data in this article can be obtained by reasonably contacting the
corresponding author.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Shi, X.Y.; Chen, H.W. Study on the pollution path and remediation of ionized rare earth ores in waste pond leaching reactor. Chin.
J. Rare Earth Sci. 2019, 37, 409–417. (In Chinese)
2. Zou, G.L.; Wu, Y.D.; Cai, S.J. Effect of ionic rare earth leaching process on resources and Environment. Nonferrous Met. Sci. Eng.
2014, 5, 100–106. (In Chinese)
3. Li, Y.; Li, H.; Xu, F. Spatiotemporal changes in desertified land in rare earth mining areas under different disturbance conditions.
Environ. Sci. Pollut. Res. 2021, 28, 30323–30334. [CrossRef]
4. Zhang, Y.H.; Zhao, X. Land-Use and Land-Cover Change Detection Using Dynamic Time Warping–Based Time Series Clustering
Method. Can. J. Remote Sens. 2020, 46, 67–83. [CrossRef]
5. Xiao, Y.; Jiang, Q.G.; Wang, B.; Li, Y.H.; Liu, S.; Cui, C. Object-oriented land use classification based on mixed feature selection of
Relief F and PSO. Trans. Chin. Soc. Agric. Eng. 2016, 32, 211–216. (In Chinese)
6. Zhang, Z.M.; Liu, B.; Guan, X. Land Use Type Change in the Upper Yellow River Basin Coal Mining Area and Its Impact on
Carbon Sequestration Services. Environ. Sci. Res. 2024, 37. (In Chinese) [CrossRef]
7. Xu, J.X.; Chen, C.; Zhou, S.T.; Hu, W.M.; Zhang, W. Land use classification in mine-agriculture compound area based on
multi-feature random forest: A case study of Peixian. Front. Sustain. Food Syst. 2024, 7, 1335292. [CrossRef]
8. Zheng, Y.; Wang, Y.; Liu, Y. Vegetation classification and recognition of mountain meadow based on seasonal characteristics of
hyperspectral data. Spectrosc. Spectr. Anal. 2022, 42, 1939–1947. (In Chinese)
9. Tu, C.R.; Li, P.; Li, Z.H.; Wang, H.J.; Yin, S.W.; Li, D.H.; Zhu, Q.T.; Chang, M.X.; Liu, J.; Wang, G.Y. Synergetic Classification of
Coastal Wetlands over the Yellow River Delta with GF-3 Full-Polarization SAR and Zhuhai-1 OHS Hyperspectral Remote Sensing.
Remote Sens. 2021, 13, 4444. [CrossRef]
10. Xing, F.; An, R.; Guo, X.L.; Shen, X.J.; Soubry, I.; Wang, B.L.; Mu, Y.M.; Huang, X.L. Mapping Alpine Grassland Fraction Coverage
Using Zhuhai-1 OHS Imagery in the Three River Headwaters Region, China. Remote Sens. 2023, 15, 2289. [CrossRef]
11. Qin, P.; Cai, Y.L.; Wang, X.L. Small Waterbody Extraction With Improved U-Net Using Zhuhai-1 Hyperspectral Remote Sensing
Images. IEEE Geosci. Remote Sens. Lett. 2022, 19, 1–5. [CrossRef]
12. Zhou, G.L.; Ni, Z.Y.; Zhao, Y.B.; Luan, J.W. Identification of Bamboo Species Based on Extreme Gradient Boosting (XGBoost)
Using Zhuhai-1 Orbita Hyperspectral Remote Sensing Imagery. Sensors 2022, 22, 5434. [CrossRef]
13. Cui, B.G.; Li, X.H.; Wu, J.; Ren, G.B.; Lu, Y. Tiny-Scene Embedding Network for Coastal Wetland Mapping Using Zhuhai-1
Hyperspectral Images. IEEE Geosci. Remote Sens. Lett. 2022, 19, 1–5. [CrossRef]
14. Mo, Y.; Zhong, R.F.; Cao, S.S. Orbita hyperspectral satellite image for land cover classification using random forest classifier.
J. Appl. Remote Sens. 2021, 15, 014519. [CrossRef]
15. Wang, Z.H.; Zhao, Z.; Yin, C.L. Fine Crop Classification Based on UAV Hyperspectral Images and Random Forest. ISPRS Int. J.
Geo-Inf. 2022, 11, 252. [CrossRef]
16. Somnath, P.; Nikhil, R.D.; Mukunda, D.B.; Bimal, K.B.; Jadunandan, D. Species-level classification of mangrove forest using
AVIRIS-NG hyperspectral imagery. Remote Sens. Lett. 2023, 14, 522–533.
17. Qin, H.M.; Zhou, W.Q.; Yao, Y.; Yao, W.M. Individual tree segmentation and tree species classification in subtropical broadleaf forests
using UAV-based LiDAR, hyperspectral, and ultrahigh-resolution RGB data. Remote Sens. Environ. 2022, 280, 113143. [CrossRef]
18. Liu, J.F.; Hu, X. Overall Planning of Mineral Resources in Jiangxi Province 2008–2015; Jiangxi Science and Technology Press: Nanchang,
China, 2011.
Sustainability 2024, 16, 3582 21 of 21

19. Zhuhai Orbita Aerospace Micro Technology Co., Ltd. Available online: https://www.myorbita.net/index.aspx (accessed on 21
February 2024).
20. BIGMAP. Available online: http://www.bigemap.com/ (accessed on 21 February 2024).
21. Yu, W.W.; Zhao, P.; Xu, K.J.; Zhao, Y.J.; Shen, P.J.; Ma, J.J. Evaluation of red-edge features for identifying subtropical tree species
based on Sentinel-2 and Gaofen-6 time series. Int. J. Remote Sens. 2022, 43, 3003–3027. [CrossRef]
22. Guo, Y.C.; Ren, H.R. Remote sensing monitoring of maize and paddy rice planting area using GF-6 WFV red edge features.
Comput. Electron. Agric. 2023, 207, 107714. [CrossRef]
23. Pei, H.; Sun, T.J.; Wang, X.Y. Object-oriented land use/cover classification based on Landsat 8 OLI image texture. Trans. Chin. Soc.
Agric. Eng. 2018, 34, 248–255. (In Chinese)
24. Wang, L.J.; Kong, Y.R.; Yang, X.D.; Xu, Y.; Liang, L.; Wang, S.G. Land use classification of agricultural areas based on feature
optimization random forest algorithm. Trans. Chin. Soc. Agric. Eng. 2020, 36, 244–250. (In Chinese)
25. Wang, C.X.; Ma, N.; Ming, Y.F.; Wang, Q.; Xia, J.F. Classification of hyperspectral imagery with a 3D convolutional neural network
and J-M distance. Adv. Space Res. 2019, 64, 886–899. [CrossRef]
26. Sun, Z.P.; Shen, W.M.; Wei, B.; Liu, X.M.; Su, W.; Zhang, C.; Yang, J.Y. Object-oriented land cover classification using HJ-1 remote
sensing imagery. Sci. China Ser. D Earth Sci. 2010, 53, 34–44. [CrossRef]
27. Zhang, X.; Chen, B.Z.; Zhao, H.; Fan, H.D.; Zhu, D. Soil Moisture Retrieval over a Semiarid Area by Means of PCA Dimensionality
Reduction. Can. J. Remote Sens. 2016, 42, 136–144. [CrossRef]
28. Wang, X.F.; Lu, C.H.; Li, J.X.; Liu, J. Generalized cosine two-dimensional principal component analysis. Acta Autom. Sin. 2022, 48,
2836–2851. (In Chinese)
29. Jason, Y.; Du, X.R. An enhanced water index in extracting water bodies from Landsat TM imagery. Ann. GIS 2017, 23, 141–148.
30. Xiang, S.Y.; Xu, Z.H.; Zhang, Y.W.; Zhang, Q.; Zhou, X.; Yu, H.; Li, B.; Li, Y.F. Construction and application of ReliEF-RFE feature
Selection Algorithm for hyperspectral image classification. Spectrosc. Spectr. Anal. 2022, 42, 3283–3290. (In Chinese)
31. Yu, X. A feature selection approach based on NSGA-II with ReliefF. Appl. Soft Comput. 2023, 134, 109987.
32. Li, H.K.; Wang, L.J.; Xiao, S.S. Land use random forest classification in southern hilly region based on multi-source data.
Trans. Chin. Soc. Agric. Eng. 2021, 37, 244–251. (In Chinese)
33. Sabat-Tomala, A.; Raczko, E.; Zagajewski, B. Comparison of support vector machine and random forest algorithms for invasive
and expansive species classification using airborne hyperspectral data. Remote Sens. 2020, 12, 516. [CrossRef]
34. Xi, Y.; Ren, C.; Wang, Z.; Wei, S.; Bai, J.; Zhang, B.; Xiang, H.; Chen, L. Mapping tree species composition using OHS-1
hyperspectral data and deep learning algorithms in Changbai mountains, Northeast China. Forests 2019, 10, 818. [CrossRef]
35. Meng, Q.Y.; Dong, H.; Qin, Q.M.; Wang, J.L.; Zhao, J.H. A plant cover index MTCARI for monitoring vegetation chlorophyll
content based on hyperspectral remote sensing. Spectrosc. Spectr. Anal. 2012, 8, 2218–2222. (In Chinese)
36. Kodimalar, T.; Vidhya, R.; Eswar, R. Land surface emissivity retrieval from multiple vegetation indices: A comparative study over
India. Remote Sens. Lett. 2020, 11, 176–185. [CrossRef]
37. Tian, X.Y.; Zang, Y.H.; Liu, R.; Wei, J.J. Study on multi-temporal Sentinel-2 large-scale winter wheat extraction considering
vegetation red edge information. J. Remote Sens. 2022, 26, 1988–2000. (In Chinese)
38. Wu, J.Y.; Liu, X.L.; Bo, Y.C.; Shi, Z.T.; Fu, Z. Identification of plastic greenhouses based on GF-2 data combined with multi-texture
features. Trans. Chin. Soc. Agric. Eng. 2019, 35, 173–183. (In Chinese)
39. Liu, J.; Wang, L.; Teng, F.; Yang, L.; Gao, J.; Yao, B.; Yang, F. Impact of red-edge waveband of RapidEye satellite on estimation
accuracy of crop planting area. Nongye Gongcheng Xuebao/Trans. Chin. Soc. Agric. Eng. 2016, 32, 140–148.
40. Hong, D.F.; Wu, X.; Ghamisi, P.; Chanussot, J.; Yokoya, N.; Zhu, X.X. Invariant Attribute Profiles: A Spatial-Frequency Joint
Feature Extractor for Hyperspectral Image Classification. IEEE Trans. Geosci. Remote Sens. 2020, 58, 3791–3808. [CrossRef]
41. Zhao, P.; Han, J.C.; Wang, C.K. Classification of wood species by hyperspectral microscopic imaging based on I-BGLAM texture
and spectral fusion. Spectrosc. Spectr. Anal. 2021, 41, 599–605. (In Chinese)
42. Zhang, C.Y.; Li, F.Y.; Li, J.; Xing, J.J.; Yang, J.Z.; Guo, J.T.; Du, S.H. Land use classification of open pit coal mine based on
DeepLabv3+ and GF-2 high-resolution images. Coal Geol. Explor. 2022, 50, 94–103. (In Chinese)
43. Rashedul, I.; Boshir, A.; Ali, H. Feature reduction of hyperspectral image for classification. J. Spat. Sci. 2022, 67, 331–335.
44. Masoud, M.; Bahram, S.; Fariba, M. The First Wetland Inventory Map of Newfoundland at a Spatial Resolution of 10 m Using
Sentinel-1 and Sentinel-2 Data on the Google Earth Engine Cloud Computing Platform. Remote Sens. 2019, 11, 43.
45. Islam, M.R.; Siddiqa, A.; Ibn, A.M.; Uddin, M.P.; Ulhaq, A. Hyperspectral Image Classification via Information Theoretic
Dimension Reduction. Remote Sens. 2023, 15, 1147. [CrossRef]
46. Hao, Y.F.; Man, W.D.; Wang, J.H.; Liu, M.Y.; Zhang, K. Wetland information extraction based on Relief-F algorithm and decision
tree method. J. Liaoning Tech. Univ. (Nat. Sci. Ed.) 2021, 40, 225–233. (In Chinese)
47. Xie, S.Y.; Fu, B.L.; Li, Y.; Liu, Z.L.; Zuo, P.P.; Lan, F.W.; He, H.C.; Fan, D.L. Research on classification method of swamp wetland in
Honghe National Nature Reserve based on multi-dimensional remote sensing image. Wetl. Sci. 2021, 19, 1–16. (In Chinese)

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

You might also like