Edge Computing and IoT Technology for Enhancing Crop Yield
Revanasiddappa Bandi
Department of CSE, BNM Institute of Technology, Bengaluru, Karnataka, India -560070.
Corresponding author E-mail address: sidduaitcse@[Link]
*
Abstract
This research focuses on predicting the yield of crops according to various factors,
including the type of crop and climatic conditions such as humidity, temperature, and
pH levels. Data from the land is gathered using a device that has sensors
and is uploaded to a remote server. The sensory information is retrieved from the
server to the integrated development environment (IDE) to forecast crop yield
employing advanced machine learning techniques. The construction of the
model utilizes a Raspberry Pi connected to an analog to digital converter
MPC3008, which allows the connection of soil moisture and pH sensors, while
the temperature sensor can be directly linked to the Raspberry Pi. Data retrieved from
these sensors is uploaded to a public application program interface (API). In
this research, a flask server is utilized to fulfill the API interface requirements. The
same data is then transmitted to an IDE where it is combined with
existing information to predict crop yield in a specific
area. Python serves as the programming language, and Spyder is used as the IDE.
Introduction
In India, agriculture serves as the main source of income for the majority of the
population. It has been observed that there has been little progress in crop
development within the last decade. The decreasing rate of crops is a significant
factor behind the ongoing increase in food prices. Since 2010, this has led to over 40
million people slipping into poverty. Multiple factors contribute to this issue, such as
improper fertilizer usage, low soil fertility, diseases, climate change, and
water wastage. Effective agricultural intervention is essential, and the integration of
IoT with sensor networks and machine learning algorithms provides a
solution. It plays a critical role in enhancing agricultural
practices and has the potential to transform agricultural progress.
The Internet of Things comprises three tiers. These consist of the network layer,
application layer, and perception layer. The sensors and devices in the perception
layer collect data in real time, which is then processed for further analysis
and predictions. Internet of Things (IoT) network layer serves as its infrastructure to
provide comprehensive service. It indicates how the application layer and perception
layer interact. The application layer integrates industry-specific technologies with the
Internet of Things. The internet of things is being utilized across various industries,
including smart parking, smart buildings, smart agriculture, smart transportation, and
many others. Among these sectors, agriculture is one of the major
industries impacting millions of individuals.
K. A. Patil and Prof. N. R. Kale [1] introduced a thoughtful agricultural model that
combines information and communication technology (ICT). ICT has consistently
played a significant role in the agriculture sector. Though village farmers may
have cultivated the same crop for generations, over time, climatic trends, soil
conditions, and outbreaks of pests and diseases have evolved. Utilizing the suggested
method, the updated information received enables farmers to adapt to and even reap
benefits from these changes. Due to the highly localized nature of agricultural
knowledge, especially distinct conditions, providing such information is a demanding
task. Comprehensive real-time and historical environmental
data is anticipated to facilitate the efficient management and use of resources.
Sally Jo Cunningham and Geoffrey Holmes [2] argued that the subjective attributes
for mushroom grading may not hold practical value,
and therefore, they might not need to be measured or recorded. Standards based on the
attributes outlined in the J4. 8 models may be beneficial in forming more
objective criteria for quality classification and pricing in the mushroom market. For
some data mining applications, the objective might be to use a model in a predictive
manner to automate the classification of new instances. In such scenarios, the learning
component will probably constitute a minor aspect of a much larger
software architecture. Since WEKA learning schemes can be accessed from
other applications, a learning module can be integrated into a larger system
with minimal extra programming.
Li Minbo, et al. [3] put forward an information service system for agriculture
IoT. This system offers superior management and utilization of the multiple sourced,
real-time, and vast data stored in distributed IoT-IS servers compared to other IoT
information systems. At the same time, every agricultural product is clearly identified
and decoded to allow for querying the information of agriculture IoT using this
system. All perception data is directly tied to business data and agricultural products,
facilitating management and query execution. The tracking and tracing of
the entire agricultural production process is made possible through the DS server and
distributed IoT-IS servers. An information discovery system composed of ONS, DS,
and IoT-IS servers is developed to enable
the capturing, standardization, management, rapid locating, and convenient querying
of all business data regarding agriculture production. A business event
model for agriculture IoT is established for unified management. Furthermore, this
system can enhance productivity and reinforce management for production,
processing, and circulation enterprises. At the same time, consumers can utilize this
system to inquire about product information to confirm the authenticity and quality
of agricultural products.
Mohanraj I et al. [4] described how farmers require assistance during various
phases of crop development, and timely guidance should be provided. Farmers
are experiencing significant difficulties economically, socially, and
politically. Numerous challenges within the agriculture sector have been
recognized, and a framework was established to address the aforementioned
issues. The Knowledgebase is organized with diverse crop information that pertains
to knowledge gathering, information flow, and various inputs such as market
availability, geospatial data, and
weather forecasting [10]. Monitoring encompasses modules
like reminders, tracking plant growth through different stages, irrigation planning,
crop profitability calculation, disaster assessment, and problem detection. The
Evapotranspiration method is utilized to determine a plant's daily water requirements
with the assistance of a developed algorithm. A
comparative analysis was conducted between various existing applications and the
current developed system, considering several factors such as the knowledge base,
monitoring modules, effectiveness, and dependability.
Nikesh Gondchawar and Prof. Dr. R. S. Kawitkar [5] described that the sensors and
microcontrollers of all three Nodes have been successfully integrated with the
raspberry pi, enabling wireless communication among various Nodes. All
experimental tests and observations
confirm that the project offers a comprehensive solution to challenges related to field
activities, irrigation issues, and storage difficulties through a remote-controlled robot,
smart irrigation system, and a smart warehouse management system respectively. The
application of such a system in the field can contribute to enhancing crop yield and
overall production.
Ciprian-Radu RAD et al. [6] indicated that the Potato crop for upcoming generations
will necessitate sustainable practices, processes, and systems (economically,
environmentally, and socially). In this framework, this paper proposed a precision
agricultural management integrated system architecture for monitoring the vegetation
condition of potato crops based on CPS architecture and design
technologies. The suggested system enables farmers
to track the progression of specific parameters of interest and make
informed decisions to boost agricultural productivity. Implementing such a
system presents challenges and demands familiarity with potato crop management
strategies, processing, and information technology visualization to be feasible in
practice. Ultimately, the concept articulated in this paper will serve as a foundation
for other researchers in the precision agriculture domain.
Chandhini. K [7] introduced an IoT-based agricultural production
system that has emerged from the enduring desire of farmers to maintain the
productivity of their land in the future. It also addresses the community's expectations
and concerns regarding safe food and environmental safeguarding. An agricultural
production system utilizing IoT technology and implemented as GUI visualization
software was developed. The IoT-based agricultural production system, through
correlation analysis between crop statistical data and agricultural environmental
data, has improved the capabilities of farmers, researchers, and government
officials in analyzing current conditions
and forecasting future harvests. Furthermore, the quality of agricultural products can
be enhanced as farmers observe the entire cycle from seeding to selling with this IoT-
based agricultural production system. The production system can be refined to
support a wider variety of products and deliver additional services. By leveraging IoT
technology, the efficiency of agricultural production can achieve substantial
improvements. With ongoing
advancements, Agriculture IoT should be positioned to usher
agricultural production into a new era.
G. V. Satyanarayana and S. D. Mazaruddin [8] outlined the particular
use that requires the soil environmental monitoring system and assessed the issues
present in the monitoring system. We developed and implemented a wireless sensor
network grounded in the soil temperature humidity monitoring system. The system
can swiftly achieve automatic networking and simultaneous data acquisition,
transmission, and display. It features low cost, low power
consumption, flexible networking, no cabling, and a user-friendly interface. By
employing GPRS technologies and Web Services technology, we can facilitate data
networking and remote monitoring, demonstrating that the system satisfies the soil
environmental monitoring requirements for temperature and humidity, along with
centralized management. In the future, we can track the groundwater level using a
suitable sensor. Moreover, this research could be expanded to create a system capable
of monitoring weather conditions across a larger region, such as a state or nation.
Anup Managave et al. [9] discussed the implementation of Wireless Sensor Network
(WSN) in Precision Agriculture (PA). Farmers can conveniently utilize the
system due to its straightforward user interface. Additionally, the system will keep
farmers alerted about every moment that takes place in the field, enabling them
to take more informed actions. The central system consists of the Raspberry Pi, which
is connected to the Arduino UNO ATmega328P/ATmega16U2 (Development Board)
and the clients to a router via a WiFi module. The router serves to access web pages
through the user interface by the clients.
Carlos Andrés González Amarillo et al. [12] introduced the design of a greenhouse
traceability model for seedling recordkeeping and monitoring utilizing IoT. A web
application has been created to manage and oversee the greenhouse, and farmers
can leverage the application for information regarding seedlings or the harvest of
products cultivated within the greenhouse. By employing this system, water usage can
be reduced.
B. Manjula Josephine et al. [13] described the prediction of millet crop
yield performed using an RF classifier, achieving an accuracy
of 99. 74%. Various inputs such as soil
moisture, maximum temperature, minimum temperature, humidity, and rainfall were
utilized for the predictions.
Meeradevi et al. [14] reported on crop yield prediction employing the linear
regression technique. Data is generated through various sensors,
including temperature, nitrogen, phosphorous, and potassium (NPK), pH, and
humidity from the actual field, and subsequent data analysis is conducted. The data
will be recorded in a dedicated system and gathered via a server for system access.
Ashok Tatapudi and P Suresh Varma [15] explored the extensive use of remote
sensing systems in machine learning. The system uses sensors such as pH,
temperature, moisture, rainfall, and humidity, and it analyzes the sensor
data with algorithms like decision trees, linear regression,
gradient boosting (GDBoost), and RF. Among these algorithms,
GDBoost provides the most accurate yield prediction [17]. This system assists farmers
in determining which crops to plant based on prevailing environmental conditions.
P. Kanaga Priya and Dr. N. Yuvaraj [16] explained that the system utilizes humidity,
temperature, soil moisture level, and pH
gathered from various sensors through IoT. Along with these, other factors such
as land type, soil type, and area sown are taken into account during the construction of
the system. These measurements are supplied to the graphical user interface
(GUI) designed for crop prediction and offer recommendations via crop suggestion
UI. A deep neural network (DNN) was employed for predicting the
crop grown, expected yield, and fertilizer applied.
Chhaya Narvekar, et al. [18] detailed that the implementation of IoT and WSN
has transformed how we gather and apply information about plants. Sensors are
capable of monitoring and forecasting crop and weather conditions. By
utilizing convolutional neural networks and datasets, potato blight
disease detection was achieved, which allows for automatic irrigation when the soil
moisture level decreases and facilitates remote monitoring of crops. By enhancing
agricultural practices and offering affordable, reliable farming strategies, farmers
will gain advantages, and the agricultural sector will experience growth.
Chiara Bersani, et al. [19] explained that the incorporation of IoT solutions in
agriculture has resulted in the creation of smart greenhouses. These greenhouses make
use of sensors and communication systems to observe and gather data in real-
time, leading to effective management of light, humidity, temperature, CO2
concentrations, and ventilation. This innovation brings various benefits
and possibilities for the agricultural sector.
V. Mamatha and J. C. Kavitha [20] introduced a hydroponics system that uses organic
coconut coir medium as a replacement for non-biodegradable rock wool. The system
is automated on a large scale, encompassing the entire greenhouse
with various crops cultivated under optimal conditions for the highest yield. The
absolute growth rate for leafy vegetables is estimated using the KNN algorithm across
different conditions. NFT combined with coconut coir medium has shown to be
the most effective hydroponic system for growing commercial crops, achieving an
accuracy of 93%. The system records an f1 score of 93% for weighted average and
33% for macro average.
Materials and methods
Machine learning (ML) is a popular technology presently
utilized in contemporary agriculture. It assists in developing healthier seeds. In
machine learning, approaches are based on the learning process. These
methodologies must be acquired through experiences to execute specific tasks. As the
ML model gains experience over time, its performance improves. After
completing the learning phase, the model can then be utilized for
prediction, classification, and data testing. This is accomplished
following the successful training process. The aim of the research study is to foster
significant advancements in the agricultural sector. The focus of the study is to
predict crop yield based on crop type, geographical area, climatic conditions
(temperature and humidity), and pH levels. Real-time data is collected via sensors
from the land using a Raspberry Pi and transmitted to a remote server. The sensory
data will be retrieved from the server to the IDE for crop yield prediction utilizing the
ML algorithm as illustrated in Figure 1.
Figure 1. Block diagram of crop yield prediction
For Precision Agriculture, we have distinct systems that analyze and forecast crop
production based on temperature, humidity, pH, and soil moisture. i) The sensory data
can be obtained via email or short message service (SMS). ii) Real-time analysis is
available for monitoring plant growth. ii) Cyber surveillance is
implemented for ongoing alerts and monitoring.
Proposed system: In this study, we are aiming to create a unified architecture
to forecast the crop yield based on factors like temperature, humidity, pH, and soil
moisture collectively for various crops and multiple regions. For
data acquisition through sensors, we have implemented a public API interface using
the Flask framework. The real-time forecasting of crop yield in a specific area utilizes
the Machine Learning algorithm known as Multiple Linear Regression. Data Mining
Various source data through comma-separated value (CSV) files have
been utilized in forecasting the yield of crops in any location.
System Architecture
A system architecture is a conceptual model that defines the structure, behavior, and
more views of a system. An architecture description is a formal description and
representation of a system, organized in a way that supports reasoning about the
structures and behaviors of the system. The system architecture consists of
components involved in the project such as raspberry pi, sensors, ADC, etc. The
arrow’s direction indicates the direction of the data flow.
Figure 2. Smart Agriculture Architecture
Hardware elements
The various hardware and sensors used in this study are outlined in
the subsequent sections.
Raspberry Pi
The Raspberry Pi is an affordable, credit-card-sized computer that connects to a
computer monitor or TV and requires a standard keyboard and mouse. It is
a versatile device that allows individuals of all ages to delve into computing and
learn programming languages such as Scratch and Python. The Raspberry Pi operates
on Linux, but it also features a set of general-purpose input/output (GPIO) pins
that enable control over electronic components for physical computing
and exploration of the IoT. Raspberry Pi 3b+ was employed in the development of the
proposed system as shown in Figure 3.
Figure 3. Raspberry Pi 3b+
DHT11 Temperature and Humidity Sensor
The DHT11 Temperature and Humidity Sensor has a calibrated digital signal output
with the capability to measure temperature and humidity. It incorporates a high-
performance 8-bit microcontroller. Its design guarantees high reliability
and remarkable long-term stability. This sensor comprises a resistive element and a
sensor for wet NTC temperature-measuring devices. It possesses
high quality, rapid response time, resistance to interference,
and superior performance. Each DHT11 sensor features a highly precise calibration of
the humidity calibration chamber. The calibration coefficients are stored in the one-
time programmable (OTP) memory, and internal sensors identify signals during the
process, for which we should reference these calibration
coefficients. The integrated single-wire serial interface system is designed for quick
and straightforward use. With a small size, low power consumption, and signal
transmission distance up to 20 meters, it accommodates various applications,
including even the most demanding ones. The product is configured as a 4-pin single-
row pin package labeled VCC, DATA, GND, and an unconnected pin, as illustrated in
Figure 4.
Figure 4. DHT11 Sensor
pH Sensors
A pH meter is an electrical device designed to measure hydrogen-ion activity (acidity
or alkalinity) in a solution. Essentially, a pH meter consists of a
voltmeter connected to a pH-responsive electrode and a reference
electrode, as shown in Figure 5. pH probes determine pH by measuring the voltage or
potential difference of the solution in which they are immersed. Thus, a pH probe
measures the potential difference generated by the solution
by assessing the variation in hydrogen ion concentration through the Nernst equation
and delivers the pH as output.
Figure 5. pH Sensor
YL-38 YL-69 Sensors
The soil moisture sensor, or hygrometer, is typically utilized to assess the humidity of
the soil. Therefore, it is suitable for monitoring the soil moisture levels of plants. The
sensor consists of two parts: the electronic board (on the right) and the probe with two
pads that detect the water content (on the left), as depicted in Figure 6a. The
sensor features an integrated potentiometer for adjusting the sensitivity of the digital
output (D0), in addition to a power LED and a digital output LED, as illustrated in
Figure 6b.
(a) (b)
Figure 6. (a) Y-38 YL-69 (b) Y-38
Soil Moisture Working
The output can be a digital signal (D0) LOW or HIGH, depending on the water
content. If the soil humidity exceeds a certain predefined threshold value, the
modules output LOW, otherwise, it outputs HIGH. The threshold value for the
digital signal can be adjusted using the potentiometer. The output can be an analog
signal and so you’ll get a value between 0 and 1023.
MPC3008
The MCP3008 is a 10bit 8-channel Analogue-to-digital converter (ADC). It is
cheap, easy to connect and doesn't require any additional components. It uses the
SPI bus protocol which is supported by the Pi's GPIO header. An ADC may also
provide an isolated measurement such as an electronic device that converts an input
analog voltage or current to a digital number representing the magnitude of the
voltage or current.
Datasets
There are 5 datasets in CSV format used in this research in which two datasets are
shown in Table 1 and Table 2.
Table 1. Different crops with temperature and pH
Table 2. Bajra crop with rainfall, temperature, pH, and production
Results and Discussion
The results shown in Figure 7 are for a particular area chosen by the user which
indicates the possible crops that can be grown in the region of Bangalore in
Karnataka state. The accuracy is shown in the rightmost column and the highest
accuracy of 0.946 is predicted for barley crops. Figure 8 shows a 4D graph
depicting the yield of the crop at different parameter values, such as soil moisture,
temperature, and pH. Figure 9 shows the predicted yield of different crops. Figures
10, 11 and 12 depict the graphs of predicted yield v/s moisture, predicted yield v/s
temperature, and predicted yield v/s pH.
Figure 8. Accuracy of different crops yield
Figure 9. Predicted Yield of different crops
Figure 10. Predicted Yield v/s Moisture
Figure 11. Predicted Yield v/s Temperature
Figure 12. Predicted Yield v/s pH
Conclusion and Future Scope
The proposed research focuses on bringing together the various implementations of
different problem statements in the domain of agriculture into a single unified system.
This can increase the overall efficiency of the system as a whole and attributes to a
lesser building cost when compared to the individual systems. The current system
makes use of a computer which is connected to the Raspberry Pi to run the IDE and
make predictions. But future scope may involve an IDE that can run directly on the
Raspberry Pi and making predictions that can be displayed through an API Interface
instead of a computer. In the future, the above-proposed system may be modified by
using wireless sensors which will improve the model’s efficiency.
References
1. Prof. K. A. Patil and Prof. N. R. Kale, A Model for Smart Agriculture Using IoT,
2016 International Conference on Global Trends in Signal Processing,
Information Computing and Communication, 22-24 December 2016.
DOI: 10.1109/ICGTSPICC.2016.7955360.
2. Sally Jo Cunningham and Geoffrey Holmes, Developing innovative applications
in agriculture using data mining, The proceedings of the Southeast Asia regional
computer confederation conference, 1999, pg.25 – 29.
3. Li Minbo, Zhu Zhu, Chen Guangyu, Information Service System Of Agriculture
IoT, automatika 54(2013) 4, 415–426.
4. Mohanraj I, Kirthika Ashokumar, Naren J, Field Monitoring and Automation
using IOT in Agriculture Domain,6th International Conference on Advances In
Computing & Communications, ICACC 2016.
5. Nikesh Gondchawar, Prof. Dr. R. S. Kawitkar, IoT based Smart Agriculture,
International Journal of Advanced Research in Computer and Communication
Engineering, Vol. 5, Issue 6, June 2016.
6. Ciprian-Radu Rad, Olimpiu Hancu, Ioana-Alexandra Takacs, Gheorghe Olteanu,
Smart Monitoring of Potato Crop: A Cyber-Physical System Architecture Model
in the Field of Precision Agriculture, Agriculture and Agricultural Science
Procedia 6 (2015), 73 – 79.
7. Chandhini. K, A Literature Study on Agricultural Production System Using IoT
as Inclusive Technology, IJITR, Volume No.4, Issue No.1, December - January
2016.
8. G. V. Satyanarayana, SD. Mazaruddin, Proceedings of the Conference on
Advances in Communication and Control Systems (CAC2S 2013), 1951-6851,
April 2013.
9. Anup Managave, Ojas Savale, Deepika Ambekar, Sushmita Sathe, Precision
Agriculture using Internet of Things and Wireless sensor Networks, IJARCET
Volume 5, Issue 4, April 2016.
10. D. Rajesh, Application of Spatial Data Mining for Agriculture, International
Journal of Computer Applications (0975 – 8887) Volume 15– No.2, February
2011.
11. Sweksha Goyal, Unnathi Mundra, Prof. Sahana Shetty, smart agriculture using
IOT International Journal of Computer Science and Mobile Computing
(IJCSMC), Vol. 8, Issue. 5, May 2019, pg.143 – 148.
12. Carlos Andrés González-Amarillo & Juan Carlos Corrales-Muñoz, An IoT-Based
Traceability System for Greenhouse Seedling Crops. Special section on new
trends in brain signal processing and analysis IEEE 2018.
13. B. Manjula Josephine and K. Ruth Ramya, Crop Yield Prediction Using Machine
Learning, International Journal of Scientific & Technology Research: 9 (2020) 2.
14. Meeradevi and Monica R Mundada, Automated Control System for Crop Yield
Prediction using Machine Learning Approach, International Journal of Applied
Engineering Research: 14 (2020) 2.
15. Ashok Tatapudi and P Suresh Varma, Prediction of Crops based on
Environmental Factors using IoT & Machine Learning Algorithms, International
Journal of Innovative Technology and Exploring Engineering (IJITEE): 9 (2019)
1.
16. Kanaga Priya & Dr. N. Yuvaraj, An IoT Based Gradient Descent Approach for
Precision Crop Suggestion using MLP, International Conference on Physics and
Photonics Processes in Nano Sciences. Conf. Ser. 1362 012038. 2019.
17. Monica Herrero-Huerta, Pablo Rodriguez-Gonzalvez, Katy M Rainey, Yield
prediction by machine learning from UAS-based multi-sensor data fusion in
soybean, Plant Methods, vol 16 (1),1-16, 2020.
18. Prof. Chhaya Narveka and Farman Haider, AI Based Crop Monitoring Using IOT,
International Journal for Research in Applied Science & Engineering Technology
(IJRASET): 10(2022) 5.
19. Chiara Bersani and Carmelina Ruggiero, Internet of Things Approaches for
Monitoring and Control of Smart Greenhouses in Industry 4.0, Energies:15 (2022)
3834.
20. V. Mamatha and J.C. Kavitha, Machine learning based crop growth management
in greenhouse environment using hydroponics farming techniques, Measurement:
Sensors:25 (2023) 100665.
