Downloaded 40 times
Remote Sensing applications for phenotyping
This document discusses remote sensing applications for plant phenotyping. It defines phenotype and phenomics as the study of observable plant traits and how genes and environmental factors influence trait expression. Remote sensing methods like proximal sensing, imaging, and spectroscopy are non-destructive ways to measure various plant phenotypes over time. Challenges include trait changes over the crop cycle and inconsistent canopy measurements. A variety of sensors exist to measure phenotypes remotely, including RGB, multispectral and hyperspectral cameras, as well as thermal, lidar and fluorescence sensors. Unmanned aerial vehicles are valuable for high-resolution phenotyping but data processing is required.
More Related Content
Similar to Remote Sensing applications for phenotyping
More from International Institute of Tropical Agriculture
Remote Sensing applications for phenotyping
- 1. www.iita.orgA member ofCGIAR consortium www.iita.orgA member of CGIAR consortium Remote Sensing applications for phenotyping.
- 2. www.iita.orgA member ofCGIAR consortium “Pheno …” - what? • A phenotype is any observable characteristic or trait of an organism, such as its morphology, development, biochemical or physiological properties. • Phenomics is the field of research studying the results from the expression of plant genes as well as the influence of environmental incl. managed factors and the interactions between the genes and these factors.
- 3. www.iita.orgA member ofCGIAR consortium
- 4. www.iita.orgA member ofCGIAR consortium Mapping field gradients Environmental data collection Phenotyping
- 5. www.iita.orgA member ofCGIAR consortium Phenomics: Genotype x Environment x Management
- 6. www.iita.orgA member ofCGIAR consortium Phenotypic properties • Size, growth • Shape, architecture • Photosynthesis • Gas exchange • Color, pigments • Metabolite content • Yield
- 7. www.iita.orgA member ofCGIAR consortium Categories of trait evaluation 1. Proximal/remote sensing and imaging 2. Sample collection and their analysis in laboratory
- 8. www.iita.orgA member ofCGIAR consortium Proximal/remote sensing for phenotyping… … is non-destructive and non-invasive … is mostly based on the reflectance or transmission of light through the plant parts ( information on “colour”) … can involve active components (like a light emitting source or a magnetic coil): RADAR, LIDAR/LASER, MRI, PET ( information on “travel time”)
- 9. www.iita.orgA member ofCGIAR consortium Full featured Phenotyping platform • Multiple sensors ??? Tools for proximal sensing (information collection near the crop) EM38: Electromagnetic soil mapping (Measures conductivity to electrical current) • soil moisture content • salt levels • soil texture ~ 17,000 USD Field Spectrometer • Vegetation/soil Reflectance • NDVI 15,000 – 50,000 USD
- 10. www.iita.orgA member ofCGIAR consortium Non-destructive plant/crop measurement Plant/Conopy Root Seed Measurement • Multi/Hyperspectral imaging • Chlorophyll fluorescence • Thermal imaging • Leaf area • Leaf angle • Stem/internode length • Plant/Canopy height • 3-D characteristics Calculated variables • Light interception • Growth rate of leaves, stems, plant or plants • Dry weight of leaves, stems, flowers or plant • Shoot biomass Estimated traits • Yield Measurements • Root area/volume • Root length/depth • Thickness • Branching Calculated variables • Growth rate • Root biomass Estimated traits • Shoot/root • Compactness Measurements • Number of panicles Total number of seeds per plant • Number of fertile seeds • Seed weight • Seed yield • Seed width, length and area
- 11. www.iita.orgA member ofCGIAR consortium Challenges in phenotype recognition Traits develop and change… … over time (throughout the crop cycle) ... in space (at the canopy level) crops don’t capture resources (radiation, water and nutrients) uniformly throughout the season Multiple measurements per season are highly recommended (Time series creation)
- 12. www.iita.orgA member ofCGIAR consortium Image based sensors (cameras) Can be categorized into three different classes: 1. Panchromatic/ single banded cameras 2. RGB/CIR cameras 3. Multi-Spectral cameras 4. Hyperspectral cameras 5. Narrow banded Cameras (like Chlorophyll Fluorescence Sensors/MIPS)
- 13. www.iita.orgA member ofCGIAR consortium Camera sensors (1/3): RGB/CIR Cameras Conventional digital cameras usually capture light from three different wavelengths: red, green and blue. For CIR cameras the blue filter is replaced with a NIR filter to enable the capture of wavelengths in the Near Infrared spectrum. Most RGB/CIR cameras use a bayer filter so the colour information for every 50% has to be interpolated.
- 14. www.iita.orgA member ofCGIAR consortium Camera sensors (2/3): Multispectral cameras Multi-Spectral sensors capture up to nine different bands from the electro-magnetic spectrum (Typically 4-5). Some wavelengths are not captured. Each band covers approximately 50-400 nm of the EMS. They either use one camera per channel or incorporate a mechanism to redirect the light to one particular CCD array per spectral band (For example moving mirrors/lenses). Top: Landsat 8 (OLI) Bottom: Landsat 7 (ETM+) Color information for 8 pixels out of 64 = 87.5% of all pixels have to be interpolated!
- 15. www.iita.orgA member ofCGIAR consortium Camera sensors (3/3): Hyperspectral cameras Hyperspectral cameras capture the reflected light in hundreds of different bands. Each band only covers a very small range of the EMS (approximately 3-5 nm). HS cameras provide collect images from nearly 200 different bands. The data is usually organized in “Hyperspectral cubes”.
- 16. www.iita.orgA member ofCGIAR consortium Phenotyping related limitations of spectroradiometric sensors • Different plant parts have their particular absorption spectrum which vary as the plant develops. • Leaf properties like roughness, trichomes, epicuticular waxes all have influence on the reflectance. Measurements need to be standardized to ensure comparability between similar development stages. Some issues that should be taken into consideration: – soil covering (due to variations in canopy density) – Plant architecture (plant height, leaf size and angle, emerged spikes)
- 17. www.iita.orgA member ofCGIAR consortium Sensors and measurable properties (1/2)
- 18. www.iita.orgA member ofCGIAR consortium Sensors and measurable properties (2/2) Sensor Parameters +/- RGB color, shape, root architecture, morphology, surface features, leaf disease assessments, … + Rapid measurements + affordable solutions - Limited physiological information Fluorescense Photosystem II, disease severity - only for rosettes - pre‐ acclimation conditions Thermal (IR) Surface temperature ( leaf and canopy transpiration) - Rapid data collection - Sound bio‐physical interpretation required Multi- /Hyperspectral Water content, seed composition, Biomass, leaf and canopy water status, disease severity, pigment composition + Large amount of information - Costs Stereo Camera Systems Shoot structure, leaf angle distribution, canopy structure + High 3D accuracy - Complex data reconstructions LIDAR Shoot biomass, leaf angle distribution, canopy structure + excellent 3D accuracy - Some instruments require specific illumination
- 19. www.iita.orgA member ofCGIAR consortium Comparison of different Phenotyping platforms
- 20. www.iita.orgA member ofCGIAR consortium Remote sensing applications A comparison Satellite image acquisition UAV mapping Repeatability Limited: (revisiting time of satellites: 1,2 days off-NADIR and ~16 days at NADIR) Unlimited (But depends on weather conditions, number of batteries/chargers available) Resolution not smaller than 30 cm/pixel ( “mixed Pixels”) Depends on altitude of UAV. Typically 4-15 cm at 20- 100 m altitude (depends on Camera) Image size Unknown (accuracy/resolution significantly decreases to the edges). Estimated 8,000 x 8,000 pixels Typically 1,280 x 960 pixels Price ~ 30-60 USD per km2 for nine spectral bands at ~ 30 cm resolution (WorldView 3) Labour costs only Ready to use images YES (But image enhancements may be performed as well) NO (image alignment necessary, can be automated if image quality is sufficient). Clouds YES (Especially during raining season) NO • Rainy season = clouds • coarse resolution
- 21. www.iita.orgA member ofCGIAR consortium UAV platforms • Balloons (“Blimps”) – Excellent payload (several kilograms) – Controlled through ropes that are pulled by people ( Limited manoeuvrability) • Fixed wing UAV (planes) – Fast – Need runways/flat surfaces to land – Limited payload • Paragliders – Slow (good for long exposures) – Limited manoeuvrability • Helicopter – Can start/land vertically (no runway needed) and maneuvered into a stationary position – High payload – Vibration issues • Multi-Rotor/Multicopter – Can start/land vertically (no runway needed) and maneuvered into a stationary position – Excellent payload (Up to 3 kg, depending on Multicopter size)
- 22. www.iita.orgA member ofCGIAR consortium UAV example (1/2) List of components: • Frame • Motors • LiPo Battery • Power module • Electronic speed controllers • Flight computer (PixHawk) • Buzzer • Safety button • Telemetry • Compass, Magnetometer • Receiver • Transmitter • Software • (FPV = First person view)
- 23. www.iita.orgA member ofCGIAR consortium UAV example (2/2)
- 24. www.iita.orgA member ofCGIAR consortium Putting it all together
- 25. www.iita.orgA member ofCGIAR consortium