Evaluation of the COSICAM hyperspectral camera with agricultural monitoring experiments S. Livens, S. Delalieux, J. Blommaert, P. Baeck, A. Sima, B Delauré M. Boonen
Mol, Belgium Remote Sensing unit: Applications global local agriculture water Technology: missions instruments image quality image processing Spatial Data Services operational processing aerial satellite value added products
Hyperspectral RPA S VITO COSICAM Results on Strawberries
Hyperspectral RPAS VITO COSICAM Results on Strawberries
THE RPAS REVOLUTION RPAS allow local mapping: Precision agriculture Limited payloads RGB cameras wanted extra info:» NIR» red edge» subtle spectral differences» Fit a hyperspectral imager onto UAS? 5
THE HYPERSPECTRAL IMAGING REVOLUTION from large high end instruments to miniature instruments novel enabling technologies: thin film interference filters direct deposition on imaging sensor Livens et al: Evaluation of the COSICAM hyperspectral camera with agricultural monitoring experiments Earsel 2016
HOW TO CAPTURE 3D HYPERCUBE USING A 2D SENSOR? scanning Spatial Spectral Spatio-spectral Snapshot Hypercube slicing concept Pushbroom Alternating bands 1 image contains: spectrum of 1 image line full res image of 1 band Spatiospectral full res image with spectral stripes Mosaic alternating spectral pixels 7 Tiled tiles with 1 band Resolution high high high low low Remarks not possible to create spatial image for UAS requires band switching full res. spatial image needs pixel precise filters complex optics
SPATIO-SPECTRAL IMAGING Tilted hypercube sampling Result in rainbow colored 2D spatial map Scanning needed to create hypercube 8
CREATE HYPERCUBE FROM SPATIO SPECTRAL IMAGING: GEOMETRIC CHALLENGE Scanning motion required to obtain full spectral information at all locations Each image is a spatial map of the scene, facilitating pointing, focusing, and data analysis. This is particularly valuable for irregular or irretrievable scanning movements. 9
Hyperspectral & UAV VITO COSICAM Results on Strawberries
COSICAM FLYING PROTOTYPE CMOSIS CMOS sensor CMV2000 direct deposited interference filters 2048 x 1088 pixels 128 bands x 8 image lines 192 band x 5 image lines (in prep.) Typical operational framerate= 30fps 11
COSI SYSTEM Complete solution COSI cam: delivers raw images COSI cube: Uses photogrammetry to align image series Creates reflectance hypercubes COSI map : Derives value added products: ranging from vegetation indices to management zone maps 12
IMAGE QUALITY Spatio-spectral 5x higher spatial res. 5x higher spectral res Mosaic Snapshot (simulated) Livens et al: Evaluation of the COSICAM hyperspectral camera with agricultural monitoring experiments Earsel 2016 13
GEOMETRIC PERFORMANCE To be presented in July 2016 at ISPRS congress 14
Hyperspectral & UAV VITO COSICAM Results on Strawberries
DATA ACQUISITION CAMPAIGNS OVER STRAWBERRY TEST FIELD 16
STRAWBERRY FIELD - FALSE COLOUR IMAGE( RGB = 801nm, 667nm, 604nm) COSI GSD = 2.0cm 0.5m RGB GSD = 2.6cm 17 17
IMAGE COMPARISON COSI cam COSI, GSD = 2cm Nex-6, 16mm lens, GSD = 0.8cm 18
TEST FIELD TESTUP Plots with different varieties Treatment: nutrients : 0 vs 80kg N Irrigation: 20% vs. 100% 19
EXTRACT VEGETATION INDICES Pigment Specific Simple Ratio PSSRb= R800/R635 100% irrigation 80kg N Livens et al: Evaluation of the COSICAM hyperspectral camera with agricultural monitoring experiments Earsel 2016 20
EXAMPLE: PREDICTION WITH VEGETATION INDICES Experiment: Plots with 20% irrigation vs plots with 100% irrigation Index: Pigment Specific Simple Ratio PSSRb= R800/R635 Good correspondence PSSRb from COSI Ground Truth 21
Conclusion : Hyperspectral & UAV viable with thin film filters on chip VITO COSICAM complete spatiospectral solution high spatial + high spectral resolution Results on Strawberries good yield prediction
COSICAM now industrialized with
COSI CAM: delivers the future of high resolution precision farming