Skip to main content
ARS Home » Northeast Area » University Park, Pennsylvania » Pasture Systems & Watershed Management Research » Research » Publications at this Location » Publication #424129
Research Project: Comprehensive Environmental Framework to Facilitate Resilient and Sustainable Intensification of Crop-Livestock Systems

Location: Pasture Systems & Watershed Management Research

Title: Using artificial intelligence to extend the spectral range of Unmanned Aircraft Systems (UAS) imagery

Author
item MASRUR, ARIF - Esri
item OLSEN, PEDER - Microsoft Research Lab
item Adler, Paul

Submitted to: Electronic Publication
Publication Type: Other
Publication Acceptance Date: 5/28/2024
Publication Date: 5/28/2024
Citation: Masrur, A., Olsen, P.A., Adler, P.R. 2024. Using artificial intelligence to extend the spectral range of Unmanned Aircraft Systems (UAS) imagery. Electronic Publication. 169.

Interpretive Summary:

Technical Abstract: Unmanned Aircraft Systems (UAS) and satellites have potential as important tools in precision crop management (e.g., detecting nutrient stresses and weed, disease, and insect infestations). However, while satellite imagery is too coarse for targeted applications, UAS’ are impractical for large areas and/or frequent coverage. Furthermore, since performance improves with spectral range in addition to spatial resolution, and sensor costs increase with spectral range, extending the spectral range for UAS’ can be cost prohibitive. Our objective was to develop an artificial intelligence model using super-resolution techniques on satellite images from Sentinel-2 to spectrally extend UAS-RGB imagery. Hyperspectral imagery of covers crops was collected using a Headwall Nano-Hyperspec [VNIR 400–1000 nm] and Velodyne VLP-16 LiDAR Puck LITE, and biomass and nitrogen (N) data collected at the time of flights. To fuse Sentinel-2 and UAS imagery we constructed an artificial intelligence (AI) model using a super-resolution convolutional neural network (SRCNN). This notebook [https://github.com/microsoft/farmvibes-ai/blob/main/notebooks/spectral_extension/spectral_extension.ipynb] demonstrates how to obtain high-resolution (0.125m) Sentinel-2 bands by combining a Sentinel-2 product with RGB UAS imagery. Although the model was trained on cover crop images in the Maryland region, it can be applied to other regions and crops.