Large scale crop canopy cover estimation using satellite-UAS synergy and machine learning

Abstract

Crop canopy cover is a key agronomic variable for understanding crop growth and crop development status. Estimation of crop canopy cover rapidly and accurately at large scale is crucial for crop monitoring in support of agricultural management. This work is to explore the potential of satellite and Uncrewed aircraft systems (UAS) remote sensing synergy in estimation of crop canopy cover. Specifically, high-resolution UAS remote sensing multispectral imagery collected from crop fields are used to derive accurate crop canopy cover percentage through machine learning-based classification and segmentation. UAS multispectral imagery-based canopy cover maps are aligned with satellite imagery, and used as ground truthing data for satellite data-based canopy cover estimation model development. Satellite data such as Harmonized Landsat Sentinel-2 (HLS) and PlanetScope multispectral imagery are used to estimate crop canopy cover using machine learning methods such as Random Forest, Support Vector Machine and Convolutional Neural Networks (CNN). Our results show that the synergy of satellite and UAS remote sensing is a valid approach and strategy for crop canopy cover estimation with high accuracy at large scale. By facilitating accurate and scalable predictions of crop canopy cover on course resolution remote sensing data, our approach offers opportunities to the broader objectives of sustainable agricultural practices and food security.