Winter Wheat Biomass and Yield Estimation using Unmanned Aerial Vehicle-based and VENµS Satellite Imagery with Machine Learning Techniques
Abstract
Monitoring crop productivity is crucial in precision agriculture, often using biomass and yield as metrics to measure crop health and growth status. This thesis aims to predict dry above-ground biomass using Unmanned Aerial Vehicle (UAV) multispectral imagery, derived vegetation indices (VI), plant height, leaf area index (LAI), and plant nutrient content ratios. Additionally, the thesis tests the viability of VENμS satellite data as an alternative to other popular multispectral satellite data for predicting winter wheat yield. Conducted in two winter wheat fields in southwestern Ontario, Canada, the study employed Random Forest (RF) and Support Vector Regression (SVR) machine learning models with various variable combinations. The results demonstrate that the approach in biomass estimation was accurate and provided valuable insights into the applicability of biochemical parameters. Furthermore, VENμS produced promising yield prediction results, proving to be a better satellite platform compared to other publicly available satellite data for yield prediction.