Snap Bean Yield Modeling and Harvest Scheduling

Overview

This doctoral research combined hyperspectral sensing and machine learning to model snap bean yield and assess harvest readiness, progressing from controlled greenhouse experiments to drone-based field studies. Greenhouse experiments examined relationships between plant spectra, yield, growth stage, and pod maturity to identify informative wavelengths and useful observation periods. The work developed models for growth-stage classification and harvest scheduling, then extended them to UAS hyperspectral imagery collected across multiple growing seasons and cultivars. Workflows included reflectance calibration, vegetation extraction, spectral denoising, feature selection through Jostar, and predictive modeling. The field studies evaluated yield estimation and pod-size classification. Across four publications, this research established a foundation for using crop spectral information to support timely, data-driven agricultural decisions.