Towards Early Detection: Physics-based Hyperspectral Models for the Detection of Tomato Plant Diseases

Smart agriculture is essential for achieving a successful ecological transition, but its progress is limited by unresolved technological challenges. This study addresses these challenges by integrating Visible and Near-InfraRed (VNIR) hyperspectral imaging with physics-based modeling to develop tools for the early and accurate detection of plant diseases. Unlike conventional RGB imaging, hyperspectral imaging offers rapid, non-invasive monitoring capable of identifying plant diseases before visible symptoms emerge. Relying on physics-based inversion models—particularly the PROSPECT model—this work focuses on retrieving detailed physico-chemical and biophysical parameters from tomato leaves affected by powdery mildew. The results demonstrate the feasibility of using PROSPECT modeling in conjunction with VNIR hyperspectral data to effectively detect and characterize disease at early stages. This methodology shows strong potential for advancing precision agriculture, with artificial intelligence integration proposed as a future enhancement.