Automatic classification of healthy / diseased plants using multispectral images

Recent advancements in drone technology have significantly increased their applicability across various domains, including search and rescue missions, environmental monitoring, topographic mapping, and general-purpose videography. Within the agricultural sector, drone imagery has emerged as a particularly valuable tool for assessing crop health. However, as agricultural fields expand in scale, the task of monitoring the condition of individual plants becomes increasingly complex. Traditional plant inspection methods, which rely heavily on manual visual assessments and, in some cases, laboratory testing, are often impractical due to their time-consuming and costly nature. This thesis builds upon the foundation laid by the Dronuts Project, which aims to develop both hardware and software solutions for plant monitoring through remote sensing. The primary objective is to evaluate the health status of individual plants (Hazelnuts trees in Piedmont, Italy) using multispectral images captured by drones. The work presented here focuses on applying a variety of software techniques to analyze drone imagery acquired at different time points, with the goal of autonomously identifying the unhealthy parts of hazelnut trees.