Low-Cost Tree Health Categorization and Localization Using Drones and Machine Learning

In this work, we propose a low-cost tree health cat-egorization and localization scheme using drones and machine learning. Crop data is collected and employed to inform agricul-tural decisions, with the ultimate goal of enhancing production. Our solution employs visual imaging obtained from unmanned aerial vehicles (UAVs), commonly known as drones, for the purpose of monitoring the well-being of fruit trees. A user activates the drone from the ground station, initiating a "snake-like" flight path through the orchard to capture images before landing. Subsequently, our models process the data, triggering an alert for any visually identified unhealthy trees. The system then furnishes the farmer with precise locations of these afflicted trees. Our artificial-intelligent (AI)-driven precision agriculture technology is poised to significantly streamline the process and reduce costs for farmers by swiftly identifying and categorizing both unhealthy and marginally healthy trees in the orchard. The proposed framework is a cost-effective solution employing an off-the-shelf visible light camera. It operates on a physical drone equipped with a Raspberry Pi 4 computer running in the background, providing realtime tree classification data. The experimental outcomes demonstrate an impressive average validation accuracy of 92.67%.