A Smart UAV System to Assess the Health of a Vineyard

This paper presents the hardware and software components of a system designed to autonomously detect the vine’s health state through the analysis of vine leaves texture, shape, and color. The detection component uses a deep neural network and its execution is supported by an i.MX 8M Plus processor. Based on the NXP HoverGames drone, that flies over the vineyard, the detection system can analyze, in real-time, the health state of the different sections of the vineyard. The system, referred to in this paper as agriHoverGames, obtained an average $84.7 \%$ classification performance when classifying leaves into three classes (i.e., healthy leaves and leaves affected by two distinct diseases). Although this project was initially implemented to identify two vine diseases, it actually provides a mobile platform capable of identifying any other type of disease as long as this one manifests through specific changes (in texture, shape, color, or size) in the fruits, flowers, or leaves of a plant or fruit tree and, more, these ones can be identified through the video information.