Identification of Fruit Tree Pests With Deep Learning on Embedded Drone to Achieve Accurate Pesticide Spraying

Abstract

Tessaratoma papillosa (Drury) first invaded Taiwan in 2009. Every year, T. papillosa causes severe damage to the longan crops. Novel applications for edge intelligence are applied in this study to establish an intelligent pest recognition system to manage this pest problem. We used a detecting drone to photograph the pest and employed a Tiny-YOLOv3 neural network model built on an embedded system NVIDIA Jetson TX2 to recognize T. papillosa in the orchard to determine the position of the pests in real-time. The pests' positions are then used to plan the optimal pesticide spraying route for the agricultural drone. Apart from planning the optimized spraying of pesticide for the spraying drone, the TX2 embedded platform also transmits the position and generation of pests to the cloud to record and analyze the growth of longan with a computer or mobile device. This study enables farmers to understand the pest distribution and take appropriate precautions in real-time. The agricultural drone sprays pesticides only where needed, which reduces pesticide use, decreases damage to the environment, and increases crop yield.

Keywords

• Edge intelligence
• unmanned aerial vehicles (UAV)
• real-time embedded systems
• slope land orchard
• object detection
• agricultural pests damage