부산대학교 도서관

Fruit flies pose a significant threat to fruit yields, necessitating immediate detection solutions for effective pest management. In this study, we present our approach using YOLOv7 and the Jetson Nano 4GB for rapid and accurate fruit fly detection. Our methodology attains an Average Precision (AP) of 0.993 at an Intersection over Union threshold (IoU) of 0.75, facilitating almost instantaneous enumeration of ensnared fruit flies in field conditions, thereby augmenting the efficacy of pest management practices. Concurrently, the system transmits this data to either a local or cloud-based server, assisting end-users in accurately identifying regions with elevated fruit fly densities to strategically administer pesticides. This targeted approach has significant implications for sustainable agriculture as it reduces the need for widespread spraying. Moreover, our versatile method adapts for detecting other pests like aphids or beetles, serving as a valuable tool in agriculture and beyond.