Sistema automatizado de detecção de mosca-das-frutas com rede neural convolucional

Abstract

This work addresses the automation of pest monitoring in fruit growing, with an emphasis on fruit fly detection, using the McPhail trap integrated with image capture and computational analysis technologies. Agriculture, especially fruit growing, faces significant challenges due to pests that affect fruit production and quality. Traditional management of these pests is often inefficient and unsustainable. The study proposes an innovative solution through the development of an automated system that combines advanced hardware, such as digital cameras and processors, with image processing algorithms to detect and manage pest infestations more effectively and with less environmental impact.The methodology includes system assembly, laboratory and field testing, and data analysis to validate the effectiveness of the automated approach. The system utilizes a Raspberry Pi 4 microcomputer, a 4MP camera, and solar power for energy autonomy. The YOLO v8 algorithm was employed for real-time detection of fruit flies, achieving a mean Average Precision (mAP50) of 95% and a recall of 90%. The system demonstrated robustness in field conditions, with the ability to log detection data, including GPS coordinates, date, and time, for further analysis.The expected results include improved accuracy and efficiency in pest monitoring, reduced operational costs, and the promotion of more sustainable agricultural practices. The system’s ability to operate autonomously with minimal human intervention makes it a promising tool for integrated pest management in fruit production.