A Pest Monitoring System for Agriculture Using Deep Learning

Abstract

The study develops a pest monitoring system for agriculture using deep learning, pi camera and environmental sensors. The developed system can continuously monitor and calculate the number of pest insects stick on the yellow sticky papers. The detected pest insects were counted using image processing and deep learning model, specifically, the You Only Look Once (YOLO), with an average accuracy of 52.3% and computation time of 8-10 minutes per picture. The environmental information of temperature and humidity were also gathered in the study site, where fruit plants were grown as the main crop. The tests revealed that humidity has the strongest correlation with pest insects. As conclusion, the developed system can effectively collect the insect counts automatically, which provides useful information for efficient pest control in crop cultivation processes.