Abstract: Traditional pest control methods require a lot of manpower and material resources, and cannot achieve high accuracy. In order to do a more scientific and efficient pest control work, this paper combined deep learning technology and internet of things technology to develop a pest detection system for remote pest detection and improve the efficiency of control work. The system mainly adopted the YOLO-v5 network model combined with transfer learning to train and learn the characteristics of common forest pests and farmland pests, so as to achieve efficient detection and identification. Based on the internet of things technology, remote control was achieved to capture images of pests and diseases, and they were transmitted to the computer for recognition through Wi-Fi. The types and quantities of pests in the field were shown through the visual interface. It has good practical value for reducing the consumption of manpower and material resources and realizing the scientific pest control.