This study built an automatic insect pest image identification system based on tiny Yolov3 deep learning model. By optimizing the tiny Yolov3 detection model, images of insect pests on scanned sticky paper can be automatically identified. The system achieves a testing accuracy of 0.93, 0.90 for whitefliesthrips respectively.

Compared to traditional insect pest identification methods using image processing, based on deep neural networks, this research can yield a better detection accuracy efficiency. Through implementing the tiny Yolov3 algorithm, we can finish the detection of a scanned sticky paper image within 14.21 second, outperforming works using Faster-RCNN with a detection time of more than ten minutes.

The most common way to prevent the outbreaks of insect pests is through sticky paper traps. Traditionally, identification of pests was done by manual inspection. Through implementing this insect pest identification system, the efficiency of insect pest management can be drastically improved. Moreover, the risk of exporting infested crops can be reduced, which can help our product gain more reputat