Real-Time Detection and Analysis System for Tea Leafhopper

Abstract

The tea leafhopper is the most common and harmful pest in a tea garden. The traditional automatic pest detection and counting method is low in accuracy and efficiency. Therefore, accurate and real-time detection of the tea leafhopper is of great significance for predicting insect situation. To this end, a set of real-time detection and analysis equipment was designed to solve the problem of the tea leafhopper. First, remote image acquisition equipment was used to collect images, which were uploaded to the cloud server for detection. Afterward, the results of detection and tea leafhopper analysis were displayed on the web. The results reveal that the designed system is available for detection using images of 4000*3000 pixels within 8s. Test set consists of 25,000 independent images, Recall is 0.9 and the number of FP is only 90. The experimental results demonstrate that the system meets the requirements for real-time (detecting pests in less than 10 s is defined as real-time) and accurate detection of tea leafhoppers.