Research on recognition of Empoasca flavescens based on machine vision

Abstract. As one of the major tea pests, Empoasca flavescens has great harm to the tea production. Without enough information about this pest situation, the traditional method for treating the pests is spraying chemical pesticides blindly and frequently, which leads to pollute the environment and threaten the food safety. Therefore, accurate observation and forecast of Empoasca flavescens is helpful to take appropriate management strategies in tea garden. In this study, the machine vision technique was introduced to automatically recognize the Empoasca flavescens on the yellow sticky traps in natural scenes. The superpixels segmentation algorithm and DBSCAN (Density-Based Spatial Clustering of Applications with Noise) cluster algorithm were employed to separate interesting target regions from background. However, the Empoasca flavescens can be easily regarded as background under complex environment. Therefore, this study obtains the target image by fusing two clustering images with different thresholds. Then, the six classification features, including mean value of L, a, and b and their standard deviation, were extracted from the marked area in target image. Last, the least squares support vector machine (LSSVM) was developed to identify Empoasca flavescens from other insects that were captured by sticky traps. The proposed algorithm achieved 98.81% of the overall recognition accuracy, and the identification accuracy of Empoasca flavescens was 91.29%. The method can provide an effective way for real-time detection of Empoasca flavescens.