Research on crop pest and disease recognition algorithm based on target recognition

Abstract

As global agriculture faces increasing challenges, the effective identification and management of crop pests and diseases is key to ensuring food security. This paper presents a new crop pest recognition algorithm based on integrated target recognition technology. By integrating the advantages of YOLOv5 (You Only Look Once version 5), EfficientNet and Vision Transformer (ViT), the algorithm designed an efficient and accurate object recognition framework YEV. Specifically, we first used YOLOv5 to conduct preliminary rapid detection to locate possible areas of pests and diseases; Then, the feature information in these regions is extracted by EfficientNet to optimize the utilization efficiency of computing resources. Finally, ViT is used to solve complex pattern recognition problems, especially the identification of pest and disease details. The experimental results show that compared with the single model, the proposed fusion model YEV has significantly improved the accuracy and recall rate, and can more effectively support the development of precision agriculture.