Research on Optimization of YOLOv5 Model for Pest and Disease Detection in Kidney Beans

In modern agricultural production, rapid and accurate detection of pests and diseases is crucial. In this study, an improved YOLOv5 object detection model is proposed to meet the real-time pest and disease detection needs of kidney bean. Efficient pest and disease detection is achieved by introducing the C3CA attention mechanism to improve the feature extraction capability and by using the BiFPN module to optimise multi-scale feature fusion. Experiments on a 680-image dataset show that the improved model achieves 90.3% mAP and 42 FPS detection speed, representing a 7.8% mAP improvement and 3 FPS reduction over the baseline model. Ablation experiments confirm the effectiveness of each enhancement module, particularly in detecting small targets in complex backgrounds. Compared to traditional manual methods, this study provides a new technical solution for intelligent monitoring of crop pests and diseases that can potentially reduce pest control costs by 30-40% and save 50% of inspection time. The model is easy to deploy on mobile devices, bringing the solution within reach of farmers. For the development of cost effective and efficient smart agriculture, this advancement has significant implications.