Detection of Corrosion Areas in Power Equipment Based on Improved YOLOv5s Algorithm with CBAM Attention Mechanism

During long-term operation, power equipment can generate various forms of rust. Manual inspection consumes a significant amount of manpower and resources, and the results are not always satisfactory. The effectiveness and precision of traditional image processing methods in rust detection also have some shortcomings. In order to improve the timeliness and reliability of mobile devices such as drones in detecting rusty areas on power equipment, we propose an improved method based on the YOLOv5s object detection algorithm, which is optimized by integrating an attention mechanism. We use two different attention modules - CBAM and SE, to improve the YOLOv5s algorithm, enabling it to automatically identify rust areas. Experimental results show that Whether it is the original YOLOv5s, YOLOv5s + SE with the SE attention mechanism, or YOLOv5s + CBAM with the CBAM attention mechanism, all of them can effectively detect rust areas. However, by expanding the dataset and optimizing parameters, YOLOv5s + CBAM exhibits significant improvements in precision, recall, and mAP.