CACS-YOLO: A Lightweight Model for Insulator Defect Detection Based on Improved YOLOv8m

Insulator defects in transmission lines can threaten the safe and stable operation of the power system. Therefore, insulator defect detection plays a crucial role in the electricity sector. Currently, the you only look once (YOLO) algorithm is widely used in insulator fault detection. However, insulator defect detection using YOLO still faces the double challenges of detection precision and real-time performance. Therefore, we construct a lightweight model for insulator defect detection named coordinate attention mechanism (CAM) and feature channel shuffle operation (CSO) YOLO (CACS-YOLO). By introducing the CAM and CSO in the original YOLOv8m, we improve the detection precision and reduce the parameters of the model, compared with existing insulator defect detection models. Meanwhile, in order to solve the problem of insufficient data and cope with different weather conditions, we use synthetic weather algorithms for data enhancement and construct a synthetic foggy and rainy insulator dataset (SFRID). The experimental results show that the mean average precision (mAP) of CACS-YOLO is 96.9%, and the parameters of the proposed model are 27.44 M. These outcomes show the high-precision and lightweight of our proposed model in insulator defect detection. The source code and dataset are available on GitHub.