Contamination detection scheme of glass insulators based on yolov8 classification and case segmentation

In the inspection of power systems, there exist problems like low automation, inefficiency, and safety hazards in the traditional manual inspection method. For the important process of insulator contamination detection, UAV target detection methods have broad application prospects. Although there are strategies for network training in traditional insulator contamination detection, the general detection accuracy and recall rate of models are relatively low. After experimental verification, the yolov8-based instance segmentation insulator contamination detection technology performs exceptionally in metrics like accuracy, recall, detection speed, and map, far surpassing traditional target detection methods. The detection map is improved by nearly 12% compared to common yolov5 target detection algorithms. The classification algorithm studied in this paper provides a reliable reference for insulator contamination detection in power systems.