Aerial insulator defect detection method based on CWSP-YOLO

With the rapid evolution of UAV technology, intelligent power equipment inspection via aerial imagery is critical. To address traditional insulator defect detection bottlenecks (high false/missed detection, insufficient multimodal fusion), this paper proposes an improved YOLOv11-based model integrated with multimodal data, featuring cross-modal collaboration, wavelet-optimized C3k2, channel attention, and PIoU v2-based dynamic gradient optimization. Experiments on a self-built dataset show it achieves 84.77% mean average precision, 94.53% accuracy, 82.38% recall, with 24 FPS meeting real-time requirements, offering reliable support for UAV power inspection.