基于ID-YOLO的航拍图像绝缘子缺陷检测方法OA

To address the challenges of insulator defect detection in UAV-based inspection scenarios,including interference from complex backgrounds,limited computational resources on edge devices and inaccurate localization of small targets,this paper proposes a lightweight and efficient detection algorithm named insulator defect-YOLO(ID-YOLO).First,a dual-branch sampling fusion(DBSF)module is constructed to preserve multi-scale feature details while reducing the number of parameters.Second,a cross-stage feature mixing with squeeze-and-excitation(CFM-SE)module is designed to enhance the semantic representation of defect-sensitive regions under cluttered backgrounds by integrating a channel attention mechanism.Furthermore,a defect-aware symmetric difference balanced loss function(DASD)is introduced,which dynamically adjusts the weight of symmetric difference sets to optimize the localization performance for small targets.The experimental results on a self-built insulator defect dataset demonstrate that ID-YOLO achieves an mAP of 94.2％,outperforming YOLOv9,YOLOv10 and YOLOv11 in detection accuracy.When deployed on the NVIDIA Jetson Orin Nano embedded platform,the model reaches an inference speed of 55.3 frames per second,exhibiting strong real-time performance and good deployment adaptability.