Abstract
<title>Abstract</title> <p>To address the challenges of complex backgrounds, large defect scale variations, and the issues of false positives and missed detections in insulator defect detection for transmission lines, this paper proposes a DTRW-YOLO11s insulator defect detection algorithm that integrates multi-scale perception and triplet attention mechanisms. First, a DSDown downsampling module is introduced into the backbone network to enhance the model’s ability to extract features from targets of different scales during the downsampling process, thereby reducing the loss of small-target feature information. Second, the Triplet Attention (TA) mechanism is incorporated into the final layer of the backbone network to strengthen the correlation between channel and spatial dimensions, improve the model’s focus on global information, and suppress background noise interference. Then, by integrating the Receptive Field Channel Attention Convolution (RFCAConv) into the C3k2 modules of both the backbone and neck networks, a C3k2_RFCAConv module is constructed to achieve effective perception of multi-scale defect features. Finally, the WIoU loss function replaces the original CIoU loss function to improve the model’s localization accuracy and generalization performance. Experimental results show that, compared with the baseline model, the improved DTRW-YOLO11s network achieves a 4.1% improvement in mAP@0.5, with Precision and Recall increased by 3.3% and 4.2%, respectively. Meanwhile, the number of Parameters and FLOPs are reduced by 10.6% and 19.7%, respectively, achieving a balanced trade-off between detection accuracy and computational efficiency. The proposed algorithm effectively meets the requirements of daily power line inspection tasks.</p>