Abstract:To address the issues of missed and false detections caused by blurred defect edges, unclear textures, complex background interference, and varying target scales in the damage detection of conductors and ground wires in concealed parts of transmission lines, this study conducts research on such damage detection and proposes a CMAF_DETR detection algorithm based on RT-DETR-Resnet18. Specifically, a Multi-scale Edge Information Selection module is designed to promote multi-scale feature extraction while reducing the parameter count and computational complexity of the algorithm. Moreover, a High-low Frequency Separation Feature Attention is integrated into the intra-scale feature interaction mechanism to optimize the feature interaction process. Furthermore, a Multi-scale Feature Focusing and Bi-directional Fusion Pyramid Network is constructed to enhance the multi-scale feature fusion capability of the algorithm. Experimental results verify that, compared with the baseline algorithm, the proposed CMAF_DETR algorithm reduces the number of parameters by 15.7% and improves the average precision by 2.8%. Practical application verification indicates that this algorithm provides an efficient and stable multi-scale detection solution for transmission line inspection.