A Multi-Scale Feature Network for Defect Detection in Substation Equipment

To address the challenges in defect detection of substation equipment, including large variations in target scale, complex background interference, and insufficient recovery of edge details, this paper proposes a defect detection method for substation equipment based on multi-scale feature perception. First, an MSEEM module is designed. It integrates multi-scale depthwise convolution, edge prior enhancement, and channel-adaptive calibration to improve the model’s ability to represent defects at different scales. Second, a C2SCSA module is constructed to select effective features from both spatial and channel dimensions. This enhances the representation of defect regions under complex backgrounds. Finally, the DySample dynamic upsampling module is introduced during the feature fusion stage to adaptively restore fine-grained information through content-aware sampling, thereby improving bounding-box localization performance. Experiments on the substation equipment defect dataset constructed in this study show that the improved model can accurately identify different types of equipment defects in complex scenarios. Relative to YOLOv11n, MSDFNet increased precision, recall, mAP@0.5, and mAP@0.5:0.95 by 3.3, 7.0, 4.9, and 7.3 percentage points, respectively. These results indicate that the method improves the detection accuracy and localization ability of defects in substation equipment while maintaining low computational overhead.