An Improved YOLOv9-Based Defect Detection Method for Catenary System Support Device Components

The detection of component defects of catenary support devices is a key task to ensure the safe operation of railway electrification systems. Aiming at the problems such as high missed detection rate of small target defects and insufficient positioning accuracy of the existing detection methods in complex industrial scenarios, this study proposes an improved detection framework of YOLOv9. The multi-scale feature extraction ability is enhanced by introducing the full-dimensional dynamic convolution module (ODConv), the Haar wavelet downsampling (HWD) is adopted to retain high-frequency detail information, and the Slide weighted Loss function (Slide Loss) is designed to optimize the bounding box regression accuracy. Experiments demonstrate that the average accuracy of the enhanced model on the catwire component defect dataset (mAP@0.5 and mAP@[0.5:0.95]) reaches 71.3% and 48.8%, respectively, representing improvements of 13.8% and 4.2% compared to the original YOLOv9 model, while maintaining a detection speed of 29 FPS. This performance surpasses mainstream algorithms such as YOLOv7 and Faster R-CNN. The proposed method offers a novel approach for achieving high-precision real-time detection in complex industrial environments.