Railway tunnel lining structural damage detection algorithm based on improved YOLOv5s

Addressing various defects in the internal structure of operational railway tunnels, such as difficulties in discovering, detecting, and identifying defects like uncompacted, hole, and void, propose the CCB-YOLOv5 algorithm for detecting defects in railway tunnel lining structures. The algorithm is based on the YOLOv5 framework and incorporates the ConvNeXt feature enhancement module into the Backbone network, improving the model's ability to extract features from fine textures of lining defects. Add the Coordinate Attention (CA) module after upsampling in the feature pyramid network of the Neck section to enhance defect feature information and reduce background interference. Introduce the Bidirectional Feature Pyramid Network (BiFPN) to fuse shallow layer defect detail textures, reducing the loss of deep network detail features during sampling. To evaluate the performance of the proposed algorithm, it was compared with SSD, Faster RCNN, YOLOv5m, YOLOv8s, and YOLOv8n. The experimental results indicate that the improved algorithm achieved 92.3% precision (P) and 94.1% mean average precision (mAP) in detecting structural defects in railway tunnel linings, outperforming the other mainstream algorithms in terms of recognition accuracy and overall performance.