Detection Model of Drainage Pipe Defect Based on Improved YOLOv5

A drainage pipe defect detection model based on improved YOLOv5 is proposed for the problem of leakage and error detection in the manual inspection of drainage pipes. First, the proposed CSSPPF module is added to the YOLOv5 network to enhance multi-scale feature learning, enhance feature map detail information extraction, and improve the detection precision of similar defect objects. Secondly, the attention mechanism CBAM is introduced to suppress the interference of irrelevant tube wall background information and improve the detection accuracy of the model. Finally, the YOLOv5 backbone network structure is modified to introduce a small object detection layer to obtain a larger feature map and improve the precision of the network for detecting small objects.. The experimental results show that the precision, recall , and mAP of the improved model are improved by 0.5%, 2% , and 2.5%, respectively, compared with the original YOLOv5 network model. The experimental results show that the precision, recall , and mAP of the improved model are improved by 0.5%, 2% , and 2.5%, respectively, compared with the original YOLOv5 network model. Compared with the mainstream detection algorithms YOLOv4, SSD , and YOLOv7, the mAP of the improved model improved by 26.6%, 12.4% , and 7.5%, respectively. And it has stronger feature extraction ability and higher recognition rate for drainage pipe small object defect detection, which meets the demand for efficient and intelligent drainage pipe defect detection and recognition.