Automatic Detection Method of Sewer Pipe Defects Using Deep Learning Techniques

Regular inspection of sewer pipes can detect serious defects in time, which is significant to ensure the healthy operation systems and urban safety. Currently, widely used closed-circuit television (CCTV) system relies mainly on manual assessment, labor intensive inefficient. Therefore, it urgent develop an efficient accurate automatic defect detection method. In this paper, improved method based YOLOv4 proposed for defects. A improvement using spatial pyramid pooling (SPP) module expand receptive field improve ability model fuse context features different fields. Meanwhile, influence three bounding box loss functions performance are compared their processing speed accuracy, effectiveness combination DIoU function SPP verified. addition, address lack datasets detection, a dataset that contains 2700 images 4 types was created, provides useful help application computer vision techniques field. Experimental results show that, with model, mean average precision (mAP) by 4.6%, mAP reach 92.3% recall 89.0%. The effectively classification accuracy defects, has advantages other methods.