Research on Inner Surface Defect Detection Device of Polyethylene Pipe based on Machine Vision

This paper proposes a defect detection system based on Mask R-CNN for the internal surface of polyethylene pipelines, addressing the issue of defects that affect the service life of these pipelines during usage. A defect detection robot designed to collect images of pipes with diameters ranging from 15cm to 20cm is presented. An improved DnCNN network is utilized to denoise the Gaussian noise present in the original images, demonstrating superior performance in PSNR and SSIM evaluation indicators compared to traditional Gaussian filtering methods. Additionally, an intelligent defect detection scheme is proposed, featuring a host computer interface designed to facilitate the application of the model. The system achieves an accuracy of over 87% in defect detection in a simulated pipeline environment.