Severity Assessment of Sewer Pipe Defects in Closed-Circuit Television (CCTV) Images Using Computer Vision Techniques

摘要

Currently, visual technologies such as closed-circuit television (CCTV) are commonly utilized for sewer pipe inspection and condition assessment. Inspectors are required to interpret CCTV videos, evaluate the defect severity, and assess the pipe condition manually, which is time-consuming, and the assessment results are subject to the expertise and experience of the inspectors. Computer vision techniques are drawing the attention for interpreting inspection results automatically. Previous studies mainly focus on classifying defect types or detecting the relative location of defects in CCTV videos, but there is a lack of approaches for automated condition assessment of sewer pipes. This study proposed an approach for efficiently evaluating the severity level of sewer pipe operation and maintenance (O&M) defects, such as deposit and tree root, based on computer vision techniques. Firstly, the codes for the severity assessment are designed with references to existing standards. The measurement of the defects is performed by extracting related features from the images, such as the relative area of the pipe cross section and the area of the defects. The edge of the pipe joint and the vanishing point of the pipe ground surface are detected on the image. Then the joint shape model is fitted based on the detected edge and the vanishing point. On the other hand, the relative area of the defects on the image is obtained after performing defect segmentation using deep learning models. Combining the relative area of the pipe cross section and the area of the defects, the severity level of defects is computed based on the loss percentage of the pipe section and the designed severity assessment codes. In the end, illustrative examples are provided using images extracted from CCTV inspection videos to demonstrate the applicability of the proposed approach.