Development of Quantifying Internal Defects in Water Pipes Based on a Monocular Image Vanishing Point Detection Technology

초록

Water pipe systems, which are a type of social infrastructure, provide essential services for human life. However, as they are buried underground, access is limited, causing considerable difficulty in internal exploration for maintenance. Previous defect detection in sewer pipes is relatively straightforward due to stable camera positioning and clear image geometry under water shutdown conditions. However, in water distribution pipes, where inspection must be performed using nondisruptive (non-shutdown) endoscopic cameras, varying camera orientation and image distortion make defect detection and quantification significantly more challenging. To address this, this study proposes a method for estimating the vanishing point from monocular endoscopic images to geometrically align the view and reduce perspective distortion, enabling more accurate defect analysis. The proposed monocular vanishing point detection (Cs-MVP) technique utilizes brightness channel extraction, binarization, Canny edge detection, and the Hough circles method to detect circular cross-sections and estimate the vanishing point. The proposed method is capable of detecting reliable vanishing points with an average measurement accuracy exceeding 96% in refined environments. Although results indicate that area errors of up to 1.5 times may occur under intense noise conditions, this approach minimizes human error typically associated with conventional secondary manual evaluations. By providing essential geometric cues for defect assessment, the method is expected to support more objective and scientific maintenance decision-making.

키워드