Inner wall defect detection in oil and gas pipelines using point cloud data segmentation

Abstract

Inner wall defects in oil and gas pipelines threaten operational safety. Traditional manual detection is vague and risky. Laser scanning technology offers precise point cloud data for accurate characterization of the inner wall. However, the pipeline's quasi-cylinder model and complex defects complicate detection. This paper proposes a method for defect detection using point cloud data segmentation. It simplifies the pipeline model with cylindrical projection and employs a bidirectional cloth simulation filtering (BCSF) for the rough segmentation, effectively handling intricate geometries, subtle slopes, and bidirectional defects. Density-based spatial clustering of applications with noise (DBSCAN) and region growing (RG) are utilized for the fine segmentation of ambiguous areas. Experimental results show superior accuracy and robustness compared to conventional methods, with a three-axis mean error of 1.9 %, mean Intersection-over-Union (IoU) of 96.6 %, and an 83.7 % reduction in computation time. Thus, this method significantly supports pipeline safety assessment.