Particle Swarm Optimization-based Shape Reconstruction of Pipelines using 3D Point Clouds

Abstract

This paper presents a Particle Swarm Optimization based algorithm for the geometric shape reconstruction of pipe bends using 3D point clouds obtained internally from vision sensors. The method utilizes the modular Pipe-Z system to model uniform pipe bends, accommodating various bend angles, radii, and orientations. Through simulations, the study evaluates the impact of the number of modules on accuracy and computational time, and demonstrates the robustness of the algorithm to practical occlusions, such as data loss beyond the field of view and absence of points on internal surfaces. Experimental validation using an Intel RealSense D435 depth camera in a real-world setup shows the practical applicability and accuracy of the system, achieving percentage error of 1 to 2 in large-diameter environments, and maintaining an error below 5 percent even in highly occluded, small-diameter commercial pipes. This research contributes to autonomous navigation and inspection of internal pipe structures, with efficient modeling achievable within approximately 3 seconds with fine-tuned parameters.