Map Construction for Underground Pipeline Based on Improved Iterative Closest Point

Accurate underground pipeline maps are crucial for infrastructure management, providing essential spatial data including the precise location, orientation, and dimensions of the pipeline. Inadequate mapping precision will severely impact subsequent construction and maintenance. However, the classic point cloud registration algorithm Iterative Closest Point (ICP) may produce substantial drift during extended pipeline mapping. This paper presents an improved ICP algorithm specifically designed for pipeline robotic mapping applications. Firstly, the Random Sample Consensus (RANSAC) method is introduced to extract linear features from local point clouds. Subsequently, the direction vectors of the feature lines are incorporated into the optimization equation. Additionally, a novel method for solving the rotation matrix is proposed, which simultaneously considers both point and line features to compensate for errors in point-to-point matching. Finally, laser mapping experiments were conducted on a simulation platform and in a pipeline environment to validate the feasibility of the proposed method in pipeline mapping.