Abstract: In mining tunnels and urban subway engineering,a large number of pipeline facilities are installed along the axial direction. During mobile laser scanning,due to scanning angles,structural occlusions,and sensor placement constraints,point clouds on the back-facing side of pipelines are often difficult to capture,resulting in missing arc segments of the cylindrical structure. This further degrades the completeness and accuracy of 3D reconstruction. To address this issue,this paper proposes a local feature-based geometric fitting method for point cloud completion of pipelines. First,local structural features of pipeline point clouds are analyzed to extract stable geometric parameters. Then,based on the geometric constraints of cylindrical structures,local fitting and point cloud reconstruction are performed in missing regions,enabling continuous recovery of the pipeline surface. Experiments are conducted on point cloud data of subway tunnel pipelines and mine ventilation ducts. The results demonstrate that the proposed method can effectively complete missing back-facing regions and local holes. The reconstructed point clouds smoothly connect with the original data and accurately recover the complete cylindrical geometry. Moreover,the method does not rely on large-scale training data and shows strong adaptability and robustness.