Research on 3D Reconstruction and Point Cloud Data Processing Algorithms and Their Automated Precision Control for Complex Steel Structure Construction

Abstract

Complex steel structure construction requires stringent geometric accuracy control, yet conventional inspection workflows are often labor-intensive, discontinuous, and insufficient for capturing localized deviations on intricate members and joints. This paper investigates a 3D reconstruction and point cloud data processing pipeline for automated precision control in complex steel structure construction. The proposed framework reconstructs site geometry from multiview sensing and generates dense point clouds that are subsequently denoised, registered, and segmented to isolate key structural components. A model-driven alignment strategy is employed to register reconstructed point clouds to design references, enabling automated extraction of geometric quality indicators such as dimensional tolerances, member straightness, surface flatness, and joint alignment errors. To ensure reliability under occlusion, clutter, and varying acquisition conditions, the pipeline integrates robust outlier suppression and adaptive registration constraints, and introduces an automated deviation evaluation mechanism that localizes error sources and supports corrective decision-making during construction. Experimental validation across representative steel construction scenarios demonstrates that the proposed approach can produce stable reconstructions and accurate deviation measurements, enabling continuous, data-driven precision monitoring and reducing reliance on manual rework cycles. The results indicate that the framework is practical for on-site quality assurance, providing scalable automated precision control for complex steel structure construction.