An Automatic BIM-Based Construction Quality Inspection and Acceptance Method Using 3D Point Cloud Deep Learning
Bin Li, Zhiwei Zhang, Deepak Ranga
Accurate and efficient quality inspection and acceptance are essential for ensuring construction performance and reducing rework in building projects; however, conventional inspection methods are labor-intensive, subjective, and difficult to scale in complex construction environments. To address these challenges, this paper proposes an automatic BIM-based construction quality inspection and acceptance method using 3D point cloud deep learning, which integrates interdisciplinary techniques from construction engineering, computer vision, and artificial intelligence. The proposed framework combines building information modeling (BIM) with high-resolution 3D point cloud data acquired from laser scanning or photogrammetry to enable data-driven and knowledge-informed quality assessment. First, the as-built point cloud is registered and aligned with the BIM model to establish precise spatial correspondence between designed and constructed components. Then, a deep learning–based 3D point cloud analysis network is employed to perform component segmentation and geometric feature extraction, allowing deviations in dimensions, position, and surface quality to be automatically detected. Furthermore, quality inspection rules and acceptance criteria derived from construction standards and engineering specifications are encoded and integrated with data-driven analysis results, enabling objective pass/fail decisions and quantitative deviation reporting. Experimental results on representative building scenarios demonstrate that the proposed interdisciplinary approach achieves higher inspection accuracy and efficiency than traditional manual inspection and rule-based geometric comparison methods, while significantly reducing human intervention. The results indicate that the proposed method provides a reliable and scalable solution for automated construction quality inspection and acceptance in BIM-enabled digital construction environments.