ScholarWorks@경상국립대학교

초록

This study proposes an automatic registration process between Building Information Modeling (BIM) and point cloud models to enhance construction accuracy analysis. Specifically, the proposed method utilizes Z-axis and centroid alignment based on RANSAC, global registration using Fast Point Feature Histograms (FPFH), and fine registration via a 2D plane-based Iterative Closest Point (ICP) algorithm. To validate the performance of the automated process, registration results were visualized using heatmaps to analyze spatial error distributions. Furthermore, the effectiveness of the proposed technique was quantitatively evaluated by comparing the global registration stage with the complete automatic registration process using point-to-point distance error histograms and statistical metrics, including Mean, Root Mean Square Error (RMSE), and Standard Deviation. Experimental results using corridor data from a target building demonstrated that the proposed method reduced the mean error by over 80% compared to existing methods, achieving high registration precision. These findings suggest that the proposed approach improves both the efficiency and accuracy of BIM–point cloud integration, thereby streamlining construction quality control tasks while enhancing inspection precision.

키워드