Abstract: The quality of line and surface fitting results has a direct impact on the accuracy of subsequent data processing, making it a crucial step in point cloud data processing. To address the limitations of traditional RANSAC (Random Sample Consensus) algorithms in the effectiveness and accuracy of plane fitting for buildings, we introduce a sequential normal vector detection mechanism and propose an improved algorithm called M-RANSAC. The M-RANSAC algorithm enhances plane and edge detection by introducing increased angle constraints to the fitting model. To validate the feasibility and effectiveness of this method, a comparison was made with the conventional total station measurement method and the traditional RANSAC algorithm. The results confirmed the superiority of the proposed approach. Furthermore, the method was successfully applied in the current situation survey of buildings along the Pearl River Delta water resources allocation project. In conclusion, the M-RANSAC algorithm offers improved accuracy and efficiency in line and surface fitting.

Key words: point cloud data, line and surface fitting, improved random sampling consensus (M-RANSAC), building status survey