Pre-processing of 3D scanning line point cloud data

On the basis of analyzing the characteristics of scanning line point cloud and comparing methods already existed, Algorithms for point cloud preprocessing based on divide-and-conquer strategy is presented. It adopts topological spatial neighborhood theory. Firstly point cloud is distributed to little cubes (square is regarded as degraded cube), then according to cube´s space 26-neighbor relationship, the maximum connected region (MCR) including scanning line point cloud is generated and the data out of MCR is deleted as noises. Using the layering feature and square´s 8-neighbor connection, the combination of point´s sensor property and minimum distance criteria is used for outliers abridging and multi-sensor data integration. Presented algorithms only need to setup the length of cube edge d and the time used for preprocessing of scanning line data including 200,000 points or so does not exceed 8s on common computer. The data preprocessing results and reconstructed surface model show that the algorithms is so effective and efficient that can meet subsequent requirements.