Point Cloud Feature Extraction And Simplification In Geographic Environment

The construction of urban spatial data infrastructures provides city function mechanism support for the analysis, dynamic monitoring management and auxiliary decision-making. 3D laser scanning and modeling system serves a bran-new technical way for 3D spatial data acquisition and 3D surface model construction, but the process of these mass and non-topological point cloud require highly for the computer resource and storage space. In the process of 3D surface modeling, embed feature lines which extracted from point cloud in 3D models, can guide and constrain the process of surface reconstruction. This method can improve the model precision and increase data compression ratio. Models with different resolution are required for use in different conditions. Modeled with the simplified point clouds, surface model construction can be more quickly.Based on the point cloud which 3D terrestrial Laser scanner system get in geographic scene, in this paper, on one hand, we studied on how to exact features quickly from objects with massive point clouds and improve modeling precision. On the other hand, we studied on point cloud simplification to improve the speed of models construction, and decrease errors brought in the simplification process of point cloud. Main research contents are as follows:1) methods of point cloud collection and pretreatment. Research on the point cloud collection flow of 3D terrestrial Laser scanner system, point cloud pretreatment, such as the separation of point cloud from the back ground, point cloud registration, filtering and resampling.2) Inside feature extraction of the surface. Establish spatial reference for point cloud with linear octree, label point clouds with surface variation, extract feature point with curvature analyze method, connect feature lines with MST. Experiment results indicate that this method can be use in point cloud feature extraction in geographic scene, and improved the speed of feature extraction.3) Boundary feature extraction of the surface. With the local projection method ,analyze the distribution of point cloud, we extract the boundary feature of surface. This method is simple and can get exactitude boundary feature.4) The influence of topology neighborhood in feature extraction. For the problem of the topology neighborhood selection in point cloud differential geometric computation, analyze on influence in feature extraction brought by k-Nearest Neighborhood5) The method of point cloud simplification. Research on typical methods of point cloud simplification and point cloud simplification error, improve the clustering method of point cloudsimplification, improve the simplification accuracy.According to the research contents mentioned above, the main research result obtain is as follows:Establish spatial reference for point cloud with linear octree, label point could with surface variation, improved the speed of feature extraction; improve the clustering method of point cloud simplification, improve the simplification accuracy.