Automatic Identification Of Rules Based On The 3d Point Cloud Shaped Surface

At present the three-dimensional object recognition is a hot point in the computer vision and the pattern recognition, as the regular object like plane, spherical surface, and column surface and so on constituted the complete surface or the part of the object in the real world, therefore, it is an important basic research that the regular surface recognition from laser scanning point clouds. The three dimensional contour line is significant in the expression methods of the scene and mountain, which is involved in this article, and is regarded as one kind of special regular shape surface.The regular surface recognition refers to the point cloud slicing, the surface parameter extraction and so on, on the foundation of existing theory, this thesis proposes the point cloud slicing method based on the octree model, the regular surface is recognized according to the relationship of the regular surface and the curvature in the slicing process, and then the regular surface parameter is extracted by the RANSAC algorithm. Regarding the contour line, the Kringing of spacial gridding method is used for repairing the empty hole of the datum, and the problem that point cloud distributed belt-shaped is processed with the filtring method in plane, thus the correct contour line is extracted from the three dimensional point cloud.In this thesis the main results is as below:(1)In the data slicing and recognition aspect, this article uses the law vector, curvature information and the octree theory, for subdividing the point cloud to realize the point cloud slicing gradually; Then based on the relations between the curvature information of the boundary unit and the regular surface, each local region block is recognized.(2) In the regular surface parameter extraction aspect, this article adopts the robust RANSAC algorithm and seeks model parameter according to the sample, which is taken out many times from the point clouds at random. These model parameter is examined through the entire point cloud for electing the optimal model. In view of the computational complexity of the algorithm, the pre-testing procedure is added, namely replaces the complete data examination with the partial data, only the pre-examinated model is carried on the entire data examination, thus the efficiency of the algorithm is enhanced.(3) In the contour line extraction aspect, the Kringing of spacial gridding method is used to interpolate empty in this article, and its interpolation effect closes to the reality. At the same time, spheroid search method is used to filter data.