Rescarch On Processing Technology Of Point Cloud From Laser Scanning System

With the development of laser scanning technology, point cloud of3D model’s surface can be collected quickly and precisely, which greatly accelerate the development of reverse engineering. Because point cloud is noisy and redundant, how to efficiently deal with the point cloud to generate3D model for remanufacture is one of the core tasks of reverse engineering. Based on this background, a series of key technologies in point cloud processing are researched.Firstly, the topology structure of point cloud is reconstructed by finding the k-nearest neighbors of every point. Furthermore, the normal and the curvature of a given point are estimated by fitting the local micro-tangent plane and the second parametric surface, which are preparing for the follow-up point cloud processing.Secondly, according to the cause of outliers in the acquisition process, the3D region growing method and the similarity metric function algorithms are used to filter outliers. Then the improved bilateral filtering algorithm is used to smooth point cloud. The improved bilateral filtering algorithm is more robust and better retention features.Thirdly, a new simplify algorithm is designed base on curvature simplify algorithm. A streamlined distance of every given point is calculated by point cloud density and the point’s curvature. Advantages of the curvature simplify algorithm can be inherited very well and "hole" can be avoided effectively by the new simplify algorithm.Finally, point clouds’initial registration is completed by using neighborhood similarity of curvature and rigid transformation invariant and other constraints. On the basis of the initial registration, point clouds’accurate registration is completed by generate the effective initial points that involved in the ICP algorithm.Experimental results demonstrate the effectiveness of the presented approaches.