Research On Algorithm Of Point Cloud Registration Based On Neighborhood Feature

With the development of computer-aided design technology, reverse engineering technology which can construct digital model via physical model is more and more widely used. In order to complete the data model of the objects, we need to identify a suitable coordinate transformation, after which data points collected from various perspective directions can be measured in a uniform coordinate system and form a complete point cloud data, then the visualization operation is convenient, this is called point cloud data Registration. In this paper, registration technology is studied deeply, and then a point clouds registration method based on invariant features of neighborhood is proposed, the main contents as following;Firstly, this paper present an improved algorithm that based on weighted target points set in the process of preliminary registration. It could be used to choose the proper corresponding point-pairs by the Gaussian curvature and the concavity and convexity of point cloud, it also can effectively reduce the amount of corresponding point-pairs and descend the time complexity for choosing the correspondent point-pairs.Secondly, an improved algorithm for finding corresponding point-pairs is proposed, it can effectively accelerate the lookup speed by using the methods of binary search and k-nearest neighbors search. Moreover, on the basis of Euclidian distance, an angle constraint is introduced into the algorithm, it can be used to eliminate the bad matches efficiently and this can raise the accuracy of the algorithm.Finally, for the exact registration, the center of gravity of two point clouds is calculated first, and then to improve the initial location of point clouds through adjusts their centroidal distance. Furthermore, two new invariant features of neighboring points are introduced into the algorithm of exact registration, it can be used to eliminate the bad matches efficiently and complete the exact registration.In the part of the experiment, the improved algorithm of preliminary registration and exact registration are implemented through computers. The experiment mainly including the following:contrast experiments of different algorithms and registration based on different viewpoints. Experiments indicate the good results of the improved preliminary registration and the better results of the exact registration, which are an effective and feasible method of registration.