| With the development of precision machining, the traditional measurement methods can not meet the needs in actual measurement anymore. Non-contact methods are often relied on to get dense point cloud of the surface of machine parts. The huge amount of three-dimensional point cloud data is difficult to store and process. furthermore, during the actual measuring process influenced by various factors, three-dimensional model point cloud data obtained by the measuring device will contain a great amount of noise points. Therefore, the requirements of processing obtained 3D point cloud data have become increasingly urgent. Meanwhile, analyzing dimension parameters of machine surface with the means of fitting that is based on point cloud is urgent too.Firstly, for the storage of the large three-dimensional point cloud data, this paper proposes a algorithm based on kd tree structure. In order to adapt to the retained characteristics of the point cloud, an algorithm of keeping a point cloud characteristics is realized. According to the result of experiment, it shows that the cloud point streamline algorithm in this paper can simplify point cloud data quickly and efficiently, so as to meet the actual demand.Secondly, point cloud noise with different attributes are classified which are processed with different methods. An algorithm is proposed to remove noise point by the recognition of the point cloud model based on random sampling consistency principle. This algorithm can process many kinds of noise fast and effectively according to the analysis of experiment.At last, for that the traditional contact measurement methods have been unable to meet the actual measurements, this paper explores regular surface fitting based on point clouds. In order to improve the disadvantage of the initial value calculation of cone, an algorithm based on point clouds normal is proposed. The feasibility and robustness of the algorithm has been verified by experiments. |
PDF Full Text Request