Research On Automatic Registration Of Point Cloud Data In Reverse Engineering

With the development of computer aided design technology, reverse engineering technology has been widely applied in aviation, aerospace, automobile, shipbuilding, mold and other fields, which produces digital model by physical model. Point cloud registration is an important part in reverse engineering, and the stand or fall of registration result directly affects the subsequent data processing and reconstruction precision. The paper focuses on improving the accuracy and efficiency of point cloud registration, and the main research results and the content are as follows:Firstly, a registration method based on fast point feature histogram was presented for the point cloud, which only has the space position information. The method firstly extracts the feature points according to the local characteristic of the normal vector of the point cloud, calculates fast point feature histogram for the each obtained feature point, and determines the initial corresponding set of points according to the feature. Secondly, using the random sample consensus algorithm removes the error corresponding points, gets the accurate corresponding set of points. Quaternion method is used to solve the transform parameters, and transforms the point cloud to the same coordinate to complete the coarse registration. Finally, the improved ICP algorithm is used to get accurate result. The algorithm has high efficiency. The experimental results verify the effectiveness of the algorithm.Secondly, a method with spatial location information and color information is put forward to register the rich color information point cloud. Firstly, the concept of feature figure is proposed and used as the character description of points. The feature points are exacted according to the one dimensional feature figure of the points and the initial corresponding set of points are gotten by the one dimensional feature figure and color information. Secondly, using high dimensional feature figure primarily eliminates incorrect corresponding points, then based on Greedy Bound.algorithm combining with the rigid transformation feature of point cloud further purify error corresponding points to get the accurate corresponding set of points. Then, quaternion method is used to calculate the initial transform parameters, and transforms the point cloud to complete the coarse registration. The proposed feature figure has the advantage of high-dimensional feature and low dimensional characteristic, at the same time, the algorithm makes full use of the color information of point cloud, which improves the efficiency and robustness of point cloud registration.Finally, a point cloud processing system was developed using the Visual Studio 2010 integrated development environment, OpenGL and PCL library. The system has point cloud filtering, sample, registration and other processing functions. Examples demonstrate the process of point cloud processing.