A Method Of Rigid Registration Based On Local Surface Approximation

With the development of computer aided manufacturing technology,reverse engineering has been more and more widely used.How to transform the sampled data of different angles of the model into the same coordinate system to form a complete digital model,that is,the registration of the scanning point cloud is the core link in the reverse engineering technology.The accuracy of the registration of the scanning data also affects the quality of the subsequent sample normal estimation,feature recognition,and surface reconstruction.However,the current registration method generally has the problems of low registration efficiency,slow convergence rate and high convergence error.Therefore,this paper focuses on these problems existing in the midpoint cloud registration of reverse engineering,How to improve the efficiency of rough registration and improve the accuracy of precision and precision of the study carried out in-depth system,proposed the minimum package of the base box to achieve rough registration,point-moving least squares surface precision registration method,the main progress is as follows:1.In order to improve the computational efficiency of rough registration,this paper presents a fast method for solving the minimum bounding box,and achieved rough registration based on the minimum bounding box.By simplifying the non-characteristic region of the sampling point set,the minimum bounding box of the simplified set of points was taken as the locating space for solving the minimum bounding box of the original sampling point set,and the minimum axial bounding box under the space was taken as the original sampling point set the minimum bounding box.The results show that the proposed algorithm can improve the efficiency of solving the problem of bounding box solution,and obtain the rough registration of the two-angle point cloud based on the minimum bounding box,which avoids the iterative process and provides a good initial Pose.2.In order to improve the accuracy of the corresponding point pairs in the fine registration,this paper proposes an adaptive moving least squares reconstruction algorithm based on the improvement of the calculation method of the compact least squares surface reconstruction taper branch radius based on the empirical formula.By using the method of principal component analysis,the compacted radius of the moving least squares surface reconstruction was extended to the principal component direction of the local sample,and the adaptive adjustment of the surface reconstruction is realized.The moving surface of the model surface was reconstructed by using the moving cube algorithm.The adaptive reconstruction of the surface was more accurate than the original moving least squares,and the algorithm was applied to the registration of the scanning data,which improves the accuracy of the corresponding point pair and the precision of the registration3.In this paper,we propose a precise registration method for point-moving least squares surfaces.By constructing the adaptive moving least squares surface of the local sample of the target point cloud,the nearest point of the least squares surface of the local point of the source point cloud was used as the matching point pair and the confidence degree of the matching point pair was defined Function to filter the initial matching relationship.Finally use the singular value decomposition method to solve the transformation matrix.The results of the final test show that the accuracy of the algorithm is higher than that of the classical ICP algorithm,and the convergence time is only about 30% of the ICP algorithm.Based on the point matching,the algorithm is more stable.