| Reverse engineering technology,as an emerging engineering design technology developed in recent years,has gradually penetrated into the development process of various products due to its unique advantages.The features of the point cloud model are the key to understanding the 3D model,and provide important reference information for model matching,segmentation,reconstruction and 3D printing.The quality of feature extraction directly affects the extraction accuracy of part design parameters.This article focuses on the 3D scanning point cloud The hole feature extraction method and contour line fitting are studied,and on the basis of fully investigating the existing hole feature extraction algorithms,the existing problems are improved and perfected.The existing hole identification methods mainly design identification algorithms for a point cloud model with a uniform sampling density distribution.At the same time,the detection results in the mechanical part model with complex surfaces often contain more noise points and in the point cloud model containing multiple holes It is not possible to distinguish the boundary of each hole.On the basis of comprehensively considering the existing algorithms,a point cloud hole recognition method for automatic repair of complex parts is proposed.By improving the neighborhood selection method to adapt to the point cloud model with uneven density distribution,after preliminary extraction of the hole boundary,the neighbor The domain-supported method removes the noise points in the detection results,and at the same time considers the deficiencies of the existing point cloud clustering algorithm,proposes an improved European clustering algorithm,and clusters out each set of boundary points,which will lay down for subsequent automatic hole repair basis.In view of the problem that the existing tensor voting algorithm has a high algorithm time complexity and is not suitable for the extraction of hole boundaries of point clouds,etc.,on the basis of studying the existing tensor voting algorithm,a hole identification for improved tensor voting algorithm is proposed,for the input undirected point cloud,theoriginal tensor voting algorithm framework estimates the undirected point cloud normal vector by ball tensor voting,and then encodes the normal vector into the form of a second-order tensor,voting again.although the algorithm framework can adjust the voting form according to the actual needs to achieve different purposes,but the ball tensor voting process algorithm complexity is too high,therefore,other methods are used to estimate the point cloud normal vector and adjust the normal direction to the outside of the model surface.The ball tensor voting process with high time complexity is avoided.At the same time,the existing tensor voting adopts the form of second-order tensor voting.Although it can spread more direction information,it will also increase the calculation time of the algorithm.The first-order tensor is sufficient to express the positional relationship between two points in the hole identification process of the model,so the first-order tensor voting method is adopted The voter voted the direction information of the voter and changed the original bar tensor voting field from pointing to the center of the close circle to adjust the direction along the close circle tangent.The experimental results show that the proposed algorithm can extract the complex point cloud model of mechanical products Holes.When studying the correlation between the eigenvalues of the point cloud neighborhood covariance matrix and the eigenvectors,a point cloud hole boundary detection algorithm based on the covariance matrix is proposed.First,a boundary detection operator is defined by the three eigenvalues of the neighborhood point covariance matrix for preliminary point cloud hole boundary extraction.and then using a computer graph theory method,a hole polygon generation algorithm that comprehensively considers boundary scores and distances is proposed to obtain ordered hole boundary points,and improved the cubic B-spline fits the outline of the hole. |
