Research On 3D Point Cloud Registration Technology For Cultural Relics Protection

Digital cultural relics protection refers to the use of computer technology,software technology and other information-based methods to protect cultural relics.With the development of laser technology,obtaining point cloud model by laser scanning has gradually become one of the main means to obtain digital information.Laser scanning can effectively solve the problem of information acquisition and storage of cultural relics,and avoid the secondary damage to cultural relics.However,laser scanning also has problems such as noise and defects.Aiming at the problems of laser scanning model,this paper focuses on point cloud filtering and point cloud registration Aiming at the problems of noises and defects in laser scanning model,this thesis focuses on point cloud filtering and point cloud registration.As a preprocessing work of registration,point cloud filtering aims to remove the noise contained in the scanning model.Point cloud registration is to adjust the models scanned from different angles to a unified position to obtain a complete point cloud model.The main work of this thesis is as follows:（1）In order to obtain higher accuracy of registration,point cloud filtering is used as its preprocessing.Aiming at the problem of small-scale noise plane close to the model contained in the point cloud model,a point cloud filtering method based on tensor voting and multi-scale normal vector estimation is proposed in this thesis.Firstly,in order to ensure that the flat areas in the model are not damaged,the tensor voting method is used to separate the high-frequency information areas and low-frequency information areas of the point cloud;Next,using the low-scale（less neighborhood points）normal vector estimation of high-frequency information points can make the direction of the normal vector at the noise plane tend to be consistent,and then the plane extraction using random sampling consistency can effectively remove the small-scale noise plane close to the model;Finally,the curvature combined with threshold is used for further filtering to eliminate the possible residual noise points,lay a foundation for the registration of cultural relics protection.（2）In order to meet the needs of digital cultural relics protection,a complete cultural relics point cloud model is obtained.In this thesis,a point cloud registration method based on improved covariance matrix descriptor is proposed to register the filtered point cloud.Firstly,in order to reduce the amount of computation,the Intrinsic Shape Signature key point extraction method is used to obtain the key points of the point cloud model;Next,in order to make full use of the neighborhood information of point cloud,a covariance matrix descriptor combining angle,Point Feature Histogram and curvature is used to calculate the corresponding descriptor for each key point;Then,the geodesic distance is calculated by using the two-way nearest distance method and the generalized eigenvalue to find the matching pairs,and the rotation angle is used as a constraint to eliminate the wrong matching pairs to improve the registration accuracy;The transformation matrix is calculated by singular value decomposition and refined by iterative nearest point,lay a foundation for the reconstruction of cultural relics.（3）For the above work,under the environment of Visual Studio 2017,a variety of cultural relics point cloud models are experimentally verified by using Point Cloud Library.The experimental results show that the filtering method in this thesis can effectively remove the noise contained in the scanning point cloud,and the registration method can obtain high-precision registration results.