Research On Structural Feature Preserving Filtering Algorithm Of Point Cloud Model

With the rapid development of science and technology,the efficiency and accuracy of 3D laser scanning equipment has been significantly improved.Through three-dimensional laser scanning equipment,the real world objects are displayed in the form of point clouds,so as to realize the rapid digital reconstruction.Therefore,it is used in engineering projects more and more widely.However,due to the effects of specular reflection on the model surface,changes in ambient lighting,and human manipulation,the obtained point cloud may contain some additional noise points.The noise and complex geometric features will affect the subsequent processing(such as segmentation and reconstruction).In recent years,a large number of scholars at home and abroad are committed to the study of point cloud filtering.However,due to the complexity of point cloud data itself and the diversity of noise,it is still a very challenging problem to maintain the structural characteristics of the model while removing the noise.Therefore,for the study of structural feature preservation filtering algorithms,two filtering methods based on structure-guided models are proposed.The proposed two algorithms first filter the normal of the point cloud,and then update the vertex position according to the filtered normal.In order to maintain the structural features,both algorithms have a structural guidance model.The first algorithm is inspired by the low rank matrix model.Firstly,the iterative closest point algorithm is used to find the similar blocks for each local block to form the global similar block normal vector matrix.Then,in order to preserve the structural feature the ₀l norm minimization based on the total variation is used to get surface with structures.Finally,the normal matrix of the original surface is combined with the structure guidance model matrix,and the normal vector field is filtered through the low rank matrix recovery.The second algorithm is inspired by the joint bilateral filtering,which first uses the block moving strategy to generate the structure guided model.Then,we follow the framework of the joint bilateral filter under the guidance of the structure guided model.At last,in order to prevent the gap near the edge,a vertex update strategy based on the isotropic neighborhood is proposed for updating vertex positions.We compare the proposed algorithms with four advanced methods in three aspects of visual effect,comparative test and quantitative analysis.The experimental results show that the algorithm in this paper is superior to other algorithms in structural feature preservation.