Research On Simplification Algorithm Of 3D Laser Point Cloud Data And Triangular Mesh Model Optimization

With the continuous improvement of 3D laser scanning accuracy,3D laser scanning technology has been widely used in deformation monitoring on subway tunnels,the inspection of industrial parts,ancient cultural relics 3D reconstruction.Because of the high accuracy and high density of 3D laser point cloud data,the amount of point cloud data is becoming larger and larger,which brings a series of difficulties to the processing and application of point cloud data.So,research on point cloud simplification with preserved features is the key to solve the application of point cloud data.However,with the improvement of reduction rate,the error of the simplified point cloud data and the original point model increases gradually,and the accuracy is decreasing when the simplifying point cloud data processed by point cloud reconstruction algorithm.In order to both simplify the point cloud and ensure the accuracy of the model,it is necessary to improve the mesh accuracy.Point cloud simplification algorithm and the precision of the simplifying point cloud reconstructing grid was researched in this paper.The important parts and research results of this paper are as follows:As the incomplete retention of features when the point cloud data processed by point cloud simplification algorithm,and data holes which caused by small-curvature point cloud simplification algorithm,a new point cloud simplification algorithm integrated k-means clustering and Hausdorff distance is proposed.The octree was used in the new algorithm to establish the topological adjacency and k-neighborhood of scattered point cloud data.Then the Hausdorff distance was used to analyze the curvature of point cloud data points to determine the features area,and simplify the point cloud preliminarily through the threshold set determined by the simplification ratio.Finally,k-means clustering is employed to process the preliminarily simplified point cloud data to derive the final simplified data.Results indicated that this method can retain the geometric features of the original point cloud data well,and the simplified results distribution is even,the data holes caused by oversimplification is solved.As the increase of error and loss of accuracy when the high-precision and high fidelity simplifying point cloud data processed by point cloud reconstruction algorithm,an object optimization method of complex geometric model for point cloud data is proposed.The octree was used in the new algorithm to build the index of the point cloud and establishes the topological relationship between the point cloud and the mesh,and the method of the original model and the reconstructed model vertex Euclidean distance metric was used to calculated the error of the reconstructed mesh.Then determine the mesh to be optimized according to the expected accuracy,and then subdivide the mesh by patch segmentation(adding-point method).Finally,the pre-insertion point is repositioned according to the insertion point formula,and the minimum error point was used as the insertion point.Results indicated that this method can make the reduction rate 90% of the rabbit reconstructed mesh error from 0.81 mm to 0.48 mm,and the reduction rate 90% of the Dragon reconstructed mesh error from 0.36 mm to 0.11 mm.