Research On Reconstruction And Smoothing Technology Of Triangular Mesh Model

In the wake of developments in 3D reconstruction technology,3D surface model has attracted more and more attention.Due to its many advantages such as intuitionism and stereoscopic,this technology has been applied to many fields.At present,3D reconstruction technology has made great progress,but there are still some problems.For example,the shape of some triangular surfaces in the model is too long and narrow,and the density of the triangular surfaces in different regions of the same model varies greatly,so the quality of the model is poor.When the triangular mesh model is optimized,the contour and edge information of the model is not well protected,which greatly reduces the refinement of the model.The 3D model of large scene consumes the memory of computer very seriously.Some models with large amount of data can not even be processed by computer.At the same time,due to the large amount of data,the efficiency of some 3D model optimization algorithms is so low that they cannot meet the requirements of real-time interaction.In view of the above problems,the main contributions to this paper are as follows:1)To solve the problem of detail loss in incremental mesh reconstruction algorithm,the feature edge detection algorithm(ESOD)is introduced,and the ESOD algorithm is improved.According to the features of feature edges in the model,the weighted term of area and distance is introduced.The improved ESOD method is used to detect the feature edge,and the feature edge is kept unchanged in the subsequent mesh reconstruction to achieve the purpose of detail protection.In this paper,Hausdorff distance is used to measure the merits and demerits of the mesh reconstruction algorithm,and the merits and demerits of the algorithm are quantitatively analyzed by comparing the deviation degree between the reconstructed mesh and the original mesh.2)Aiming at the problem that the normal vector smoothing method has poor effect on small noise points,this paper adds noise detection,finds out the noise surface containing noise points by calculating the sharpness of the vertex,and adopts different smoothing strategies for the noise surface and the general triangular surface in the subsequent corrected normal vector calculation.At the same time,a solution to the low efficiency of the normal vector smoothing method is proposed,which adopts the GPU multi-thread acceleration technology to calculate and correct the normal vector and the vertex coordinate offset in parallel.Under the premise that the result is completely consistent with the original algorithm,the efficiency of the algorithm is significantly improved.3)For the problem that the model data is too large to be processed by computer,this paper adopts an algorithm of equal data volume partitioning.This algorithm can divide the mesh into a specified number of subnets,while ensuring that the number of vertices of different subnets is relatively close,so as to avoid the problem that the model irregularity leads to too big difference between subnets.After the chunking process is complete,the subnets are combined.