一种基于面片的三维重建算法

摘要

The PMVS algorithm does not add space geometric constraint during expanding spare points to dense points, and the Poisson surface reconstruction algorithm outputs a closed mesh model which contains wrong vertices. In order to solve this problem, this paper considers patches around the expanding one at the same time, detects the outlier vertices of triangle meshes according to k Nearest Neighbor(kNN) and removes extraneous portions of the mesh. Experimental result shows that the reconstruction algorithm is more smooth and more accurate.%传统基于面片的多视角立体视觉算法在扩展面片时缺少空间几何约束,导致Poisson表面重建时生成错误顶点.为此,提出一种基于面片的三维重建算法.利用相邻面片初始化扩展面片,根据顶点k近邻的距离检测并去除离群顶点,从而得到物体表面的三维模型.实验结果证明了该算法的有效性和准确性.