基于图的网格分割

Graph based mesh segmentation

为了简化模型反求过程，提出了一种基于图的网格分割算法，分别以网格中的顶点和三角面片作为图中的元素，生成2种未定向加权全连通图的表达．对于由顶点生成的图把网格顶点的几何信息转化为颜色信息，对曲率计算引入的噪声采用中值滤波和均值滤波的方法滤除，并利用顶点之间颜色的差别作为图中连接相邻元素的边的权．对于由三角面片生成的图，利用相邻面片的二面角作为图中相邻元素的边的权；然后利用一个不相交集合的森林分割图，进而实现网格的分割．结果表明，这种分割方法可以快速、有效地实现网格的分割．

To simplify the reverse engineering process of model, a graph based mesh segmentation algorithm was proposed. Two kinds of undirected, weighted graphs were constructed to represent a mesh. Vertices and triangles of the mesh were taken as elements of the graphs respectively. For the graph constructed from vertices, the geometry information of mesh vertices was transformed to color information, and the noise introduced by curvature calculation was filtered by median filter and mean filter. The color difference between two neighboring vertices was taken as the weight of the corresponding graph edge. For the graph constructed from triangles, the dihedral angle of two adjacent triangles was taken as the weight of the corresponding graph edge. The graph was segmented by a disjoint-set forest, so the mesh. Experiment results indicated that the method can quickly and efficiently segment meshes.