Research On Mesh Segmentation Algorithm Based On Salient Point

The 3D mesh segmentation algorithm is a basic research topic in computer aided design and computer graphics,which is divided into semantic components by processing the data of the input 3D mesh model.The 3D mesh is composed of points,lines and surfaces,and the salient points have obvious protruding and convenient for observation and extraction.In this thesis,two 3D mesh segmentation algorithms based on salient points are mainly studied.Based on MFC framework and OpenGL display technology,the input model can be read,displayed,rotated and zoomed.The graph of input mesh model is constructed to make data structure save processing data better.Two algorithms for finding the shortest path of single source points are compared,and the Dijkstra algorithm is used to obtain the geodesic distance.The characteristics of the convex function are studied,according to the differences between center point values and the convex points function value,extract the convex points of the mesh model using the convex function.After extracting convex points,clustering algorithm is used to group the extracted convex points.Thus,the preprocessing of input mesh model data is realized.Segmentation algorithm for 3D Mesh with local protrusions is researched in this thesis.The input 3D mesh models need to have obvious protrusion properties,and the models can be segmented into semantic components.Firstly,the salient points of the 3D model are extracted,clustered and grouped.Then the central area of the mesh is estimated.Finally,the segmentation boundary is found by using the minimum cutting algorithm to further refine the segmentation area.It is showed in experiment that segmentation algorithm in this thesis improves the robustness and segmentation efficiency of the program.And at the same time makes the components more explicit.The K means clustering segmentation algorithm with prior conditions is introduced to carry out a unified multidimensional scaling transformation of the 3D mesh model.The significant points and the feature patches within a certain range of distance are extracted as the prior information of the algorithm.And the fast presegmentation of an iterative K means clustering is carried out on the mesh model.The segmentation efficiency is improved in order to enhance the semantic meaning of segmentation results.On this basis,a post segmentation algorithm based on hierarchical band pushing regional growth is studied.Experiments show that the segmentation boundary is smooth by using this algorithm and the segmentation effect is obvious.Compared with presegmentation,segmentation semantics are improved.