Algorithm For Feature Extraction Of 3D Model Based On Topological Structure

With the rapid development of 3D model,3D model retrieval has become a research hotspot in the field of computer vision.In the feature processing technology of 3D model,topological structure feature as a method of obtaining the shape information,holds a very important position,the most frequently used topology information including the branch and connectivity of 3D model.The skeleton can express the topological information of the model very well,preserve the topological connectivity and shape of the model,and ignore the redundant information of the original model.The skeleton is an important topological descriptor of the object.This paper presents a skeleton extraction algorithm of mesh model based on Reeb graph,according to the principle of Reeb graph,an improvement is presented on the basis of the algorithm.The proposed method modified the definition of the μ function value and uses the geodesic distance to calculate the classification value of the model vertices.This paper subdivides the mesh model,using Dijkstra algorithm to solve the geodesic distance from vertices,using Dijkstra algorithm to calculate the geodesic distance,not only use of the shortest path effectively,but also help to obtain more accurate results.This paper extracts the skeleton of 3D model as the whole topological feature,on this foundation calculates the similarity between the models and implements the 3D model retrieval system based on Reeb graph.Based on the traditional Reeb graph representation method,this paper further presents a new Reeb graph description method,which can be used to parameterize the RG,to strengthenthe topological consistency of the RG representation,and to highlight the geometric characteristics of the model.After constructing the RG,the shape is segmented into Reeb charts,and then the topological map is used to form the standard plane domain.Then,two stretching signatures,corresponding to the area and angle distortion,are taken as three-dimensional-shape descriptor called hybrid signature.The method is used to retrieve the3 D model and improves the accuracy of similarity estimation between the 3D models.