LAUSR

Extensive use of three dimensional models in various areas indicates the importance of 3D data retrieval accuracy. In this paper, a directional graph, is introduced for 3D model retrieval. The proposed directional graph is isometric invariant and extracts the features of 3D model vertices in a way that both provides a sample of the features of each salient part and illustrates how various parts are linked and connected with each other. Directionality of the graph does not refer to a certain physical direction, but it shows the arrangement of different points in the graph. The proposed method commences with determining the salient points and continues with constructing a directional graph. Each branch of the graph starts from one salient point in the 3D model and ends in another salient point of that model. The points between the beginning and end of each branch are the vertices of 3D model from which the shortest geodesic path between these two salient points crosses. The whole graph of each model is constructed out of the accumulation of multiple branches, each of which is associated with a pair of salient points. After constructing of the directional graph, WKS is calculated in the path of graph points as the geometric feature of the model. In fact, the points of the directional graph do not provide any information on the object features and they only specify the location where the features should be calculated. The features are calculated and placed in the graph points so, the final graph is built. After completing the feature extraction process, the difference between various models is estimated using Munkres Assignment problem. The experimental results indicate the effectiveness of the directional graph in object description for non-rigid 3D model retrieval. Comparing the proposed method with other approaches by computing the evaluation parameters as well as investigating the computational complexity substantiates the superior performance of it.