3D Models´ retrieval system design based on Poisson´s histogram of 2D selective views

In this paper, we provide an effective new 3D models´ retrieval system based on Poisson equation. Usually, 3D models have considered in two major kinds: directly in 3D space and indirectly by a set of 2D views that extracted from that model. Compared with the obtained results of different kind of these kinds of methods, we try to design our 3D models´ retrieval system on special 2D views that select by using Kmeans clustering method that uses a set of geometric features of each silhouette to result the most discriminating views. This will be great helpful to improve any 3D model retrieval system performance. After finding the best views of each 3D model, we will use the Poisson equation to define 2D views´ shape signature that used in a retrieval system in histogram form. We use the 3D shapes of McGill database to verify the performance of our proposed 3D models´ retrieval. The simulation results show that proposed method is better performance rather than some existing 2D view and histogram based shape descriptors in retrieving correctness.