3D Model Recognition via Local Sparse Representation

To classify 3D models whose classification information is unknown prior, this paper proposes a recognition algorithm for 3D models.Firstly, the algorithm extracts feature vectors for each 3D model based on an improved shape diameter function (SDF) feature descriptor.Secondly, each 3D model, whose classification information is unknown, is regarded as the test model.And then the algorithm finds k models, which are similar with the test model, in the 3D models database where each model's classification information is known in advance.Finally the sparse representation classifier is applied to the test model and the k models to determine the classification information of the test model in the 3D models database.Experimental results show that the algorithm is simple and easy to implement.Besides, the algorithm is highly accurate and robust.