3D model identification method based on topographic sparse coding

Deep learning is a hot research topic at home and abroad. In view of the massive 3D model data, a 3D model identification method based on topographic sparse coding is proposed. Firstly, the 3D model is projected to get its three views, and the feature vector is used as the training sample. A complete dictionary is constructed by using the greedy algorithm to train the dictionary, which makes the normal samples show good sparsity in the constructed dictionary. Finally, the feature learned by the topographic sparse coding is used to represent 3D model. The experimental results on the Princeton Shape Benchmark dataset showed that compared with the principal component analysis and self-learning neural network, the proposed method has better recognition performance.