Point-BLS: 3D Point Cloud Classification Combining Deep Learning and Broad Learning System

3D object recognition and detection based on point clouds is an important research topic in computer vision and autonomous navigation. Nowadays, deep learning algorithms have significantly improved the accuracy and robustness of 3D point cloud classification. However, deep learning networks usually suffer from complex network structures and time-consuming training process. In this paper, we proposed a 3D point cloud classification network Point-BLS, which combines deep learning and broad learning system together. Specifically, we first extract point cloud features through a deep learning-based feature extraction network, and then classifies them with the broad learning system. Experiments on the ModelNet40 dataset showed that our proposed network can achieve high 3D point cloud recognition accuracy of over 87%, which is better than that of a pure deep learning network with an identical backbone. In addition, the shortest training time of Point-BLS is 10.31 seconds in our experiments.