2D&3DHNet for 3D Object Classification in LiDAR Point Cloud

Accurate semantic analysis of LiDAR point clouds enables the interaction between intelligent vehicles and real environment. This paper proposes a hybrid 2D 3D Hough Net by combining global features local with classification deep learning network. Firstly, object are mapped into space to extract features. The generated input convolutional neural network for training Furthermore, multi-scale critical sampling method is designed points in views projected from reduce computation redundant points. To features, grid-based dynamic nearest neighbors algorithm searching Finally, two networks connected full connection layer, which fully layers classification.