BICANet: LiDAR Point Cloud Classification Network Based on Coordinate Attention and Blueprint Separation Involution Neural Network

With the advent of the era of Industry 4.0 and the continuous development of point cloud data acquisition technology, point cloud data have been widely used in the unmanned distribution of intelligent logistics. Applying deep neural networks to accurate light detection and ranging (LiDAR) point cloud classification results is considerably significant for unmanned transport. This article designs a 3-D point cloud classification model coordinate attention blueprint separation involution neural network (BICANet) with multidimensional feature extraction. First, to extract more point cloud features, 3-D point clouds are projected to a 2-D plane for calculating point cloud feature values, and multidimensional point cloud feature information is fused from different views. Second, the involution network is introduced to reduce the amount of redundant data for neural network computation and improve the whole network computation efficiency. At the same time, to further enhance the network feature learning capability, the blueprint separation convolution is combined with coordinate attention (CA). Finally, to draw our conclusions more rigorously, we conducted error analysis experiments and experimented with the generalization ability of our proposed BICANet model. The overall accuracy of BICANet in the Vaihingen and GML_B datasets was experimentally demonstrated to reach 86.0% and 98.8%, respectively. It is highly competitive with the currently available methods.