Point cloud identification method and system based on channel attention

Abstract

The invention belongs to the technical field of three-dimensional data processing, and particularly discloses a point cloud recognition method and system based on channel attention, and the method comprises the steps: carrying out global position coding through employing an input point cloud, taking the point cloud position information and the global position coding as the input of a first-layer hierarchical point cloud feature extraction module, and carrying out the global position coding; the output of the first layer of point cloud feature extraction module is used as the output of the second layer of hierarchical point cloud feature extraction module, multiple layers are stacked in the same manner, the last layer outputs global features, in the hierarchical point cloud feature extraction module, point clouds are sampled and grouped, hierarchical local position coding is performed, and the local position of the point clouds is coded; and carrying out attention feature fusion to extract spatial features of the point cloud. By adopting the technical scheme, on the basis of the original point cloud data, position coding is carried out based on the global position and the local position of the point cloud, so that the global position information and the local position information are fused into the point cloud features, the channel attention is fused to extract the channel features of the point cloud, and the point cloud feature extraction effect is improved.

Description

technical field [0001] The invention belongs to the technical field of three-dimensional data processing, and relates to a point cloud recognition method and system based on channel attention. Background technique [0002] In recent years, with the development of robotics and autonomous driving technology, 3D sensors have become popular, 3D data has become easier to obtain, and large-scale 3D data sets have begun to appear. How to correctly process and analyze 3D data has also become a concern. topic of. [0003] However, it is not feasible to directly use the method of processing two-dimensional image data to process three-dimensional data, because two-dimensional images are represented by pixels point by point, and three-dimensional data is usually a set of unordered and discrete points in three-dimensional space. . On two-dimensional images, complex convolutional neural networks are constructed through convolution operations to process two-dimensional data, which greatl...

AI Technical Summary

Problems solved by technology

However, this method does not make full use of the channel data of the point cloud, nor does it make full use of the global and local position information of the point cloud, resulting in that the position information is not fully considered in the process of extracting features from the point cloud data.

Images