Point cloud classification method and device

Abstract

The embodiment of the invention discloses a point cloud classification method and device. A specific embodiment of the method comprises: obtaining initial point cloud data; randomly extracting first point cloud data from the initial point cloud data based on a preset first downsampling rate; randomly extracting second point cloud data from the first point cloud data based on a preset second downsampling rate; respectively inputting the first point cloud data and the second point cloud data into a pre-trained spatial aggregation network to obtain first feature data and second feature data; extracting feature data corresponding to the second point cloud data from the first feature data, and aggregating the extracted feature data and the second feature data into third feature data; and inputting the third feature data into a pre-trained point cloud classification network to obtain category information corresponding to each point in the second point cloud data. According to the embodiment,uniform sampling with different densities is carried out on the point cloud data, the richness of extracted features is improved, and the performance of carrying out small target segmentation on thepoint cloud is improved.

Description

technical field [0001] The embodiments of the present application relate to the field of computer technology, and in particular to a point cloud classification method and device. Background technique [0002] The 3D point cloud semantic segmentation technology can automatically analyze the acquired 3D point cloud scene data, judge the category of each point through the spatial position information, spatial structure information, color information, etc. between points, and then segment different objects in the scene. 3D point cloud semantic segmentation technology has a wide range of applications, mainly in urban-level surveying and mapping, autonomous driving, scene navigation, virtual reality, augmented reality and other fields. [0003] Since point clouds are composed of sparse and disordered points, traditional convolutional neural networks are not suitable for feature extraction of point clouds. In response to this problem, the researchers proposed the following three ...

AI Technical Summary

Problems solved by technology

However, this operation is likely to cause uneven distribution of sampling points. There are almost no small target points in the sampling points, and large target points occupy a large part.

Therefore, small targets are easily affected by large targets

Images