Point cloud semantic segmentation method based on point global context relation reasoning

Abstract

The invention discloses a point cloud semantic segmentation method based on point global context relation reasoning. The method comprises the steps of obtaining a training set T and a test set V; constructing a point cloud data semantic segmentation network of deep learning and global context reasoning; using a multi-classification cross entropy loss function as a loss function of the point cloudsemantic segmentation network; performing P rounds of supervised training on the point cloud data semantic segmentation network by using the training set; and inputting the test set into the trained network model for semantic segmentation to obtain a segmentation result of each point. The method has the beneficial effect that the problem of insufficient global information extraction of 3D point cloud semantic segmentation is solved by utilizing a method based on deep learning and global context reasoning. On the basis of deep learning, an added global context inference module models the relationship between feature channels by using a channel attention mechanism, global information of the relationship between the channels is further transmitted and aggregated through graph convolution, andthe global information can be obtained so as to refine the result of point cloud semantic segmentation.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a point cloud semantic segmentation method based on point global context reasoning. Background technique [0002] In recent years, with the widespread application of depth cameras and LiDAR, a large amount of 3D point cloud data has become more and more accessible, and 3D vision tasks based on deep learning have attracted a lot of attention. 3D point cloud data is a collection of vector representations of scattered points in three-dimensional space. The vector can be composed of three-dimensional coordinates xyz, color information rgb, light intensity r and other information. These points are irregularly distributed and have the characteristics of disorder, unstructure, rotation invariance, etc., so it is impractical to directly extend the existing image segmentation methods to 3D point clouds. Applying deep learning methods to point cloud data faces many problems. [0003] ...

AI Technical Summary

Problems solved by technology

[0005] Although PointNet++ can capture local fine-grained and global context information, it is still unable to capture the long-distance context dependencies between points because it uses stacked MLP layers to learn the features of point clouds.

In the segmentation results, some specific categories are easily confused with other categories, such as windows and board objects, doors and wall areas, and their segmentation accuracy is very low

In order to learn a more class-discriminative feature representation, we need to fuse more information, and PointNet++ does not consider enough category context information

Images