Point cloud segmentation method and system, medium, computer equipment, terminal and application

Abstract

The invention belongs to the technical field of computer vision, and discloses a point cloud segmentation method and system, a medium, computer equipment, a terminal and an application, and the method comprises the steps: constructing a feature extraction module based on a layered graph Transformer, wherein the feature extraction module comprises a feature down-sampling network and a feature up-sampling network based on the graph Transformer; constructing a point cloud segmentation network based on the hierarchical graph Transformer, using a cross entropy loss function with weight and a training set to perform supervised training on the hierarchical graph Transformer point cloud segmentation network, and adjusting network parameters according to a loss value in each round of training to obtain a network model; and predicting the point cloud in the test set by using the trained network model to obtain a segmentation result of each point in the point cloud. According to the invention, the initial characteristics of the point cloud are enriched; the graph Transformer can effectively extract the relationship between the neighborhoods of the point cloud; and the precision of point cloud segmentation is improved.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a point cloud segmentation method, system, medium, computer equipment, terminal and application. Background technique [0002] In recent years, the emergence of acquisition equipment such as camera matrix and lidar has reduced the difficulty of obtaining 3D point cloud data. How to obtain useful information from a large amount of 3D point cloud data and solve the problems in scene understanding is an important content of current computer vision research. Target detection, classification, and recognition based on 3D point cloud data are the main technologies for solving scene understanding at present, and 3D point cloud segmentation is the basis of these technologies. [0003] The traditional point cloud segmentation technology is mainly based on non-deep learning methods. This type of method mainly performs simple segmentation on the point cloud. Each object ...

AI Technical Summary

Problems solved by technology

[0006] (1) The existing PointNet series methods do not consider the relationship between point pairs, and cannot construct relatively rich features, resulting in low segmentation accuracy

[0007] (2) The existing PointNet series methods can only process small-scale point cloud data. For a large-scale scene, the scene needs to be divided into multiple uniform areas, which limits the application scene

Images