Point cloud dynamic region graph convolution method, classification method and segmentation method

Abstract

The invention discloses a point cloud dynamic region graph convolution method, and a point cloud dynamic region graph classification method and a point cloud dynamic region graph segmentation method using the point cloud dynamic region graph convolution method. According to the method, a new convolution operation form for point clouds is adopted, point feature information of a plurality of different neighborhoods is aggregated through a nonlinear method according to a constructed point cloud picture structure, so that neurons can adaptively select the size of a region. Compared with existing technical schemes such as PointNet for analysis on a single point, the method has the advantages that a plurality of different local neighborhood graph structures are constructed, each neuron can adaptively select a proper neighborhood receptive field, and then similar convolution operation is performed by using the relation between each point and a neighborhood point to obtain local features, the surrounding neighborhood information can be better combined, the local geometric information can be more effectively extracted, and finally the classification or segmentation accuracy of the point cloud data is improved.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a point cloud dynamic area map convolution method, a classification method and a segmentation method. Background technique [0002] Point cloud data contains rich semantic information and has the characteristics of high density and high precision. However, due to the irregularity and disorder of point cloud data, semantic analysis based on point cloud data is still a difficult challenge. Some early methods used features with complex rules for manual extraction to solve such problems. In recent years, with the rapid growth of deep learning and machine learning technology, deep learning methods have also been introduced for the analysis and processing of point cloud data. The data that needs to be processed by the deep network is of regular shape, and the point cloud data is fundamentally irregular, and the spatial distribution of the point cloud data is not af...

AI Technical Summary

Problems solved by technology

Although 3D convolutions have achieved good results in point cloud classification and segmentation tasks, their high requirements for storage performance and high computational costs make them still have insufficient accuracy on large-scale datasets and large scenes.

Images