Point cloud and multi-view fused vehicle-mounted laser point cloud multi-target identification method

Abstract

The invention relates to a point cloud and multi-view fused vehicle-mounted laser point cloud multi-target identification method, which comprises the steps of constructing a depth model PGVNet based on an independent point cloud object to perform surface feature category prediction by performing point cloud local feature extraction on the independent point cloud object by utilizing a point cloud feature extraction module; generating a multi-view image of the independent object, and extracting an optimal view feature by using a view feature extraction module through view grouping and group feature fusion; fusing the optimal view feature and the point cloud feature by using a point cloud view feature fusion module based on an attention mechanism to obtain a point cloud global feature fused with attention; and finally, using the classifier MLP is to predict the category of the independent ground object target on the vehicle-mounted laser point cloud surface. According to the method, on one hand, the problem of information redundancy between similar views is reduced, on the other hand, the optimal view features can be used for guiding the model to learn point cloud local features, themodel classification precision is improved, and a new research method is provided for vehicle-mounted laser point cloud roadside multi-target fine classification.

Description

technical field [0001] The invention relates to the field of vehicle-mounted laser scanning point cloud data processing, in particular to a vehicle-mounted laser point cloud multi-target recognition method that integrates point clouds and multi-views. Background technique [0002] The complex and diverse street trees and pole-shaped targets (street lamps, traffic signs) on both sides of the road constitute important infrastructure in urban construction and management. Accurate and high-precision roadside target recognition is very important for urban road planning, urban modeling and automatic important role in driving. As a rapidly developing high-tech surveying and mapping technology, the vehicle-mounted laser scanning system can quickly and accurately obtain high-precision three-dimensional space information of roads and features on both sides, and is widely used in the collection and update of urban traffic information. [0003] The traditional method mainly focuses on ...

AI Technical Summary

Problems solved by technology

However, when point cloud objects are converted into images, feature expressions or voxelization, the classification accuracy and efficiency of ground objects are easily affected by image resolution and voxel size.

Therefore, some scholars work directly on the 3D point cloud based on the original point cloud method, such as Pointnet, Pointnet++, Pointsift, PointCNN, SO-Net and DGCNN, etc., but for the point cloud, its distribution in 3D space is not continuous characteristics, making the neighborhood structu

Images