Laser radar point cloud target grading identification method based on Gaussian mixture sorting

Abstract

The invention relates to a laser radar point cloud target grading identification method based on Gaussian mixture sorting. The method comprises the steps of collecting laser radar point cloud data, performing preprocessing, performing ground point cloud removal to screen out non-ground point clouds, performing non-ground point cloud density clustering, performing sorting according to importance degrees, sequentially performing identification according to the importance degrees, outputting identified targets and the like. According to the method, the number of iterations can be greatly reduced, the real-time performance of a ground fitting algorithm is improved, and the stability of a ground fitting module is improved; layered sorting can be carried out according to the importance degree of to-be-recognized targets, it is guaranteed that the importance degree of each target is unique, and computing resources can be preferentially allocated to targets with more important identification when the computing resources are limited; and the method has high real-time performance and high stability, and has good applicability for being deployed on equipment with limited computing resources, such as an automatic driving vehicle, and coping with a complex and changeable real environment.

Description

technical field [0001] The invention relates to a target recognition method, in particular to a laser radar point cloud target classification recognition method based on mixed Gaussian sorting. Background technique [0002] Object detection is one of the main tasks of environment perception in the field of autonomous driving and intelligent transportation, and the real-time and effectiveness of detection algorithms still need to be further optimized. The sensors that target detection relies on include lidar, industrial cameras, etc., among which lidar can provide massive point cloud data for the target detection system. [0003] LiDAR point cloud is a special data form of 3D image, which contains the spatial geometric information of the observed scene. Compared with the planar information provided by 2D images, the acquisition of depth information enables 3D images to easily separate scenes at different distances. Target recognition based on lidar point clouds is gradually ...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problems of large data volume, complex data, and long calculation time caused by information redundancy in the existing laser radar point cloud, the present invention provides a hierarchical recognition method for laser radar point cloud targets based on mixed Gaussian sorting, including the following steps :

Images