Three-dimensional target sensing method in vehicle-mounted edge scene

Abstract

The invention discloses a three-dimensional target sensing method in a vehicle-mounted edge scene, which realizes three-dimensional target sensing and tracking in a vehicle-mounted system by utilizing point cloud projection and two-dimensional image fusion. According to the method, under the optimization of a parallel computing algorithm, filtering and segmentation operations are performed on point cloud image data, then point cloud classification and feature value extraction are performed, a two-dimensional image is combined, the point cloud is projected to the two-dimensional image for clustering, and finally matching of information points and association of targets are performed by combining related data of front and back frames, so that the matching and tracking effects are achieved. According to the method, the problems of how to carry out target identification through the combination of images returned by the laser radar and the camera and how to deploy on a small terminal device are solved at the same time, after the method is applied to the vehicle-mounted device, an accurate identification and tracking effect is achieved, and high generalization and real-time performance are achieved.

Description

technical field [0001] The invention relates to the vehicle-mounted field of intelligent identification and multi-sensor fusion, in particular to a three-dimensional object perception method in a vehicle-mounted edge scene. Background technique [0002] In recent years, with the continuous growth of car ownership, the road carrying capacity of many cities has reached full load, and traffic safety, travel efficiency, energy saving and emission reduction have become increasingly prominent. Intelligent and networked vehicles are usually considered to solve the above traffic problems important way. [0003] With the development of artificial intelligence and computer vision becoming more and more mature, in many practical applications on the perception layer or sensors of the Internet of Vehicles architecture, the demand for visual tasks such as object detection and object tracking has increased dramatically. At the same time, the research on target detection technology of mult...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to provide a three-dimensional object perception method in the vehicle edge scene, to solve the problems of poor generalization and easy loss of target information in the existing traditional point cloud perception method, and to improve real-time detection and tracking accuracy

Images