Vehicle multi-target tracking method based on radar vision fusion under severe environment roadside view angle

Abstract

The invention discloses a vehicle multi-target tracking method based on radar visual fusion under a severe environment roadside visual angle, which integrates radar and a visual sensor, designs a calibration cost function combining a target distance and a radar scattering area, realizes radar visual data level fusion processing, improves the dimension and reliability of perception information, andimproves the reliability of the perception information. By introducing a data enhancement and anti-interference encoder decoder structure, a target detection network architecture with high anti-interference capability is designed, and the identification capability of a target tracking system in a severe environment is improved; and finally, different tracking strategies are given to the space targets in different states by calculating the similarity between the target position and the target distance, thereby realizing the dynamic accurate tracking process of the static recognition target. Bymeans of fusion enhancement processing of the data and the model, the scheme has high reliability and accuracy, and the position and speed information of the vehicle target can be continuously and accurately acquired in scenes such as backlight and nighttime.

Description

technical field [0001] The technical field involved in the present invention is the field of intelligent transportation and information perception, and mainly relates to a vehicle multi-target tracking method from a roadside perspective in a harsh environment oriented to intelligent traffic scenes, especially the multi-target tracking method of a vehicle in a multi-sensor fusion manner method. Background technique [0002] As an important means to solve traffic congestion, traffic accident rescue, traffic management, environmental pollution and other problems, the development of intelligent transportation system has been attracting much attention. In a broad sense, intelligent transportation systems include traffic information collection systems, information processing and analysis systems, and information release systems. Among them, the traffic information collection system is the data source and infrastructure of the intelligent transportation system, and its core underl...

AI Technical Summary

Problems solved by technology

[0006] In view of the serious defects of poor reliability in harsh environments and poor recognition accuracy of target speed and position in the existing multi-target vehicle tracking method from the roadside perspective, the present invention intends to disclose a vehicle multi-target tracking method based on the fusion of radar and vision from the roadside perspective

Images