Intelligent driving target re-identification method used under multi-target tracking condition

Abstract

The invention discloses an intelligent driving target re-identification method under the condition of multi-target tracking, and the method comprises the following steps: setting a target, setting three parallel detection heads based on Center Net, and respectively detecting a heat map, an object center offset and a surrounding frame area of the target; the heat map is provided with a response coefficient from 0 to 1, the response coefficient is 1 if the target coincides with the center of the real object on the ground, and the response coefficient is exponentially attenuated along with the length of the distance between the position of the heat map and the center of the real object; according to the method, on the basis of re-identification of the same target, the problem of continuous tracking of multiple types and multiple targets in different view angles and different pictures in unmanned driving is solved. Through cooperative calculation of the heat map, the object center offset and the bounding box, a specified target can be re-identified so as to help the intelligent driving module to make a better driving scheme selection.

Description

technical field [0001] The present invention relates to the field of automatic driving, in particular to a method for intelligent driving target re-identification in the case of multi-target tracking Background technique [0002] With the further development of environmental awareness and the intelligentization of vehicles and roads, the form of vehicles on the road in the future will gradually be dominated by intelligent driving vehicles. Before the realization of fully automatic driving, for a long time to come, there will be a stage where human-driven vehicles, semi-intelligent vehicles, and network-connected automatic driving vehicles will be mixed on the road at the same time. [0003] The vehicle-road coordination method can effectively improve the organization of the group, the intelligence of the driving of the vehicle, relieve the pressure of urban traffic, and improve the traffic capacity of the road. [0004] The existing multi-target tracking methods are all bas...

AI Technical Summary

Problems solved by technology

[0006] But in unmanned driving, we need to care about all road participants and their trajectories, so the current target re-identification technology cannot meet the needs of this scenario

[0007] Moreover, in the unmanned driving scene, the problem is that the moving camera is very common, and the occlusion is very common, so this increases the requirement for target re-identification, so there is no good technology to realize the unmanned driving based on the target re-identification method. Multi-category multi-object tracking method for

Images