The Research On Extended Target Tracking Method Based On Cooperated Target

Target tracking is widely used in robots,intelligent vehicles,and other fields.With the rapid development of sensor technology,extended target tracking technology has also continued to develop.In particular,extended target tracking methods with auxiliary information have greatly improved the performance of target tracking.However,cooperated target,as an important kind of auxiliary information,can be used to improve the existing extended target tracking methods.Therefore,it is of great theoretical value and practical significance to study the extended target tracking method based on cooperated target.In the existing extended target tracking methods,the extended target tracking model and data association are two core issues.Compared with the existing extended target tracking models,the improved extended target tracking model introduces the target's width and height features,heading angle feature and cooperated target's identity feature into the state space.Based on the improved model,an improved design method of tracking gate is presented.Firstly,use the width,height features and heading angle feature of the target to design a target tracking gate to select all the measurements.Secondly,use the position feature of the target to design another target tracking gate to obtain valid measurements.Finally,valid measurements are used for data association and filtering.This method can effectively select invalid measurements and improve the subsequent data association efficiency and tracking accuracy.Aiming at the problem of using a single model for the point target tracking method based on cooperated target,an improved interactive multi-model algorithm framework is presented and applied to two scenes of sparse targets and dense targets.In the sparse scene,the probabilistic data association algorithm is improved by using the width,the height,the identity,and the heading angle feature provided by all measurements.Firstly,use the identity feature of the cooperated extended target to reconstruct the one-step predicted value.Secondly calculate the probability that the valid measurements originated from the extended target.Finally,the filtering is performed to get the estimated state value.Corresponding experiments also prove the effectiveness of the improved extended target tracking method in the sparse scene.In the dense scene,the joint probability data association algorithm is improved under the framework of interactive multi-model algorithm.Firstly,the method uses the identity feature of the cooperated extended target to reconstruct the one-step predicted value.Secondly,the confirmation matrix is given based on the geometric relationships between the valid measurements and the extended target.Meanwhile,the feasible events and associated events within the current aggregation are obtained to calculate the association probability of the valid measurements and the extended target.Thirdly,split the confirmation matrix into feasible events and related events to calculate the association probability.Finally,the filtering is performed.Corresponding experiments also prove the effectiveness of the improved extended target tracking method in the dense scene.