Research On Extraction Method Of Traffic Target Track Based On Random Finite Set

The multi-target track extraction of traffic in actual scenes faces many challenges,such as time-varying number of targets,unknown target motion models and dense clutter.Among the various multi-target tracking algorithms that have been developed,the multi-target tracking algorithm based on random finite set can avoid complicated data association,and the labeled random finite set tracking model distinguishes the target track by adding tags to the target.Based on this,the generalized labeled multi-Bernoulli filter algorithm(δ-Generalized Labeled Multi-Bernoulli filter,referred to as δ-GLMB)has been widely used.After reviewing the development history of multi-target tracking technology and the research status of multi-target tracking algorithms based on random finite set,this paper develops the δ-GLMB filter algorithm based on labeled random finite set model to study the method of multi-target track extraction.However,when there are multiple measurements of some targets in the traffic scene,the standard δ-GLMB filter algorithm can’t achieve track matching between multiple measurements and a single target,which leads to tracking errors.Therefore,an improved algorithm of adaptive threshold determination based on δ-GLMB tracking results is proposed.The research work of this paper is as follows:(1)In view of the fact that the actual traffic road conditions are complicated and clutter-intensive causes the sensor to obtain some high false alarm rates and unclear dynamic characteristics of the noise,this paper selects appropriate thresholds for each characteristic parameter of the target in the 77 GHz millimeter wave radar monitoring data in different scenarios to improve tracking accuracy.Then,theδ-GLMB filter algorithm is used to extract multiple target tracks from different data.It is found that there are two problems in the application of the standard δ-GLMB filter algorithm in the multi-target track extraction of traffic scenes: single target multi-track problem and the track gap problem.In view of the above two problems,different filtering parameters are selected respectively for the distribution characteristics and motion characteristics of targets in different scenes.The experimental results show that the tracking performance has been improved,but the tracking error problem cannot be completely solved.(2)Aiming at the problems of single target multi-track and the track gap when the δ-GLMB filter algorithm extracts traffic multi-target tracks,this paper analyzes the limitations of the standard δ-GLMB filter algorithm applied to traffic multi-target tracking,and proposes an adaptive threshold decision algorithm based on the δ-GLMB tracking result is proposed.In order to improve the accuracy of the algorithm,this paper has determined the optimal available historical motion state time through experimental comparison,and derives the adaptive threshold expression.The experimental results show that the adaptive threshold determination algorithm effectively achieves the deletion of redundant tracks and intermittent tracks.