Research On Multi-target Tracking Of Vehicle Active Anti-collision Warning System Based On Data Association Algorithm

Vehicle safety has get more and more attention，and can be divided into active safety andpassivity safety. Based on the innovation project of Science and Technology Department ofJilin Province------"Vehicle Active Anti-collision Warning System Research andDevelopment", this paper has researched on the multi-target tracking of vehicle activeanti-collision warning system based on data association. And this system is aim at activesafety. Multi-target tracking is important section of Vehicle Active Anti-collision WarningSystem. Multi-target tracking belong to radar data processing after the radar signal processing.Multi-target tracking can make sure detection target, and estimate motion state and motiontrail of detection target to come true vehicle active anti-collision warning.This paper puts forward the performance requirements of Vehicle Active Anti-collisionWarning System. Vehicle Active Anti-collision Warning System need estimate motion stateand motion trail. So we need choose linear filtering and prediction theory which can besolved by data association to process multi-target tracking data. Then the Kalman filter andfixed gain filter were reaserched, and comparing two filter to use in this paper. And thenIntegrate Probabilistic Data Association (IPDA) algorithm, Joint Probabilistic DataAssociation (JPDA) algorithm and maneuvering target tracking algorithm were reaserched,and Compared with the performance characteristics of tracking multi-target on the road. Andthen target motion models were established, combined with the performance characteristics ofdriving environment on the road, And based on the range of speed and acceleration designedmulti-target tracking filter.This paper will introduce three questions about research on multi-target tracking ofvehicle active anti-collision warning system based on data association below this paragraph.At first, multi-target tracking theory of vehicle active anti-collision warning system wasreaserched by combining with the performance characteristics of tracking and data source,and established multi-target tracking coordinates which is rectangular Cartesian coordinate on a two-dimensional plate. Filter theory was reaserched and we choosed linear filter to cometure vehicle active anti-collision warning system. And then the Kalman filter and fixed gainfilter were choosed. Correlation gate theory was reaserched which can improve detectionprobability and decrease false-alarm probability. Basic target motion theory was reaserchedand some actual target motion models are established on the road.Secondly, data association algorithm theory was reaserched and we analyzes objects ofdata association which can be divided into trajectory maintain and trajectory primary. TheBayesian data association algorithm was choosed to come ture trajectory maintain whichcontain Integrate Probabilistic Data Association (IPDA) algorithm and Joint ProbabilisticData Association (JPDA) algorithm. These two kinds of data association algorithms can trackmulti-target well in the dense clutters. And the modified logic algorithm was choosed to cometure trajectory primary.Thirdly, maneuvering target tracking theory was reaserched and we analyzes targetmoving condition which is changed timely. Maneuvering target tracking algorithms containdetection thresholds for maneuvering multi-target tracking algorithm and the activemaneuvering multi-target tracking algorithm. And then Adjustable white noise algorithm,Singer model algorithm and multi-model algorithm were reaserched and we analyzes theperformance characteristics of tracking maneuvering target.Finally, on the basis of the previous reaseach, it uses the MATLAB do the simulation ofmulti-target tracking. Some basic target motion models are uniform motion model, uniformlyaccelerative motion model, coordinate turn model and maneuvering turn model which areestablished on the actual environment. And then compares and analyzes the maneuveringtarget tracking effect of Integrate Probabilistic Data Association (IPDA) algorithm, JointProbabilistic Data Association (JPDA) algorithm and maneuvering target tracking algorithm.And we analyzes performance of accuracy, real-time, and reliability.The research results of the collision warning radar system and its multi-target trackingmake sure detection target, and estimate motion state and motion trail of detection target.These lay the theoretical foundation for the development of improving detection probabilityand decreasing false-alarm probability in anti-collision warning radar and also have a certain guiding significance and practical value in its future application and promotion.