The Algorithm Study Of Multi-sensor Multi-target Tracking

Multi-sensor multi-target tracking technology mainly include two parts: multi-sensordata fusion and multi-target tracking. Multi-sensor target tracking effectively combine theinformation obtained from multiple sensors to estimate the target state，which has betterperformance than single sensor tracking. It is a application paradigm of data fusiontechnology in the field of target tracking. Multi-target tracking problem is that it classifiesnumbers of measurement values which the sensor received according to their origin in thedense environment, and using adaptive filtering and data association algorithm to estimate thetrajectory of every target tracking,then analyze the accuracy and reliability of the trajectory.First of all， this thesis describes the significance and theoretical foundation ofresearching information integration which is an emerging discipline, and research status quoon domestic and overseas, the model and structure of multi-sensor information fusion, themain techniques and methods of information fusion, theoretical system, the advantages ofinformation integration, the ranks and levels of information fusion, classify the trackingsystem of fusion, and technology applications of information fusion and so on.And then we study the basic principles of multi-maneuvering target tracking, the targetmodel of maneuvering, including the CA model, the CV model, the Singer model, the Jerkmodel, the current statistical model and the curve model, and compare the advantages anddisadvantages of various models, discuss the basic tracking filtering and prediction methods:the standard linear Kalman filter(KF), Kalman filter(EKF) and Unscented Kalman filter(UKF)which is extended for nonlinear system, and a brief introduction of new technologies ofseveral target tracking.Finally, for the problem of multi-sensor multi-target tracking, we bring forth themulti-sensor multi-target tracking algorithm which based on random finite set, introduce thebasic theory of random set, the Statistics of random finite set(FISST), and the probabilityhypothesis density(PHD) of random set. We explore the distribution function of random setsand the derivation process of derivative and integral of set, give out the theoretical frameworkof information fusion for random set, to extend the specification of modeling method Bayesof single-sensor single-target to multi-sensor multi-target. Then indicate the meaning of theprobability hypothesis density, use the PHD filter to achieve the multi-sensor multi-targettracking, then through the simulation to show the effectiveness of the algorithm.