Multi-sensor Target Tracking Algorithm Based On Probability Of Target Existence In Complex Environment

In moder war,the multi-sensor target tracking in complex environment with high target missed detection and high clutter interference is a practically challenged problem.In order to address the multi-sensor target tracking problem,there are usually two kinds of fusion systems,centralized and distributed fusion.Compared with the centralized architecture,distributed fusion system has been more widely used in the real application attribute to its better system robustness with the de-centralization deployment,as well as its substantially reduced communication bandwidth requirement between nodes(sensors)and fusion center.In this paper,under the framework of distributed fusion system,the multi-sensor target tracking technology in complex environment with high target miss detection and high clutter interference is studied,which mainly consists of the single sensor target tracking and multi-sensor track fusion.This work focuses on the target existence based single sensor automatic target tracking technology and multi-sensor track fusion method,which utilize the probability of target existence as the track quality measure for track management.The main contents and contributions of this paper are as follows:(1)The Integrative Probabilistic Data Association(IPDA)algorithm is deeply studied.IPDA is a single sensor automatic target tracking technology,which describes the target existence as a random event and utilizes the Markov chain to model the event evolution,what's more,the probability of target existence is recursively calculated as a track quality measure used for track management.The track initialization,track management(including true track confirmation and maintenance,false track recognition and elimination)are extensively exploited via experiments simulation,besides,some other IPDA related implementation techniques are also introduced.(2)An improved Track-to-Track association and fusion(IT2TAF)algorithm is proposed.Track-to-track association and fusion(T2TAF)algorithms are widely used in the tracking community,but ignores the track management.In this paper,T2TAF algorithm is improved by incorporating the probability of target existence in the fusion center as a track quality measure used for track management in the complex environment.However,in the IT2TAF algorithm,the fusion center utilizes a hard one-to-one track association method to find the tracks originated from the same target,whose association performance deteriorates badly in complex environment.(3)An improved all-neighbor track fusion(IANTF)algorithm is proposed.The key point of the all-neighbor track fusion(ANTF)algorithm is to transmit each sensor track which satisfies the upload threshold to the fusion center,and to associate all tracks in the fusion center by using the idea of data correlation,which delivers stronger association fault tolerance than the T2TAF method in the complex environment.However,in the ANTF algorithm,the upload threshold value of each sensor track is usually low,which leads to a large number of tracks needed to be associated in the fusion center and increases the computation of track fusion tremendously.What?s worse,the ANTF method requires each sensor to transmit the Kalman gain matrix to fusion center in order to calculate the correlation covariance among different sensor tracks,which obviously increases the node-to-fusion center communication bandwidth.Motivated by addressing the problems above,this paper employs the tracking gate technology in the fusion center to reduce the number of associated tracks for fusion,what's more,the covariance intersection fusion method is utilized to avoid calculating the correlated covariance among sensor tracks aimed to avoid calculating the correlated covariance.As a result,the improved algorithm greatly saves communication bandwidth and reduces computing costs while ensuring the performance of the algorithm.