Multiple Targets Tracking For Infrared Passive Sensors Based On Motion Models

Multiple targets tracking is a key problem in the field of multi-sensor information fusion system, in which tracking and data association are two important aspects and have been widely investigated for decades.This paper deals with motion models based multi-target tracking problem for passive sensors. Firstly, the basic principle of multi-target tracking is introduced and the measurement characteristics of infrared passive sensor are analyzed in detail. Secondly, two motion models of targets are presented. With the distance information assumed, a linear time variant filter model of target motion is established. Further a linear approximation model of target trace is derived according to linear approximation theory. When applied to target tracking, two target tracking algorithms based on the linear time variant filter model and linear approximation trace model are proposed respectively. Using assumed distance imformation, the former has a higher convergence rate and predictive precision of target states than conventional tracking algorithms, which adopt a Thaler-Series based linear filter model, such as EKF model. While the latter can track non-linear moving targets by using a linear approximation trace model. For the data association problem in the multi-target tracking, we domerstrate that most incorrect association can be elimilated rapidly and efficiently through rough data association. Such preprocessing scheme can reduce the complixity of builing candidate association trees greatly for native S-D assignment algorithm, thus improve the data association speed significantly.The simulation results verify the validity and efficiency of the proposed algorithms.