一种基于二阶Markov目标状态模型的多帧关联动态规划检测前跟踪算法

传统的动态规划检测前跟踪(Dynamic Programming Track-Before-Detect,DP-TBD)算法在每一阶段的数据关联中,仅用当前帧的观测数据与前一帧的指标函数进行关联积累,对目标状态在连续相邻帧间的相关性以及目标运动特征的考虑不充分,这样在低信噪比时,容易发生目标关联错误,严重影响了DP-TBD算法的检测和跟踪性能。针对此问题,该文提出了一种基于二阶Markov目标状态模型的DP-TBD算法,该算法以目标状态的条件概率比最大为准则,采用二阶Markov模型描述目标状态的相关性,并根据目标运动特征给出了一种与目标转弯角度相关的状态转移概率模型。在此基础上,实现了多帧数据关联的DP-TBD算法。通过仿真实验与传统的DP-TBD算法进行了比较,验证了该算法的检测及跟踪性能。

A Multi-frame Association Dynamic Programming Track-before-detect Algorithm Based on Second Order Markov Target State Model

Traditional Dynamic Programming Track-Before-Detect(DP-TBD) algorithms use only observation data of current frame to associate with merit function and accumulate energy at each stage of data association.The ignorance of target’s state relevance among successive frames and its own kinematic characters results in false state association at low Signal-to-Noise Ratio(SNR),which reduce detecting and tracking performance profoundly.To solve this issue,a DP-TBD algorithm based on second order Markov target state model is proposed.Taking maximum of the target’s state conditional PDF ratio as the optimal criteria,this algorithm makes use of second order Markov model to describe the target’s state relevance and defines a state transition probability model according to target’s kinematic characters,which relates to target’s turning angle.On these bases,a multi-frame data association DP-TBD algorithm is implemented.Compared to traditional DP-TBD algorithm through a simulation experiment,the proposed algorithm turns out to have better detection and tracking performance.