异类传感器动态数据关联算法

Dynamic Data Association Algorithm for Heterogeneous Sensors

数据关联是异类传感器系统中最核心且最重要的内容之一,典型的数据关联算法可以归结为特定的分配为问题,然而现有的S维分配算法只考虑同一时刻的每个传感器量测的互联。将此静态关联推广到动态关联中,提出了一种适用于异类传感器的(S+1)维动态数据关联算法。该算法首先将同一时刻各传感器的量测与目标轨迹的一步预测值合并,把问题转化为(S+1)维分配问题,然后将各传感器量测估计的位置信息与目标航迹的预测值的差值作为关联代价,并利用LP-SOLVE工具包解决多维分配问题,最后利用求得的全局最优关联解进行滤波和航迹的更新。仿真实验表明提出的关联代价能更精准地反映数据关联的可能性,能够对多目标进行稳定的跟踪。

Data association is one of the most important and the most primary contents in heterogene- ous sensor systems. A typical data association algorithm can be formulated as a special assignment prob- lem. However, the existing S-D （S Dimensions） assignment algorithm only considers the interconnection of each sensor at the same time. The static S-D assignment algorithm is extended into the dynamic S-D assignment algorithm,and a dynamic （ S ＋ 1 ） D data association algorithm for heterogeneous sensor sys- tems is proposed. In this algorithm, firstly the measurement of each sensor at the same time is combined with one step prediction value of the target trajectory, and the problem is transformed into a （S ＋ 1 ） -di- mensional assignment problem, then the difference between the estimated position information of each sensor and the predicted value of the target track is taken as the associated cost, and the S-D assignment problem is solved by using the LP-SOLVE toolkit, finally the global optimal solution is used to filter and updating the track. The results show that the cost function can reflect the association probability more ac- curately and can steadily track multiple targets.