采用随机矩阵的多扩展目标滤波器
A Multi？？Target Filter Based on Random Matrix

针对杂波环境下多扩展目标的运动状态和形状信息的联合估计跟踪的问题,提出了一种基于随机矩阵的扩展目标跟踪算法。该算法采用具有噪声基于密度的空间聚类(DBSCAN)划分与预测划分相结合的联合划分算法对量测集进行划分,然后采用联合概率数据关联(JPDA)的软关联思想建立量测簇与扩展目标之间的对应关系,最后采用随机矩阵法对扩展目标进行估计获得运动状态和形状信息,特点是:将量测集划分为互不相交的几个簇,以使每个簇中的量测源于同一目标或杂波;建立量测簇与扩展目标之间的关联关系及状态更新。联合划分算法与DBSCAN划分的比较仿真实验表明,在有距离相近目标时采用联合划分算法比采用DBSCAN划分的滤波器的跟踪效果好得多。所提多扩展目标滤波器与ET？？GMPDH滤波器的仿真实验表明,所提算法有较高的跟踪精度、较大的检测概率及较小的虚警概率。
A multiple extended target filter based on random matrix is proposed to track the kinematic states and shape information of multiple targets in the presence of the clutter measurements. The proposed filter employs a joint partitioning algorithm, which combines the DBSCAN(density based spatial clustering of applications with noise) and the prediction partitioning algorithm, to partition the measurement set. Then, the JPDA(joint probabilistic data association), which is a soft association algorithm, is applied to description of the association between measurement clusters and extended targets. Finally, the method of random matrix is employed to estimate the kinematic states and shape information of extended targets. Simulation results which compare the joint partitioning algorithm with DBSCAN partitioning show that the filter by using joint partitioning algorithm could achieve much better tracking performance than that by using DBSCAN partitioning when there are spatially close extended targets. Moreover, simulation results in comparison with the ET？？GMPHD filter show that the proposed multiple extended target filter has higher tracking accuracy, higher detection probability, and lower false alarm probability