中文摘要：

针对联合综合概率数据关联算法（JIPDA）存在的航迹合并问题，将目标建模为随机有限集（RFs）提出改进的JIPDA算法。传统JIPDA首先产生初始概率密度函数（PDF），之后对该PDF进行近似来估计目标状态。为了使目标状态估计PDF与初始PDF之间的相似性最大化，当目标标签无意义时，提出对JIPDA的初始PDF进行优化。将KL散度作为相似性的衡量标准，建立起优化过程的代价函数。仿真实验表明，所提方法可有效地抑制传统JIPDA引起的航迹合并。

英文摘要：

To avoid the track coalescence of the Joint Integrated Probabilistic Data Association （JIPDA）, a modified version of JIPDA is proposed by modelling targets as Random Finite Set （RFS）. The JIPDA first generates the original Probability Density Fhnction （PDF） and then makes an approximation of the PDF to estimate target states. To maximize the similarity between the state estimate PDF and the original PDF, the original PDF is optimized when target label is irrelevant. Using the KL divergence as a measure of the similarity, the cost function is developed. The experimental results show that the proposed method can effectively avoid the track coalescence.