一种改进的多伯努利多目标跟踪算法

针对粒子势均衡多目标多伯努利滤波的粒子实现形式所需粒子数多、粒子退化严重的问题，将均方根容积卡尔曼滤波与粒子势均衡多目标多伯努利滤波相结合，提出均方根容积卡尔曼粒子势均衡多目标多伯努利滤波算法．该算法利用均方根容积卡尔曼滤波构建重要性密度函数，再对其进行采样获得预测粒子状态，从而提高粒子的准确性，减轻粒子退化．与基于无迹卡尔曼的粒子势均衡多目标多伯努利滤波相比，该算法更稳定，且算法性能不受目标状态维数的限制．仿真实验表明，所提算法与粒子势均衡多目标多伯努利滤波算法和基于无迹卡尔曼的粒子势均衡多目标多伯努利滤波算法相比，其跟踪精度更高．

Multi-target tracking with the cubature Kalman multi-bernoulli filter

The particle cardinality-balanced multi-target multi-bernoulli(P-CBMeMBer) filter needs large numbers of particles and has serious particles degradation. To solve this problem, we combine the square-rooted cubature Kalman filter(SCKF) with the P-CBMeMBer filter, called square-rooted cubature Kalman P-CBMeMBer(SCP-CBMeMBer) filter. The SCP-CBMeMBer filter obtains the predicted particles by sampling the importance density function generated by the SCKF in order to alleviate particles degradation. Compared to the P-CBMeMBer filter based on the unscented Kalman filter(UP-CBMeMBer), the proposed method is more stable and its performance is unrestricted by the dimension of the target states. The results show that the proposed method has a higher accuracy than the P-CBMeMBer filter and the UP-CBMeMBer filter.