Rao-Blackwellized粒子势均衡多目标多伯努利滤波器

发布时间：2018-04-11 11:01

本文选题：多目标跟踪 + 多伯努利　；参考：《控制理论与应用》2016年02期

【摘要】：由于多伯努利滤波器直接近似递推了多目标状态的后验概率密度,使得多目标跟踪问题在基于随机有限集理论框架下的求解及目标状态的估计显得更为直观.本文针对一个状态可分解(线性/非线性)的状态空间模型,分析基于Rao-Blackwell定理的滤波估计方法,结合噪声的去相关构造线性状态的滤波方程.文中详细推导并提出Rao-Blackwellized粒子势均衡多目标多伯努利滤波器的一般实现形式,包括给出多伯努利非线性状态粒子滤波的实现形式,并结合非线性滤波结果给出多伯努利线性状态的递推滤波公式.本文提出的滤波器实现方法能够在更低维的状态空间上进行采样,滤波器的整体跟踪性能得到提高.多目标跟踪的仿真实验结果验证了该算法的有效性.
[Abstract]:Because the multi-Bernoulli filter directly approximates the posterior probability density of the multi-objective state, it is more intuitive to solve the multi-target tracking problem under the framework of stochastic finite set theory and estimate the target state.In this paper, for a state-decomposable (linear / nonlinear) state space model, the filtering estimation method based on Rao-Blackwell theorem is analyzed, and the filtering equation of linear state is constructed by combining the de-correlation of noise.In this paper, the general realization form of multi-target multi-Bernoulli filter with Rao-Blackwellized particle potential equalization is derived and presented in detail, including the realization of multi-Bernoulli nonlinear state particle filter.Combined with the nonlinear filtering results, the recursive filtering formula for the linear state of multiple Bernoulli is given.The filter implementation method proposed in this paper can sample in the lower dimension state space and improve the overall tracking performance of the filter.Simulation results of multi-target tracking show that the algorithm is effective.
【作者单位】：西安交通大学智能网络与网络安全教育部重点实验室;兰州理工大学电气工程与信息工程学院;
【基金】：国家重点基础研究发展计划(“973”计划)(2013CB329405) 国家自然科学基金创新研究群体项目(61221063);国家自然科学基金项目(61370037;61005026;61473217) 甘肃省高等学校科研项目(2014A 035) 甘肃省自然科学基金(1506RJZA090)资助~~
【分类号】：TN713

【相似文献】