自适应门限GM-CPHD多目标跟踪算法

Adaptive Gating GM-CPHD Filter for Multitarget Tracking

带有势估计的高斯混合概率假设密度滤波(GM-CPHD)作为一种杂波环境下目标数可变的检测前跟踪方法,将复杂的多目标状态空间的运算转换为单目标状态空间内的运算,有效避免了多目标跟踪中复杂的数据关联问题,但该方法的计算复杂度与观测数的3次方成正比,在密集杂波情况下计算量十分巨大。针对该方法计算复杂度高的问题,提出利用一种最大似然自适应门限的快速算法,该算法首先利用自适应门限对观测进行处理,然后仅利用处于门限内的有效观测进行GM-CPHD算法的更新步计算,大大降低了算法的计算复杂度。实验结果证明,本文方法在有效降低计算复杂度的同时,在多目标跟踪效果方面与GM-CPHD相当,优于GMPHD滤波算法。

The Gaussian mixture cardinalized probability hypothesis density filter （GM-CPHD）is a recursive Bayesian filter for track-before-detect multitarget tracking algorithm in clutter,which propagates the first moment of the multi-target posterior density,incorporating track initiation and termination without consideration of measurement-to-track association.Due to the fact that GM-CPHD filer has a great computational complexity：O（nm3）,where n is the number of targets and m is the cardinality of measurement set,an adaptive gating algorithm is proposed.The algorithm reduces the measurement set by using a maximum likelihood adaptive gate,and only the measurements falling into the gate are used to update the PHD estimation.Simulation results show that the proposed algorithm reduces the computational complexity obviously,and obtains a similar performance.