面向快速多目标跟踪的协同PHD滤波器

考虑到存活目标与新生目标在动态演化特性上的差异性，提出了面向快速多目标跟踪的协同概率假设密度(collaborative probability hypothesis density, CoPHD)滤波框架。该框架利用存活目标的状态信息，将量测动态划分为存活目标量测集与新生目标量测集，在两个量测集分别运用PHD组处理更新基础上建立了处理模块的交互与协同机制，力图在保证跟踪精度的同时提高计算效率。该框架由于采用PHD组处理方式而具有状态自动提取功能。进一步给出了该框架的序贯蒙特卡罗算法实现。仿真结果表明，该算法在计算效率以及状态提取精度上具有明显优势。

Collaborative PHD filter for fast multi-target tracking

Considering the difference of dynamic evolution between the survival target and the newborn target,a collaborative probability hypothesis density (CoPHD) filter framework for fast multi-target tracking is proposed. The framework strives to improve the systematic implementing efficiency as well as guarantee the tracking accuracy by dynamically partitioning the measurement set into two parts,survival and newborn target measurement sets in which PHD groups are updated respectively,and constituting an interactive and collaborative mechanism for the processing modules. In addition,the framework has the ability of state self extracting by utilizing PHD group processing, and the implementation via the sequential Monte Carlo (SMC) method is presented. Simulation results show that the proposed SMC-CoPHD filter has greatly reduced computation cost and significantly improved state extraction accuracy.