基于均值漂移的穿墙雷达多目标跟踪

Mean-Shift Based Multiple Targets Tracking for TWR

针对穿墙成像雷达中多人体目标跟踪的问题,提出了一种基于均值漂移的多目标图像域跟踪方法。首先,通过基于双轴投影的变尺度模板创建方法进行航迹起始;其次该文利用均值漂移算法进行航迹更新,并对不同幅度像素点的加权处理使得迭代过程能收敛于目标图像中心;最后,结合M/N准则,实现了对航迹的管理。通过对多组实测数据的处理可以看出,该算法具有稳健的跟踪性能。

This paper deals with the tracking problem of multiple human targets hidden behind walls using through-wall-imaging (TWI) radar. We propose a mean-shift based multi-target tracking algorithm in image domain. Firstly, an adaptive method of target template construction based on biaxial projection is presented. Secondly, the mean-shift algorithm is used to update the trajectories, and the different amplitudes is weighted in order to make the iterative process converge to the center of the target. The M/N logic is finally considered to manage trajectories. The proposed algorithm is evaluated by the experiments.