锥扫型光学传感器像平面多目标轨迹跟踪

Tracking multi-target on image plane for the cone-scanning optical sensor

积分概率多假设跟踪(IPMHT)是一种基于期望极大化(EM)的准最优贝叶斯多目标迭代跟踪算法,研究了该算法在锥扫型光学传感器像平面多目标轨迹跟踪中的问题。为提高算法的跟踪性能和计算效率,利用逻辑概率数据关联滤波(PDAF)方法进行目标初始状态估计,并利用目标幅度信息和波门技术对IPMHT进行优化。针对锥扫型传感器非线性观测下的多目标跟踪,将扩展无味卡尔曼滤波(AUKF)与优化的IPMHT算法相结合,实现像平面多目标轨迹的起始、维持和终结。蒙特卡洛仿真实验表明,该算法成功地解决了锥扫型传感器的像平面多目标轨迹跟踪问题,在提高目标跟踪性能的同时改善了计算效率。

Integrated probabilistic multi-hypothesis tracking（IPMHT） for multi-target was an iterative suboptimal Bayesian method based on exception maximum（EM）,which was studied for tracking targets on image plane for the cone-scanning optical sensor.In order to improve the algorithm performance and efficiency,probabilistic data association（PDAF） under logical ruler was utilized to estimate the target initial state and IPMHT was optimized as to the target intensity and association gate.To track multi-target under nonlinear observation of the cone-scanning sensor,augmented unscented Kalman filter（AUKF） was introduced to combine the optimized IPMHT for the multi-target initiation,maintenance and termination.As the Monte Carlo experiments show,the presented algorithm successfully tracks the multi-target with better tracking performance and efficiency as well.