基于优化AGA与辅助边缘粒子滤波的目标跟踪方法

目标跟踪是当今的重要研究课题，广泛应用于通信导航、计算机视觉与自动控制等领域。针对现有的边缘粒子滤波算法目标跟踪可靠性低的问题，提出了一种基于优化自适应遗传算法（Adaptive Genetic Algorithm，AGA）和辅助边缘粒子滤波的目标跟踪方法。在状态空间降维的基础上，推导出崭新的辅助边缘粒子滤波框架，有机地将目标运动的状态划分成线性分量和非线性分量。对于线性分量，沿用卡尔曼滤波估计；对于非线性分量，植入辅助变量构建显式概率分布函数。另一方面，提出了一种的优化AGA实时调节交叉概率与变异概率具有非线性特性，以期筛选出优越的粒子拟合目标的运动状态。实验结果表明，所提出的方法能有效跟踪常见目标，具有估计准确的优点。

Target tracking method based on optimized AGA and auxiliary marginal particle filter

Target tracking has become an important research topic, which applies in communication navigation, computer video and auto control fields. Aiming at the low tracking reliability of the existing marginal particle filter (MPF) algorithm, a target tracking method based on optimized adaptive genetic algorithm (AGA) and auxiliary MPF is presented. According to reduced dimension in state space, a novel auxiliary MPF is derived firstly, where the target motion state is organically separated into both linear component and nonlinear component. In view of linear component, the Kalman filter is utilized. As for nonlinear component, the explicit proposal probability distribution function is achieved based on auxiliary filtering variable. Besides, an optimized AGA is presented to adjust both crossover probability and mutation probability for drawing robust particles that can approximate target motion state. The example results indicate that the proposed method can effectively track normal targets with accurate estimation.