一种基于模型选择的PF-TBD算法

在目前的基于粒子滤波检测前跟踪(PF-TBD)算法中,对粒子的预测通常是根据单一直线运动模型实现的,在目标机动时由于与运动形式相差较大,影响了跟踪效果。为此,提出一种基于模型选择的粒子滤波检测前跟踪(MM-PF-TBD)算法。该算法由已估计出的目标位置,计算相对偏转角,并以此判定目标当前的运动模式,进而选择相应的运动模型对下一时刻的粒子进行预测,显著提高了对粒子预测的精度。理论分析和仿真实验表明,文中所提算法适用于目标不同的运动形式,有效提高了目标机动时的检测和跟踪性能。

A Multiple-Model Choosing Particle Filter Tracking Before Detection Algorithm

In the current particle filter tracking before detection algorithm（PF-TBD）, the prediction of particle is usually based on a single straight line motion model, the difference between the forms of exercise and motion model affects the tracking results when target maneuver. To solve this problem, this paper presents a multiple-model choosing particle filter tracking before detection（MM-PF-TBD） algorithm. By the estimated target location, the algorithm calculates the relative deflection angle, and determines the current moving patterns of the target, and then selects the appropriate motion model to predict the particle at next time, which significantly improves the prediction accuracy of the particle. Theoretical analysis and simulation experiments show that the proposed algorithm can be applied to different forms of exercise, and effectively improves the detection and tracking performance when target maneuvering.