一种基于自适应粒子滤波的目标跟踪方法

A Object Tracking Method Based on Adaptive Particle Filter

提出一种利用最近的目标轨迹信息自适应调整运动模型的粒子滤波方法,根据背景地形或道路信息建立若干目标轨迹模式,然后利用目标轨迹模式将最近的目标轨迹进行分类,通过与当前目标最近段轨迹匹配的目标轨迹类,获得当前目标在下一时刻状态后验概率分布对应的粒子。实验结果表明该方法具有较强的鲁棒性,能有效实现复杂场景下的目标跟踪。

A particle filter method whose moving model is adaptively adjusted using the information of recent object trajectories is proposed. Some object trajectory patterns are built based on the information of background terrain and road, then the recent object trajectories can be classified based on the built object trajectory patterns. The particles of current object＇ s next time posterior probability distribution can be obtained from a certain object trajectories class which matchs the current object＇ s recent part of trajectory. Experimental results show that the proposed method in this paper is robust and can effectively realize object tracking under eomplex background.