一种基于粒子滤波的多目标跟踪算法

来源期刊:中南大学学报(自然科学版)2011年第z1期

论文作者:许伟村 赵清杰

文章页码:805 - 810

关键词:多目标;跟踪;粒子滤波;子空间

Key words:multi-target; track; particle filtering; sub-space

摘    要:提出一种新的基于粒子滤波的多目标跟踪算法。该算法根据每个目标的状态估计值从全局观测空间中为其分配一个独立的子空间作为其生存空间,通过分别估计每个子空间中的后验概率分布,以更低的计算复杂度得到多模态的后验概率分布估计,从而实现多目标跟踪。另外,提出一种基于加速度估计的动态模型和基于该模型的自适应粒子数策略,能够进一步提高在非线性非高斯状况下跟踪的准确度。在实验中,使用提出的算法对实际拍摄的视频序列中的多个目标进行跟踪,并针对一种非线性非高斯系统分别采用3种不同多目标跟踪算法所用的动态模型进行实验。结果证明:所提出的算法能够在实际环境中实时稳定地实现目标数不固定的多目标跟踪,且提出的动态模型也明显优于其他2种动态模型,更加适用于非线性非高斯情况下的跟踪。

Abstract: A new multi-target tracking framework was proposed based on particle filtering. It first supplies a sub-space from the global observation space for each target according to its state value, and then computes the posterior density of every sub-space from randomly generated particles to get an approximation of global multi-modal posterior density with much lower computational complexity. A new dynamic model based on acceleration estimation and a new adaptive particle strategy based on this dynamic model were also proposed to approve the tracking precision in non-linear non-Guassian situation and decrease the computational cost. The proposed tracking algorithm was tested by tracking targets in real visual series captured in real world and the results indicate that the proposed algorithm can track multiple targets robustly almost in real time when the number of targets is not fixed and really large. Another experiment was carried out to compare the efficiency of the proposed algorithm and another multi-target tracking framework in detecting new targets and in the experiment the proposed algorithm performs much better than the other one. The dynamic model was also tested with two other different models and proved to be better than the other two models.

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号