摘要
主要研究了视频图像目标跟踪准确性问题。在基于核的颜色特征统计描述及以此建立视觉目标观测概率方法的基础上,提出了一种改进的粒子滤波视频图像目标跟踪算法。首先,本文给出了基于标准粒子滤波的单特征、单目标跟踪算法,然后针对加权样本参数的选择不同,提出改进思路,最后通过与基于均值移位视觉目标跟踪算法的实验结果对比。提出的改进的粒子滤波跟踪算法在稳健性方面有显著地提高,而且若适当选择视觉跟踪参数,在实时性方面能得到有效地保证。
This paper mainly studies the video image target tracking accuracy.In the kernel based on color feature statistical description and to establish a visual target observation probability method,this paper proposes an improved particle filter algorithm for target tracking in video image.First,this paper gives the standard particle filter based on a single feature,single target track algorithm,and then the weighted sample parameter selection is different,put forward improvement ideas,finally through with visual target tracking based on mean shift algorithm experimental results,this paper puts forward the improved particle filter tracking algorithm in robustness is obviously improved,and if the appropriate selection of parameters in real-time visual tracking,can be effectively guaranteed.
出处
《科技通报》
北大核心
2012年第8期72-73,76,共3页
Bulletin of Science and Technology
关键词
粒子滤波
目标跟踪
视频图像
单目标
particle filter
target tracking
video image
single target
