混合线性模型在模板跟踪中的应用  

Mixture hyperplanes approximation for template tracking

作　　者：顾菘[1] 马争[1] 解梅[2] 

机构地区：[1]电子科技大学通信与信息工程学院,成都617731 [2]电子科技大学电子工程学院,成都617731

出　　处：《计算机应用研究》2016年第8期2500-2503,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(61271288);国家高技术研究发展计划资助项目(2012AA011503)

摘　　要：模板跟踪在计算机视觉上已经有了很广泛的应用。利用图像的单应性特征,构建目标的运动模型,找出观察数据与目标运动参数的关系,估计目标的运动参数,实现目标跟踪的目的。提出了一种基于混合线性模型的模板跟踪方法,提取目标的运动参数和表观特征,建立数据集,利用全监督学习的方法计算出两者的映射关系,从而实现对目标的有效跟踪。这种方法既克服了由单一线性模型造成的非线性误差,又减少了非线性模型的计算量,提高了目标跟踪的精度。此外,还提出了一种快速学习的计算方法,克服了混合模型中每个子空间由于学习样本少而容易受到噪声影响的缺点,不仅增加了系统的鲁棒性,而且减少了计算量。实验结果表明该算法具有良好的跟踪效果。Template tracking has been extensively studied in computer vision with a wide range of applications. The motion model is constructed by homography, and the object is tracked by approximating the motion model based on the relationship of the observed data and the motion parameters. This paper proposed a method based on finite mixtures of generalized linear re- gression models to perform robust, real-time tracking from a stationary camera. The model was learnt by motion parameters and appearance data of the object. Moreover, it discussed a fast learning strategy as well, which would improve robustness against noise. It also demonstrated and evaluated the performance and the stability of mixture hyperplanes approximation on a set of challenging image sequences. Experimental result shows that the algorithm has better tracking effect.

关 键 词：模板跟踪 混合线性模型 快速学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]