红外与激光工程, 2002, 31 (2): 143, 网络出版: 2006-04-28
基于奇异值分解的特征跟踪方法
Eigen-tracking method based on singular value decomposition
摘要
在传统的基于模板匹配的跟踪方法中,均是给定一个模板,然后从图像中各个位置取出一个个与模板大小一致的区域进行相似性度量,找出与模板距离最小的一个区域作为当前模板,以便进行下一步的匹配跟踪工作.在景象匹配和相关跟踪过程中,由于所面临的大多数是变化的场景,实时获取的图像与预存模板之间存在比较大的差异,传统相关匹配方法的应用就会受到限制;而且在跟踪过程中,随时更新模板会造成跟踪性能对扰动过分敏感,从而产生漂移.首先拍摄目标不同角度的图像(尽可能包含目标可能出现的所有情况),构成目标图像训练集合,抽取出特征矩阵,对它进行奇异值分解,构成一个关于目标的多维空间.然后再用匹配方法在全局范围搜索,找出目标的大致位置,并利用收敛方法在确定的大致位置内进行搜索,确定目标的仿射变换参数,从而得到一个目标位置的确切描述.
Abstract
Traditional matching will be degraded in scene-matching and correlative tracking system because of the presence of the difference of the gray value and the distortion between the object image and the reference image. And the risk of template drift will increased by the method of simply updating the template. Eigen tracking approach for tracking rigid objects is presented which builds on and extends work on SVD method. It is a promising candidate for an appearance-based object representation. Firstly the test object images are decomposed to get the eigne space of the object. And then they are applied as the characters of the object image in the matching process. The objects over long image sequences are tracked by using the techniques. This method could bear affine image motion and image distortion to some extent and it is very promising.
周晓, 任仙怡, 张桂林, 张天序. 基于奇异值分解的特征跟踪方法[J]. 红外与激光工程, 2002, 31(2): 143. 周晓, 任仙怡, 张桂林, 张天序. Eigen-tracking method based on singular value decomposition[J]. Infrared and Laser Engineering, 2002, 31(2): 143.