摘要
保持局部图嵌入的流形鉴别分析方法将图像集所属子空间看作流形上的点,并使流形变换前后局部结构关系不变.然而在构造局部区域相似图矩阵时,用于描述节点局部区域范围的近邻节点个数会极大地影响算法的准确率,并会出现变换后流形的可分辨性相比变换前提升很小甚至更低的情况.针对该问题,提出了一种低秩描述下的Grassmannian流形鉴别分析方法.通过对图像集的低秩描述,流形变换中局部嵌入时仅保持同类别节点的最近邻局部结构以及所有节点间的相异类别信息,从而避免了对近邻节点个数的选择,并增强了变换后流形的可分辨性.由15类复杂自然场景和Caltech101图像数据集的实验结果表明,该方法是可行的,并且极大地提高了图像集分类的准确率.
Graph embedding discriminant analysis on manifold approach represents each image set as a subspace on manifold. It maps the manifold to a more discriminative one with geometrical structure and local information preserved. However, its accuracy critically depends on the number of local neighbours when constructing similarity graph. This paper presents a novel approach with fixed neighbour numbers to implement graph embedding Grassmannian discriminant analysis based on low-rank representation （LRR） for each image set. After the low-rank components of each set being recovered, to preserve the nearest neighbour structure of nodes with the same label and all the different label information during the manifold mapping can always achieve the best performance. Experiments on two image datasets （15-scenes categories and Caltechl01） show that the proposed method greatly improves the classification accuracy of image sets.
出处
《同济大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2013年第2期271-276,共6页
Journal of Tongji University:Natural Science
基金
国家自然科学基金(61103070)
"十二五"国家科技支撑计划(2012BAF10B12)
上海市科委项目(12dz1125400)
中央高校基本科研业务费专项资金(0800219171)
关键词
流形鉴别分析
低秩分解
图像集
局部图嵌入
manifold discriminant analysis
low-rank representation
image set
graph embedding
