K2DPCA Plus 2DPCA: An Efficient Approach for Appearance Based Object Recognition

Abstract

In this paper, we propose a new object recognition algorithm called two-directional two-dimensional kernel-based principal component analysis(K2DPCA plus 2DPCA). This approach mainly analyzes the object in the two dimensional principal component analysis (2DPCA) transformed space. Firstly, decorrelation in the row direction of images by through the standard K2DPCA method, then using 2DPCA way to further decorrelation in the column direction of images in the K2DPCA subspace. To overcome the shortcoming of massive memory requirements of the 2DPCA and 2D-FPCA, we introduce K2DPCA plus 2DPCA method, which needs smaller memory space and has higher discernment rate, and computational efficiency is higher than the standard KPCA /K2DPCA/(2D) 2 FPCA method. Finally, we verify this method in the finger vein database.