Dual subspace discriminative projection learning. (March 2021)

• Record Type: Journal Article
• Authors: Belous, Gregg; Busch, Andrew; Gao, Yongsheng
• Is Part Of: Pattern recognition. Volume 111(2021)
• Publication Date: 2021-03
• Subjects: Pattern recognition -- Feature extraction -- Subspace learning -- Image classification -- Subspace discriminative projection; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2020.107581
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: We propose a novel feature extraction algorithm called dual subspace discriminative projection learning (DSDPL) for multi-class image classification with low sample size training data. Our approach serves to decompose original high dimensional data, via learned projection matrices, into class-shared and class-specific subspaces. Comprehensive experimental analysis is performed across five publicly available databases for face, object and scene classifications. Our experimental results demonstrate the effectiveness of DSDPL over current benchmark subspace learning methods and deep learning models. Abstract: In this paper, we propose a dual subspace discriminative projection learning (DSDPL) framework for multi-category image classification. Our approach reflects the notion that images are composed of class-shared information, class-specific information, and sparse noise. Unlike traditional subspace learning methods, DSDPL serves to decompose original high dimensional data, via learned projection matrices, into class-shared and class-specific subspaces. The learned projection matrices are jointly constrained with l 2, 1 sparse norm and LDA terms while the reconstructive properties of DSDPL reduce information loss, leading to greater stability within low dimensional subspaces. Regression-based terms are also included to facilitate a more robust classification approach, using extracted class-specific features for better classification. Our approach is examined on five different datasets for face, object and scene classifications. Experimental results demonstrate not only the superiority and versatility of DSDPL over current benchmark approaches, but also a more robust classification approach with low sample size training data.