Infrared-to-Visible Face Recognition Using Sparse Learning Systems
Abstract
This project will address cross-modal infrared-to-visible face recognition using a supervised learning system. The proposed system will consist of feature extraction, feature mapping based on sparse dictionary learning, and a sparse representation-based classification. For feature extraction, a gray scale face image is transformed into a pattern map where edges and lines characterizing the textural information are classified by pattern matching, and the feature vector of the image is comprised of a histogram of patterns. A bi-level sparse dictionary learning approach will be used to acquire the mapping between infrared and visible face features. A Laplacian prior will be added to the sparse dictionary learning model to enhance the correlation between images of the same subject. The recognition problem will be cast as classification based on sparse representation in the visible face feature space. Feature selection will be embedded in the classification to alleviate the problem introduced by lack of sufficient training data. New algorithms for infrared-to-visible face recognition and other cross-modal applications will be developed. The focus will be developing effective and computationally efficient algorithms that are applicable in a real time system. The project will also contribute to the education of historically underrepresented students in computer science, electronics engineering, and mathematics, who will be mentored to pursue graduate study for eventual research and development (R&D) careers. The results of this project will be applicable to the needs of the Army and other Department of Defense (DOD) agencies, and the Department of Homeland Security (DHS) in nighttime surveillance and intelligence gathering, which has unique challenges due to operational factors such as clutter, illumination, and occlusion. It will broaden the research and education capabilities of Fort Valley State University in support of national defense and security and other critical science and technology applications; afford opportunities for collaboration with Defense-related researchers, especially in the Army, and academia; and promote education and mentoring of computer science, mathematics, and engineering undergraduate students for subsequent doctoral studies.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 25, 2019
- Source ID
- W911NF1810457
Entities
People
- Xiangyan Zeng
Organizations
- Army Contracting Command
- Fort Valley State University
- Office of the Secretary of Defense