Discipline: Computer Sciences and Information Management
Subcategory: Computer Science & Information Systems
Session: 3
Room: Hoover
Nhat Phan - Fort Valley State University
Co-Author(s): Chunhua Dong, Fort Valley State University, Fort Valley, GA.
Thermal face recognition is a more refreshing part of the facial recognition area, using the implementation of the growing artificial intelligence technology. Thermal face recognition has the potential to play a crucial role in the modern field of security, with the ability to recognize identities of individuals using their thermal infrared facial images that can be taken when the light is absent. In this study, we propose a thermal face recognition algorithm using the cross-modal support vector machine (SVM). First, band-pass filters called Difference of Gaussians (DoG) are used to emphasize edges and other details in addition to removing high and low-frequency noise. Then the texture features of images are extracted using the histogram of oriented gradients (HOG), which reduce the modality discriminant between visible and thermal facial features. Finally, we incorporate the thermal information into a one-vs-all SVM-based framework to enhance cross-model recognition. A binary SVM model is built for each subject in the training dataset, in which visible face images of this subject are positive samples and visible face images of the remaining subjects are negative samples. To improve the cross-modal discriminability between the subjects, a set of unrelated thermal face images are added as cross-examples to all the SVM models as negative samples. The trained SVM models were used to identify the corresponding subjects of test thermal images. Experimental results have demonstrated that the thermal cross-examples provide useful cross-modal information and improve recognition performance.
Funder Acknowledgement(s): The research was sponsored by the Army Research Office and was accomplished under Grant Number W911NF-18-1-0457
Faculty Advisor: Xiangyan Zeng, zengx@fvsu.edu
Role: I participate in writing the source code and running the program using different parameters to provide the results for discussion. I also provide feedback on the performance of the program and be involved in the discussion of how to improve the program.