Face Authentication System of Thermal Image with Gabor Filter

Objectives: A new unique feature extraction technique for face authentication is implemented in thermal imaging. The proposed work extracts vasculature information creates a signature of thermal image and authenticates the individual. Methods/Statistical Analysis: A biometric system for face authentication with a better accuracy and reliability is proposed with an algorithm that uses morphological operators for feature extraction, Gabor filter for registration and correlation coefficient, similarity measures are used for matching. The operators used in image morphology are opening and top-hat segmentation. In order to ensure changes in vasculature at various time intervals four images are taken at different instants and thermal signature template is developed. Due to the consistent thermal features the biometric process is not affected. Findings: 22,784 images of eighteen subjects are used for testing the facial authentication system. An accuracy of 91% is obtained for skeletonised signatures using correlation coefficient which is better than Euclidean Distance that gives 88%. False Acceptance Rate (FAR) of the system is 0.0% for both Euclidean Distance and Correlation Coefficient. False Rejection Rate (FRR) of the system is 0.18% in the case of Correlation Coefficient and 0.24% in the case of Euclidean Distance. Application/Improvements: The enhancement of Face Authentication System to reduce the False Rejection Rate and improve the accuracy. The future scope of this work is the enhancement in military environment, where authentication is more important and the development of multimodal system.