A Novel AI-Based Framework Real-Time Facial Recognition within Low-Light Conditions Using Infrared Imaging

Abstract

Facial recognition, a critical tool in security and surveillance, faces challenges in low-light conditions. Traditional methods struggle with poor illumination, but infrared imaging offers a solution by capturing heat signatures. However, thermal images lack detail. This research proposes an AI-driven framework that integrates thermal and visual data (when available) to enhance facial recognition accuracy. The framework employs a pre-trained ResNet-50 convolutional neural network (CNN) for feature extraction and classification. Transfer learning is utilized to adapt the model to the specific task of facial recognition in low-light environments. To improve the quality of thermal images, histogram equalization is applied to enhance contrast and visibility of facial features. Experimental results demonstrate the effectiveness of the proposed framework, outperforming traditional methods such as the Local Binary Patterns Histogram (LBPH) and Eigenfaces. The model achieves high accuracy, precision, recall, and F1-score in recognizing faces in low-light conditions. This research contributes to the advancement of facial recognition technology, enabling reliable and efficient identification in challenging lighting scenarios.