Face recognition in low light conditions

Abstract

This paper is devoted to an urgent problem of biometric authentication - face recognition in conditions of poor visibility and low light conditions. The article analyzes the main difficulties, such as sensory noise and loss of key image features. The Head Pose Image Database dataset was used to study the effectiveness of the algorithms. Special attention is paid to the method of expanding the training sample using data augmentation in the OpenCV library. The process of programmatically adjusting the brightness and contrast of images is described, which allows simulating extreme shooting conditions to improve the accuracy of modern computer vision systems.