Efficient face recognition in video for uncontrolled environment

Abstract

In this research work, an efficient Face Recognition in video for uncontrolled environment is proposed to effectively recognize a face in real time for smart surveillance and secure authentication services. Intelligent video surveillance systems equipped with face recognition facilitates an organization to automate the process of person identification and prevent unauthorized access. An efficient Face Recognition system comprises of Face Detection, Face Alignment and Normalization, Face Liveness Detection, Feature Extraction and Feature Matching. This research work proposes an efficient Face Recognition System (FRS) to enhance the performance of video surveillance systems and to identify the presence of an unauthorized person. A robust and accurate face recognition system should address challenges like variations in pose, illumination, facial expression, aging, scaling and occlusion. The success of a face Recognition system depends on the robustness of the face detection module in terms of speed and accuracy. Though Viola-Jones based face detector is widely used, it is not efficient in real time due to high false positives. This research work proposes Viola-Jones with CbCr color model for efficient face detection. To address the need for fast detection rate, the proposed VJ-CbCr face detector is implemented upon NVIDIA Quadro K2000 with 384 CUDA cores and Hybrid Scale Invariant Feature Transform based feature descriptor is proposed to recognize face in video under different illumination and pose variation. newline newline newline