Research on Face Recognition under Complex Illumination Conditions for Express Cabinet Applications

There are many hidden dangers in traditional identity verification methods in express cabinets such as false claiming and false taking. In this paper, face recognition technology is used to improve the security and efficiency of identity verification in express cabinet. Three deep learning networks based on SeetaFace model are used to realize face detection, face alignment and face verification. A Fust funnel cascade network is applied for face detection. A coarse-to-fine CFAN network is used to realize face alignment. And a VIPFaceNet network is used to achieve face verification. In order to estimate the influence of uneven illumination which is a major disturbance on face recognition, an adaptive gamma correction method (AGC) is proposed. Experiments on Yale B and a face database of our laboratory prove the efficiency of the strategy.