Deep Face Recognition for Intelligent Video Surveillance at Electrical Substations

Abstract

Robust face recognition (FR) in real-world surveillance videos is a challenging but important issue due to the need of practical applications such as security monitoring at electrical substations. While the performance of current FR systems has been significantly boosted by deep learning technology due to its high capacity in learning discriminative features, they still tend to suffer from variations in pose, illumination, occlusion, scale, blur or low image quality in real-world surveillance videos. In this paper, we propose a novel framework which integrates face detection and recognition with tracking. Extensive experiments validate the effectiveness of the proposed framework. Our method outperforms previous SOTAs on three public datasets, i.e., LFW, CFP and AgeDB. Moreover, on the challenging testing datasets collected from the electrical substation surveillance system, the proposed method achieves an average accuracy of 91.4%.