2019/02/23 by Sumit Agarwal, Agarwal, Sumit, Harshit Sikchi +7
Computer Science · #Face recognition and analysis #Biometric Identification and Security #Face and Expression Recognition
paper · pdf · doi:10.48550/arxiv.1902.08802
Face recognition in real life situations like low illumination condition is\nstill an open challenge in biometric security. It is well established that the\nstate-of-the-art methods in face recognition provide low accuracy in the case\nof poor illumination. In this work, we propose an algorithm for a more robust\nillumination invariant face recognition using a multi-modal approach. We\npropose a new dataset consisting of aligned faces of thermal and visual images\nof a hundred subjects. We then apply face detection on thermal images using the\nbiggest blob extraction method and apply them for fusing images of different\nmodalities for the purpose of face recognition. An algorithm is proposed to\nimplement fusion of thermal and visual images. We reason for why relying on\nonly one modality can give erroneous results. We use a lighter and faster CNN\nmodel called MobileNet for the purpose of face recognition with faster\ninferencing and to be able to be use it in real time biometric systems. We test\nour proposed method on our own created dataset to show that real-time face\nrecognition on fused images shows far better results than using visual or\nthermal images separately.\n