Real-Time Face Attendance Marking System in Non-cooperative Environments

Face recognition achieves good performance in attendance marking system, but most of face attendance marking systems need people cooperate with the camera. In this paper, we propose the real-time face attendance marking system. It works well in non-cooperative environments. Firstly, detected face regions are tracked to help detection algorithm detect occluded and deformed faces. Then, the features of tracklets (track fragment) formed by detected faces are extracted to realize face recognition. By using tracklets instead of a single image, it realizes robust recognition in non-cooperative environments. Finally, we create a reference gallery of multimodal facial features, which improves the accuracy and speed of multimodal face recognition. Experiments show that our system can detect and recognize multimodal faces in non-cooperative environments and run in real-time (25FPS).