Through video object deep analyzing based on deep neural network, we have broken through the technical problem that face recognition is hard to be used in various complex environment, and realize face recognition with high accuracy, fast speed, and strong adaptability. The developed systems have been deployed and applied in Chongqing, Xinjiang and Shanghai etc.

1. High accuracy: The accuracy rate of automatic face acquisition can reach 99.5%, and the accuracy rate of face alignment can reach 98%;

2. High real-time: Dynamic face detection and tracking time is less than 20ms, and face recognition time is less than 40ms;

3. Powerful transplantable capability: The system can overcom the problem that the traditional face recognition systems need onerous work of data collection and manual annotation once used in new environment.

The system can be applied in the safety protection, criminal investigation and public utilities fields such as banking, insurance, and transportation etc.