Abstract: Aiming at the problem of difficult identification of personnel in mine video surveillance system, a video identification system based on light signal of miner's headlamp was developed. By embedding near infrared LED light signal in the headlamp, the light signal was designed to be easy to identify and track.In this study, an optical signal coding method based on Manchester code was designed, and a series of detection methods including Hough circle/ellipse detection method and dynamic optimization decoding method were proposed.The recognition identification system transmits the unique ID optical signal through the miner's headlamp, and the video surveillance system collected and processed the light signal to parse the miner's headlamp ID. The effectiveness of the video recognition system was verified by underground experiment in coal mine. The research results provide an innovative personnel identification technology for mine safety production.