Abstract: Targeting the insufficient performance problem of urban rail transit face recognition gate system, a performance optimization solution based on pre-filtering dynamic face database is proposed. Through the pre-filtering of face data, a pre-filtering dynamic face database is established, so as to shorten the time-consumption of face comparison and search in system, thus to achieve the performance optimization goal. It has been verified by experiments that this optimization solution can greatly improve the performance of the face recognition gate system and has relatively high feasibility.