Compared with the fixed timing, the signal timing algorithm based on the dynamic change characteristics of traffic flow has better road state adaptability. Therefore, this paper proposes an adaptive control strategy based on traffic flow identification, firstly the self-organizing mapping network(SOM) neural network is used to cluster the historical traffic flow state, and the environment feature analysis of the intersection time segment and the road segment is combined to realize the traffic flow mode division. On this basis, the probabilistic neural network (PNN) is introduced to train the traffic flow pattern of the intersection. Finally, according to the traffic flow of different state types, the threshold service polling signal timing and Webster signal timing strategy is dynamically selected to calculate the signal lamp timing period, and the matching between the control strategy and the traffic flow dynamic change characteristics is realized. The simulation results show that the mixed service intersection signal control method, which distinguishes the traffic flow mode, has better adaptability to the random variation of traffic flow.