With the acceleration of urbanization, traffic congestion has become one of the major issues impacting the urban efficiency and the life quality of the residents. Particularly during the peak traffic flow at intersections, the unreasonable signal timing schemes frequently result in severe traffic congestion and queue overflows. In order to reduce the number of vehicles experiencing secondary queuing and the maximum queue length during peak hours at the intersections, and enhance the efficiency of traffic flow, the DQN (Deep Q Network) algorithm in the deep intensive learning is used to optimize signal timing schemes at intersections. Firstly, the simulation model of intersection is established as the learning environment of DQN algorithm. Secondly, the actions are accepted and the feedback is given. Finally, the traffic operation indicators are extracted to assess the optimization effects of the DQN algorithm in intersection signal control. The research result shows that compared to the Webster signal timing method, the DQN algorithm can reduce the maximum queue length by 14% and the intersection vehicle delay by 16%, demonstrating that signal control schemes under the DQN algorithm can more effectively adapt to fluctuations in traffic flow, thus improving the operational efficiency of vehicle flow at intersections.