The invention discloses a deep learning and intelligent street lamp-based traffic signal lamp intelligent control system, and belongs to the field of intelligent public traffic. The system comprises a roadside unit and an intelligent traffic signal lamp, wherein the roadside unit and the intelligent traffic signal lamp are arranged at a traffic road intersection. The roadside unit comprises an image acquisition module, a vehicle detection module, a congestion judging module and a communication module. The image acquisition module collects road traffic images in each intersection direction and sends the images to the vehicle detection module. In this way, the vehicle flow in each intersection direction is obtained. Meanwhile, the congestion judging module determines the congestion level in each intersection direction in combination with the vehicle flow, and transmits the congestion level to the intelligent traffic signal lamp through the communication module. The intelligent traffic signal lamp adjusts and displays the transit time in real time based on the congestion level. According to the technical scheme of the invention, the road traffic condition is detected in real time. For different road sections, different congestion level threshold values are set flexibly. Therefore, the congestion condition can be automatically judged and the transit time in each direction at the intersection is adjusted. As a result, traffic signal lamps are more flexible and more intelligent to use.