Research on Adaptive Adjustment Method of Intelligent Traffic Light Based on Real-Time Traffic Flow Detection

In this paper, YOLOX algorithm is used to detect the traffic flow at traffic intersections by using UA-DETRAC dataset. The motor vehicle detection mode and vehicle classification detection mode are used to analyze the experimental results, and according to the experimental results, the adaptive scheduling strategy of intelligent traffic lights at traffic intersections is put forward. In the motor vehicle detection mode, our model obtains a map of 0.844 on the test set. In the vehicle classification detection mode, our model obtains an average map of 0.607 on the test set, in which the map of cars and buses reaches 0.799 and 0.863 respectively, which shows that our model has good detection and positioning performance for these two categories of motor vehicles. We propose two different traffic flow calculation strategies according to whether it is in the morning and evening peak hours: in the morning and evening peak hours, we use the vehicle classification detection model for classification detection and calculate the weighted traffic flow, and in the off-peak period, we use the vehicle detection model to count the actual traffic flow.