Research on Traffic Flow Detection Based on Yolo V4

With the continuous development of science and technology and transportation industry, intelligent transportation is the main trend of urban development, and traffic flow monitoring is an important part of intelligent transportation, which is of great significance for traffic guidance and path planning. In this paper, the Yolo V4 network model is modified to increase the convolution times after the feature layer. On the one hand, it can deepen the network depth and increase the network capacity and complexity. On the other hand, it can obtain more global and higher semantic level feature information. Through the test of the test set, it is found that the average detection time of the improved model is 8 ms slower than that of the original model, but in terms of test accuracy, the improved model is 2.16% higher than the original model, and the FPS is 25, which can meet the requirements of real-time detection. Combined with vehicle tracking and virtual line detection method to count the vehicles and verify the traffic video, the statistical accuracy of the improved model is 93%, which is 2% higher than that of the original model.