Research on Urban Traffic Flow Detection and Forecasting Based on YOLO Model and LSTM

Abstract

Urban traffic congestion is becoming increasingly prominent, and accurate traffic flow detection and short-term forecasting are crucial for achieving intelligent traffic scheduling and alleviating congestion. To enhance the real-time performance and accuracy of traffic flow detection in transportation scenarios while enabling effective traffic flow prediction, this paper employs an improved lightweight YOLO object detection model for real-time road vehicle recognition and traffic flow statistics. It further integrates a Long Short-Term Memory (LSTM) network to construct a traffic flow time-series prediction model. Experimental results demonstrate that the YOLO model efficiently performs vehicle detection on edge devices. When integrated with the LSTM model, it achieves precise traffic flow predictions for the next two hours. This provides data support for intelligent scheduling by traffic management authorities and travel planning for users, offering practical application value for smart transportation development.