Predicting Future Traffic Conditions in Urban Areas

Traffic jams are fast becoming a severe problem in urban areas, impacting day-to-day life, consumption of fuel, and the surrounding environment. The paper proposes an efficient and real-time traffic prediction system that analyzes traffic conditions using video streams. In our scheme, vehicles are detected and counted from live video using background registration and contour detection using OpenCV. The actual count of vehicles is obtained, and a Linear Regression model is trained to predict future traffic density. One can also integrate XGBOOST with it for better accuracy in prediction. The purpose of this work is to develop a solution that can be low-cost and scalable for use on existing surveillance infrastructure without additional hardware or sensors. By adding vehicle detection to a machine-learning-based prediction, the system can facilitate effective traffic control, signal optimization, and urban planning. Experimental results show that the model performs reliably in different scenarios, enhancing its chance for deployment in smart city traffic management systems.