Published Paper Details:

With the rapid growth of urban population and the increasing number of vehicles on roads, traffic congestion has become a major issue in many cities. Conventional traffic management systems, which mainly rely on fixed signal timings and manual monitoring, often fail to respond to changing traffic conditions. This results in longer waiting times, increased fuel consumption, and unnecessary traffic buildup. In our project, we aim to address this problem by developing an intelligent traffic management system that uses computer vision and deep learning to analyze real-time traffic situations and make better decisions. The proposed system takes traffic video input, either live or recorded, and processes it frame by frame to detect vehicles and other road entities. For this purpose, a YOLO-based object detection model is used, which is capable of identifying different types of vehicles such as cars, bikes, buses, and trucks. From our observations during testing, the model was able to detect most of the visible objects with good accuracy, even in moderately crowded scenes. This forms the foundation for further analysis in the system. After detection, the system calculates traffic density and occupancy percentage. Instead of relying only on the number of vehicles, the system also considers how much area they occupy on the road. This gives a more realistic understanding of the actual traffic condition. Based on these values, the traffic is classified into categories such as low, medium, or high. In our testing, this classification matched well with the actual visual conditions in the video, which shows that the approach is practical and reliable. Another important feature of the system is adaptive signal timing. Rather than using fixed durations, the signal time is adjusted dynamically based on the current traffic condition. For example, if the traffic is low, the signal duration is reduced, and if it is high, the duration is increased. This helps in improving traffic flow and reducing unnecessary waiting time at intersections. We noticed that this feature makes the system more efficient and suitable for real-world applications. In addition to this, a simple dashboard is developed to display key information such as vehicle count, density percentage, congestion level, and signal timing. This allows traffic authorities to easily monitor and understand the traffic situation. The system can also be extended to include emergency vehicle prioritization for better handling of critical situations. Overall, the results indicate that the proposed system provides an effective and practical solution for traffic management. Although there are some limitations, such as reduced performance in highly congested scenarios, the system performs well under normal conditions. In the future, the model can be improved further and integrated into smart city systems for large-scale deployment.