Towards Smarter Cities: Real-Time Traffic Optimization Using Computer Vision and Data Intelligence

Abstract

Congestion in urban roads is one of the largest dilemmas in contemporary cities which causes an increment in the travel. it is cheaper, less consuming fuel and causing pollution. The standard traffic control systems are cycle-based. Existing intelligent traffic systems rely on manual monitoring and thus could not cope with dynamic traffic conditions, a fact that is augmented by vast current intelligence traffic systems. The drawbacks of the use of these data sources are the lack of the real- time responsiveness, the multi-source data integration, and scalability. To respond to capitalism, In this project, an Advanced Traffic Management System will be achieved using real-time data, predictive variables. Electronic merchandising, analytical tools and computer vision. The system combines live video transmissions of mobile cameras and other general cameras. Validation of traffic journeys, access to traffic plans, planning in timetables, and weather information, preprocessing, and object detection of the vehicles. z counting and congestion estimation, map based interface, with an overlay traffic status and route. Suggestions on Google Maps. The system was trained and tested on open-source and simulation data sets such as AI City Challenge traffic flows, GTFS APIs, OpenWeatherMap information, and past traffic information. The d-dimensional Cosine Similarity and the d-dimensional Jaccard Similarity can both achieve an accuracy of 92%, and possession a precision of 89.4 and an F1-score of 90.1, respectively, with a processing time of only 0.084 and 0.048 seconds, respectively. 105 milliseconds. The results show greater accuracy, efficiency and reliability. Firm base to further improvements such as adaptive signal control, edge computing and complete smart city integration.