An intelligent traffic management system based on the Internet of Things for detecting rule violations

One of the top priorities in smart cities is having an effective traffic management system. Traffic management systems greatly aid in the planning of traffic flow, the prevention of traffic accidents, and the reduction of traffic congestion. However, numerous challenges must be overcome due to the high utilization of many vehicles, a lack of sufficient workers to manage traffic flow, and the fact that traffic violations are sometimes not captured. Drivers who travel at excessively high speeds and those who violate traffic laws by making unwarranted lane changes and other manoeuvres are the most significant contributors to increased collisions. This problem must be addressed immediately to reduce the number of deaths that have occurred for no apparent reason. Because they are not designed to prevent excessive congestion, today's traffic control systems, mainly developed and directed by human specialists, must be revised. Tensor Flow, a machine learning platform and you only look once (YOLO), an object identification technique, is used in this study to propose a hybrid model for real-time vehicle recognition. The hybrid model is "you only look once" (Yolo). The proposed technique determines the improvement of the YOLOv3 algorithm in-vehicle detection systems over the previous model by integrating these two dependencies with other requirements and using Python as the programming language. The proposed hybrid model computes the data from a surveillance camera near the traffic signal. This camera is linked to the city's traffic servers. Suppose any vehicle crosses the device while violating the traffic above regulation. The information must be detected and transmitted to the server immediately. Furthermore, it will be easier to determine who was responsible for violating the traffic regulation, making it easier for the traffic department to enforce the laws strictly. Signal jumps are determined using the region of interest and the vehicle's location throughout the frames. After being optimized for accuracy, the proposed model has an approximate accuracy of 93.83%which will help reduce the number of daily accidents.