Real-time Traffic Management Model using GPUenabled Edge Devices

Auto management and controlling road traffic while identifying abnormal driving behavior is one of the key challenges faced by the traffic authorities. In most of the cities, the traffic violations are detected manually by placing sergeants at various regions on the road. Placing sergeants is not economical and does not cover all the metropolitan area. Only in modern countries, traffic authorities have developed systems that use static road cameras to monitor real-time city traffic for identification of major traffic violations. However, these cameras just cover limited areas of the cities, such as, intersections, signals, roundabouts, and main streets. Therefore, in this paper, we have proposed a real-time traffic violation detection model by using vehicular camera along with the edge device in order to control and manage the road traffic. The edge device is equipped with the graphics processing unit (GPU), deployed inside the vehicle, and directly attached to the vehicle camera. The camera monitors every vehicle ahead, whereas, the edge device identifies the suspected driving violation. As a use case, we have tested our model by considering a wrong U-turn as a traffic violation. We designed a wrong U-turn detection algorithm and deployed it on the GPU-enabled edge device. In order to evaluate the feasibility of the system, we considered the efficiency measurements corresponding to the video generation rate and data size. The results show that the system is able to identify violations far faster than the video generation time.