Image processing based smart intersection application in Malatya province

Due to the increase in migration from rural to urban areas in recent years, urban resources are insufficient with the maximum increase in the population in the centers. It has caused many problems in the fields of transportation, security, environment, health, management and economy. By solving these problems, the concept of smart city has been developed to make cities more livable and manageable. With technological developments such as sensors, internet connection speeds, big data, cloud computing, artificial intelligence, internet of things, smart transportation, smart parking, smart stop, smart waste collection, smart lighting etc. smart city applications have been realized and continue to be implemented. One of the most important components used in smart transportation systems is smart intersection systems. In normal intersection systems, the duration of traffic lights is determined as fixed. Vehicles and pedestrians have to wait within the specified time even if there are not many or no vehicles in the traffic. In smart intersection systems, the duration of the traffic lights is determined automatically according to the vehicle density. Two methods are used in these systems. In the first method, the camera system and the number of vehicles at the intersection are calculated by image processing methods. Optimize the waiting time in signaling according to the number of vehicles. In the second method, the number of vehicles is determined by the sensors placed on the roads at the intersections. Then, the traffic density is prevented by giving the right of way to the side with high vehicle density by the system. In addition, in this system, it is possible to intervene remotely with the central management. Passing privileges to the roads can be given manually. The smart intersection system, which is used with a special permission from the Malatya Metropolitan Municipality transportation unit, has been examined. The operation of this system and the benefits it provides are emphasized. In addition, DARKNET's real-time object detection system YOLOV3 deep learning model, which is an open source neural network framework, was used for vehicle detection from data images at intersections. Counting and classification of vehicles were instantly detected and recorded in the database. With the program interface, it is ensured that the signaling is dynamic with the instant information here. Thus, fuel savings, time and emission of less toxic gases from vehicle exhausts to the environment are ensured. It contributes to the creation of a clean environment.