DEVELOPMENT OF AN INTELLIGENT TRAFFIC CONTROL SYSTEM BASED ON COMPUTATIONAL VISION
OpenCV. MOG2. YOLO SSD. MobileNetV2. CNN. Artificial intelligence. Deep Learning. Smart traffic light. Redundancy.
A vehicle fleet in Rio Grande do Norte has increased by approximately 250 thousand vehicles in the last 5 years, or 7% per year. Considering that 80% of the population lives in urban areas, traffic management is becoming one of the most important issues. Traffic lights that operate with fixed time (STF) to control the flow of vehicles are not efficient in all databases. At this time, in the literature, many studies have advanced traffic light control based on vehicle density as a solution to improve traffic flow. With the advancement of Computer Vision (VC) technologies, such as detection and classification techniques for moving objects and the requirement of little computational power to perform these tasks, it was possible to develop an intelligent traffic control system based on VC. This low-cost solution was implemented to take advantage of the FTS system, cameras and logical network infrastructure already present in municipalities in Brazil. An application developed and implemented on a computer was used to capture photo(s) of the traffic light(s), count the vehicles and calculate the time needed for them to make the crossing. The Raspberry Pi 3 controlled like traffic lights. Compared to the STF, there was a gain of up to 33% in the fluidity of traffic. The VC was used to count the vehicles crossing or traffic lights, thus, it was possible to warn about traffic jams and also to create a database that can be used for decision-making by the bodies with circumscription about the lanes.