Intelligent Traffic Light Control System Based On Traffic Environment Using Deep Learning

Abstract

Abstract At present, the traffic control frameworks in India, need insight and go about as an open-loop control framework, with no input or detecting system. Present technologies use Inductive loops and sensors to detect the number of vehicles passing by. It is a very inefficient and expensive way to make traffic lights adaptive. Using a simple CCTV camera can improve the conditions. The visual tracking of objects is amongst the most critical areas of computer vision and deep learning. The objective of this work was to develop the traffic control framework by presenting a detecting system, which gives an input to the current system, with the goal that it can adjust the changing traffic density patterns and provides a vital sign to the controller in a continuous activity. Using this method, improvement of the traffic signal switching expands the street limit, saves time for voyaging, and prevents traffic congestion. The framework additionally goes for consolidating exceptional arrangements for clearing the path for emergency vehicles. In this paper, we will detect and track vehicles on a video stream and count those going through a defined line and to ultimately give an idea of what the real-time on-street situation is across the road network. Our real target is to advance the deferral in the travel of vehicles in odd hours of the day. It uses the YOLO (“You Only Look Once”) object detection technique to detect objects on each of the video frames And SORT (Simple Online and Real-time Tracking algorithm) to track those objects over different frames. Once the objects are detected and tracked over different frames, a simple mathematical calculation is applied to count the intersections between the vehicle’s previous and current frame positions with a defined line. However, accuracy drops when vehicles are either close together or have large shadows, dark vehicles do not always meet the detection criteria, and night scenes are challenging to resolve as headlight beams can create large areas that meet threshold criteria. We are focusing on the Indian roads where the traffic conditions are harsh and abrupt. We have already tested it on a live feed from various traffic prone roads and have found satisfactory results. Other Adaptive Traffic Light Systems were not able to work on the Indian traffic conditions. Our model has the edge over the other by performing at par on Indian roads. We’ll increase or decrease the timer according to the conditions of the roads. This will tremendously improve the traffic conditions at a very low cost. Inductive loops are a feasible but expensive method. So, this system reduces costs and provides quick results.