A real-time density-based traffic signal control system

Abstract

Traffic congestion and accidents have become two major issues in Sri Lanka today. These issues cause to create many social, economic and environmental problems. Lack of effective Traffic Light Control System is one of the reasons for it happen. This research proposed an approach to develop an effective real-time density-based traffic light control system. This research consists of two major parts; Image processing model for capture real-time data and ANN model for predict the results considering real-time data. Identify the best features from gathered data and minimize dimensionality between the features, by principal component analysis (PCA) to train a Neural Network model. Using cameras, lanes are monitored and capture image of its. Detection and counting of number of vehicles in each lane and length of queue is done by using image processing. The data from each lane is sent to the ANN unit. According to the count of vehicles, the trained model will be decided the lane and time limit that will need to allow green phase. The NN model has achieved 0.9274 accuracy in the training phase. Thus, the traffic lights at the intersections will have changed isolated and dynamically according to the conditions of real-time traffic when using this traffic light control system than the existing fixed time traffic light control system or traditional computation algorithms. This system reduces the average waiting time and increases the efficiency of traffic clearance. New adaptive traffic management also reduces the pollution due CO2 emission and also social and economic problems.