DefinePK

Abstract

Traffic which is the main backbone of a city plays an important role in social stability and development. The traffic signals in our country are standalone systems where the duration of the signal is determined by a traffic officer or a constant value of signal duration is pre-programmed. Sometimes this timing is too short in high traffic density and too long for lesser number of vehicles. These lead to economic losses and unnecessary delays. Therefore we go for Adaptive Traffic Control (ATC) systems which change the signal duration continuously based on the changing arrival pattern of the vehicles. In this paper we design an ATC using heuristic approach where the system is trained to many images with various possibilities of vehicles using Probabilistic Neural Network (PNN). Before training each image is processed using background subtraction with morphological filtering and connected component analysis to detect the number of vehicles. The real time input image from the camera is forwarded to the system which uses heuristic technique to match the current image with one of the trained images and displays the duration of the traffic signal based on the counted number of vehicles thereby regulating traffic easily without any congestion.