Real time vehicle counter system for Intelligent Transportation System

Abstract

This paper presents real time vehicle counter system for Intelligent Transportation System. A stream of video frames is processed using a sequence of procedures: foreground extraction, object segmentation and labeling, and object classification to differentiate between motor- cycle and car. Foreground extraction utilizing a simple method, background subtraction, and the segmentation utilizing methods of mathematical morphology: erotion, dilation and connected component labeling. Classification process is based on the size of connected component. In the image where no shadow of unwanted objects, the system delivers the success rate of a maximum 97% to recognize motorcycle and a maximum 95% to recognize car.