TITAN: a traffic measurement system using image processing techniques

Abstract

Presents the vehicle detection algorithms which have been developed using the traffic images collected from various scenes covering 100-250 m of a three three-lanes freeway. Pictures can be of front or rear views. The algorithms cope with the possibility of using remote controlled cameras. This means that the system includes features which are able to re-calibrate the algorithms and to detect the lanes if an operator has moved the camera. Different luminance and traffic conditions are considered. Emphases are made upon the new 'lane detection' module and analysis of dense traffic and nocturnal traffic. >