Research seminar "OD Matrix Estimation of a Highway Network"

Abstract

Today’s road traffic control problems include traffic situations with many congestions, increased traffic demand in peak hours, need for high mobility and fast response in case of an incident. Video sensor or camera combined with state of the art image processing algorithms is more and more becoming the approach to road traffic monitoring. Advantage of obtained traffic video footage processing is that a significant amount of high level traffic information can be extracted. Extraction of traffic data using computer vision technology is nowadays related to creation of traffic statistics, measurement of traffic parameters and to improvement of the traffic conditions for drivers. High level traffic information includes incident detection, vehicle classification, origin-destination (OD) matrix estimation, etc. This research work deals with the possibility to track vehicle trajectories in a highway traffic network using license plate recognition (LPR). The presented results include developed image processing system, vehicle detection and LPR to create an application that enables vehicle detection in video footage, extracting an image part with the vehicle and it license plate, LPR, archiving extracted license plates and vehicle origin country distribution analysis. Commercially available SDK CARMEN is used for LPR. The first developed application testing has been carried out on two different real world highway traffic videos and the data extracted were used to compile traffic statistics on a real world traffic highway and establish the principles of automatic highway network OD matrix estimation which will be developed in the future. Presentation ends with experimental results obtained from real world highway traffic video footage and a summary of the tasks done including future work proposals.

CV

Kristian Kovačić was born on 16th January 1986 in Zagreb, Croatia. He received his bachelor and master degree on Faculty of Transport and Traffic Sciences University of Zagreb. During his education he showed particular interest in programming, image processing and 3D graphics applications. Currently he is employed at Faculty of Transport and Traffic Sciences as a research engineer on the project VISTA. His interests are related to application of computer vision in transport and traffic sciences, particularly vehicle detection and tracking, vehicle origin-destination matrix estimation using license plate recognition and measurement of traffic parameters.