University of Cambridge > > Darwin College Science Seminars > Vision-Based Over-Height Vehicle Detection

Vision-Based Over-Height Vehicle Detection

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Evros Loukaides.

Over-height vehicle drivers continuously ignore warning signs and strike onto bridges despite the number of preventative methods installed at low clearance bridges. Asset owners are left to remedy the damages resulting in operation losses due to inspections, temporary road closures, and traffic delays. IT systems are needed that address all three key areas of over-height vehicle strike management: (1) prevention, (2) detection and, (3) reporting. In this paper we present a new method for over-height vehicle strike prevention that replaces the transmitter, a receiver, and loop detectors with a single calibrated camera mounted on the side of the roadway. The camera is installed at the height of the “over-height plane” formed by the average of the maximum allowable heights across all lanes in a given traffic direction. At that height, the over-height plane can be safely approximated as a line in the camera view. Any vehicle exceeding this line is consequently over-height. The camera position and orientation is determined via a calibration process proposed. Instances of over-height vehicles are detected via optical flow monitoring. A prototype was implemented to evaluate the method’s performance. A total station was used to collect ground truth data for validating the height accuracy. Evaluation of the system yielded a height accuracy of ±2.875 mm and detection accuracy of 96.9 %. While its accuracy is comparable to existing laser beam systems, it outperforms them on cost which is an order of magnitude less due to eliminating the need for new permanent infrastructure.

This talk is part of the Darwin College Science Seminars series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


© 2006-2023, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity