A novel technique to dynamically measure vehicle speed using uncalibrated roadway cameras

The Washington State Department of Transportation (WSDOT) has a network of several hundred CCTV traffic surveillance cameras deployed on the freeways and arterials around Seattle for congestion monitoring. However, these cameras are not calibrated and can be panned, tilted and zoomed. This paper presents a novel method of automatically computing enough calibration information for these cameras so that they can produce reliable speed estimates. The method presented uses a straightening technique to remove perspective effects and a correlation technique to establish the necessary scale factor. Temporal correlation between sequential straightened frames of video is used to make robust speed estimates.