Abstract
In order to build the Electronic Toll Collection System (ETC), a system having a vehicle detection rate of more than 99.9% is needed to detect and to count all the passing vehicles running freely on ordinary roads at a maximum speed of 120km/h. The authors studied ways to monitor from above the road surface by linear image sensors, with cyclic marks of black/white drawn on the road, normalized corelation and differential-integrating method were employed as a means to compare video signals with reference image signals from such sensors' output signals, and generate two-dimensional image data to analyze a spatio-temporal image to detect a vehicle.