Methods of traffic flow measurement using spatio-temporal image

Traffic flow measurement sensors, which measure vehicle velocity and detect traffic accidents, traffic congestion and so on using image processing, have been installed. For the sensors we utilized spatio-temporal images, which are used to process sequential images; as a recognition method for vehicles, we developed Directional-Temporal plane Transform (DTT) which transforms spatio-temporal image to 2-D data onto a directional-temporal plane. The rear-end-collision prevention system using this method decreased accidents to 1/3 as compared to the last few years before installation at a place on a national highway where many rear-end collisions had occurred. However, in the case where cameras are set up at low positions such as in tunnels, it is difficult to measure vehicle velocity in heavy traffic conditions because one vehicle often blocks the view of another in the images. To overcome this drawback, we propose a method for measurement of velocity for a group of vehicles by way of measuring moving time of the group by matching patterns, of DTT data at several points. We describe DTT-including its application-and the method for measurement of velocity for a group of vehicles, and then show the experimental results using road images.