Indexing of traffic flow image sequences for traffic information services

The aim of our project is to forecast the traffic state like a weather forecast and to show the state with dynamic image. For forecasting we plan to construct the dynamic image database management system. The system needs a lot of dynamic images with indices, i.e. time, date, place, congestion parameters and so on. It is a key technique for the DBMS to extract the parameters from observed image automatically. In this paper, we present a method to acquire the parameters from real traffic images by image processing techniques. In order to show that our program works as good as human, we compared the ordering by computer with the ordering by human drivers. The result shows that computer's ordering is almost same as human ordering within the variation between human drivers' ordering.