With the explosive growth and wide spread of traffic cameras deployed on the road networks, the amount of Automatic Number-Plate Recognition (ANPR) data captured daily by traffic cameras is very substantial. In this paper, we apply data-mining techniques to discovering vehicle activity patterns from ANPR data. We propose some quantitative indicators of vehicle trace features, and vehicle activity classification method based on the feature of vehicle trace. Evaluations base on collected ANPR records show the capability and efficiency of the proposed approach.
