The model of urban cabs moving is one of the key issues for model building. The model needs reflecting moving state information of cabs. More importantly, users can quickly query the moving cabs. Under the condition of real time information of cabs which is hard to obtain, we should model the moving points and forecast the state information according to the moving history. A method which applies Hidden Markov theory to model of the moving trajectory is proposed. Through an analysis of real trajectory data of San Francisco, the caps moving model which is used to query the caps by users was constructed. Experiment with real datasets shows that the method proposed can simulate the moving state of caps. Users can also quickly obtain the useful location information from the model. 