Abstract: Traffic flow models are important means to describe various complex traffic procedures and to establish intelligence traffic systems of urban freeway. Generally the formation of a traffic flow model requires massive observation as well as in-depth analysis of real road traffic. Empirical data are indispensable to identify the parameters of the corresponding model. It is drawn in this paper that 210920 "car velocity-headway distance" data pairs are obtained from the long time video recordings of urban freeway in four Chinese cities. Three traffic flow data-samples are made up of parts of these measuring data and are applied to an optimization study of several manifold well-known traffic flow velocity-density models. Some qualitative and quantitative results of the optimal free velocity value in the models have been obtained. The superiorities of modified Kerner-Konhauser equilibrium function and Payne's one are found in simulating the measuring data under the traffic condition of be free and be crowd, respectively, while Greenshields model (i.e. 1D pipe fluid model when m=3) can give the satisfying results for the measuring data when the traffic condition is between free and crowd.