In the face of the frequent occurrence of urban traffic congestion, only part traffic sensors can be utilized in the urban freeway, in order to more accurately estimate traffic densities and timely identify the congested road segments, a novel estimated algorithm of proportional integral state observer is proposed. Firstly, by embedding the Cell Transmission Model (CTM) into the Dynamic Graph Hybrid Automata (DGHA) framework, we model traffic flow over the urban freeway network. In modeling procedure, an urban freeway network is partitioned in term of three hierarchical structures which are freeway networks, links and cells, respectively. On the one hand, both the dual digraph of road segment structure and hybrid automata respectively are adopted to describe the freeway topology and the multi-modes of dynamic densities in road segments. On the other hand, a deterministic and reasonable constraint is used to reduce the number of combination modes, and thus the modeling procedure is modularized, rule-based and easily-extensible with the help of a combination algorithm. Therefore, the Piecewise Affine System (PWAS) model of the freeway network can be obtained. Secondly, In order to estimate the traffic densities of the urban freeway network, based on the dynamic model, we design a switched proportional integral state observer, and the observer design problem is transferred into the stability problem of the switching system. So the proportional gains and integral gains can be computed via introducing the Lyapunov function and using LMI technique, and thus the traffic densities can be estimated by the designed observer. Especially, the traffic congestion can be identified by checking whether the estimated densities are greater than the critical ones. Finally, a section of Beijing south third ring freeway as an example to verify the proposed approach, and the effectiveness and the feasibility are shown in the experiment results.