Understanding collective mobility patterns is of great importance for various applications such as emergency detection, location based services and urban planning. As for existing researches, bank notes and mobile phone data are used for this goal. Here, we attempt to use traffic data to discover collective mobility patterns. A large number of sensors are distributed on the whole city roads to capture the traffic data in a real-time manner. We propose a new feature to characterize the state of the whole sensor network, and represent the evolution of traffic data with a trajectory in the feature space. Then, chain similarity and hierarchical clustering are applied to explore the trajectory patterns, which can reflect human mobility in urban environment. The traffic data from the Twin Cities is used in the experiment. The experimental results show that there exist two regular mobility patterns for the local residents. This can improve our understanding of human outgoing patterns.