Finding features of users' trajectories in a period of time is one of the key point to realize user's personalized recommendation service.In this paper, how to find the interests in a period from the large amount of user's trajectories is presented with a filter-refinement strategy.In the filter step, the user's trajectories in the same period for several certain days are clustered based on density to obtain the user's stops;in the refinement step, the stops are clustered to obtain the user's interests.Finally, experiments show the effectiveness of this work.