This paper describes an autonomous outdoor navigation method based on view-based localization. We develop a new view-based localization method by extending Seq-SLAM which exhibits a robust place recognition performance across seasons. In mobile robot navigation, traversable region detection, as well as localization, is important. We therefore combines a view-based localization with a traversable region detection using 3D range sensors such that subgoals towards the destination are continuously chosen inside the local traversable region. We developed an autonomous navigation system by implementing the proposed method on a mobile robot, and demonstrated the validity of the method by carrying out autonomous navigation experiments in own campus.