Vision-based localization of leader-follower formations