Text this: A method based on visual saliency for vehicle-mounted monocular camera ego-motion estimation and vehicle scale estimation