Exploiting distinctive visual landmark maps in pan-tilt-zoom camera networks