Detecting Morphing Attacks via Continual Incremental Training