An asteroid explorer (Hayabusa, Hayabusa2, …) takes many images of the target asteroid, and the images are used for the navigation around the asteroid. Some of representative optical navigation methods detect feature points of the asteroid in the images such as rocks and craters. Hayabusa succeeded to touch down on the asteroid by use of the optical navigation method called GCP-NAV. However, these kinds of navigation methods need to construct the database of the feature points beforehand by the precise measurement of the asteroid ground. In this research, we constitute virtual feature points from the topographical information captured in the images, and propose an efficient, autonomous optical navigation method without large-scale measurement in advance.