複数アプローチによる月面クレータ自動抽出アルゴリズム

抄録

There are large volumes of lunar images that are archived and to be archived by missions such as Clementine and SELENE (SELenological and Engineering Explorer). We have developed an algorithm to automatically detect and classify craters in massive lunar images. Craters are of scientific interest because their density can yield the relative ages of the surface units. However, the manual extraction of craters remains difficult because it requires a great deal of man power. Several automated crater detection algorithms have been developed so far but none are yet practical nor have been sufficiently tested.
Our algorithm locates craters using four different approaches. These are extractions of : (1) the shade and sunny patterns of craters in case of low sun elevation, (2) the circular features from the edge image, (3) the circular lines from thinned and connected line segments, and (4) discrete or separated circular edge lines which consist circles using fuzzy Hough transform. We applied the proposed algorithm to Clementine, Lunar Orbiter and Apollo images acquired over both mare and highland under different solar elevation. As a result it is shown that the algorithm can detect craters from various images at 80% detection rate without parameter tuning.