Automatic Crater Recognition on Digital Terrain Model

Introduction: ESA, SMART-1 has orbited around the Moon [1], Space Agencies of Japan, China, and India have expressed subsequent moon missions, and USA announced a return to the moon in New Vision for Space Exploration Program [2]. We have a Moon Rush. Japanese SELENE mission [3] has 14 instruments, and is going to provide a dataset more than 20TB. These massive and diverse data raise awareness about the need for automatically extracting/dictating geologic information. Here we remark " crater " on digital terrain models, this is just preparation for analysis of SELENE DTM products. Automatic crater recognition has been studied [4][5][6], and they are almost categorized into a type of template matching. Although their results have shown good results, this type requires diverse templates and a heavy load for processing. Moreover this matching tends to have trouble in overlay craters and a round pattern with small craters. We adopt a generalized Hough transform after extracting edges from DTM, and establish algorithm for detecting craters. This goal is a software development to detect craters automatically with following parameters; diameter, coordinates, depth, and direction of incidence. Sample Data: We adopt Mars' topography (MOLA data) as sample data for verifying this developed software. Its spatial resolution, bit resolution, and dynamic range are 1.864 km/pixel, 1m/DN, 16bit, respectively. Procedures: This main processing notifies two; tangent's properties and the generalized Hough transform [7]. This algorithm can detect ellipsoids robustly even if craters show lack of larger parts of edge/rim. The first step is to detect candidates of crater rims as edges on DTM. This step disregards deficiency of ellipsoids, and brings binarized pictures with outlines of the candidates. Next each ellipsoid is characterized. The generalized Hough transform can complement the outlines' deficiency up to 50%. This transform method provides following properties of each ellipsoidal outline; coordinates of the center, length of major axis and minor one, and rotation angle of the major axis from east-west direction. Next, the ellipsoids are divided into isolated ones and overlay ones. The isolated ellipsoids are determined as craters with no hesitation. Another