Automated Crater Delineation

Introduction: Crater detection algorithms (CDA) have greatly evolved in the last decade and are detecting much smaller structures with higher performances [1-3] which is permitting their use in the construction and upgrade of crater catalogues, namely for Mars [4] and Phobos [5]. Although improvements are still required in the CDA, namely for detecting metric craters in diameter with the same high performances, the algorithms already available are robust enough for large scale detections. Thus, the focus of attention can now be moved towards crater characterization, namely to establish a degree of preservation or erosion for each individual crater. The first step we are considering for establishing this issue consists in the delineation of the real contour of the crater, since CDA outputs describe each crater by perfect circular shapes, no matter its degree of erosion. Therefore, our current objective is to propose a robust algorithm able to deal with the automated delineation of impact craters of any size and degree of preservation, on a wide variety of terrains, and adequate for large scale delineations. Previous algorithms: The delineation of the crater rim has been manually performed [6] and there are almost no algorithms in the literature addressing this problem. The exceptions are two exploratory approaches of ours: one based on a judicious sequence to find and link the crater edges in polar coordinates ('Po-lar') [7], the other based on the watershed transform and other mathematical morphology operators ('Mor-phologic') [8]. The results obtained on a small dataset from Mars achieved good rates [9] but when the da-taset was enlarged, encompassing a wider diversity of terrain ages and textures, the performances were not preserved. The evident degradation of the performance in the most difficult examples showed us that there was still room for improvement. Algorithm based on Edge Maps and Dynamic Programming: The algorithm devised processes the images in polar coordinates and is constituted by two main steps: 1. Edge enhancement, constructing an Edge Map based upon the intensity transitions along radial lines intersecting the center of the crater, and 2. Crater delineation, determining an optimal path from the minimization of an energy functional by Dynamic Programming. The formalism of the algorithm is described in detail in [10].