Impact crateer detection on optical images and DEMs
Kim, Jan‐Peter Müller · UCL Discovery (University College London) · 2003
Introduction: Impact craters are crucial landmarks for geodetic control as well as being interesting geological research targets themselves . Several methods have been developed to automatically detect impact craters but none are as yet practical or have been sufficiently tested[1][2][3]. We have developed an automatic impact crater detection algorithm for optical images [4]. It is based on a focusing strategy using texture analysis and ellipse fitting [5] on local edges. Currently, the technique works well for craters of medium size but fails for very large impact craters. There is also some confusion between local edges derived from the images and the crater rim due to of solar illumination condition. Therefore, the detection process for real applications has been developed using fusion techniques exploiting both DEMs and optical images. This new algorithm introduces crossover checking on multi-images and a verification stage using templates so that a correct crater numbering is possible according to their size. Examples are shown of its assessment. Overall algorithm: Inputs currently consist of MOC ISIS level 2 images and MOLA based DEMs. Individual detection is performed using each input separately and merged by a registration process so that impact crater counting and 3D reconstruction is possible. The most important factor in optical image based processing is the illumination. Sometimes, the boundary edge line of an impact crater is very ambiguous with low sun elevation angle. There is no good solution for such intrinsic problems. Consequently, a crossover checking process in a set of multiple images is necessary. The overall structure of the algorithm is shown in Fig 1).