Automatic Delineation of Impact Craters on HRSC Images of the Surface of Mars

Abstract

Impact craters are the most studied structures in planetary sciences, due to the large amount of information that they can provide about the history, geology and physical processes of a planet. The need for investigations about impact craters is increasing, as many missions are being launched into space, acquiring increasingly rich and detailed data. Therefore, to develop an algorithm that automatically detects craters on planetary surfaces through images has been a focus of research worldwide. In this regard, we propose a methodology to detect and delineate automatically impact craters on the Martian surface. It is based on mathematical morphology operators and consists of three main steps: (i) filtering based on connected operators to noise removal and to highlight crater rims, (ii) segmentation based on the watershed transform and dynamics of contours to delineate the real contours of impact craters, and (iii) post-processing to exclude non-crater structures. The image dataset used for testing is constituted by craters depicted from high spatial resolution imagery from the surface of Mars. Specifically, images of High Resolution Stereo Camera (HRSC) with a spatial resolution of 12,5 m/pixel. The global values of the true detection rate was 83,51% and false detection rate was 13,11%, considering 1000 craters delineated with radius within 500 to 1000 meters. Thus, we concluded that the application of mathematical morphology and image processing techniques in a well-designed sequence can contribute to the solution of a current problem in the context of crater detection on planetary surfaces, especially in high spatial resolution images.