Automatic Delimitation of Regions with Aeolian Features on Mars

Introduction: The number of currently available images of the surface of Mars is in the order of the hundreds of thousands. These have been captured over the years by many different instruments on board automatic probes. They have diverse spectral characteristics, varying spatial resolutions and cover areas of different dimensions, providing global, regional and sometimes local views. To manually browse all these images in search of a specific structure of interest is a hard and time consuming process. Thus, our objective is to create a tool that can learn the characteristics of a certain structure in an image and be able to automatically search for it over a huge quantity of data. Currently, we are interested on dune fields, being our approach developed, tested and calibrated for searching those aeolian features on Mars. Dunes fields serve as unique indicators of the interaction between the atmosphere and surface, being among the most widespread aeolian features present on Mars. The study of dune processes contributes to both atmospheric and sedimentary science. Previous works that characterized aeolian features in regional terms conducted manual surveys on image datasets in order to find dune fields [1]. Also, the well established Mars Digital Dune Database [2] was constructed and updated without the use of automated procedures.