Potential Underground Water Source Mapping with Convolutional Neural Network on Mars
The studies have shown, that due to similar pasts, Mars has similar rocks formed by the same forces as Earth . Therefore, in this research, we aimed to find and locate sedimentary rocks, which indicate potential underground water sources, on Mars with help of the convolutional neural network that trained with Earth-based types of rocks dataset.
The rocks of the earth have been formed by three different forces: pressure (Metamorphic), magmatic activity (Igneous), and natural forces (Sedimentary) . Therefore, it is possible to predict the history of the environment where rocks have been found. Specifically, Sedimentary rocks are reliable evidence of water force and the potential underground water sources in the environment where they were found . Although the knowledge about the rock types and their formation processes were earth-based, studies  have shown that terrestrial planets such as Mars, whose past is similar to the earth, were also under the same influences and shaped their rocks similarly to the earth. Therefore, with the developments in the field of machine learning, it has become possible to classify the known earth-based rocks  by comparing them autonomously with the rocks on the Martian surface .
We collected pictures of the sub-rock types of Metamorphic, Igneous, and Sedimentary rocks found in visual search engines as a data set. We then trained the dataset for object detection with the convolutional neural network Then, with the help of the convolutional neural network, which we trained with earth-based rock images, we found the matches on the Mars surface photographs we collected from MSL Analyst’s Notebook. Finally, by marking the matching rocks on the Mars map, we determined the areas where the rock formations are concentrated.
 Matthew J. J., Weathering: Types, Processes and Effects : Types, Processes and Effects, edited by Colon, Nova Science Publishers, Incorporated, 2011.
 M. Giller, “Igneous Rocks / Metamorphic Rocks / Sedimentary Rocks,” School Library Journal, vol. 49, (7), pp. 112, 2003.
 D. J. DesMarais, “Exploring Mars for Evidence of Habitable Environments and Life,” NASA Center for AeroSpace Information (CASI). Conference Proceedings, 2014
 X. Ran et al, “Rock Classification from Field Image Patches Analyzed Using a Deep Convolutional Neural Network,” Mathematics, vol. 7, (8), pp. 755, 2019.
 J. Li et al, “Autonomous Martian rock image classification based on transfer deep learning methods,” Earth Science Informatics, vol. 13, (3), pp. 951–963, 2020.