Applying Deep Learning in Mars Exploration: A Neural Network-based Study to Classify Martian Terrain Features

The red planet has persisted in being the top destination for space exploration and has also fascinated researchers to experiment with frequently launched its dataset of images. NASA has launched various missions on the track of space expeditions to explore the surface and the environment of the potential life-sustaining planet. Among all the missions for capturing images of the Martian terrain, the HiRISE experiment has been successful in coming up with the best datasets featuring the surface of Mars. In this research study, the authors have targeted the classification of this high-quality image dataset produced as the outcome of the HiRISE using a deep learning model. The proposed model is convoluted and built with the help of neural network layers (neuron layers). Based on the classification, the authors were able to classify the images of 7 terrain features of Mars, namely Crater, Impact Ejecta, Bright Dune, Dark Dune, Slope Streak, Spider, and Swiss Cheese. And the classification accuracy achieved by the model is 94.8%.