Seminar

The study of sand dunes is important to understand the history of the environment and atmospheric trends on surface of mars. This study focuses on the detection of the Martian sand dunes. Previously, the detection of sand dunes on surface of mars is done by a manual method with the help of geologists, which is laborious and time-consuming. Our study serves the demand for automated detection of the surface features by using the application of supervised maximum likelihood classification, unsupervised classification and CNN for the detection of sand dunes on the surface of Mars. The results obtained shows that the supervised classification depicts good accuracy with high Kappa coefficient while the unsupervised classification results are noisy and cannot be termed as reliable for classification. The CNN results are better than both of these techniques, showing higher accuracy with high precision, recall rate, and f1 score. Overall, it can be concluded that one can rely on the CNN for Mars sand dunes detection. Moreover, this study can be extensively utilized for all of the Mars for different surfaces feature detection.

Additional information: Mr. Syed Hassan Iqbal Ahmad Shah, hassan12@connect.hku.hk