Predicting HiRISE-equivalent Rock Density on Mars Using CTX Image Features

In this paper, a Bayesian framework is used to infer HiRISE-equivalent rock density from CTX image features without needing to explicitly identify rocks. The statistical relationship between HiRISE rock density and CTX image features is modeled using Bayesian Networks. The model is enriched by including geological features—specifically, geomorphic units. The geomorphic units are identified by geologists based on different surface features visible in a variety of orbital data. The inclusion of geomorphology in the Bayesian framework makes it possible to not only infer HiRISE-equivalent rock density directly from CTX image features, but also to make a preliminary assessment of the representative geomorphic units. HiRISE and CTX images of the Mars Phoenix mission landing area are used in this study. Rock density estimates from HiRISE images are used as ground truth to train the Bayesian Network using features extracted from corresponding CTX images. The approach is evaluated on a few test cases using a probabilistic estimate of landing safety based on inferred rock density from CTX image features. The initial results of this exploratory study show that predicted landing safety from the inferred rock density is very close to that obtained using actual HiRISE rock density estimates.