Abstract
The analysis of rockfall distribution and magnitude is a useful tool to study the past and current endogenic and exogenic activity of Mars. At the same time, tracks left by rockfalls provide insights into the mechanical properties of the Martian surface. While a wealth of high-resolution spaceborne image data are available, manual mapping of displaced boulders with tracks is inefficient and slow, resulting in: 1) a small total number of mapped features; 2) inadequate statistics; and 3) a suboptimal utilization of the available big data. This study implements a deep learning-driven approach to automatically detect and map Martian boulders with tracks in high resolution imaging science experiment (HiRISE) imagery. Six off-the-shelf neural networks have been trained either on Martian or lunar rockfall data, or a combination of both, and are able to achieve a maximum overall recall of up to 0.78 and a maximum overall precision of up to 1.0, with a mean average precision of 0.71. The fusion of training data from different planets and sensors results in an increased detection precision, highlighting the value of domain generalization and multidomain learning. Average processing time per HiRISE image is ∼45 s using an NVIDIA Titan Xp, which is more than one order of magnitude faster than a human operator. The developed deep learning-driven infrastructure can be deployed to map Martian rockfalls on a global scale and within a realistic timeframe.
Highlights
M ARS is a dynamic planet and recent surface activity has been shown by a large number of studies using data returned from ground-borne and spaceborne missions, such as the Mars Science Laboratory (MSL or Curiosity) or the Mars Reconnaissance Orbiter (MRO)
We investigated whether the result of such a multidomain learning approach is beneficial for the detection performance of a convolutional neural network (CNN) or rather reduces it
The potentially improved generalization does not appear to improve the recall of the respective CNNs
Summary
M ARS is a dynamic planet and recent surface activity has been shown by a large number of studies using data returned from ground-borne and spaceborne missions, such as the Mars Science Laboratory (MSL or Curiosity) or the Mars Reconnaissance Orbiter (MRO). Dynamic aeolian features include dunes (e.g., [1], [2]) and dust devils (e.g., [3], [4]) that migrate across the Martian surface. Observed mass wasting processes include polar ice and dust avalanches (e.g., [5], [6]), rockfalls (e.g., [7]–[9]), slope streaks (e.g., [10]–[12]) and potentially recurring slope lineae (RSL) (e.g., [13]–[15]). The occurrence, frequency, and magnitude of mass wasting phenomena, and rockfalls in particular can be indicative of seismic activity of planetary bodies or moons in general, as several recent studies have shown ([7], [8] [19]–[21]). The tracks created by extraterrestrial rockfalls, i.e., boulder tracks, are a valuable tool to estimate the basic mechanical properties of the surface substrate present [22], [23]
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call