Deep Learning-Driven Detection and Mapping of Rockfalls on Mars

Bickel, Valentin Tertius; Conway, Susan J.; Tesson, Pierre-Antoine; Manconi, Andrea; Loew, Simon

doi:10.1109/jstars.2020.2991588

Cited by 33 publications

(23 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We found that the CNNs performance on new areas was biased towards higher precision scores. This is similar to observations from other deep learning work done in the area of domain generalization and multidomain learning 37 . To rule out any negative contributions from any of the four training inventories used by the combined learning models, we adopted a jackknife approach by conducting more experiments to train CNNs from only three study areas, while systematically leaving out one each time.…”

Section: Discussionsupporting

confidence: 87%

A new strategy to map landslides with a generalized convolutional neural network

Prakash

Manconi

Loew

2021

Sci Rep

Self Cite

View full text Add to dashboard Cite

Rapid mapping of event landslides is crucial to identify the areas affected by damages as well as for effective disaster response. Traditionally, such maps are generated with visual interpretation of remote sensing imagery (manned/unmanned airborne systems or spaceborne sensors) and/or using pixel-based and object-based methods exploiting data-intensive machine learning algorithms. Recent works have explored the use of convolutional neural networks (CNN), a deep learning algorithm, for mapping landslides from remote sensing data. These methods follow a standard supervised learning workflow that involves training a model using a landslide inventory covering a relatively small area. The trained model is then used to predict landslides in the surrounding regions. Here, we propose a new strategy, i.e., a progressive CNN training relying on combined inventories to build a generalized model that can be applied directly to a new, unexplored area. We first prove the effectiveness of CNNs by training and validating on event landslides inventories in four regions after earthquakes and/or extreme meteorological events. Next, we use the trained CNNs to map landslides triggered by new events spread across different geographic regions. We found that CNNs trained on a combination of inventories have a better generalization performance, with a bias towards high precision and low recall scores. In our tests, the combined training model achieved the highest (Matthews correlation coefficient) MCC score of 0.69 when mapping landslides in new unseen regions. The mapping was done on images from different optical sensors, resampled to a spatial resolution of 6 m, 10 m, and 30 m. Despite a slightly reduced performance, the main advantage of combined training is to overcome the requirement of a local inventory for training a new deep learning model. This implementation can facilitate automated pipelines providing fast response for the generation of landslide maps in the post-disaster phase. In this study, the study areas were selected from seismically active zones with a high hydrological hazard distribution and vegetation coverage. Hence, future works should also include regions from less vegetated geographic locations.

show abstract

Section: Discussionsupporting

confidence: 87%

A new strategy to map landslides with a generalized convolutional neural network

Prakash

Manconi

Loew

2021

Sci Rep

Self Cite

View full text Add to dashboard Cite

show abstract

“…We found that the CNNs performance on new areas was biased towards higher precision scores. This is similar to observations from other deep learning work done in the area of domain generalization and multidomain learning 32 . To rule out any negative contributions from any of the four training inventories used by the combined learning models, we adopted a jackknife approach by conducting more experiments to train CNNs from only three study areas, while systematically leaving out one each time.…”

Section: Discussionsupporting

confidence: 87%

Mapping Landslides With a Generalized Convolutional Neural Network

Prakash¹,

Manconi²,

Loew³

2020

Preprint

Self Cite

View full text Add to dashboard Cite

“…In order to further study the characteristics of rockfalls on a local scale we selected and performed measurements in 13 areas of interest (AoIs) across the Moon (Bickel, 2021). These specific locations have been selected because they are rockfall hotspots and belong to one of the four relevant lunar geomorphic (sub) classes as outlined by Bickel, Jordan, et al (2020): crater (impact-caused, -induced, and -ejected), volcanic vent, tectonic structure, or other, unclassified geomorphic context.…”

Section: Local Geologic and Geomorphic Analysismentioning

confidence: 99%

“…A summary of all used global auxiliary data sets and their references is located in Table S1 in Supporting Information S1. The local, detailed rockfall catalog used for this study (Bickel, 2021) can be accessed here: https:// edmond.mpdl.mpg.de/imeji/collection/I86ZoEoyZLTRkeSJ.…”

Section: Conflict Of Interestmentioning

confidence: 99%

“…Recently, Bickel et al (2018) and Bickel, Conway, et al (2020) developed a method to automate rockfall mapping in satellite imagery using a convolutional neural network (CNN), and used that tool to map the global distribution of lunar rockfalls in more than 250,000 NAC images. For the first time, a global study of lunar rockfalls and their drivers was possible (Bickel, Jordan, et al, 2020).…”

mentioning

confidence: 99%

See 1 more Smart Citation

Global Drivers and Transport Mechanisms of Lunar Rockfalls

et al. 2021

Self Cite

View full text Add to dashboard Cite

Lunar rockfalls are ubiquitous mass wasting features that were first observed in Lunar Orbiter high-resolution photographs in the late 1960s (Eggleston et al., 1968;Filice, 1967;Moore, 1970). Rockfall events, which have been previously referred to as block falls or rolling boulders, involve the detachment of a boulder or rock mass from an elevated source region, which then slides, bounces, and rolls down the local topographic gradient, ultimately depositing downslope. Despite the fact that these features were recognized in the late 1960's, all early studies have only examined rockfalls in relatively small geographic areas on the lunar surface, because of a lack of available high resolution imagery (e.g., Eggleston et al., 1968;Hovland & Mitchell, 1973). Due to this, the characteristics of source regions, as well as the mechanisms that govern the failure and runout of lunar rockfall remain largely unknown.

show abstract

Deep Learning-Driven Detection and Mapping of Rockfalls on Mars

Cited by 33 publications

References 37 publications

A new strategy to map landslides with a generalized convolutional neural network

A new strategy to map landslides with a generalized convolutional neural network

Mapping Landslides With a Generalized Convolutional Neural Network

Global Drivers and Transport Mechanisms of Lunar Rockfalls

Contact Info

Product

Resources

About