Computer vision model for monitoring block falls in the Martian north polar region

The north polar region of Mars is one of the most active places of the planet with avalanches and ice block falls being observed every year on High Resolution Imaging Science Experiment (HiRISE) data. Both phenomena originate at the steep icy scarps, which exist on the interface between two adjacent geological units, the older and darker Planum Boreum Cavi unit, also called Basal Unit (BU) and the younger and brighter Planum Boreum 1 unit, which is a part of the so called North Polar Layered Deposits (NPLD). These exposed layers of ice and dust contain important information about the climate cycles of the planet. We are primarily interested in monitoring the current scarp erosion rate (quantified through analyzing ice debris) at the same time differentiating between the activity originating in the NPLD [1] from that originating in the BU The large scale of the region of interest, combined with a growing amount of available satellite data makes automation key for this project. To achieve the latter we propose a computational pipeline consisting of three consecutive steps, namely: scarp segmentation, single image super-resolution and ice-block detection. For the final analysis Mean Average Precision (mAP.95) was used as a benchmark metric. The performance value of 93.6% was obtained on a test dataset, leading us to conclude that the network is able to perform even on small ice fragments (which comprise the majority of the debris). On a system running 4 RTX3090 GPUs the finished pipeline processes a single HiRISE product in just under 20 minutes, returning the scarp outline and precise ice boulder coordinates. Using this pipeline, we next plan to robustly monitor the mass wasting activity in the whole north polar region and throughout the entire Mars Reconnaissance Orbiter (MRO) mission. [1] Su, S. et al., 2024. EGU 2024.