Automated detection of block falls in the north polar region of Mars

Abstract

Highlights: A change detection method for block falls' identification on Mars is proposed. The images and a training data set are prepared. A combination of SVM/HOG and blob detection is used. The results show a true positive rate of ~75% and a false detection rate of ~8.5%. AbstractWe developed a change detection method for the identification of ice block falls using NASA's HiRISE images of the north polar scarps on Mars. Our method is based on a Support Vector Machine (SVM), trained using Histograms of Oriented Gradients (HOG), and on blob detection. The SVM detects potential new blocks between a set of images; the blob detection, then, confirms the identification of a block inside the area indicated by the SVM and derives the shape of the block. The results from the automatic analysis were compared with block statistics from visual inspection. We tested our method in 6 areas consisting of 1000×1000 pixels, where several hundreds of blocks were identified. The results for the given test areas produced a true positive rate of ~75% for blocks with sizes larger than 0.7 m (i.e., approx. 3 times the available ground pixel size) and a false discovery rate of ~8.5%. Using blob detection we also recover the size of each block within 3 pixels of their actual size.

Deep Learning-Driven Detection and Mapping of Rockfalls on Mars

“…For the lunar environment, a Convolutional Neural Network (CNN) was able to achieve near-human detection performance while reducing the required processing time 2 (human vs. machine) by more than one order of magnitude. On Mars, [27] developed a method to automatically detect ice falls in HiRISE images of the northern polar caps, using support vector machines. Due to its atmosphere and dynamic geologic environment, Mars generally poses additional challenges for the detection and mapping of rockfalls from orbit.…”

Automated NPLD scarp mapping for the block falls monitoring pipeline

“…Our first step towards fully monitoring the activity of the north polar region resulted in a detailed and automatically produced map of the boundary between BU and the NPLD based on the significant difference in intensity between them. Fanara et al (2020a) already used traditional machine learning techniques to detect newly appeared ice blocks, achieving a good performance and estimating the current erosion rate of a steep scarp [8] . We are now planning to complement this study with our scarp mapping algorithm.…”

An automatic approach for ice-block falls detection in the Martian north polar region

“…Fanara et al developed an automatic target detection algorithm based on the histogram of oriented gradients (HOG) features and the support vector machines (SVM) to detect changes associated with ice-block falls. The proposed algorithm achieved a local precision rate of 91.5% and a recall rate of 75.1% for blocks larger than 0.5m 2 by using HiRISE images [5]. Then Fanara et al developed an ice-block falls change detection algorithm to estimate their volume from the HiRISE images.…”