Mountain summit detection with Deep Learning: evaluation and comparison with heuristic methods

Comparison of digital elevation models through the analysis of geomorphic surface remnants in the Desatoya Mountains, Nevada

“…Histograms and dot plot diagrams were also generated using RStudio v1.2.5033. For another quantitative assessment of the feature extraction performance, the confusion matrix (Sărășan et al., 2018; Stehman, 1997; Torres et al., 2020) approach was used, where all 244 digitized peaks served as reference.…”

“…As a consequence of continuous erosional processes, previously significant and extensive surfaces degraded into surface peaks, therefore, those eroded peaks can indicate the altitudinal position of paleosurface remnants (Bugya, 2009; Ebert, Hättestrand, Hall, & Alm, 2011; Rowberry, 2012; Soares & Riffel, 2006) (Figure 1). Several methods have been developed to detect surface peaks (Deng & Wilson, 2008; Kirmse & Ferranti, 2017; Podobnikar, 2012; Torres, Fraternali, Milani, & Frajberg, 2020). However, the Geomorphons algorithm can provide a fast, adjustable, and open‐source method for peak detection.…”

Automated Earthwork Detection Using Topological Persistence

“…(2019), image processing and pattern techniques have also been applied for terrain analysis, but these algorithms tend to be poorly suited to terrain data, since they are generally not smoothly‐varying (Wang & Li, 2021). Machine learning has been recently introduced for feature detection with great success (e.g., Hsu et al., 2021; Steinfeld et al., 2013; Torres et al., 2020; Wang & Li, 2021; with true positive detection rates of about 60%–98%); data preparation for model training, though, can be time‐consuming and labor‐intensive since training data requires pre‐existing data sets that are generally developed using semi‐automated or fully manual methods.…”

Toward geomorphometry of plains - Country-level unsupervised classification of low-relief areas (Poland)

“…However, the most significant limitation of supervised methods is that they cannot reveal new patterns in the data (Iwahashi et al, 2021), and the measure of success is the degree of compliance with existing, usually manually created maps. For that reason, supervised methods, including deep learning (Torres et al, 2019;Shumack et al, 2020;Xie et al, 2020) focus on the search for most effective classification methods reconstructing the complicated ground-truth (Du et al, 2019;Li et al, 2020;Janowski et al, 2022) rather than searching for alternative designs.…”