Machine Learning for Classification of an Eroding Scarp Surface Using Terrestrial Photogrammetry With Nir and RGB Imagery

Abstract

Abstract. Increasingly advanced and affordable close-range sensing techniques are employed by an ever-broadening range of users, with varying competence and experience. In this context a method was tested that uses photogrammetry and classification by machine learning to divide a point cloud into different surface type classes. The study site is a peat scarp 20 metres long in the actively eroding river bank of the Rotmoos valley near Obergurgl, Austria. Imagery from near-infra red (NIR) and conventional (RGB) sensors, georeferenced with coordinates of targets surveyed with a total station, was used to create a point cloud using structure from motion and dense image matching. NIR and RGB information were merged into a single point cloud and 18 geometric features were extracted using three different radii (0.02 m, 0.05 m and 0.1 m) totalling 58 variables on which to apply the machine learning classification. Segments representing six classes, dry grass, green grass, peat, rock, snow and target, were extracted from the point cloud and split into a training set and a testing set. A Random Forest machine learning model was trained using machine learning packages in the R-CRAN environment. The overall classification accuracy and Kappa Index were 98% and 97% respectively. Rock, snow and target classes had the highest producer and user accuracies. Dry and green grass had the highest omission (1.9% and 5.6% respectively) and commission errors (3.3% and 3.4% respectively). Analysis of feature importance revealed that the spectral descriptors (NIR, R, G, B) were by far the most important determinants followed by verticality at 0.1 m radius.

An intelligent framework for end‐to‐end rockfall detection

“…Specifically, recent studies, including References [4,24-30] use point clouds as their main data source for their analysis. Moreover, advanced research studies use both images and point clouds portraying a speciality and differentiate themselves from the majority of the related studies 31‐33 . Furthermore, Reference [34] use images and TLS point cloud data to model slopes.…”

“…Moreover, advanced research studies use both images and point clouds portraying a speciality and differentiate themselves from the majority of the related studies. [31][32][33] Furthermore, Reference [34] use images and TLS point cloud data to model slopes. Last but not least, Reference [35] utilize data from wireless sensor networks and machine learning algorithms to monitor and forecast landslides in real time.…”

“…Machine learning-based methodological workflows are the most used techniques, as there are various research articles utilizing them. Recent research studies such as References [25][26][27][28][29][30]32,33,35] utilize mainly machine learning algorithms for rock-slope and landslide monitoring and analysis. More elaborate machine learning methods outperform the majority, examples include Reference [40], who developed a landslide monitoring approach for TLS point cloud data, which integrates a specialized machine learning classification involving topological rules in an object-based framework of analysis.…”

End-to-End Intelligent Framework for Rockfall Detection

“…Specifically, recent studies, including (Mayr et al (2018); Weidner et al (2019Weidner et al ( , 2020b; Kong et al (2020); Huu Phuong (2020); Fanos et al (2020); Weidner et al (2020a); Wang et al (2020b)) use point clouds as their main data source for their analysis. Moreover, advanced research studies use both images and point clouds portraying a speciality and differentiate themselves from the majority of the related studies (Fanos & Pradhan (2019); Bernsteiner et al (2020); Loghin et al (2020)). Furthermore, Salvini et al (2013) use images and TLS point cloud data to model slopes.…”

“…Machine learning based methodological workflows are the most used techniques, as there are various research articles utilizing them. Recent research studies such as (Weidner et al (2019); Hemalatha et al (2019); Weidner et al (2020b); Kong et al (2020); Huu Phuong (2020); Bernsteiner et al (2020); Loghin et al (2020); Weidner et al (2020a); Wang et al (2020b)) utilize mainly machine learning algorithms for rock-slope and landslide monitoring and analysis. More elaborate machine learning methods outperform the majority, examples include Mayr et al (2017), who developed a landslide monitoring approach for TLS point cloud data, which integrates a specialized machine learning classification with topological rules in an object-based analysis framework.…”