The analysis of three-dimensional point cloud data is becoming one of the most used approaches to assess instabilities processes affecting rock slopes. With the increased collection of point cloud data, there is an increasing demand for rapid computational point cloud segmentation techniques to format data for rock fall risk analysis. However, there is a scarcity of semantic segmentation research focusing on geological data and the practical issues they pose. In this study, we propose a novel application to automatically define the spatial extent and intensity of weathered rock slope areas. The studied rock slopes are in Hegra Archaeological Site (al-Hijr / Madā ͐ in Ṣāliḥ), the first UNESCO World Heritage area inscribed in the Kingdom of Saudi Arabia (KSA) in 2008. Here, cavernous weathering processes, with honeycombs and tafoni features, affect the carved rock slope and increase the rockfall hazard. With the spread and improvement of close-range methods, such as Terrestrial Laser Scanner (TLS) and Unmanned Aerial Vehicle Digital Photogrammetry (UAV-DP), it is possible to obtain fast and accurate high-resolution point cloud data of the studied scenario, i.e., rock mass. This paper presents the automated mapping results obtained by a combination of explainable artificial intelligence (XAI) algorithms and three-dimensional geometric analysis of a point cloud using the geometric feature tools available for free on CloudCompare (CC) open-source software. Three-dimensional data were obtained using the UAV-DP technique. The proposed workflow allows, using random forest modelling (RF), to quantify the cavernous weathering extent and estimate the relative intensity on the rock slope, faster than using traditional time-consuming manual segmentation processes. The RF model was able to accurately forecast the cavernous weathering regions (accuracy of up to 85%), with a slight propensity to underpredict, as indicated by the somewhat low recall (∼ 63%), and a limited number of false positives, as indicated by the high precision (up to 84%). This approach is easily reproducible at low cost in terms of computational, human, and financial resources. The results highlight the reliability of the proposed method for the preliminary assessment of cavernous erosional features with the aim of future assessment, monitoring, and deeper analysis of rock-carved slopes.
Classification of rock slope cavernous weathering on UAV photogrammetric point clouds: the example of Hegra (UNESCO world heritage site, Kingdom of Saudi Arabia) / Beni T.; Nava L.; Gigli G.; Frodella W.; Catani F.; Casagli N.; Gallego J.I.; Margottini C.; Spizzichino D.. - In: ENGINEERING GEOLOGY. - ISSN 0013-7952. - ELETTRONICO. - 325:(2023), pp. 107286.1-107286.27. [10.1016/j.enggeo.2023.107286]
Classification of rock slope cavernous weathering on UAV photogrammetric point clouds: the example of Hegra (UNESCO world heritage site, Kingdom of Saudi Arabia)
Beni T.
;Gigli G.;Frodella W.;Casagli N.;Margottini C.;
2023
Abstract
The analysis of three-dimensional point cloud data is becoming one of the most used approaches to assess instabilities processes affecting rock slopes. With the increased collection of point cloud data, there is an increasing demand for rapid computational point cloud segmentation techniques to format data for rock fall risk analysis. However, there is a scarcity of semantic segmentation research focusing on geological data and the practical issues they pose. In this study, we propose a novel application to automatically define the spatial extent and intensity of weathered rock slope areas. The studied rock slopes are in Hegra Archaeological Site (al-Hijr / Madā ͐ in Ṣāliḥ), the first UNESCO World Heritage area inscribed in the Kingdom of Saudi Arabia (KSA) in 2008. Here, cavernous weathering processes, with honeycombs and tafoni features, affect the carved rock slope and increase the rockfall hazard. With the spread and improvement of close-range methods, such as Terrestrial Laser Scanner (TLS) and Unmanned Aerial Vehicle Digital Photogrammetry (UAV-DP), it is possible to obtain fast and accurate high-resolution point cloud data of the studied scenario, i.e., rock mass. This paper presents the automated mapping results obtained by a combination of explainable artificial intelligence (XAI) algorithms and three-dimensional geometric analysis of a point cloud using the geometric feature tools available for free on CloudCompare (CC) open-source software. Three-dimensional data were obtained using the UAV-DP technique. The proposed workflow allows, using random forest modelling (RF), to quantify the cavernous weathering extent and estimate the relative intensity on the rock slope, faster than using traditional time-consuming manual segmentation processes. The RF model was able to accurately forecast the cavernous weathering regions (accuracy of up to 85%), with a slight propensity to underpredict, as indicated by the somewhat low recall (∼ 63%), and a limited number of false positives, as indicated by the high precision (up to 84%). This approach is easily reproducible at low cost in terms of computational, human, and financial resources. The results highlight the reliability of the proposed method for the preliminary assessment of cavernous erosional features with the aim of future assessment, monitoring, and deeper analysis of rock-carved slopes.File | Dimensione | Formato | |
---|---|---|---|
Beni et al ENGINEERING GEOLOGY 2023.pdf
accesso aperto
Tipologia:
Pdf editoriale (Version of record)
Licenza:
Open Access
Dimensione
49.33 MB
Formato
Adobe PDF
|
49.33 MB | Adobe PDF |
I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.