An intelligent framework for end‐to‐end rockfall detection
Rockfall detection is a crucial procedure in the field ofgeology, which helps to reduce the associated risks.Currently, geologists identify rockfall events almostmanually utilizing point cloud and imagery data ob-tained from different caption devices such as TerrestrialLaser Scanner (TLS) or digital...
| Autores: | , , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión aceptada para publicación |
| Fecha de publicación: | 2021 |
| País: | España |
| Institución: | Universidad de Barcelona |
| Repositorio: | Dipòsit Digital de la UB |
| OAI Identifier: | oai:diposit.ub.edu:2445/178934 |
| Acceso en línea: | https://hdl.handle.net/2445/178934 |
| Access Level: | acceso abierto |
| Palabra clave: | Esllavissades Fotogrametria Landslides Photogrammetry |
| Sumario: | Rockfall detection is a crucial procedure in the field ofgeology, which helps to reduce the associated risks.Currently, geologists identify rockfall events almostmanually utilizing point cloud and imagery data ob-tained from different caption devices such as TerrestrialLaser Scanner (TLS) or digital cameras. Multitemporalcomparison of the point clouds obtained with thesetechniques requires a tedious visual inspection to iden-tify rockfall events which implies inaccuracies that de-pend on several factors such as human expertize and thesensibility of the sensors. This paper addresses this issueand provides an intelligent framework for rockfall eventdetection for any individual working in the intersectionof the geology domain and decision support systems.The development of such an analysis framework pre-sents major research challenges and justifies exhaustiveexperimental analysis. In particular, we propose an in-telligent system that utilizes multiple machine learningalgorithms to detect rockfall clusters of point cloud data.Due to the extremely imbalanced nature of the problem,aplethoraofstateoftheart resampling techniques ac-companied by multiple models and feature selectionprocedures are being investigated. Various machine learning pipeline combinations have been examinedand benchmarked applying wellknown metrics to beincorporated into our system. Specifically, we developedmachine learning techniques and applied them to ana-lyze point cloud data extracted from TLS in two distinctcase studies, involving different geological contexts: thebasaltic cliff of Castellfollit de la Roca and the con-glomerate Montserrat Massif, both located in Spain. Ourexperimental results indicate that some of the abovementioned machine learning pipelines can be utilized todetect rockfall incidents on mountain walls, with ex-perimentally validated accuracy. |
|---|