Rockfall source location and volume from autonomous UAV-acquired photos: A case study from a railway slope in northern Ontario

Dave Gauthier, D. Jean Hutchinson, Matt Lato, David F. Wood, A.J. Morris

In the proceedings of: GeoRegina 2014: 67th Canadian Geotechnical Conference

Session: Transportation Geotechnics

ABSTRACT: Terrestrial photogrammetry techniques have been applied widely and successfully to modeling open pit slopes, large rock faces, etc., in order to develop precise three-dimensional (3D) terrain models. Unfortunately, the geometry and scale of many steep, hazardous rock slopes is not conducive to either terrestrial or downward-looking aerial cameras. In this study we generated detailed 3D photogrammetric models of a slope along the CP railway adjacent to Lake Superior, from hundreds of digital photographs collected by a side-looking camera-equipped autonomous UAV, in August 2012 and October 2013. We then compared the models to look for changes. We were able to identify the source and volume of a single rockfall event. In the paper we discuss the data collection and development of the three-dimensional models, the change-detection and volume calculation methodology and the major limitations in the analysis and applications of these methods. RÉSUMÉ La géométrie et la taille de nombreux raides, pentes rocheuses dangereuses n'est pas propice à la modélisation 3D à l'aide de photogrammetruc terrestre ou vers le bas à la recherche de caméras aériennes. Dans cette étude, nous avons généré des modèles de photogrammétrie 3D détaillées de la pente le long de la voie ferrée du CP adjacent au lac Supérieur, des centaines de photographies numériques recueillies par un drone autonome côte regardant la caméra équipée, en Août 2012 et Octobre 2013. Nous avons ensuite comparé les modèles à la recherche de changements. Nous avons pu identifier à la fois l'emplacement et le volume d'un événement de chutes de pierres source unique. Dans cet article, nous discutons de la collecte de données et le développement des modèles en trois dimensions, le changement de détection et de la méthodologie de calcul du volume de l'éboulement de 9m3 unique et les limitations majeures dans l'analyse et les applications de ces méthodes. 1 INTRODUCTION Terrestrial remote-sensing methods, such as LiDAR, have been used extensively to develop digital terrain models and characterize geomorphological or geomechanical features (e.g. Olariu et al., 2008; Lato et al., 2009; Sturznegger and Stead, 2009; Brodu and Lague, 2012; Gigli et al., 2014). Multiple LiDAR data may be compared, and changes detected (e.g. Rosser et al., 2007; Oppikofer et al., 2008, 2009; Abellan et al., 2009, 2010, 2011; Stock et al., 2011, 2012; Royan et al., 2013). (See Andrew et al. (2013) and Abellan et al. (2014) for excellent reviews of the applications of LiDAR techniques to rock slope problems). More recently, ‚multi-view stereo' (MVS) or ‚structure-from-motion' (SFM) photogrammetric modeling has advanced to the point where in some cases the 3D terrain models generated in this way can match or exceed the fidelity and utility of terrestrial LiDAR (e.g. Westoby et al., 2012; James and Robson, 2012; Fonstad et al., 2013; Hugenholtz et al., 2013). Consumer-grade camera equipment and modern software has allowed for the development of rich 3D terrain models from a set of overlapping oblique photographs taken from ground level, or with traditional vertical aerial photography, for rock slope characterization (e.g. Haneberg, 2008; Sturznegger and Stead, 2009; Wolter et al., 2014), landslides (e.g. Ganzalez-Diaz et al., 2013), glacial mapping (Whitehead et al., 2013), and other applications in the geosciences (e.g. Javernick et al., 2014). The geometry and scale of many hazardous rock slopes adjacent to highways or railways are often not conducive to either terrestrial or downward-looking aerial cameras and laser scanners (i.e. the traditional approach to LiDAR). For example, terrestrial LIDAR or photographic data for transportation corridors may be difficult or impossible to collect in narrow through-cuts, or for any location with both tall slopes and some restriction in possible scan sites with a useful vantage-point (e.g. along steep valleys, or lakeshores and river valleys). Furthermore, the downward-looking aerial techniques, like traditional airborne LiDAR, are typically unable to resolve very steep or vertical slopes. In these cases, the newest generation of autonomous, unmanned aerial vehicles (UAV) equipped with high-resolution side-looking digital

RÉSUMÉ: all source location and volume from autonomous UAV-acquired photos: A case study

Access this article:
Canadian Geotechnical Society members can access to this article, along with all other Canadian Geotechnical Conference proceedings, in the Member Area. Conference proceedings are also available in many libraries.

Cite this article:
Dave Gauthier; D. Jean Hutchinson; Matt Lato; David F. Wood; A.J. Morris (2014) Rockfall source location and volume from autonomous UAV-acquired photos: A case study from a railway slope in northern Ontario in GEO2014. Ottawa, Ontario: Canadian Geotechnical Society.

@article{GeoRegina14Paper370,author = Dave Gauthier; D. Jean Hutchinson; Matt Lato; David F. Wood; A.J. Morris,title = Rockfall source location and volume from autonomous UAV-acquired photos: A case study from a railway slope in northern Ontario,year = 2014}