SUPERVOXEL-BASED MULTI-SCALE POINT CLOUD SEGMENTATION USING FNEA FOR OBJECT-ORIENTED ROCK SLOPE CLASSIFICATION USING TLS

In mountainous regions around the world rockfalls pose a hazard to infrastructure and society. To aid in our understanding and management of these complex hazards, an inventory can be compiled. Three-dimensional remote sensing data can be used to locate the source zones of these events and generate models of areas which detached. We address the way in which the shape of a rockfall object can be measured. The shape of a rockfall has implications for forward modelling of potential runout zones.

The work carried out for this study is part of a collaborative research program studying the impact of ground hazards on transportation infrastructure in Canada. The focus of the paper is the testing and application of a new simulation technique which can model the movement of falling rock material. These initial tests show that our simulation technique is capable of reproducing material accumulations from rockfall events which occurred above a section of railway in British Columbia, Canada.

Rockfalls present a hazard to railways in mountainous terrain. 3-D remote monitoring data can be used to identify events that occurred between data collections. Using a case study from British Columbia, we present a method combining 3-D rockfall event data with spatial rockfall simulations to provide a refined estimate of the frequency of rockfalls presenting a direct hazard to passing trains and railway infrastructure, which is often less than the total number of rockfalls that occurred.

We developed and tested an automated terrestrial laser scanning (ATLS) system with near-real-time change detection at the Séchilienne landslide. We monitored the landslide for a 6-week period collecting a point cloud every 30 min. We detected various slope processes including movement of scree material, pre-failure deformation of discrete rockfall events and deformation of the main landslide body. This system allows the study of slope processes a high level of temporal detail.