f A workflow for the automatic characterization of geological surfaces from terrestrial LIDAR data

Point clouds acquired with terrestrial LIDAR are used as a digital support to accurately and precisely georeference outcrop characterizations; as well as to resolve accessibility problems, and improve outcrop characterizations. The LIDAR data allows for an efficient visualization and analysis of the outcrop in the computer, and is also useful for revisiting field data in the office or for teaching purposes. The common practice for virtual outcrop interpretation is visual identification and manual digitalization of pointsets or polylines by using 3-D CAD-like modules. Other, less generic, approaches are oriented towards the automated or semi-automated extraction of geological features, either based on the processing of intensity or other attributes of the virtual outcrop (RGB, hyperspectral) or on geometric parameters calculated from positions. In this presentation, we propose a workflow for the automatic characterization of planar surfaces (typically stratigraphic bedding or fractures) from LIDAR data. The workflow directly uses the point cloud; therefore no decimation, smoothing, intermediate triangulated or gridded surface are required; and is designed aiming to minimize user interaction to allow for a simple, fast, objective and semi-automated use. The result of the workflow is the reconstruction of planar surfaces identified in the point cloud by means of TIN surfaces, organized into families according to their orientations. These surfaces can be used as seeds for building surface-based models of the outcrop, or can be further analysed to investigate their characteristics (geometry, morphology, spacing, dimensions, intersections, etc.). The workflow is based on planar regressions carried out for each point in the point cloud. Which allow the subsequent filtering of points based on normal vector orientation, planar regression quality, relative locations of points or their relative normal vectors differences. This is aimed at individualizing planar patches with geological signification. A coarse grid search strategy is implemented to speed up neighbouring points searches and allow handling multimillion point clouds. The workflow is illustrated using synthetic and natural examples.<br>