Collection, processing, interpretation and modelling of digital outcrop data using VRGS: An integrated approach to outcrop modelling

Abstract

Much focus has been given to the hardware and data collection techniques for digital outcrop analogue work. The software development has, however, been left behind, with many geoscientists relying on applications designed for civil engineering or surveying purposes. Though these approaches have yielded interesting and often impressive results, without dedicated software applications the true power of digital outcrop data will never be realised. For the past 7 years software has been under development at Manchester University dedicated to 3D digital outcrop work, with a focus of being able to use very large data sets (collected from LiDAR or other digital sources) effectively and efficiently, but importantly to integrate these approaches with more traditional data collection approaches such as sedimentary logging and field mapping. The software developed facilitates processing of point cloud data from LiDAR and satellite sources (such as Digital Elevation Models), the triangulation of that data into meshes, and interpretation on both point clouds and meshes where appropriate. Interpretation tools include typical polyline mapping tools, structural measurement tools, sedimentary logging tools as well as more automated interpretation and mesh/point cloud classification approaches. Due to the nature of the rapid and large scale data collection possible using modern surveying systems and abundance of publically available satellite imagery and DEM data, digital outcrop datasets can be very large in size. This presents problems in the time taken to interpret and extract surfaces, structures and geostatistics from these data. One solution to reduce the time needed for interpretation and classification is the application of artificial intelligence to the problem. Artificial Neural networks try and replicate the same learning process used by humans and other animals. These Artificial Neural Networks (ANN) potentially provide very powerful ways of classifying data. Examples will be given showing the application of these ANN approaches to the classification of point cloud and mesh data, in particular addressing the problems of extracting structural data on plane orientations such as fracture and bedding planes. The applications of other soft computing and artificial intelligence approaches will also be presented. Integration of multiple data sources into one environment facilitates the development of new modelling approaches. A predictive approach to surface modelling will be presented relying on the use of structural data from dip-azimuth measurements from bedding planes, and polyline interpretations from key stratal surfaces. This modelling approach relies on converging a triangulated mesh, based on the control data, onto a solution matching that input data, rather than using traditional interpolation/extrapolation approaches. With the rapid evolution of computer hardware, particularly the development of high power graphics-card based computing, the application of modern graphics-card features to the processing visualisation and rendering of large digital outcrop datasets will be demonstrated. These hardware advancements will prove of significant benefit to the geosciences, but only if software applications are written to take advantage of them.