BUILDING 3D GEOLOGICAL MODELS WITH THE SUPPORT VECTOR MACHINE

3D geological models are a powerful way of visualization, analysis and interpretation of geological information. The problem is normally approached on a layer-by-layer basis through various interpolation techniques implemented in GIS tools. However, this is a challenging and time-consuming task, which also requires reasonable input data coverage. Here we propose the use of the Support Vector Machine (SVM) in order to automate the construction of 3D models from sparse input information. The SVM algorithm is based on the Statistical Learning Theory developed by Vapnik (1995). It uses a set of samples characterized by a number of features with attached class information (known as a training set) in order to build a mathematical function that can further be used to classify test samples with unknown classification. Therefore, we re-formulate the problem as a lithological classification task and use a multi-class SVM implementation to solve it. We demonstrate the application of the method in two different contexts where we compared the SVM classification results with a model produced by a more traditional interpolation technique. Overall, our experiments showed that the SVM can be efficiently applied in 3D geological reconstructions.