Automatic Fault Detection Based on Seismic Data Correlation Analysis

We describe the problem and suggest multi-step technique for fault surfaces and zones automatic detection and tracking using 3D seismic data. We examine as fault indicator the multivariate seismic attribute, which components are reconstructed on the base of seismic cube correlation structure analysis. The algorithm of correlation matrix reduction to principal axes is applied. Further we consider the set of parameters, analyzed a behavior of correlation matrix eigenvectors and eigenvalues; such set identifies fault zones. In order to associate fault zones as separated connected fields (facies), we use the multivariate classification algorithms, in particular, cluster analysis methods. It allows us to introduce, to formalize and to assess in classifying procedures a similarity level for the multivariate 3D seismic fault indicators. The special step of fault detection and tracking procedure is fault boundaries and surfaces tracking. In image processing such operations are called contour detection and skeletonization. We use available approaches with respect to our task