Automatic migration velocity analysis via deep learning

Migration velocity analysis is a crucial seismic processing step that aims to translate residual moveout in common-image gathers (CIGs) into velocity updates. However, this is often an iterative process that requires migration and significant human effort in each iteration. To derive the residual moveout correction accurately and efficiently, we propose a new method that combines a newly designed residual moveout (RMO) normalization and RMO identification. To make training successful, the former is designed to normalize the residual moveout from reflectors with different slopes and different depths to a non-dipping case. To replace manually picking the velocity spectrum, the latter is arranged to recognize normalized frown and smile patterns in CIGs and translate them into velocity updates via convolutional neural networks. Two numerical and field data examples demonstrate that the proposed method can effectively and efficiently flatten CIGs. The proposed method improves the quality of the velocity where there exists manual-picking error in comparison with traditional methods.