Separating primaries and multiples using hyperbolic Radon transform with deep learning

We propose a deep neural network (DNN) framework for computing hyperbolic Radon transform for separating primary reflections and multiples. The basic idea is to compute the weights associated with the inverse Hessian using the DNN for training data sets, which can then be applied to the adjoint transform of test data sets to obtain an initial model close to the true model. The output of the DNN can then be used as an input to the least-squares framework to obtain an output equivalent to the least-squares solution, but at a significantly reduced cost. Field data examples verify the effectiveness of the proposed approach. Presentation Date: Tuesday, October 13, 2020 Session Start Time: 8:30 AM Presentation Time: 10:10 AM Location: 351F Presentation Type: Oral