AUTHOR=Huang Yunbo , Huang Jianping , Ma Yangyang 

TITLE=A fast least-squares reverse time migration method using cycle-consistent generative adversarial network

JOURNAL=Frontiers in Earth Science

VOLUME=Volume 10 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/earth-science/articles/10.3389/feart.2022.967828

DOI=10.3389/feart.2022.967828

ISSN=2296-6463

ABSTRACT=With high imaging accuracy, high signal-to-noise ratio, and good amplitude balance, least-squares reverse time migration (LSRTM) is an imaging algorithm suitable for deep high-precision oil and gas exploration. However, the computational costs limit the large-scale industrial application of this method. The difference between traditional reverse time migration (RTM) and LSRTM is whether to eliminate the effect of Hessian operator or not, while solving Hessian matrix explicitly or eliminating the effect of Hessian matrix implicitly has a very high requirement on computation or storage capacity. We simulate the inverse Hessian by training a cycle-consistent generative adversarial network (cycleGAN) to construct a mapping relationship between the RTM results and the true reflectivity. The trained network is directly applied to the RTM imaging results, which improves the imaging quality while significantly reducing the calculation time. We select three velocity models to generate the training data sets and two other models to produce the validation sets, where the data in the validation sets are not involved in the training process. The prediction results of the trained network on the validation sets show that the proposed method significantly improves the imaging quality with almost no increase in computational effort. In addition, we applied the network trained entirely on synthetic data to the field data, and the predicted result confirms the effectiveness and good generalization of the proposed method.