Internal multiple elimination with an inverse-scattering theory guided deep neural network

Abstract

Deep neural networks can automatically mine specific features from seismic data, which can be used in the process of multiple elimination. Surface-related multiple elimination (SRME) can provide good labels for the neural network based multiple elimination. Different from SRME, it is difficult to distinguish internal multiples from primary reflections when the subsurface is complex. An extended single-sided autofocusing guided by inverse-scattering theory is introduced to remove internal multiples in a data-driven manner. In this study, we explore the potential of neural networks in identifying the internal multiples with the guidance of inverse-scattering theory. We feed the neural network with training data, consisting of the shot records with internal multiples, and the primary-only datasets as labels, which are generated by an extended single-sided autofocusing method. The primary-only labels can be beneficial to the U-net framework. The test results show that internal multiple elimination via the neural network takes the advantage of the extended single-sided autofocusing method and is cheaper when the neural network is well-trained. The corresponding reverse time migration (RTM) results show the validity of our work Note: This paper was accepted into the Technical Program but was not presented at IMAGE 2022 in Houston, Texas.