Reconstruction of missing near-offset data using multiples based on deep learning

Le résumé fourni par la source

Reconstructing missing near-offset seismic data in marine exploration presents significant challenges, particularly in complex geological settings where traditional methods often fail to deliver accurate results. This study proposes an novel approach that combines the focal transform and a deep learning framework based on a U-Net architecture. The focal transform utilizes multiples to map seismic data into the focal domain, effectively reducing the order of multiples while preserving essential kinematic characteristics for recovering missing traces. The deep learning model establishes a robust mapping between muted focal domain data and the original seismic data using a hybrid SSIM+L1 loss function. Numerical examples conducted on both flat-layer and complex models demonstrate that the proposed method achieves superior reconstruction of near-offset gaps, offering enhanced reflection continuity and accuracy. The approach is particularly effective under conditions of strong lateral variations or shallow water imaging complexities, outperforming conventional methods.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.