Insight Into Sparse Autonomous Seismic Acquisition With Compressive Sensing and Gradients

Abstract

Technological advancements in different scientific fields may allow seismic acquisition to be fully autonomous. Due to the high cost, it is still not plausible for this technology to be used. Autonomous marine seismic acquisition requires many expensive platforms and vessels to acquire high-quality seismic data. A straightforward way to reduce the cost of such seismic surveys is by using fewer platforms. However, this will result in acquiring fewer traces due to the spatial under-sampling caused by using fewer sources and receivers. This project aims to revisit new developments in compressive sensing techniques to handle missing data by reconstruction from sparse seismic data. Using data from both field and numerical settings, we looked into the role of wavefield gradients in recovering missing data from highly subsampled observations. This included experimentation with subsampling, different sparsity-promoting solvers, and the role of transformation of data to different sparse domains. The reconstruction of different datasets with their gradients using compressive sensing techniques showed promising results for regularly subsampled data. For the randomly subsampled data, the results were not as promising since the F-K domain is not an ideal sparse domain for such data.