Generative Inpainting Network Applications on Seismic Image Compression and Non-Uniform Sampling
The use of deep learning models as priors for compressive sensing tasks presents new potential for inexpensive seismic data acquisition. An appropriately designed Wasserstein generative adversarial network is designed based on a generative adversarial network architecture trained on several historical surveys, capable of learning the statistical properties of the seismic wavelets. The usage of validating and performance testing of compressive sensing are three steps. First, the existence of a sparse representation with different compression rates for seismic surveys is studied. Then, non-uniform samplings are studied, using the proposed methodology. Finally, recommendations for non-uniform seismic survey grid, based on the evaluation of reconstructed seismic images and metrics, is proposed. The primary goal of the proposed deep learning model is to provide the foundations of an optimal design for seismic acquisition, with less loss in imaging quality. Along these lines, a compressive sensing design of a non-uniform grid over an asset in Gulf of Mexico, versus a traditional seismic survey grid which collects data uniformly at every few feet, is suggested, leveraging the proposed method.
PDF Abstract
