Abstract:Facing the increasingly complex and sophisticated situation of exploration targets, many new seismic acquisition technologies such as wide-azimuth, wide-bandwidth and high-density seismic data acquisition methods have been widely used, leading to the acquisition period is becoming longer and longer and the acquisition cost is becoming higher and higher than before. How to solve such a problem is a current focused issue in this field. This paper presents a study on methods and theories of highly efficient seismic data acquisition, focusing on seismic data sparse representation theory, highly efficient acquisition and reconstruction methods. We present an improved piecewise-random sampling method for seismic data sampling, and then develop a high-precise reconstruction approach for sparse acquisition. Based on the introduction of blended seismic acquisition, the blended seismic data separation is studied thoroughly. A new data separation method based on small window median filtering and sparse constraint joint random denoising is proposed to improve the practicability of the blended seismic acquisition method. We combine the above two useful efficient seismic acquisition methods and propose three highly efficient acquisition network designing methods, including 1) dense blended sources and random geophones (DmsRg) scheme, 2) random blended sources and dense geophones (RmsDg) scheme, and 3) random blended sources and random geophones (RmsRg) scheme. The corresponding reconstruction procedures are proposed which can reconstruct the seismic data on the highly efficient acquisition network to the regular and dense acquisition network. It satisfies the requirements of subsequent processing in accordance with the existing conventional means. Furthermore, the influence of simultaneous shooting on seismic data's imaging results is analyzed. We use the international standard Marmousi model to test the proposed data acquisition methods. The numerical examples show that the theoretical framework of the joint highly efficient seismic data acquisition established in this paper can improve the efficiency of the current seismic data acquisition and reduce the exploration costs. High-precise imaging result of the reconstructed data is compatible to that of conventional methods. Although the proposed methods have achieved great progress in numerical tests, more work of field seismic data acquisition is needed to validate their effectiveness in the future.

Low-Frequency Extrapolation of Prestack Viscoacoustic Seismic Data Based on Dense Convolutional Network

Layered low-frequency extrapolation with deep learning in full-waveform inversion

Deep Learning for Low-Frequency Extrapolation and Seismic Acoustic Impedance Inversion

Highly efficient methods of seismic data acquisition based on sparse constraints and blended-source excitation

Deep Learning-Based Low-Frequency Extrapolation and Impedance Inversion of Seismic Data

Extrapolated full-waveform inversion with deep learning

Global-Feature-Fusion and Multiscale Network for Low-Frequency Extrapolation

Prediction of Low-Frequency Seismic Data Without Relying on Wavelet Using Deep Learning

Learning Seismic Low-Frequency Extrapolation in a Latent Space Using Limited Data

A self-supervised learning framework for seismic low-frequency extrapolation

Transformer For Low-frequency Extrapolating of Seismic Data

Low-Frequency Prediction Based on Multiscale and Cross-Scale Deep Networks in Full-Waveform Inversion.

Data-driven low-frequency signal recovery using deep-learning predictions in full-waveform inversion

Compensating Low-Frequency Signals for Prestack Seismic Data and Its Applications in Full-Waveform Inversion

Porosity Prediction from Prestack Seismic Data Via Deep Learning: Incorporating a Low-Frequency Porosity Model

Low-frequency data analysis and expansion

Efficient progressive transfer learning for full waveform inversion with extrapolated low-frequency reflection seismic data

Prestack seismic random noise attenuation using the wavelet-inspired invertible network with atrous convolutions spatial pyramid

Sparse Blind Deconvolution Based Low-Frequency Seismic Data Reconstruction for Multiscale Full Waveform Inversion

Porosity prediction from prestack seismic data via deep learning: Incorporating low-frequency porosity model

Low-frequency constrained seismic impedance inversion combining large kernel attention and long short-term memory