基于稀疏反演的高效混采数据分离方法

混叠采集技术能极大地提高地震采集效率，但是由于连续激发，混采数据中存在严重的邻炮干扰噪声，而且混采数据规模巨大，这要求后续的信噪分离算法精确、稳定、快速，为此提出一种基于稀疏反演的高效混采数据分离方法。在三维共检波点道集上，首先通过三维快速傅里叶变换，将时空域数据变换到频率-波数-波数域（FKK）；然后采用硬阈值函数提取相干信号，利用相干信号预测混叠噪声；最后以数据残差为驱动，不断收缩阈值，迭代更新模型，实现高效混采数据分离。模拟数据和实际数据应用结果表明，本文方法能够精确、稳定、快速地分离高混叠度的地震数据。

High-productivity blended acquired data separation by sparse inversion

The vibroseis blending acquisition improves greatly acquisition efficiency.However,this acquisition brings serious noise in massive blended seismic data due to the continuous shooting.Therefore accurate,stable,and fast noise-separation algorithms are needed.We propose in this paper an inversion-based de-blending method applied on 3D common receiver gathers.Firstly,temporal-spatial seismic data is transformed into the frequency-wavenumber-wavenumber(FKK) domain data by 3D fast Fourier transform.Then,the hard thresholding algorithm extracts coherent signals,and noise is predicted with the extracted coherent signals.Finally,driven by data residuals,signals are iteratively updated with shrinking thresholds until full noise separation from data is achieved.Tests on both synthetic and field data show that the proposed method can separate high-productivity blended acquired data in an accurate,stable and fast way.