Abstract: In marine towed-streamer seismic data acquisition, the absence of geophones near the source due to the distance between the source and the nearest geophone results in the inability to obtain near-offset seismic data. This limitation affects subsequent seismic processes such as wave equation migration imaging and multiple wave suppression. Therefore, to reconstruct high-quality near-offset seismic data is of great significance. Seismic interferometry has shown promising applications in near-offset reconstruction; specifically, cross-correlation operations are performed on the wave field records received by different geophones to generate pseudo primaries with kinematic characteristics similar to the original seismic signals. However, the pseudo primaries differ from real data in terms of dynamic characteristics, such as amplitude and phase. To address this problem, considering the structural self-similarity of seismic data, a near-offset seismic data recursive reconstruction algorithm based on non-local self-similarity priors is proposed. The method uses pseudo primaries as priors to guide the reconstruction process, combining them with the non-local self-similarity of seismic data and employing low-rank reconstruction techniques to achieve the interpolation for near-offset seismic data. Experimental results demonstrate that compared with traditional methods such as rank-reduction-based multichannel singular spectrum analysis, pseudo primaries and matched filter correction methods, and prediction error filtering methods, the proposed approach achieves superior performance in terms of data signal-to-noise ratio and reconstruction quality.

Key words: near-offset, seismic interferometry, non-local self-similarity, low-rank matrix completion, data reconstruction

摘要： 在海洋拖揽地震数据采集中,由于震源与最近的检波点存在距离,无法布置检波器,导致近偏移距地震数据无法获取,影响了后续的地震处理,如波动方程偏移成像和多次波压制等。因此,高质量重建近偏移距地震数据有着重要意义。地震干涉法在近偏移距地震数据重建上有很好的应用,通过对不同检波器接收的波场记录进行互相关运算,可以生成与原始地震信号运动学特性相似的虚拟一次波。然而,虚拟一次波与真实数据之间存在动力学特征(如振幅和相位等)差异。为此,考虑到地震数据的结构自相似性,提出了一种基于非局部自相似先验的近偏移距地震数据递推重建算法。利用虚拟一次波作为先验,指引并结合地震数据的非局部自相似性,利用低秩重建方法实现近偏移距地震数据插值。实验结果表明:相较于传统的基于降秩的多道奇异谱分析方法、基于虚拟一次波和匹配滤波校正的方法以及预测误差滤波法,提出的方法在数据信噪比和重建质量方面展现了更优的性能。

关键词: 近偏移距, 地震干涉法, 非局部自相似, 低秩矩阵补全, 数据重建