地震数据重建的谱投影梯度算法中的参数选取

摘要

Irregular sampling can result in artificial noise and spatial aliasing in the processing of seismic data.In this paper,we se-lect a spectral projection gradient L1 (SPGL1)algorithm to reconstruct the seismic data based on Jitter sampling,random sam-pling and Fourier transform.We further discuss parameter selection in this algorithm,such as different noise levels and numbers of iterations.Numerical experiments show that reconstruction results depend on the selection of different parameters,when the noise level is small,and the number of iterations is moderate,we could obtain effective reconstruction data with high SNR and computa-tional efficiency.%地震数据不规则采样会导致在后续处理工作中引入人为误差、噪声及空问假频等.为此基于Jitter下采样、随机下采样和傅里叶变换,通过L1范数谱投影梯度(SPGL1)算法对不规则下采样地震数据进行重建,并探讨了该算法中的参数选取问题.在重建过程中选取不同的噪声水平σ 值和迭代次数,比较不同参数下的重建结果.数值试验结果表明,SPGL1算法能获得较好的重建结果,不同的参数对重建效果影响较大,当σ 值较小、迭代次数适中时,重建数据的信噪比高且计算时问较短.