基于稀疏反演的多震源地震混合采集数据分离技术

摘要

多震源地震混合采集数据(混叠数据)目前还不能直接采用传统的地震资料处理流程处理,必须先进行分离.基于滤波思想的分离算法可以初步实现对混叠数据的分离,但是当有效信号完全淹没在强噪声中时,该算法的分离效果较差,有效信号会受到损害.提出了一种基于稀疏反演的多震源数据分离算法,实现步骤是:首先输入野外采集的混叠数据,将其从炮集抽取为共检波点域或共炮检距域道集,使得来自邻炮的记录数据变成非相干噪声;然后通过稀疏反演,采用L1模的阈值噪声压制算法估计出混叠数据的噪声;最后从混叠数据中减去估计得到的噪声,从而分离出有效信号.模拟和实际数据测试结果表明,该方法能对混叠数据进行有效分离,很好地压制混叠噪声和保护有效信号.%Traditional seismic data processes cannot be applied to simultaneous-source blended data without deblending.However,the filtering-based deblending method can damage effective signals when applied to complex data,completely drowning effective signals in strong noise.Therefore,a sparse inversion-based iterative deblending method is proposed.First,blended common-shot data are converted into the common receiver domain,enabling the recorded data from neighboring shots to be considered as incoherent noise based on dithering code.Next,based on sparse inversion,this method estimates the incoherent noise using the L1 norm threshold filter in the frequency domain.Finally,the estimated noise is subtracted from the original blended data using a self-adaptive shaping method.Tests of both synthetic and field datasets show its feasibility and effectiveness in deblending while protecting signals.