基于三维稀疏反演的混合震源数据分离与一次波估计

混合震源采集（下称混采）技术是当前地震勘探的潮流。但是由混采获得的数据中包含相互重叠的由多个震源激发产生的炮记录，会对后续的地震数据处理产生严重干扰。本文针对现有的基于混采数据的稀疏反演一次波估计（EPSI）方法，提出了一种改进的基于三维稀疏反演的混采数据分离与一次波估计方法。我们将混采EPSI方法的地下一次波响应估计过程转化为基于L1范数的双凸优化问题，并用基于L1范数的谱投影梯度（SPGL1）算法进行求解，确保取得全局极值，从而稳定反演过程。此外，我们还用二维曲波变换和一维小波变换组成三维联合稀疏变换对反演过程进行约束，能在确保求解精度的同时较以往的三维曲波稀疏约束大大提高计算速度。将本文方法应用于模拟混采数据和海上实际混采数据，将试算结果与传统混采数据EPSI方法对比，全面验证了本文所述方法的有效性和优越性。

Separation and Primary Estimation of Blended Data by 3D Sparse Inversion

The blended acquisition of seismic data is widely used in the industry area; however, the seismic data acquired by such a method contain overlapping shot records of multiple sources, which is not conducive to the subsequent seismic data processing. A modified separation and primary estimation method for blended data based on 3D sparse inversion is proposed in this paper. We introduce the L1 norm bi-convex optimization into the solution process of estimating primary impulse responses by conventional EPSI and SPGL1 algorithm to get the global minima, so that the inversion process is stable. Besides, 2D curvelet transform and 1D wavelet transform are combined into a 3D sparse constraint to improve the calculation speed while ensuring the inversion accuracy. Compared to the conventional EPSI for blended data in the standard industry workflow, the effectiveness and superiority of this proposed method is verified in the application in synthetic data and marine field data.