一种自适应数据驱动的时空域地震数据鬼波压制方法

In marine seismic exploration, seismic deghosting is an old and longstanding data processing challenge. Deghosting methods in time-space (t-x) domain have attracted a lot of attentions since their flexibility for various source/receiver configurations. In this work, we propose an adaptive data-driven seismic deghosting method in t-x domain, which is flexible for arbitrary variable depths of the receivers, by using the t-x variant ghost-generating model, without inputting accurate streamer depth and rough sea surface model. The proposed method estimates these t-x variant ghost's parameters by a 2D grid search method, which is based on the super-Gaussianity of the deghosting outputs. In order to improve the accuracy of these parameters' estimations, the proposed method sets predefined discrete solution spaces, which are used by the 2D grid search method, in the t-x domain adaptively, and introduces the smoothness of the ghost's parameters as a constraint. The deghosting filters may amplify the noise, since some frequency components around the notches' centers are of low signal-to-noise ratio (SNR). The proposed method reconstructs the desired signal of these low SNR frequency components by introducing its super-Gaussianity as a constraint. For some local areas consisted of more than one seismic event with different dips, the proposed method separates them at first, deghosts the separated seismic events respectively, and then merges the deghosting results together. In summary, this work combines all these techniques mentioned above to obtain a broadband seismic data.

海上地震勘探中，鬼波压制问题是地震数据处理中一个长期存在的挑战。在时空域中进行鬼波压制的方法，由于其可以适应灵活多样的炮点和检波点的布置，从而受到许多研究者的关注。本项目研究一种自适应数据驱动的时空域鬼波压制方法，采用时空变鬼波生成模型，可以有效解决实际地震数据采集过程中缆深任意变化的问题，且不需要输入精确的缆深和海面起伏模型。本方法利用地震数据的超高斯性，采用网格搜索方法估计时空变的鬼波参数，并通过自适应地确定时空变的解搜索空间和施加参数光滑性约束，提高参数估计的精度。针对鬼波陷波带中心附近频率成分信噪比低，鬼波压制时噪声被放大的问题，本方法利用地震信号的超高斯性进行约束，采用信号重构技术重构陷波带内低信噪比成分中的有效波。针对局部区域存在的不同倾角地震同相轴的交叉问题，本方法采用先分离，再鬼波压制，最后合并的策略。本项目综合利用上述技术，实现地震数据有效频带的拓宽。

在海洋地震勘探中，鬼波对海洋地震数据的分辨率和信噪比造成严重影响。传统的鬼波压制方法对缆深和海面反射系数等参数敏感，且计算量大，还存在复杂实际数据的适应性不足的问题。为解决实际地震数据采集过程中缆深任意变化的问题，本项目实现了一种自适应的数据驱动的鬼波压制方法，同时所研究的方法不需要精确的缆深数据和海面起伏模型。本方法基于非高斯性，采用网格搜索的方法估计时空变的鬼波参数。鬼波陷波带区域有效信号振幅较弱，频率成分信噪比低，使用鬼波逆滤波算子进行鬼波压制时会对噪声进行放大，针对以上问题，本方法认为传统算法得到的鬼波压制结果陷波带周围的频率成分具备较高的信噪比，利用稀疏重构算法对低信噪比的频率成分进行重构。本项目利用数据驱动的方法自适应估计缆深和海面反射系数参数，利用更加准确的参数得到初步提升信噪比的鬼波压制结果，在此基础上利用稀疏重构算法进一步有效提升了鬼波压制结果的信噪比，实现了地震数据有效频带的拓宽。