连续缺失地震数据的高阶流式预测滤波插值方法  

Seismic data interpolation beyond continuous missing data using high-order streaming prediction filter

作　　者：吴庚 刘财[1] 刘殿秘 刘洋[1] 郑植升 WU Geng;LIU Cai;LIU DianMi;LIU Yang;ZHENG ZhiSheng(College of Geo-exploration Science and Technology,Jilin University,Changchun 130026,China;Jilin Oilfield,CNPC,Songyuan 138003,China)

机构地区：[1]吉林大学地球探测科学与技术学院,长春130026 [2]中国石油吉林油田勘探开发研究院,松原138003

出　　处：《地球物理学报》2023年第3期1220-1231,共12页Chinese Journal of Geophysics

基　　金：国家自然科学基金项目(41974134,41874125)资助。

摘　　要：由于观测系统实施以及经济因素的限制,采集到的勘探地震数据在空间方向上总是不规则分布的,并且往往会出现数据的大范围连续缺失情况.许多后续地震数据处理方法(例如:多次波压制和波动方程偏移等)都需要空间上规则分布的数据.插值技术是一种解决地震数据缺失问题的有效手段,但是传统的数据插值方法在进行连续缺失数据重建时往往会出现失效的情况,尤其在处理非平稳地震同相轴时精度不高,并且大多数的方法需要迭代计算,在处理高维大规模数据时效率较低.针对连续缺失地震数据的快速插值问题,本文提出了一种非迭代的时空域高阶流式预测滤波插值方法,通过使用高阶限制条件来提高连续缺失数据的滤波器估计精度,提高局部约束条件的稳定性,改善低阶流式计算由于滤波器系数无法连续更新所造成的插值失效情况.同时,空间非因果滤波器和蛇形插值路径的设计方案可以有效减小大范围连续缺失数据和数据边界对于预测滤波器的计算误差,本方法能够有效处理包括近炮检距缺失情况在内的连续缺失数据插值重建.通过与工业标准傅里叶凸集投影(POCS)方法进行比较,理论模型和实际数据处理结果表明,本文提出的高阶流式预测滤波插值方法对高维连续缺失地震数据有较好的重建效果,在插值精度和计算效率两个方面有更好的平衡性.Due to the limitations of geometry implementation and economic factors, the acquired seismic data are always irregularly distributed along the spatial directions and often have a large range of continuous missing traces. Many subsequent data processing methods(e.g., multiple wave suppression and wave equation migration, etc.) require spatially regularly distributed data. Interpolation techniques are effective approaches to solve the missing seismic data problem, but traditional data interpolation methods occasionally fail when performing continuous missing data reconstruction, especially when dealing with non-stationary seismic events with low accuracy. Meanwhile, most interpolation methods require iterative computation, which is less efficient when dealing with high-dimensional large-scale data. To address the fast interpolation problem of continuous missing data, this paper proposes a noniterative interpolation method using high-order streaming prediction filter in the space-time domain, which improves the filter estimation accuracy of continuous missing data by using high-order constraints. The proposed method improves the stability of local constraints and the invalid situation caused by low-order streaming computation due to the inability to continuously update the filter coefficients. At the same time, the spatially noncausal filter and snaky interpolation path are designed to reduce the computation error of prediction filters from the large range of continuous missing data and data boundaries. Therefore, this method can effectively handle the reconstruction of continuous missing data including the missing near-offset case. Compared to the industry standard Fourier Projection onto Convex Sets(POCS) method, the proposed interpolation method based on high-order streaming prediction filter has a better reconstruction effect on the high-dimensional continuous missing seismic data, which shows a better balance in both interpolation accuracy and computational efficiency according to the synthetic model and field

关 键 词：地震数据插值 高阶流式预测滤波 连续缺失数据 空间非因果 蛇形插值路径 

分 类 号：P631[天文地球—地质矿产勘探]