基于非结构化网格的最小二乘逆时偏移  

Least-squares reverse time migration based on unstructured gird

作　　者：杨凯[1,2] 张剑锋[1] 

机构地区：[1]中国科学院地质与地球物理研究所中国科学院油气资源研究重点实验室,北京100029 [2]中国科学院大学,北京100049

出　　处：《地球物理学报》2017年第3期1053-1061,共9页Chinese Journal of Geophysics

基　　金：国家自然科学基金面上项目(41574135);国家自然科学基金重点项目(41330316);中石油弹性波地震成像项目(2015A-3613)联合资助

摘　　要：相对于传统的逆时偏移,最小二乘逆时偏移具有更高的成像质量,这种改善是通过迭代反演来获得的,另外其精度与效率依赖于求解波动方程算法的精度与效率.本文给出了基于非结构化网格的最小二乘逆时偏移的方法,该方法能够充分结合最小二乘算法与非结构化网格精细刻画地下界面以及随速度自适应剖分的优点;并采用带补偿的拉普拉斯滤波算法,来消除梯度计算中的低频噪声,从而加速目标函数的收敛速度.通过简单倾斜模型以及复杂Marmousi模型测试,显示了该方法的有效性和潜力.Compared with conventional reverse time migration, the least-squares reverse time migration can improve image quality by using the inversion-based algorithm. In addition, the accuracy and efficiency of this method mainly depends on the accuracy and efficiency of solving acoustic wave equation. The unstructured gird method describes interfaces more precisely and can use the velocity related adaptive mesh generation algorithm. In this paper, we propose a least- squares reverse time migration algorithm based on unstructured gird. The purpose is to combine the advantages of the least-squares algorithm and unstructured gird method, and then to use Laplacian filter with compensation to remove the low-frequency noise in the calculation of gradient. Thus we can accelerate the speed of convergence of the objective function. This method is successfully applied to a simple inclined layered model and the Marmousi model, which demonstrates its efficiency and potential.

关 键 词：最小二乘逆时偏移 非结构化网格 格子法 拉普拉斯滤波 

分 类 号：P631[天文地球—地质矿产勘探]