地球物理学进展 ›› 2018, Vol. 33 ›› Issue (6): 2318-2323.doi: 10.6038/pg2018BB0547

• 应用地球物理学Ⅰ(油气及金属矿产地球物理勘探) • 上一篇    下一篇

联合构造导向滤波和包络反演的初始速度建模方法研究

崔婵婕1,2,侯卫生1,2,*(),杨翘楚1,2,刘恒光1,2,郑天成1,2   

  1. 1. 中山大学地球科学与工程学院,广州 510275
    2. 广东省地球动力作用与地质灾害重点实验室,广州 510275
  • 收稿日期:2018-03-27 修回日期:2018-09-06 出版日期:2018-12-20 发布日期:2019-03-03
  • 通讯作者: 侯卫生 E-mail:houwsh@mail.sysu.edu.cn
  • 作者简介:崔婵婕,女,1994年9月生,山西忻州人,硕士研究生,主要从事混合域全波形反演研究.(E-mail: 15602391063@163.com)
  • 基金资助:
    国家重点研发计划资助(2017YFC1500103)

Study on initial model construction method combining envelope inversion and structure oriented filter

CUI Chan-jie1,2,HOU Wei-sheng1,2,*(),YANG Qiao-chu1,2,LIU Heng-guang1,2,ZHENG Tian-cheng1,2   

  1. 1. School of Earth Sciences and Engineering, Sun Yat-sen University, Guangzhou 510275, China
    2. Guangdong Provincial Key Laboratory of Geodynamics and Geohazards, Guangzhou 510275, China
  • Received:2018-03-27 Revised:2018-09-06 Online:2018-12-20 Published:2019-03-03
  • Contact: Wei-sheng HOU E-mail:houwsh@mail.sysu.edu.cn

摘要:

充分利用地震记录的走时、相位、振幅等运动学和动力学信息,全波形反演可构建高精度的地下介质模型.然而,反演结果严重依赖于初始模型的质量,且因其所固有的非线性性导致反演过程极易陷入局部极小.从低频到高频的逐级多尺度反演方法能在一定程度上避免反演陷入局部极小,但受限于原始数据的低频成分缺失,而难以得到模型的大尺度信息.包络反演能在原始数据缺失低频成分的情况下获取模型的低波数信息,可以在一定程度上弥补多尺度反演的不足.构造导向滤波从图像本身提取构造信息,有效去除随机噪声的同时,还能提高大尺度模型反演质量.本文提出了一种联合构造导向滤波和包络反演的初始速度模型构建方法.该方法采用构造导向滤波方法对包络反演梯度作预处理,以构建稳健的初始速度模型.Marmousi2模型的测试结果表明:与基于高斯滤波的包络反演结果相比,联合构造导向滤波的包络反演具有更好的抗噪性,可进一步提高包络反演的精度,使反演结果更具地质意义.

关键词: 全波形反演, 包络反演, 构造导向滤波, 初始模型, 非线性性

Abstract:

The Full Waveform Inversion (FWI) can reconstruct high precision underground model by fully exploiting kinematics and dynamics information from seismic data, including travel time, phase, and amplitude. However, inversion results highly rely on the quality of initial model. And inversion easily falls into local minima due to high inherent non-linearity. Multi-scale inversion that reconstructs subsurface model from low to high frequency can partly avoid falling into local minima. The absence of ultra-low frequencies in field data is still the limitation of obtaining the large scale information. Envelope inversion is able to recover low wavenumber model by extracting ultra-low frequency information from field data, although the low frequency data cannot be captured in the seismic records. Multi-scale inversion jointed with envelope inversion can effectively mitigate the dependency of initial model. Structure oriented filter can suppress random noise effectively while enhancing important structure information. Thus it can improve the inversion results of large-scale model. This paper presented an initial model construction method combining envelope inversion and structure oriented filter. With structure oriented filter, the method preconditions gradients calculated from envelope misfit function to reconstruct initial model. Testing results of Marmousi2 model illustrated that the envelope inversion results by the presented method can be noise-resistant compared to envelope inversion method with Gaussian filter. The results obtained by the presented method show higher precision and more geological information.

Key words: Full Waveform Inversion (FWI), envelope inversion, initial model, structure oriented filter, non-linearity

中图分类号: