张鹏^1,,
刘洋^2,,,
刘鑫明¹,
刘财²,
张亮³
1. 华北理工大学矿业工程学院, 河北唐山 063210
2. 吉林大学地球探测科学与技术学院, 长春 130026
3. 中国石油吉林油田公司乾安采油厂, 吉林松原 138000

基金项目: 国家自然科学基金项目（41522404，41774127）资助


详细信息

作者简介: 张鹏, 男, 1989年生, 讲师.2018年毕业于吉林大学固体地球物理学专业, 主要从事地震数据处理方面的研究工作.E-mail:peng871224@126.com

通讯作者: 刘洋, 男, 教授, 博士生导师.主要从事开源地球物理数据处理和地质-地球物理综合研究等工作.E-mail:yangliu1979@jlu.edu.cn

中图分类号: P631

收稿日期:2018-11-06
修回日期:2019-09-10
上线日期:2020-05-05

Suppressing seismic random noise based on Seislet-TV dual regularization

ZHANG Peng^1,,
LIU Yang^2,,,
LIU XinMing¹,
LIU Cai²,
ZHANG Liang³
1. College of Mining Engineering, North China University of Science and Technology, Tangshan Hebei 063210, China
2. College of Geo-exploration Science and Technology, Jilin University, Changchun 130026, China
3. Qian An Oil Factory, Jilin Oilfield, CNPC, Songyuan Jilin 138000, China


More Information

Corresponding author: LIU Yang,E-mail:yangliu1979@jlu.edu.cn

MSC: P631

--> Received Date: 06 November 2018
Revised Date: 10 September 2019
Available Online: 05 May 2020


摘要
摘要:人工地震数据总是受到随机噪声的干扰，地震数据时-空变的特性使得常规去噪方法处理效果并不理想，容易导致有效信号的损失.目前广泛应用的预测滤波类方法存在处理时变数据能力不足的问题.随着压缩感知理论的不断完善，稀疏变换阈值算法能够解决时变地震数据噪声压制问题，但是常规的稀疏变换方法，如傅里叶变换，小波变换等，并不是特殊针对地震数据设计的，很难提供地震数据最佳的压缩特征，同时，常规阈值算法容易导致去噪结果过于平滑.因此开发更加有效的时-空变地震数据信噪分离方法具有重要的工业价值.本文将地震数据信噪分离问题归纳为数学基追踪问题，在压缩感知理论框架下，利用特殊针对地震数据设计的VD-seislet稀疏变换方法，结合全变差（TV）算法，构建seislet-TV双正则化条件，并利用分裂Bregman迭代算法求解约束最优化问题，实现地震数据的有效信噪分离.通过理论模型和实际数据测试本文方法，并且与工业标准FXdecon方法进行比较，结果表明基于seislet-TV双正则化约束条件的迭代方法能够更加有效地保护时-空变地震信号，压制地震数据中的强随机噪声.
关键词: 地震数据信噪分离/
压缩感知/
全变差/
seislet-TV双正则化/
分裂Bregman迭代

Abstract:Artificial seismic data are always disturbed by random noise. In conventional denoising methods, the temporal-spatial variation of seismic data makes results unsatisfactory and leads to loss of effective signal. At present, the widely used predictive filtering methods have shown their disadvantages when dealing with time-varying data. With the continuous improvement of compressed sensing theory, sparse transforms which combine with threshold methods show the ability to suppress random noise of time-varying seismic data. But the traditional sparse transforms like Fourier transform, wavelet and so on, which were not specially designed for seismic data. So it is difficult for them to provide the best compression characteristics for seismic data. At the same time, the traditional threshold methods can easily lead to over smooth denoising results. Therefore, it has great industrial value to develop a more effective time-space-varying seismic signal and noise separation method. In this paper, we treat seismic data denoising as basic pursuit problem. Based on the frame of compressed sensing (CS), we combine VD-seislet transform which is specially designed for seismic data with total variation to construct seislet-TV dual regularization. And a split Bregman iteration algorithm is used to solve the constrained minimization problem in order to separate signal and noise effectively. According to the Tests on synthetic and field data test and comparison with the industry standard FXdecon method show that the denoising results of the proposed iteration method based on seislet-TV dual regularization can protect time-space-varying seismic data and suppress the strong random noise in seismic data more effectively.
Key words:Seismic signal and noise separation/
Compressed sensing/
Total variation/
Seislet-TV dual regularization/
Split Bregman iteration



PDF全文下载地址:

http://www.geophy.cn/data/article/export-pdf?id=dqwlxb_15466