摘要
依据L1和L2范数自适应相减方法的特性,提出了联合L1/L2范数的表面多次波自适应相减方法,并引入GPU并行加速,在充分发挥两种方法优势的同时,有效缓解了两种自适应相减方法限制性条件引发的问题,在较短的时间内获得收敛的维纳滤波器,并且较好地拟合多次波模型和原始记录中的多次波。文中方法无需L2范数方法的假设条件,相比于L1范数方法提高了计算效率。理论模型和实际海洋地震数据测试表明,基于L1/L2范数的GPU并行加速的表面多次波自适应相减方法可有效压制地震数据中的表面多次波。
We propose in this paper L1/L2 norm adaptive subtraction method to suppress surface-related multiples(SRME),and introduce at the same time the GPU parallel acceleration.Taking advantages of the two methods,L1/L2 norm adaptive subtraction method eases effectively problems caused by their restrictive conditions,and it obtains a convergent Wiener filter in a short time.At the same time,this method can better fit multiple models and real multiples.Compared with the L1-norm method,the proposed method does not need assumptions of L2 norm and improves greatly the calculation efficiency.Tests on theoretical models and real marine seismic data show that the adaptive subtraction method based on L1/L2 norm and parallel GPU acceleration can effectively suppress surface-related multiples on seismic data.
出处
《石油地球物理勘探》
EI
CSCD
北大核心
2015年第4期619-625,2,共7页
Oil Geophysical Prospecting
基金
国家自然科学基金青年基金项目(41474118
41104088)
国家"863"高技术研究发展计划项目(2012AA061202)
大连理工大学海岸和近海工程国家重点实验室开放基金项目(LP1509)共同资助
关键词
SRME
自适应相减
L1范数
L2范数
GPU
surface-related multiples elimination(SRME),adaptive subtraction,L1 norm,L2 norm,GPU
