摘要
地面微地震数据的信噪比很低,严重地影响初至拾取的精度及反演结果的可靠性。为此,研究了基于字典训练的稀疏表示去噪方法,通过曲波变换估算了剖面中的噪声方差,从而将该法用于实际地面微地震资料去噪中;为了改善在低信噪比时的去噪效果,研究了小波域的稀疏表示去噪方法,并与普通的稀疏表示去噪方法进行了定量分析。理论模型及实际资料的处理结果表明:1迭代次数及字典原子的大小会对去噪结果产生较大影响,去噪后数据的信噪比随着两者的增加而增加,但这也会导致计算效率降低。因此,在处理时对于较大的数据可以选择中等大小的字典原子及迭代次数,以保证在得到较高信噪比的同时,具有较快的运算速度;2该方法可以去除传统稀疏表示方法在去噪后引入的"背景斑块",且去噪后的信噪比也得到极大提高。因此,相对于传统的方法,本文的方法具有显著的优势及较好的应用价值。
Surface micro-seismic data is characterized by lower signal-to-noise ratio.This affects severely the accuracy of first break picking and the reliability of inversion.We proposed in this paper a microseismic data denoising method based on sparse representations over learned dictionaries in the wavelet domain.The method calculates the noise variance through Curvelet transform and then it is applied to suppress random noise in real data.In order to improve denoising effects on low signal-tonoise ratio data,the sparse representation denoising method in the wavelet domain is put forward.Quantitative analysis is also carried out in the new and common sparse representation method.Tests results on theoretical and real data show that the number of iterations and the size of dictionary atoms have great influence on the final denoising result.The signal-to-noise ratio of denoised data is improved with both a big number of iterations and a large size of dictionary atoms,but it also leads to decrease computational efficiency.A medium size of dictionary atoms and a reasonable number of iterations for larger data should be needed,so a higher signal-to-noise ratio and a faster calculation can be achieved at the same time.In addition,this method can remove the background patch introduced in convenditional sparse representation,so it will greatly improve the signal-to-noise ratio of denoised data.Therefore,the proposed method has obvious advantages compared to conventional method.
出处
《石油地球物理勘探》
EI
CSCD
北大核心
2016年第2期254-260,204-205,共7页
Oil Geophysical Prospecting
基金
国家自然科学基金项目(41504097
41374123)
国家重大专项(2011ZX05006-002)联合资助
关键词
微地震
随机噪声压制
稀疏表示
字典学习
小波变换
micro-seismic
random noise suppression
sparse representation
dictionaries training
wavelet transform
作者简介
邵婕硕士研究生，1991年生；2013年毕业于中国石油大学（华东）勘查技术与工程专业，获学士学位；2013年至今在中国石油大学（华东）攻读硕士学位，研究方向为地震数据处理及地震属性分析。
山东省青岛市开发区长江西路66号中国石油大学（华东）地球科学与技术学院，266580。Email：tangjie@upc．edu．cn
