地震信号线性与非线性时频分析方法对比  

Comparison of linear and nonlinear time-frequency analysis on seismic signals

作　　者：黄昱丞 郑晓东[1] 栾奕[2] 杨廷强 

机构地区：[1]中国石油勘探开发研究院油气地球物理研究所,北京100083 [2]香港中文大学理学院,香港999077

出　　处：《石油地球物理勘探》2018年第5期975-989,共15页Oil Geophysical Prospecting

基　　金：国家重点研发计划项目(2016YFC0601107);国家油气重大专项(2017ZX05001)联合资助

摘　　要：基于模拟信号与实际数据,系统分析了地震勘探领域目前流行的多种类时频分析方法的时频分辨率、计算效率和抗噪性能。模拟信号分析表明:线性方法的时频聚集性普遍较低,但计算效率很高,且不受交叉项干扰,其中连续小波变换法抗噪性较强;非线性方法时频聚集性较高,但计算效率普遍低于线性方法,抗噪性能方面平滑伪Wigner-Ville分布法和广义线性调频小波变换法相对稳健。因此,低信噪比情况下宜采用线性方法,信噪比较高时非线性方法时频分辨率更高,可识别薄层,亦有助于揭示地质体空间展布等潜在信息。实际地震资料的短时Fourier变换、连续小波变换、平滑伪Wigner-Ville分布和广义线性调频小波变换法处理结果表明:与线性方法相比,稳健的非线性时频方法可以更细致地刻画储层顶、底反射界面和沉积相带展布等特征。In this paper,we discuss the resolution,computational efficiency,and robustness of currently used seismic time-frequency analysis methods based on synthetic signals and real seismic data.In synthetic signal analysis,the time-frequency energy concentration of linear methods are generally weak,yet they are preferred in the low signal-tonoise(SNR)situation for their high computational efficiency,among which the continuous wavelet transform(CWT)is considerably robust;nonlinear methods are unstable with noisy data except for smoothed pseudo Wigner-Ville distribution(SPWVD)and general linear Chirplet transform(GLCT).However,nonlinear methods may reach much higher time-frequency resolution with high SNR.Thin-beds and the distribution of geologic features could be clearly identified using these methods.In real seismic data analysis,compared with linear methods like short time Fourier transform and CWT,reservoir reflection interfaces and sedimentary facies belts are better delineated on common frequency sections and time slices obtained by robust nonlinear methods like SPWVD and GLCT.

关 键 词：地震信号 时频分析 分辨率 计算效率 抗噪性能 

分 类 号：P631[天文地球—地质矿产勘探]