Dictionary Based Representation of Seismic Data for Tasks of Seismic Signals Analysis

摘要：

　　The article describes different methods of spectral decomposition on model seismic traces via machine learning approach.The main methods are considered: windowed Fourier transform,continuous wavelet transform and algorithms of spectral inversion.As a result of performed study it was shown that spectral inversion methods based on the solution of the problem of approximation of an input alignment a library of wavelets with the help of regularization,provide a much more accurate spectral representation of the signal.Several possible ways of geological interpretation are shown in article,including simple amplitude analysis(with RGB-mixing)and more complicated algorithm for estimation relative acoustic impedance and attenuation.

关键词： spectral decomposition spectral inversion OMP Lasso Thresholding machine learning

作者: A.V.Butorin F.V.Krasnov

作者单位: LLC Gazpromneft Science & Technology Centre,Saint-Petersburg,Russia

会议类型: 国内会议

会议名称: 2017年第五届数字油田国际学术会议(DOFIAC2017)

会议地点: 青岛

会议语种:英文

页码: 271-274

在线出版日期: 2017-10-15（万方平台首次上网日期，不代表论文的发表时间）