A Time-frequency Focalization Oriented Inverse Spectral Decomposition And Its Application In Seismic Interpretation

The spectral decomposition(SD)method has been widely used in the identification of geological anomalies such as faults,rivers,alluvial fans and karst caves.However,the time-frequency focusing property of classical SD methods are often limited by the selection of window functions.The limitation should be responsible for smeared seismic interpretations,seismic signals interfere together when two or several geological bodies are relatively close,which will also affect the frequency components of local signals,especially for the deep.Due to the low time-frequency resolution,the isofrequency amplitude slice obtained by using the conventional SD methods readily suffers from strong frequency mixture issue.Consequently,the isofrequency amplitude slice is still the superposition of different frequency components,which is not easy to interpret close geological bodies.To further address this issue,we study a time-frequency focalization oriented inverse spectral decomposition(TFFOISD)method.Based on the online spectral decomposition,the l_p(0<p<1)norm is selected as the sparse constraint to establish the objective function,which is solved by the fast generalized iterated shrinkage algorithm.Compared with the CWT and inverse SD(ISD)using a typical l₁-norm constraint,TFFOISD can yield a more compact spectrum contributing to detect the distributions of close geological features.This is because the l_p(0<p<1)norm plays a role in time-frequency focusing property.There are less frequency mixing for spectral amplitude computed by using TFFOISD,which is helpful for obtaining precise spectral amplitude,even when both the time variation of wavelet and the interference of thin beds complicate the time-frequency spectrum.That is to say,the isofrequency amplitude slice contains illusions are less,which contributes to detect the distributions of close geological features.The synthesized data and physical model are used to test the ability of TFFOISD,which further verifies the method has a higher time-frequency resolution and better time-frequency focusing property.The application of 4D spectrum on a 3D real dataset with an area of approximately 230 km² illustrates its potential for detecting channels and the karst slope fracture zone in the deep target reservoir.Moreover,the phase spectrum of high time-frequency focusing obtained by TFFOISD is used to identify some small channels.