Research And Application On The Method Of Seismic Signal Processing Based On Blind Source Separation

The technique of Blind Source Separarion come out with the development of informationand computer technique. The problem it solved is that: how to estimate or recover the originalsources from the observation signal on condition that both the sources and the parameters ofchannel are unkowned. As a hot spot in the signal processing areas, the technique of BSS wasresearched by more and more scholars and widely applied to many areas such as biomedicalengineering, speech signal and communication system, image processing, seismicprospecting etc.In the area of seismic exploration, the target gradually move to subtle oil and gasreservoir such as small sand body, thin layer and micro-crack etc. The features of thereservoir targets in these abnomal geological body are weak seismic response, waveformdistortion and so on. And it’s difficult to track the reflection events in the sections of thesereservoir targets. To solve the problem mentioned above, this paper researched the applicationof BSS in the separation of weak seismic signal. We completed Seismic Blind Deconvolutionbased on Preconditioned Conjugate Gradient method and improved the method of seismicsignal noise removal based on FastICA. Then we proposed an algorithm for weak seismicsignal separation. Through processing of the model and real data, we got many quite idealresults. These results proved that the algorithm is reasonable and effective. The main contentsof this thesis are consists of the following parts:(1) Based on the analysis of Blind Deconvolution theories and algorithms, constructedthe objective functions by combining least-squares constraint with Cauchy criterion. Toimplement Seismic Blind Deconvolution, used perconditon conjugate gradients algorithm toiterative invert reflction cofficient and seismic Wavelet.(2) Mainly studied the application of linear and instantaneous mixed Blind Source Separation in seismic signal noise removal. Suppose the useful signals and noise are keepstatistical independence to each other, and they are not subject to Gaussian distribution.According to ICA theory, take two observations of the same seismic trace or observations oftwo adjacent traces and then the useful signals and noise can be separated by ICAprocessing.In order to solve the uncertainty inherent in ICA, proposed an algorithm based on correlationanalysis to determine the order and phase of the separated signals.(3) Investigated the separation algorithm of weak seismic signals based on Blind SourceSeparation. In the basic of analysising the theory of Blind Source Separation and the featuresof seismic reflection signals in subtle traps, constructed the BSS mixed model which issuitable to recover the weak signals in target reservoirs, then taken iteration algorithm toimplement the weak seismic signal’s recovery. The results of the model and real dataprocessing indicated the rationality and effectivity of the algorithm.