Research On Seismic Wavelet Decomposition And Reconstruction Technology

With the development of national economy, the demand of resources isincreasing, so the oil and gas exploration target has been transformed from largegeological structures to complex subtle reservoirs, such as small fault blocks, thinlayers and thin interbed. We need more accurate strata information. At present,seismic wavelet is always assumed to be single and fixed in conventional seismicsignal processing, however, the seismic wavelet in the actual situation changes overtime and space, therefore, this assumption is a great gap between the actual situations,which may lead to loss of weak effective information, and sometimes may form someillusions. The exploration target in the past is large geological structures, so this errorcan be ignored. However, with further improve of the accuracy of seismic exploration,this limitation of conventional seismic trace model stands out. After using a variety ofreservoir prediction means, we found that the success rate did not improve, always inthe same position. If we want to improve the success rate of the reservoir prediction,we must radically improve the seismic trace model based on single waveletassumption. Therefore, seismic wavelet decomposition and reconstruction techniquebased on the multi-wavelet seismic trace model is adopted.Seismic wavelets are affected by many factors in the dissemination process, suchas source, the earth absorption filter, and field recording process and formation lateralchanges. Seismic wavelet with the shape performed in the records is a superpositionof many wavelets. A seismic trace can be superimposed by the convolutions of eachseismic wavelet and the corresponding reflection coefficient. This is a multi-waveletseismic trace model.Seismic wavelet decomposition is a technique which decomposes a seismic traceinto series of wavelets with different shapes (dominant frequency) and differentamplitude. Then according to different interpretation targets, select the best waveletswhich can express the changes and the distribution of the target, keep their positions, and re-superimpose them to form a new seismic trace. This process is called seismicwavelet reconstruction. The new synthetic seismic gathers are the maximum to reflectthe changes of the target.This paper studies the seismic wavelet decomposition and reconstructionalgorithm based on continuous wavelet transform, which decomposes a original signalinto a series of Morlet wavelets with different scales, which is also equivalent todifferent frequency due to exactly corresponding relationship between the scale andfrequency. Select all of the wavelets to reconstruct the original signal, select a specificpart of the wavelets to reconstruct part of the wavelets reconstructed signal. On thesynthetic records of decomposition and reconstruction, and the accuracy of thismethod is verified.And then the method of wavelet decomposition and reconstruction based on thematching pursuit algorithm is studied. On the basis of previous studies, in order toimprove the calculation efficiency instantaneous attribute is used to construct atomiclibrary, which makes the original search range greatly decreased and avoid blindsearch. And the scale parameter is set to free variable, which makes the atomiclibrary has a strong suitability. On the synthetic records of decomposition andreconstruction, and the accuracy of this method is verified. By comparing theprocessing effects of actual data, the wavelet decomposition and reconstruction basedon the matching pursuit is much better than the method based on continuous wavelettransform both in computation precision and efficiency.Application examples prove that the seismic wavelet decomposition andreconstruction technique has a good application effect in recognition of weakreservoir reflection, identifying faults and thin out, depicting thin layers and thereservoir prediction. Therefore, the seismic wavelet decomposition and reconstructiontechnology can provide an effective means for the interpretation of seismic data andreservoir prediction.