Research On Seismic Attributes Fusion Method Based On Multi-resolution Analysis And Its Application

Currently, many methods of seismic attribute forecast of oil and gas are limited to using a single seismic attribute to predict reservoir. These methods usually cause a serious multi-solution problem, and make the reservoir prediction lack of credibility. The problem of multi-solution can be solved by multi-attribute fusion technology. MRA(Multi-Resolution Analysis) is a time-frequency analysis tool with self-adaptability, multi-directionality and localized time-frequency characteristics, which is widely used in the field of science. Therefore, there is a very important practical significance in researching on multi-attribute fusion technology and application. In this thesis, on the basis of multi-resolution analysis, a deep study of NSCT(nonsubsampled Contourlet transform) and Shearlet transform is made, and methods of seismic multi-attributes fusion based on multi-resolution analysis are proposed, which improves the accuracy of reservoir prediction greatly and will pay great effect on the future research work. The major work and results are as follows:(1) Research the basic theory of sparse representation, and make a deep study of multi-resolution analysis, also present the basic properties and theory of usual X-let transform.(2) Research the basically geological features, characteristics and classification of seismic attributes, and propose two kinds of preprocessing methods, including neighbor substitution method and range normalization method. And we analyze the correlation between seismic attributes.(3) On the basis of multi-resolution analysis, and through deep research on adaptive fusion rules, we propose the SML(sum modified Laplacian) fusion rule which is used for the fusion of decomposed low-frequency sub-bands, and the multi-scale products rule for high-frequency sub-bands which obtains the weighted coefficients by the correlation of each scale coefficients. We have used practical seismic multi-attributes data to test several experiments, and verify the validity and advantages of the two fusion rules.(4) Derive the construction procedures of resampling matrix and quincunx filter bank of the Contourlet transform. Analyze the theory of Contourlet transform and its rapid implementation process. A method for seismic multi-attribute fusion based on Contourlet transform is presented, and the software module is deeloped.(5) In order to solve the offset problem which is caused by sampling process of Contourlet transform, we propose a new method for seismic multi-attribute fusion based on NSCT(Nonsubsampled Contourlet Transform). And some experiments and analyses have been made with practical data.(6) In order to solve the problem of complexity and large calculating quantity of NSCT, we propose a novel method for seismic multi-attribute fusion based on Shearlet transform, which avoids the sampling procedure, and uses shear filter for directional decomposition directly. The fusion algorithm improves the efficiency.Experimental results show that the proposed methods are superior to other fusion methods. Practical data application indicates that the NSCT fusion method can retain effective information, and the Shearlet fusion method can eliminate the noise and the redundancy of attributes, which can provide reliable basis for reservoir prediction.