| Seismic exploration is an important way to detect oil and gas resources.The collected seismic data can be converted into intuitive graphics or images by visualization technology.Improving the quality of data can enhance the visualization effect and make accurate geological interpretation.Aiming at the problems of noise interference and low resolution of seismic data,this thesis proposes seismic data optimization algorithm and super-resolution reconstruction algorithm,and integrates them into the data visualization system for seismic facies interpretation,which improves the effect of seismic data visualization.The main work of this thesis includes :(1)Aiming at the problem of seismic data noise,a data optimization algorithm based on autoencoder and similar graph loss is proposed.By improving the autoencoder,the mixed attention,feature enhancement,multi-scale feature extraction and PRes(Res based on PRe LU,PRes)modules are integrated to enhance the feature extraction ability of the model.A loss calculation method based on Interpolation-Random Sampling(I-RS)similarity graph is proposed.Noise2 Noise algorithm is introduced to solve the optimization problem of seismic data,and an I-RS similarity graph sampling method for seismic data is designed.The calculation loss of similarity graph obtained by I-RS method is used as a regularization term to enhance the robustness of the model.The seismic data with different levels of noise are effectively optimized.(2)Aiming at the problem of low resolution of seismic data,Edge-Enhanced Super-Resolution Generative Adversarial Networks(EESRGAN)is proposed.The generation network of EESRGAN consists of feature extraction reconstruction subnet and edge enhancement subnet.Feature extraction and reconstruction subnet efficiently extracts features and generates high-resolution data.Edge enhancement subnet enhances the real edge information.Finally,the high-resolution data with clear texture edges is obtained by fusing the edge information and one-stage reconstruction data.A loss function combination is proposed,including two-part pixel loss,confrontation loss and spatial loss based on total variation model.EESRGAN realizes high resolution seismic data reconstruction under different amplification factors.(3)A seismic data visualization system is designed and developed based on Py Qt Graph,and the proposed seismic data optimization algorithm and super-resolution reconstruction algorithm are integrated into the system.Through the system analysis and design,the data analysis,data visualization enhancement,two-dimensional visualization,three-dimensional visualization,cross-section interaction and graphic identification modules are realized,especially in the data visualization enhancement module to optimize and reconstruct high resolution seismic data,which improves the visual display effect of other modules and meets the needs of engineering applications. |
