Study Of Constrained Inversion Methods For Geophysical Imaging And Medical Imaging

The land and shallow marine sesimic data processing often encounters the challenges of the near surface complexity.Nonlinear methods for near-surface velocity model building such as first-arrival taveltime tomography and full waveform inversion have problems about multiple solutions.Similar to geophysical imaging methods,some medical imaging methods such as electrical impedance tomography are also highly ill-posed.Constrained inversion methods have been widely applied to geophysical and medical imaging to solve the multi-solution problems.For near-surface velocity model building,first-arrival traveltime tomography cannot image complex structures such as velocity reversal and thin bed because of ray assumptions and the limited traveltime information.The calculated statics from the result of traveltime tomography may not be sufficiently accurate.Tomographic inversion with statics(TIS)method has been developed to solve jointly for a velocity model and short static corrections.Statics-optimized traveltime tomography(SOTT)produces a velocity model that ensures an effective long-wavelength statics solution at the same time.We combine the two methods to produce a velocity model with both long-and short-wavelength statics accounted for,and develop a first-arrival traveltime tomography with long-and short-wavelength statics constraints.Applying the TIS method,we first obtain the short-wavelength statics solution and apply that to correct the picked traveltimes and then we invert corrected traveltimes with long-wavelength statics optimization for constraints.The final velocity results fit the data corrected by short-wavelength statics and also ensure effectiveness of the long-wavelength statics solution at the same time.We apply the new method to synthetics and real data.The results show our method has the improvement in velocity model building and statics calculation.Regularization have been applied to reduce the ill-posed problems of the inversion as a constrained inversion method.Tikhonov regularization is commonly employed to first-arrival traveltime tomography.However,the spatial resolution of the model solutions may be decreased due to the smooth effect of the regularization.In this study,we develop a constrained first-arrival traveltime tomography based on the Generative Adversarial Network(GAN).The GAN is applied to generate a prior model.We make the assumption that both the GAN model and the solutions of traveltime tomography must be sensing the same cross-gradients relationship.The prior model generated by GAN is incorporated as a structural constraint by a modified cross-gradient function in the objective function.The performance of the new method is tested with both synthetics and a real dataset acquired from Daqing,China.Compated with first-arrival traveltime tomography,the new method sharpens velocity interfaces and improves the spatial resolution of the velocity model.In the stacking result by using our method,the lateral continuity of the reflectors also has improvements.Waveform inversion should allow complex structures to be resolved;however,due to the comlexity of the near-surface wave filed,waveform inversion may tend to cycle skipping and fall into local minima.The Chuandong structural belt is located in the eastern of Sichuan,China.The belt consits of ejective folds with high dip angle.The rugged topography and large near-surface velocity variations challenge the near-surface velocity model building in the Chuandong structural belt.We apply the joint travetime and waveform inversion to image the Chuandong structural belt.The results show that the traditional waveform inversion produces artifacts near the rugged topography beacause of the cycle skipping issues.The joint invetsion can resolve the complex near-surface structures and constrain the velocities near the rugged topography.Computed tomography(CT)can accurately reconstruct the density structure of the region being scanned for medical imaging.Electrical impedance tomography(EIT)can detect electrical impedance abnormalities to which CT scans may be insensitive,but the poor spatial resolution of EIT is a major concern for medical applications.A cross-gradient method has been introduced for oil and gas exploration to jointly invert multiple geophysical datasets associated with different medium properties in the same geological structure.In this study,we develop a CT image-guided EIT(CEIT)method.The cross-gradient relationship between the CT image and the electrical conductivity distribution is applied to constrain the EIT.The cross-gradient based method allows the mutual structures of different physical models to be referenced without directly affecting the polarity and amplitude of each model during the inversion.We apply the CEIT method to both numerical simulations and phantom experiments.The effectiveness of CEIT is demonstrated in comparison with conventional EIT.The comparison shows that the CEIT method is robust to noise and can significantly improve the quality of conductivity images.We apply the constrained inversion method to geophysical imaging and medical imaging.The ill-posed and multi-solution problems are improved by applying the constraint of prior information during the inversion process.The accuracy of the imaging results is also improved.