Differential Semblance Optimization Based Migration Velocity Analysis

Wave equation Pre-stack depth migration(PSDM)is the mostly accepted method for imaging the complex structure because it can accurately depict the propagation of seismic waves in complex subsurface situation.In PDSM,the strong coupling between depth and velocity makes the migration method relies too heavily on velocity.How to obtain an accurate velocity,therefore,has always been the key and difficult problem in PSDM.The sensitivity of PSDM method to velocity restricts the quality of the image result.On the other hand,however,it suggest us to estimate the velocity using PSDM in which we optimization imaging result is considered the criterion to judge the validity of migration velocity.By building the quantitative relation between velocity error and image residual,an inversion problem is solved to obtain velocity perturbation for updating the image result.PSDM transforms the multi-offset redundant data observed at surface into redundant image result in image domain,for specific case,different types of image gathers by using corresponding extended image conditions.The velocity error is contained in the common image gathers as the focusing or the flatness of image gather directly reflects whether the velocity is correct.Wave Equation Migration Velocity Analysis(WEMVA)is based on the inversion method,which updates the velocity by optimizing the image result(a focused or flat image gathers).Different from RCA,WEMVA focuses more on wave equation itself.It decomposes wave equation through local linearizing approximate(small scale linearizing while large scale non-linearizing),then building objective function by analyzing the image residual and velocity perturbation and solving the inverse problem by gradient guided methods,minimizing the error function to obtain the best focused image result and updated velocity.At present,WEMVA can fall to Stack Power Maximization,Minimum Image Perturbation and Differential Semblance Optimization,according to the various criterions that objective functions are built by.In this thesis,we choose Differential Semblance Optimization(DSO)method as the criterion to update velocity and this method maintains better convexity and smoother gradient when there is larger velocity error from the true velocity model.What's more,DSO based MVA method does not need any picking,which makes DSO more attractive among other WEMVA methods.