Seismic Data Reverse-time Migration And GPU Parallel Algorithm

Seismic data reverse-time migration is a depth migration method which is basedon the vector wave theory. Its principle is conducting the reverse-time continuation onthe wave field recording by the geophone, then choosing appropriate image-formingcondition, and stacking to get the section of the reflecting interface. Reverse-timemigration has an advantage in the multi-wave and multi-component seismic dataprocessing than the regular migration methods. Pre-stacking reverse-time migration isbased on the two-way wave equation, therefore it would not be restricted by thechanging of the dip angle and the transverse velocity during the migration, antreverse-time migration can keep the vector feature of the wave field by which theresult can get higher precision image. Reverse-time migration need huge amount ofcomputation and it will bring low frequency noise during the computation, whichwould affect the practical application of reverse-time migration.This article starts with the theory of reverse-time migration. Use staggered gridhigh-order finite difference method to solve the one-order velocity-stress equation.Derived the reverse-time continuation scheme, use the perfect match layer to suppressthe boundary reflection, and then completed reverse continuation of the elastic waveequation.One of the key factors to affect the reverse-time result is the image-formingcondition. This article researched regular image-forming conditions, especially thecross-correlation image-forming condition.Another fault of reverse-time migration is the huge computation. This article useGPU high performance computing to solve this problem. Firstly researched theprinciple and the programming model of GPU parallel computing，then optimized theGPU parallel program, finally accomplished the parallel program of the elastic wavepre-stack reverse-time migration, and proved the result by model computation. Themodel computation shows:1) elastic wave pre-stack reverse-time migration has a favorable effect in imaging complex structure.2) the improved image-formingcondition have a better effect in suppress the noise.3) GPU parallel computation cangreatly enhance the computational efficiency of reverse-time migration, which couldpropel the widespread use of reverse-time migration.