Three dimension pre-stack surface-related multiple prediction based on nonlinear KNN algorithm for data search
Surface-related multiple prediction(SRMP)is an important part of surface-related multiple elimina-tion(SRME)and imaging.Although SRME technology is effective,it requires a regular and dense seismic data acquisition in theory.However,the spatial distribution of actual shot points and receiver points is sparse,and the seismic data fails to meet the requirements of SRME.The conventional method is to regularize seismic data before SRME.To avoiding the pre-SRME data interpolation,3D pre-stack data has been organized by a K dimensional index data tree,then a nonlinear K-nearest neighbor(KNN)algorithm comprehensively utilizes spa-tial location coordinates of sources and receivers,offset,azimuth to search from field data an approximate trace in real time,which is closest to an ideal trace.After that,a partial normal move-out correction is used to correct travel time difference because of the offset difference between the approximate trace and the ideal trace.Through the above two steps,the two seismic traces related with any downward reflection point(DRP)in the aperture of a single trace can be obtained and be used in SRMP.By convolution the two traces and stacking the results of all DRPs in the aperture of a single trace,a stable multiple model for that trace can be obtained.The method has been proven effective by testing on synthetic data of a modified 3D Pluto model and field seismic data in Northwest China.
surface-related multiplepredictioneliminationindex data treenonlinear K-nearest neighbor
国家科技期刊平台
NSTL
万方数据
