A Data Driven Approach For Time Migration Of Reflection Seismic Data

The basic idea of data-driven migration is to calculate the horizontal position and vertical travel time of the imaging points based on the slope of the traveltime and directly map the seismic data to the imaging space.Because the slope of events is a geometric attribute of the seismic observation data,it can be estimated directly,so the velocity model is not needed.Compared with the traditional model-based migration method,in the process of migration,it has the characteristics of velocity-independent,that is to say,this method gets rid of the dependence of traditional migration methods on velocity models.It has the characteristics of flexible and efficient.However,when the SNR is not high,it is not easy to estimate the slope of events,and it is difficult to guarantee its accuracy.Therefore,the data-driven migration effect is not good at low SNR.Therefore,in order to play the advantage of data-driven direct mapping with considerable computational efficiency,we need to improve the existing estimation method of slope,so that it can also achieve good results under low SNR conditions.In other words,how to accurately estimate the slope of the seismic events under the condition of low SNR is the key to the practical time migration of data-driven.In order to improve the effect of data-driven time migration imaging under low SNR,the method of data-driven estimation is improved in this paper.Due to the lack of both efficiency and accuracy in the calculation of slope,the adaptive weighted generalized inverse vector directional filtering method is proposed to calculate the slope,which is a vector filtering method.On the basis of ensuring the accurate calculation of the slope property,the calculation efficiency is improved.In order to contain noise conditions,the accurate estimation of slope,this paper improved the diffusion tensor,the design of a new continuous operator is applied to the diffusion tensor,not only suppress noise,but also in the event of the fault to ensure slope estimate surely strengthen the coherence enhancement of seismic image texture.The accuracy of the slope of the calculated data is improved.The time migration based on data-driven is calculated in a dual domain: a common-shot-common receiver point or a common center point-common offset domain.In this case,we need to carry out the collection of the shots.It can also be calculated in a single domain: a common-shot field.The direct migration of a common-shot field does not require separate the shot data.In the calculation of data-driven time migration in the common-shot domain,we need to calculate the first derivative of the slope on horizontal direction,but this calculation result is not stable.In this paper,the offset formula of common shot gather is improved.By introducing predictive painting technology,the estimated data can be converted into the equivalent slowness value by event slopes,and it can be directly generated into the calculation of the first derivative of the migration formula of the shot domain without calculating the slope.In order to verify the correctness and calculation effect of the prestack time migration method based on data-driven,several numerical models and field seismic data are processed.The test results show that the data-driven slope calculation is stable,accurate and efficient.Data-driven prestack time migration method has strong imaging ability and strong adaptability to seismic data.Because the migration time of this method is almost negligible,it can be used as a prior detection method in seismic data processing,and has great potential in real-time seismic imaging applications.