The Deblending Methods For Simultaneous Source Seismic Data

The simultaneous source technology,which allows more than one source firing almost at the same time and receiving the responses from the multiple sources,breaks the limit of the traditional seismic data acquisition.When the survey time is fixed,simultaneous source acquisition increases the number of fired sources,which can enhance the fold in order to improve the illumination quality of the underground structures.When the fold is fixed,this technology can make more than one source placed at different positions fire at the same time,which can greatly reduce the survey time in order to enhance the acquisition efficiency.Till now,the simultaneous source acquisition technology has attracted much attention from academia and industry.Different from conventional seismic data,simultaneous source data contain intense overlap,which compromises the field application of this technology.The reason is that conventional seismic data processing cannot well process simultaneous source data.There are two main ways to deal with the blended data.The first method is direct imaging,which migrates the blended data with some additional constraints to suppress the blended interference.The second method is so-called deblending,which first separates the blended data into individual records as acquired by conventional seismic acquisition,then processes the deblended data using the traditional seismic data processing workflow.In simultaneous source acquisition,the dithering code is usually implemented to gather the blended data.So,in some domains,such as common receiver domain,common offset domain,common midpoint domain,the record from the main source is coherent,the records from other sources are incoherent.This paper proposes a deblending method that is based on the sparse constraint.Due to the coherency difference,useful signals and blended interferences have different coefficients in a transform domain,thus a thresholding function can suppress those small coefficients,which usually correspond to the blended interference,to separate the blended record.Since the blended interference will increase the rank of Hankel matrix constructed from monochromatic frequency.According to this property,this paper proposes a deblending method based on rank-reduction constraint.The advantage of the proposed method is that it does not require windowing and can be easily implemented,what's more,the optimal rank is easy to determine.Combined with a thresholding function,the method can better separate the blended data.Most deblending methods are based on the coherency difference,however,when the coherency difference is small,those methods may fail.So,this paper proposes another deblending method based on the dip difference that can overcome the dependence of dithering code.The deblending problem is viewed as an inversion problem,which can be solved by the least-squares sense.Usually,the seismic data are contaminated by random noise,which prevents the deblending of simultaneous source data.In order to separate very noisy blended data,this paper proposes a hybrid constraint deblending method,which combines the dictionary learning and sparse constraints.The proposed method can well suppress random noise and separate blended data simultaneously.