Seismic Random Noise Attenuation By Structure-Oriented TFPF Algorithm

Oil and gas play an im portant role in the national econom ic development,but they are the non-renewable resources.So exploring the unknown under ground structure has become imminently.Seismic exploration has developed quickly as an effective means for searching for the under ground oil and gas.Collecting seism ic exploration data often m ixes with st rong random noise,which reduces the signal-to-noise ratio(SNR)of the data.Therefore,it is d ifficult to identify effective signals.Suppression of random noise in se ismic data and improvem ent the SNR of seismic data are particularly important.Time-frequency peak filtering(TFPF)can reco ver signal effectively under low SNR.The unbiased es timation condition of this algorithm is that the s ignal is linear.In order to m eet this condition,we take use of the seismic data lateral continuity to construct temporal-spatial trace to accom plish the tem poral-spatial TFPF filtering.Due to the trace of tem poral-spatial TFPF and the seismic reflection events are in the same direction.Thus,the signal of f iltering is approxim ate linear.So the temporal-spatial TFPF can keep signal amplitude well and suppress the strong random noise.However,temporal-spatial time-frequency peak filtering us ually adopt single trace to deal with the seism ic data.And the the seism ic reflection events of actual seismic exploration data are often complex.Thus,with fixed trace m odels and parameters,temporal-spatial time-frequency peak filtering can't recover the ir regular seismic reflection events successfully and can not preserve the im portant structural features,such as the edges and end points,and so on.The refore,it is necessary to establish the corresponding trace model according to the seism ic reflection even ts.And adjust the window function of TFPF on the basis of the features of seism ic reflection events.To solve the problem of temporal-spatial TFPF,the structu re-oriented time-frequency peak filtering(SO-TFPF)algorithm is proposed to suppress random noise of seismic data in this pape r.Combined with the s mallest univalue segment assimilating nucleus(SUSAN),the proposed algorithm can measure the structures of the seismic data,and recognize the features edges and end points of seismic reflection events according to the m easured structures,then build th e temporal-spatial trace,which is consistent with seismic reflection events.Then resampled the signal alon g the trace to accomplish the temporal-spatial time-frequency peak filtering.Because of the linear degree of the resam ped signal is hi gher,so it is benefi cial to preserve the structures of signal,which gain of SO-TFPF.In addition,the choosing m ethod of structure-oriented windows is p roposed.In signal area,to ensure th e accuracy of signal recovery we used traditional symm etric short window.At the e nd points of seismic reflection events,we used the asymmetric window,which filters along the seismic reflection events to avoid the end points slurry.In the noise area,we used symmetric long window,which along the time direction,because this type of window can suppress noise effectively and avoid the influence of the spatial coherence of the noise.We applied this m ethod to the synthe tic seismic records and the field seism ic data.The results of the experim ents shows that th e algorithm of this pape r can recognized the seismic reflection events successfully.Compared with the radial-trace time frequency peak filtering(R T-TFPF),SO-TFPF can suppress rand om noise and enhance the constant of seism ic reflection events,at the sa me time it also can avoid the blur at the end points.