Desert Seismic Noise Suppression Based On Multimodal Low Rank Processing

As one of the main methods of modern geophysical exploration and one of the important methods in the field of oil and gas resources exploration,seismic exploration has always received close attention by scholars from various countries.In recent years.In recent years,with the increasing demand for oil and gas resources in human production and life,and the reduction of easily exploitable oil and gas resources,the exploitation of unconventional oil and gas resources has become a hot spot,which has made the application of seismic exploration more difficult.The smooth development of seismic exploration projects relies heavily on a good quality seismic record.Reducing the random noise of seismic exploration and improving its signal-to-noise ratio is a key step to improve the quality of seismic records.Therefore,looking for an effective seismic exploration random noise suppression method has far-reaching significance to improve the quality of seismic records.In recent years,some scholars from various countries have proposed a series of random noise reduction methods,such as f-x deconvolution,wavelet denoising algorithm,time-frequency peak filtering,empirical mode decomposition etc.These classical algorithms have been used in the field of seismic signal processing.Although all of them have achieved certain effects,their respective limitations make the processing effect for the desert seismic events unsatisfactory.Though f-x deconvolution is more common in engineering applications,it requires linear or local linearity of the reflected wave in the same axis,and it cannot suppress irregular noise.Moreover,when processing low signal-to-noise ratio data,there will be obvious errors.wavelet transform as a kind of the multi-scale time-frequency analysis method is widely used in seismic exploration,but its denoising effect depends largely on the choice of threshold,which limits the flexibility and applicability of the method application;time-frequency peak filtering can deal with low SNR data,but it is restricted by nonlinearization and window length selection.Traditional empirical modal decomposition method does not depend on threshold selection,it is flexible to use.However,the existence of problems such as modal aliasing and endpoint effects make the amplitude of the effective signal after denoising unsatisfactory.These existing traditional filtering techniques play a role in reducing the random noise of the middle and shallow seismic exploration and improving the signal-to-noise ratio of seismicrecords.However,in the face of the "three high" requirements of energy seismic exploration,complex and difficult surface and near-surface geological conditions,deeper and deeper and more complex exploration targets.The conventional seismic data denoising technology has gradually become difficult to adapt to the current high-precision exploration needs.In order to find a denoising method that is not limited by noise properties and thresholds.But can reduce the strong and complex random noise.This paper applies complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)algorithm combined with the Greedy GoDec algorithm and applied it to the reduction of random noise in the desert.The CEEMDAN decomposition algorithm was first proposed by Torres ME et al.at the 2011 ICASSP conference.To some extent,it overcomes the problem of modal aliasing in empirical mode decomposition.It is widely used in medical signal and bearing fault detection as soon as it was proposed,and achieves good results.Although the single CEEMDAN algorithm can also achieve the purpose of denoising,some modes need to be discarded in the process of noise reduction,which greatly reduces the amplitude of the effective signal.GreGoDec is a low rank method proposed at the AISTATS conference in 2013 and has strong feature recognition ability.However,in the face of the complex and variable desert seismic data,the single low rank method is not able to separate the effective signal and noise.Therefore,the ingenious combination of the two not only avoids the problem that the conventional denoising method is not applicable due to the low frequency and complex nature of random noise in the desert zone,but also does not need to discard the modal,which greatly improves the amplitude preservation effect of the effective signal.This paper applies this idea to desert seismic signal.Simulation and actual data processing also prove that it is indeed applicable to random noise in desert seismic signals.