Noise Attenuation And Weak Signal Improvement Method Research On High-density Seismic Data

High-density seismic data are rich in signal frequency bandwidth and wave field information,and they can reflect the underground information more truly.However,they also record a large amount of noise so that the signal-to-noise ratio(SNR)is low and weak signals are always mixed in noise.The traditional filtering method often suppresses the weak effective signal while suppressing the noise,which makes the potential advantages of highdensity data negated.Therefore,the noise suppression and weak signal Improvement methods for high-density data need to be studied,and the information accuracy of seismic data is better by improving the SNR.It is of great significance to mountainous exploration,reservoir fine description and identification of thin reservoir,small fault,sand body or porous structure.In the aspect of random noise suppression of seismic data,three different denoising algorithms have been studied for the purpose of reducing the damage to weak signals during denoising.The proposed Radon domain and the curvelet domain denoising algorithm can filter significant noise by large-scale filtering in the Radon domain,and then gradually subdivided.The “multi-domain” and “multi-scale” processing is used to reduce the loss of weak signals.The K-SVD denoising method based on sparse representation adopts an adaptive learning dictionary,which is more suitable for the data itself than the deterministic basis function of the curvelet transform,and thus has a superior data representation capability.The 3D block matching denoising method based on non-local algorithm utilizes the selfsimilarity of data to protect the effective signal,especially to maintain edge features such as fractures,avoiding artifact interference,and has a very good denoising effect.In the aspect of weak signal enhancement,compressed sensing theory and its sparse promotion inversion algorithm have been studied,and non-local algorithms are introduced into compressed sensing to improve the reconstruction effect by make full use of the similarity of data sub-blocks.Combining them with enhancement functions can enhance the amplitude of weak signals and improve the visual display of weak events.