The Research Of Multiple Attenuation In The Radon Domain Based On SVD

Multiple attenuation is always a hot and difficult problem in processing seismic signal. As the nature of primary significant wave and Multiple on seismic profile is very similar, it is difficult to accurately predict the Multiple of the profile; therefore, how to separate the Multiple is a difficult problem of seismic signal processing. The quality of Multiple suppression not only affects the subsequent processing of seismic data (such as velocity analysis, migration imaging, etc), but also directly affects the interpretation of seismic profile (producing false structures, etc).Over the years, geophysicists at home and abroad have made unremitting efforts to solve the problem of Multiple suppression, proposed a variety of methods to suppress Multiple. Generally they can be divided into two types: one is the filtering method based on the differ of the characteristics and nature of the primary significant wave and Multiple,;the other is predicting subtraction based on wave equations. Although these traditional methods are practical, but there are also some disadvantages,such as the Radon transform, in order to effectively suppress Multiple, requiring the energy of significant wave and Multiple in the p / q domain can be accurately separated, however, when the real seismic data transformed to the Radon domain, there is serious "tail" phenomenon, which gives a great influence on Multiple attenuation. Because the real seismic signal is complicated, there are some overlap section of same phase axis of effective signal and Multiple. However many traditional methods have a generalize deficiency of bad removing Multiple effect in same phase axis overlap section of effective signal and Multiple. A method using a filter function in High Precision Parabolic Radon domain to suppress Multiple is presented in this paper. it improved some traditional way of multiple attenuation based on Radon Transform, Compared with these traditional ways of multiple attenuation, This method overcomes the phenomenon of Radon domain with bad tail, and it obtained a good effect with a high precision. The process of this method in this paper is roughly divided into the following steps: firstly, forecasting Multiple model from partial NMO seismic record by using SVD(singular value decomposition) technique, we make the Multiple obtained by SVD as the initial Multiple model, In fact using such prediction, the Multiple contains effective signal, while effective signal also contains multiple. secondly, transforming original seismic records and prediction model of the initial Multiple to the parabolic Radon domain by using high resolution parabolic Radon transform. lastly, suppressing Multiple by using a filter function in the high resolution parabolic Radon domain. For some real seismic signal is more complicated, we can consider introducing the repetitive filtering thoughts. making Multiple obtained by the last filtering processing as the next time filtering initial Multiple model, the processing effect will be better, but the computation will increase, whether the repeating filtering should be made, it is decided by the preceding treatment effect.Through the process of synthetic signal and real seismic signal, we find that this method can suppress Multiple effectively, meanwhile, it has a good performance on keeping the amplitude of effective signal, lastly, it can also remove the random noise in the profile after NMO and has a good performance in the area that the overlap of primary significant wave and Multiple.