Adaptive Fission Particle Filter For Seismic Random Noise Attenuation

Seismic exploration is an important means to detect oil, gas and mineral resources. However,there exist a lot of noise in seismic data acquired from complex exploration environment, severely reducing the quality of seismic data and influencing the researchers to extract useful seismic information. Random noise suppression is an effective means to improve the signal to noise ratio(SNR) of seismic data and the quality of seismic data. Conventional de-noising methods in dealing with seismic data at low SNR are unsatisfactory. Therefore, proposing a filtering method that can effectively suppress seismic random noise at low SNR has a very important practical significance.The seismic state-space model can be described as a nonlinear system. The particle filter(PF) method is a common nonlinear estimating method under a Monte-Carlo simulation and recursive Bayesian framework. PF can use the nonlinear system for nonlinear transformation in the filtering process. Furthermore, PF is not limited to model characteristics and noise distribution, dealing with no nlinear problems more effectively. Therefore, we adopt PF for seismic signal denoising. However, when the seismic data at low SNR, the sample impoverishment problem in PF leads to incomplete seismic information contained by the particles. The particles cannot accurately represent the reflected signal, causing estimating errors. In order to maintain the particle diversity and obtain the vital seismic information, and improve the quality of the particle, we propose a method to suppress seismic random noise, namely adaptive fission particle filter(AFPF).In the novel method, we establish a state-space model for seismic records firstly. The time-varying autoregressive(TVAR) model is adopted to describe the characteristics of the seismic signal. Secondly, we propose the adaptive fission thought. Specifically, particles breed ―offspring‖ particles by adaptive fission. To implement the adaptation, we apply a fission factor to monitor the degree of fission. The fission factor takes into account weights that indicate the quality of the particles. This fission process not only retains the good features of high weighted particles, but also produces more new fine particles. Therefore, particle diversity is well maintained and the effective seismic information is obtained by particles. At the same time, the proportion of highly weighted particles is increased, thus improves the particle quality. As a result, the particles provide a better representation of the desired signal and reproduce the true signal more reliably, and the filtering accuracy gets improved. Finally, we discuss four resampling methods and select the optimal method.In order to verify the effectiveness of the proposed algorithm, we apply the proposed APPF to noisy synthetic seismic records with white gaussian noise and colored noise, respectively. Experimental results illustrate the superior performance of AFPF in noise attenuation and reflected signal preservation compared with the PF. At last, the proposed method is applied to real seismic data. Results show that the method can remove the background noise well, and recover events more clearly and more continuously, which proves the effectiveness of the proposed method in practical application.