Spatiotemporal Time-varying-trace TFPF Based On SW Test For Seismic Random Noise Attenuation

Oil and gas resources are closely related to national economic development andpeople’s life. Seismic exploration is an important means of analysis the unknowngeographical structure of underground and exploring the unknown oil and gasresources. However, the surface environment and underground structure is verycomplex, which seriously reduces on the signal to noise ratio (SNR) of seismic dataand makes the processing result of seismic data not accurate. As a result, we can’texact position of the oil and gas resources, and wasted of manpower and materialresources. Therefore, effectively improve the SNR of the seismic data is a challengingproblem in seismic exploration. In order to break through this difficulty, we need toput forward better method, which can both clearly preserve the effective signal andsuccessfully suppress the random noise.Time frequency peak filtering (TFPF) has gained great progress in preserve theseismic reflection events for seismic records, when SNR is below a certain threshold.However, TFPF has two conditions for unbiased estimation. First, signals should beapproximate linear. Second, noise should be Gaussian noise. Seismic signal is verycomplex and does not always satisfy the linear condition of the unbiased estimation ofTFPF. Therefore, we adopt the Wigner-Ville distribution of a time window to improvethe linearity of the signal, and reduce the bias. But a single window length (WL) can’tget balance in the noise suppression and signal linearity increase. Short windowlength can effectively improve the linearity of the signal and preserve signal, butcannot successfully attenuate the random noise. On the contrary, long WL cansuccessfully attenuate the random noise, but can’t outperforms in improve thelinearity of the signal and preserve signal. Considering preserving signal and seismicnoise reduction, the paper propose a novel algorithm that time-varying-trace TFPF(TVT-TFPF) based on Shapiro–Francia test and Shapiro–Wilk test (SW test).In application of the proposed method for seismic random noise attenuation,firstly, we construct the time-varying filtering traces, which match the bend seismicreflection events very well. Because of the lateral continuity of the seismic reflectionevents, the seismic signal re-sampled along the time-varying filtering traces have the same amplitude. And the linearity of the signal is effectively improved. As a result, wesolve the problem that the TFPF is biased for nonlinear seismic signal. Secondly, wecombine with the SW test in statistics in order to extract the time-varying filteringtraces. The main idea is: we detect the different bend of seismic reflection eventsbased on the difference of values of SW statistic between seismic random noise andeffective seismic signals. And then we extract the time-varying filtering traces. Finally,we re-sampled the seismic data along the time-varying traces, and application ofTFPF processing the re-sampled data. So we can improve the signal recoveryprecision.In order to test the effectiveness of the proposed method in this paper, we apply itboth on the synthetic seismic records and the field seismic data. Moreover, wecompare SNR and signal amplitude of the filtering result with TFPF and RT-TFPF.The results show that the proposed method outperforms TFPF and RT-TFPF. So itdemonstrates that the proposed method can better restore the details of the effectivesignal and enhance the continuity of the event. At the same time, it can suppressrandom noise.