Self-adaptive Radial-trace TFPF For Seismic Random Noise Attenuation

Seismic exploration technology is a main mean of the exploration of stratum structure and mineral resources such as oil, natural gas. It mainly uses artificial technology to blast seismic wave. According to the reflected signals received by detectors, we can determine whether seismic record contains mineral resources. But in the process of collecting signals, because of the complexity of underground structure and the inevitable with a large number of human and environmental noise, it can severely reduce the signal-to-noise ratio(SNR) of seismic data and influence the interpretation work of researchers. Therefore, how to effectively suppress random noise in seismic data and improve the signal-to-noise of the seismic data is very important.Time-frequency peak filtering(TFPF) can effectively suppress random noise of seismic data in time and frequency domain. The condition for the unbiased estimation is that the signal is linear and it is submerged in the Gaussian white noise. But in the actual seismic data, the signal is usually nonlinear. So TFPF use the Pseudo Wigner-Ville distribution(PWVD), a version of windowed Wigner-Ville distribution(WVD) to realize the local linearity and reduce the estimation error. However, a fixed window is very difficult to take both suppressing random noise and effective signal recovery into account. If select a short window, the linearity of the signal in the window will be high, and it is close to the unbiased condition. However, short window is bad for denoising and it results that effective signals are still covered by noise. If select a long window, the linearity of signals in the window is low. In this case, although the random noise is removed well, the amplitude of effective signals decays seriously. And the TFPF filters along the time without considering the correlation between seismic channels, which is bad for signal recovery. So synthetically considering the random noise suppression and the effective signal recovery, we propose the Self-adaptive Radial-trace TFPF for seismic random noise attenuation.The Self-adaptive Radial-trace TFPF fully considers the correlation between channels and the actual direction of reflected event. Because of the locally linear of reflected event, we build the local scanning model and innovatively build the self-adaptive function. The new method finds the direction of the reflected event in the local area by utilizing the self-adaptive function and then maps the event to radial-trace area along the direction by radial-trace transform(RTT) which can improve the linearity of the signal. We do TFPF along the direction of the event. Consequently, it is more close to the unbiased TFPF and can reduce the error caused by the low linearity and the window is not the factor that can influence the filtering result. The processing results of synthetic seismic records show that the method in this paper performs better in suppressing random noise and signal recovery than traditional TFPF and Radial-trace TFPF. Comparing with Local parallel radial-trace TFPF, the method in this paper can avoid the misjudgment which can make effective signals distortional so the details of signals can be recovered better. When applied to the real seismic data, the method in this paper also has obvious advantages than the other three algorithms in deniosing and signal recovery and the reflected events are more clear and coherent after filtering.