The Time-frenquency Peak Filtering Based On Radon Transform And Its Applications In Seismic Noise Attenuation

The seismic exploration is a way to study the structure of the earth’s interior,which can help us know if there are gas and oil inside the earth. This method oftenapplies an impulse signal as input to the ground, and then we collect the reflectionsignal with seismic sensor, from which we can estimate the property of the interfaceand the structure of the geological layer. However, due to the over-exploitation, wecan not explore the gas and oil in shallow surface, so the exploration develops intodeeper layer. But the longer the signal transmits, the more serious the signal isattenuated, and the signal is even submerged in noise. So it is important to reduce thenoise in the seismic record.Time-frequency peak filtering (TFPF), a new de-noising method, is effective tosuppress the random noise in seismic record and arouses the concern of geophysicists.The conventional TFPF filters seismic data only along the time, ignoring the spatialcharacteristics of the reflection events, which results in the loss of directionalinformation of the signal. For another thing, the method uses a fixed window length(WL). A short WL can not suppress the random noise effectively. Contrarily, a longWL leads to a comparatively serious loss of valid signal. In order to improve theperformance of TFPF, we propose spatiotemporal2-D TFPF, which is doing TFPF inhigh-resolution Radon domain with varying-window-length. The events can be highlyfocused to energy points in different positions of the high-resolution Radon domain,and they are easy to identify. Then we can set a suitable threshold to distinguish thesignal parts and the noise parts. Next, we process the signal parts by TFPF with shortWL and process the noise parts by TFPF with long WL. By virtue of this new method,we can preserve the valid signal better and suppress the random noise moreeffectively.However, when processing the high-frequency signal the method metioned abovedoes not have advantage in signal amplitude preservation. Then we propose a newmethod to solve that problem. We separate the events with different curvatures intosome sub-records applying Radon transform. Then we select optimal sampling tracesfor each sub-record and filter the data using TFPF. The new method lowers thefrequency of the signal and further reduces the error brought by TFPF. The SNR of the filtered record by our method is improved compared with conventional TFPF.Through experiments on the synthetic seismic records and the field seismic data, thenovel method possesses a superior performance in random noise attenuation andseismic events preservation compared with the conventional TFPF methods.