Study On Application Of S Transform Template Filtering Technology In Active Source Data De-Noising

Whether in seismology or exploration seismology, we are facing the problem of increasing the signal-to-noise ratio (SNR) of seismic data when we describe the underground construction and the variations of the underground medium. People usually choose filtering method according to the differences (such as apparent velocity, frequency, etc.) between signal and noise. However, different filtering methods have their application conditions; they could achieve good effect only when seismic records meet the conditions. In recent years, using various sources to research the variations of underground medium becomes a hot issue in the development of seismology. In this article, we develop relevant filtering method according to the characteristics of the active source data.This thesis describes the basic theory of S transform, and its formula derivation and characteristics in detail. S transform is a combination of short-time Fourier transform and continuous wavelet transform, which not only combines the advantages of the two methods, but also avoids the shortage of them. Its time-frequency resolution is related to the signal frequency, and basic wavelet does not have to meet the permit conditions. S transform and Fourier transform have an instinct link because that S inverse transform is Fourier transform. In practice, we can transform the signal from the time domain to time-frequency domain, and then transform to frequency domain, finally transform it to the time domain. It provides a rapid, nondestructive reversible transformation without any loss of information. S transform with multiple resolutions overcomes the fixed resolution defects of the short time Fourier transform. It has the phase factor which wavelet transform lacks and reserves the absolute phase characteristics of each frequency. S transform is a linear transformation with high time-frequency resolution, and it has no cross-terms compared to bilinear transformation such as the Wigner-Ville distribution and Cohen classes. Generalized S transforms, with higher frequency aggregation ability, are proposed by scientists to make up for some deficiencies of S transform. We applied S transform into time-frequency domain de-noising based on the unique advantages of S transform to improve the SNR of seismic data. We analyze the distribution of the signal and noise over time in the time-frequency domain after transforming the signal from the time domain to the time-frequency domain and then design an appropriate filter to remove the noise and interference from the signal. Finally we transform the signal after filtering to the time domain by inverse transform.Active source detection technology is a promising technique to detect deep earth structures. The airgun source, developed from exploration seismology, breaks though some limitations of natural earthquakes, and provides a good technical platform for active detection of underground structures. Compared to some traditional artificial sources, airgun source has high signal-to-noise ratio, accurate positioning, measurable source characteristics and the advantages of low cost. Airgun source is rich in low frequency components, which makes it suitable to detect deep crust. It has precise firing time because of GPS timing. Airgun signal can propagate as far as one hundred kilometers from the source. We can stack the source to improve SNR based on the repeatability of the airgun source. Although multiple stacking of seismic data can improve the SNR, in order to improve the SNR of single-shot air gun signal, we designed a filter based on S transform time-frequency filtering.This paper proposes a new S transform template filtering method. According to the method, we stack the airgun signals received from the same station to obtain a signal with high SNR, and then design a filter in time-frequency domain based on the template from the stacked signal to de-noise each airgun signal. We demonstrate the feasibility and practicality of our method by applying the S transform template filtering technology to the simulated data processing. Then we process the real airgun data that received from the station 112km from the source, and obtain single-shot airgun signal with high SNR. Compared with the results of the band pass filtering and wavelet filtering, the S transform template filtering can not only suppress the noise effectively but also not weaken the effective signal. The high waveform similarity of the filtered signal and stacked signal also proves the effectiveness of our method.