S变换模板滤波及其在主动源数据去噪中的应用研究

发布时间：2018-05-31 22:24

本文选题：S变换 + 主动源　；参考：《中国地震局地球物理研究所》2015年硕士论文

【摘要】：无论是天然地震学还是勘探地震学,要实现对地下介质变化与地下构造的精确刻画,我们都面临着提高地震资料信噪比的难题。人们通常根据信号与噪声在时间或者空间方面的差异来确定选取合适的滤波方法,以提高地震资料的信噪比。但是各种滤波方法都有其使用条件,通常只有满足某种滤波条件的地震资料,才能取得良好的去噪效果。近年来,利用各种震源研究地下介质变化成为地震学发展的一个热点问题。本文中,我们针对主动源数据的特点,发展了相关的滤波方法。本文详细介绍了S变换的基本理论,并对其公式推导及特性做了详细阐述。S变换是短时傅里叶变换和连续小波变换的结合,它的基本小波不必满足容许性条件,时频分辨率与频率有关,并且反变换与傅里叶变换有直接的联系。在实际应用中我们可以把信号从时间域转化到时频域,再变换到频率域,最后转化到时间域,其变换快速且无损可逆,没有信息的丢失。S变换克服了短时傅里叶变换固定分辨率的缺陷,具有多种分辨率。它相当于对小波变换的进行了相位校正,具有小波变换没有的相位因子,保留了每个频率所对应的相位信息。S变换又是线性变换,和Wigner-Ville分布和Cohen类等双线性变换比较,不会存在交叉项的干扰,具有较高的时频分辨率。但是S变换的基本小波固定,因此许多学者对S变换进行了改进,得到了各种形式的广义S变换。为了提高地震资料的信噪比,我们基于S变换的独特优势,将其应用于时频域的滤波去噪。通过把信号从时间域变换到时频域,在时频域分析信号和噪声随时间的分布情况,设计合适的滤波器保留有效信号滤除噪声和干扰,再反变换到时间域从而达到信噪分离,提高地震资料的信噪比。主动源探测技术是一种很有发展前景的地球深部构造探测技术,从勘探地震学发展出来的气枪震源,为人们突破天然地震的一些限制,实现对地下结构进行主动探测提供了很好的技术平台。相较于一些传统的人工震源,气枪震源具有高信噪比、精确定位、震源特性可测量和低成本等优势。气枪震源的低频成分丰富,适合探测地球深部地壳；并且它由GPS授时,激发时间精确；气枪信号传播距离远,最远在离气枪源百公里内可以发现很强的气枪信号；而气枪源的高度可重复性可以让我们通过源的叠加来提高信噪比。虽然地震资料的多次叠加可以提高地震信号的信噪比,但是我们希望结合S变换时频滤波设计一种专门处理主动源气枪信号的滤波方法,用来提高单次激发气枪信号的信噪比。在本文中我们提出了一种S变换模板滤波去噪新方法,通过对同一台站接收到的气枪信号进行多次叠加从而得到高信噪比的信号,然后用叠加的高信噪比信号作为模板,在时频域根据模板设置滤波器并对每一次激发产生的气枪信号进行滤波去噪。通过应用S变换模板滤波技术对模拟数据的处理,验证了我们方法的可行性和实用性。然后又对距离气枪源112km的台站接收到的实际气枪数据进行了处理,得到了高信噪比的单次激发气枪信号,并且与带通滤波和小波变换滤波结果做了简单比较,说明了S变换模板滤波不仅能有效压制噪声并且不会削弱有效信号,而且经过滤波的单次激发气枪信号和叠加信号的波形相似性也很高,证明了我们方法的有效性。
[Abstract]:Whether it is natural seismology or exploration seismology, in order to realize the accurate characterization of underground medium change and underground structure, we all face the problem of improving the signal to noise ratio of seismic data. People usually determine the suitable filtering method according to the difference of time or space between signal and noise, in order to improve the signal to noise of seismic data. However, all kinds of filtering methods have their use conditions, usually only the seismic data that satisfy some filter conditions can obtain good denoising effect. In recent years, using various sources to study the change of underground media has become a hot issue in the development of seismology. In this paper, we have developed the related filter for the characteristics of the active source data. In this paper, the basic theory of S transformation is introduced in detail, and the derivation and characteristics of the formula are described in detail. The combination of.S transformation is the combination of short time Fu Liye transform and continuous wavelet transform. The basic wavelet does not have to satisfy admissibility, the time frequency resolution is related to the frequency, and the inverse transform is directly related to the Fu Liye transform. In practical applications, we can transform the signal from time domain to time domain, then transform it to frequency domain, and then transform it into time domain. Its transformation is fast and lossless. The loss of.S transform without information can overcome the defect of fixed resolution of short time Fourier transform. It has a variety of discrimination rate. It is equivalent to the phase correction of wavelet transform. It has the phase factor that the wavelet transform does not have, preserving the phase information.S transformation corresponding to each frequency and the linear transformation. Compared with the bilinear transformation of the Wigner-Ville and Cohen classes, there is no interference in the cross term and has a higher time-frequency resolution. But the basic wavelet transform of the S transformation is fixed, so many scholars have the S transformation In order to improve the signal-to-noise ratio of seismic data, based on the unique advantage of S transform, we apply it to the filtering de-noising of time and frequency domain in order to improve the signal-to-noise ratio of seismic data. By changing the signal from time domain to time domain, we design a suitable filter guarantee in time and frequency domain to analyze the distribution of signal and noise with time. The active signal can be filtered to filter noise and interference and then reverse to the time domain to achieve the signal to noise separation and improve the signal to noise ratio of seismic data. Active source detection technology is a very promising technology for deep tectonic detection of the earth. The gas gun source developed from the exploration seismology has achieved some restrictions for people to break through natural earthquakes. The underground structure provides a good technical platform for active detection. Compared with some traditional artificial sources, the gas gun source has the advantages of high signal to noise ratio, accurate location, measurable source characteristics and low cost. The low frequency components of the gas gun source are rich in the exploration of the earth's deep crust, and it is given time by GPS and the time of excitation is accurate; The distance of the gun signal is far away, and the strong air gun signal can be found in a hundred kilometers away from the source of the air gun. The repeatability of the air gun source can improve the signal to noise ratio by the superposition of the source. Although the multiple superposition of the seismic data can improve the signal to noise ratio of the seismic signal, we hope to combine the S transform time frequency filter design. A filtering method specially dealing with the active source gas gun signal is used to improve the signal to noise ratio of the single excitation air gun signal. In this paper, a new method of S transform template filtering de-noising is proposed. By superimposing the air gun signals received at the same station many times, the signal of high signal to noise ratio is obtained, and then the superimposed high signal to noise ratio is used. As a template, the filter is set according to the template in the time and frequency domain and the air gun signal produced by each excitation is filtered and de-noised. The feasibility and practicability of our method are verified by using the S transform template filtering technique to simulate the simulated data. Then the actual air gun data received from the station of the distance air gun source 112km is also obtained. The single shot gas gun signal with high signal to noise ratio is obtained, and it is compared with the band pass filter and the wavelet transform filtering results. It shows that the S transform template filter not only effectively suppress the noise and will not weaken the effective signal, but also the waveform similarity of the filtered single shot air gun signal and the superimposed signal is also similar. It is very high, which proves the effectiveness of our method.
【学位授予单位】：中国地震局地球物理研究所
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P631.44

【参考文献】