【作者】 林红波；

【导师】 杨宝俊；

【作者基本信息】 吉林大学 ， 地球探测与信息技术， 2007， 博士

【摘要】 基于时频峰值滤波技术消减随机噪声的原理是利用解析信号的时频分布沿着其瞬时频率集中的特性,采用时频分布的峰值估计瞬时频率,从而得到信号的无偏估计。本文从理论和实践的角度分析了时频峰值滤波技术消减地震勘探资料中的随机噪声的技术细节,详细论述了时频峰值滤波消减随机噪声的原理、方法和问题,并将其应用到人工合成记录和实际地震勘探资料中的随机噪声消减,为地震勘探资料信噪比提高提供了一个有效的工具。根据解析信号时频分布的理论和瞬时频率估计的方法,研究了时频峰值滤波技术处理地震子波的可能性;在低信噪比条件下,噪声、信号的非线性引起的时频峰值滤波的偏差和方差适用于地震子波;指出零均值高斯色噪声条件下时频峰值滤波技术能够无偏估计地震子波的条件;研究了具有不同层速度,不同优势频率子波,相位振幅比的人工合成记录在不同强度的高斯白噪声和高斯色噪声时频峰值滤波结果,滤波结果表明时频峰值滤波技术消减随机噪声的能力不受同相轴形式的影响,为时频峰值滤波技术应用于实际地震勘探资料去噪提供充分的数据知识;将时频峰值滤波技术应用于共炮点地震勘探资料,叠后地震剖面资料和具有各向异性地震勘探资料中,滤波后的记录随机噪声被消减,有效信号和同相轴的连续性得到改善,信噪比得到提高;为了解决时频峰值滤波应用于实际地震勘探资料处理中分辨率和噪声衰减的矛盾,给出窗长与地震子波频率和采样频率的关系,给出了适合复杂地震勘探记录的技术,理论模型和共炮点地震勘探资料处理结果表明时变窗长时频峰值滤波结果比定窗长时频峰值滤波结果具有更好的分辨率,为地震勘探资料中的随机噪声衰减提供了行之有效的工具。更多还原

【Abstract】 Seismic exploration is an important tool for oil-gas reservoirs and mineral resources exploration. With the development of method of seismic exploration and increasing demand for oil-gas, the prospecting has turn to research the complex structure and deep crustal structure. The seismic exploration data obtained from the complex earth’s surface, such as desert, hill and boondocks etc., is rather complex and has strong random noise. It is difficult to distinguish the seismic events under the strong noise , even if the intensity of the reflected signal is strong. The problem of attenuating random noise in low signal-to-noise ratio (SNR) seismic data has puzzled the seismic signal processor and producer; even make the data as the waster. Attenuation of noise and improvement of SNR is significant for researching the structure of subsurface and for prospecting oil-gas reservoirs.The summarization of status of attenuation of random noise in seismic data has been made in this paper. Based on the difference between the signal and noise in transformed domain, amplitude rule or statistical characteristic, the current methods attenuate the random noise, but they are limited by the type of events and the level of SNR of data, even are invalidated at the SNR less than 0.5dB. Aiming at this problem, we introduced the time-frequency peak filtering (TFPF) technique to attenuate the random noise. The attenuation of random noise in seismic data with TFPF is researched on theory, synthetic model simulation and seismic data in this paper. TFPF is a signal enhancement method based on the principle of instantaneous frequency estimation. In TFPF, the observed noisy signal is firstly transformed to be processed into instantaneous frequency (IF) of the analysis signal by frequency modulation, and then the peak of time-frequency distribution of analysis signal is taken as the estimation of IF for the concentration of time-frequency distribution of analysis signal along with IF. The paper researched principle, properties and realization of TFPF systematically; analyzed applicable condition. The results of theory analysis and simulation shown that the seismic wavelet which is band limited,deterministic signal can be recovered from random noise.The paper researched the theories, steps of realization and details of techniques about application the TFPF for attenuation random noise in seismic data. The paper proposed that TFPF must be realized on single trace of seismic data which are scaled first for preventing filtered signal aliased. The TFPF for seismic data utilized windowed Wigner-Ville distribution (pseudo Wigner-Ville distribution, PWVD) to compute the time-frequency distribution of analysis signal for decreasing bias induced by nonlinearity of IF. The paper analyzed the bias and variance from the random noise, as well as the bias from the nonlinearity of signal. Under the condition of Gaussian white noise, the filtered signal by TFPF is unbiased. On this basis, the paper analyzed the colored noise with zero mean and indicated that the bias induced by colored noise is related to the type of power spectrum. When Wigner-Ville spectrum of the FM colored noise is symmetrical, the bias is zero. When the mean of random noise is zero, the variance is zero; nonlinear seismic wavelet make the IF of analysis signal is nonlinear, so the TFPF based on PWVD is used to decrease bias for the peak of WVD of analysis signal deviating from the IF. PWVD can ensure the IF in the window is nearly linear by smoothing in frequency domain, but which results in the trade off between the time resolution and smoothness of noise. The paper proposed the time-varying window length TFPF, which has varying window length along with time of a single trace of seismic data according to the rule of propagation of seismic wave, to attenuate random noise in seismic data. The proposed method can get balance between improvement of the time resolution of filtered seismic wave and attenuation of random noise. According to the rule of from simple to complex condition, the paper firstly applied TFPF to the synthetic seismic data, then to actual seismic data. The constructed seismic data with single event considered the different factors of seismic wave, including the velocity of event, the frequency of seismic wave, and the attenuation of frequency with in an event, phase-to-amplitude radio of seismic wavelet as well as the mixed phased wavelet. The white random noise is added into the above models to simulate TFPF, the results show that the random noise in the seismic data with low SNR are attenuated validly, the reflective signal is enhanced and SNR of seismic data is increased. The wave of recovered signal is similar to the true wave form with the correct position at wave peak and wave valley, but it is aliased at the inflexion approaching zero. When the noise is Gaussian colored noise, the simulation results also show the good properties at improvement of SNR and attenuation of noise; Comparing TFPF with single trace Wiener filtering indicates that TFPF has advantages at holding wave form and attenuation of random noise. The synthetic seismic models with multiple events are constructed to simulate the complex subsurface media such as multi-interfaces, lamina and faultage etc. TFPF is utilized to the models and get the results of the attenuation of random noise existed in synthetic seismic data at low SNR. The results show that each event can be recovered without affecting each other and the factors of events such as the arrival time and shape of event; the lamina can be detached on the whole; the models with different faultages have the similar filtering results at attenuation random noise and improvement SNR. The conclusion from synthetic seismic data by TFPF is that TFPF is suited method for attenuating the random noise in low SNR seismic data, and the filtered results don’t affect by the shape of wavelets and shape of events which means the complexity of underground media. TFPF technique is the flexible, easy to realize on attenuating added random noise in seismic data. The simulations on synthetic seismic data provide abundant data for processing actual seismic data.Because TFPF is flexible tool for attenuating added random noise, it is applied to attenuate added random noise in pre-stacked and post-stacked and anisotropic seismic data. The fixed window length and time-varying window length TFPF are applied on pre-stacked and stacked common shot point seismic data respectively. From the results we can see that the strong noise concentrated in some area is attenuated, the reflected events are enhanced and the continuity of event is improved. The time-varying window length TFPF shows better resolution in time domain than fixed window length TFPF. Profile data processed by TFPF shows more events in filtered data than those in data before TFPF processing for the attenuation of random noise and enhancement of signal. We also try to attenuate the random noise in the anisotropic seismic data by TFPF. In the simulation results, the random noise is restrained and the valid information in seismic data be distinguished easier.On the basis of theories analysis and seismic data processing, we can conclude that TFPF can attenuate the random noise in low SNR seismic data, increase SNR, improve the continuity of event and recover valid information by adjusting the modulated factor and window length. The filtered results don’t be affected by the shape of events. The proposed time-varying window length TFPF can improve the time resolution of filtered wavelet. It is promised to be applied in pre-stacked, post-stacked and anisotropic seismic data.更多还原