基于SW检验的时空时变轨线TFPF消减地震勘探随机噪声

发布时间：2018-03-07 02:05

本文选题：地震勘探资料　切入点：时频峰值滤波　出处：《吉林大学》2015年硕士论文　论文类型：学位论文

【摘要】：油气资源与人民生活和国民经济发展息息相关，，而地震勘探正是剖析、透视地下未知地理结构和探寻未知油气资源的重要手段。然而，地表环境和地底下构造复杂，严重降低了地震勘探资料的信噪比，影响了研究者对地震勘探资料的解释，进而不能准确定位油气及矿藏资源的位置，最终浪费了人力、物力。因此，有效的提高地震勘探资料的信噪比是地震勘探中具有挑战性的难点问题。为突破此问题，就需要我们提出合适的方法，既清晰地保留地震勘探信号又有效地压制地震勘探随机噪声。时频峰值滤波(time-frequency peak filtering, TFPF)在恢复低信噪比地震勘探资料同相轴方面已经取得了一些成效，其无偏估计的条件为：信号线性且噪声为高斯白噪声。但实际地震勘探信号具有较强的非线性，不能满足时频峰值滤波无偏估计的线性条件。为此，在实际应用中我们采用加时域窗的Wigner-Ville分布来提高信号的局部线性度，从而减小时频峰值滤波偏差。但是单一窗长很难在噪声压制和信号恢复两方面得到权衡。选取小窗长时，信号的线性度得到了提高，信号恢复效果较好，但噪声压制效果不理想；选取大窗长时，噪声能得到有效压制，但信号线性度不能有效提高，信号幅值衰减严重。所以综合考虑信号恢复和噪声压制，本文提出了基于Shapiro Wilk和Shapiro Francia检验(SW检验)的时空时变轨线时频峰值滤波消减地震勘探随机噪声。在基于SW检验的时空时变轨线时频峰值滤波算法中，首先构建了与不同弯曲程度的同相轴相匹配的时变滤波轨线。因为同相轴具有横向连续性，所以，沿时变轨线对不同弯曲程度同相轴重采样后，得到的同一信号幅值几乎相同，最大限度地提高了信号的线性度，解决了时频峰值滤波对非线性地震勘探信号估计存在偏差的难题。其次为了得到时变滤波轨线，结合了统计学中的SW高斯性检验，其主要思想为：根据地震勘探随机噪声和地震勘探信号的SW高斯性统计量值差异，检测出地震勘探记录中不同弯曲程度的同相轴，从而提取出与不同弯曲程度的同相轴相匹配的时空时变轨线。最后沿时变轨线对地震勘探记录进行重采样，并对重采样后的信号应用时频峰值滤波，提高了地震信号恢复的精度。为了验证本文提出算法的有效性，将其应用到了人工合成地震勘探记录和实际地震勘探记录的处理，并与传统时频峰值滤波和径向道时频峰值滤波结果进行了对比。实验发现：本文提出算法在滤波后信噪比、信号峰值等方面都优于其他两种算法，说明本文算法在压制随机噪声的同时，能更好地恢复有效信号的细节，增强同相轴的连续性。
[Abstract]:Oil and gas resources are closely related to people's daily life and the development of national economy, and seismic exploration is an important means of analyzing, looking through the unknown geographical structure of underground and exploring unknown oil and gas resources. However, the surface environment and underground structure are complex. It seriously reduces the signal-to-noise ratio of seismic exploration data, affects the interpretation of seismic exploration data by researchers, and then fails to accurately locate the location of oil, gas and mineral resources, which ultimately wastes manpower and material resources. Improving the signal-to-noise ratio of seismic exploration data effectively is a challenging and difficult problem in seismic exploration. In order to break through this problem, we need to put forward a suitable method. The random noise of seismic exploration can be suppressed effectively as well as retaining the seismic exploration signal clearly. Time-frequency peak filtering (TFPF) has made some achievements in restoring the cophase axis of seismic exploration data with low signal-to-noise ratio (SNR). The condition of unbiased estimation is that the signal is linear and the noise is Gao Si white noise, but the actual seismic exploration signal has strong nonlinearity and can not satisfy the linear condition of time-frequency peak filter unbiased estimation. In practical application, we use the Wigner-Ville distribution with time-domain window to improve the local linearity of the signal, thus reducing the peak filtering deviation of the hourly frequency. However, the single window length is difficult to be balanced between noise suppression and signal recovery. The linearity of signal is improved, the effect of signal recovery is better, but the effect of noise suppression is not ideal. When the window is large, the noise can be suppressed effectively, but the linearity of signal can not be improved effectively. The amplitude of the signal attenuates seriously. Therefore, considering the signal recovery and noise suppression, a time-frequency peak filtering method based on Shapiro Wilk and Shapiro Francia test is proposed to reduce the random noise in seismic exploration. In the time-frequency peak filtering algorithm of time-varying track based on SW test, the time-varying filter rail line matching with the in-phase axis with different bending degree is first constructed. Because the in-phase axis has transverse continuity, so, The amplitude of the same signal is almost the same after sampling the same phase with different bending degree along the time-varying rail line, which improves the linearity of the signal to the maximum extent. In order to obtain the time-varying filtering trajectory, the time-frequency peak filter is combined with the SW Gao Si test in statistics to solve the problem of the deviation of time-frequency peak filtering to the estimation of nonlinear seismic exploration signal. The main idea is: according to the difference of SW Gao Si statistical value between random noise of seismic exploration and seismic exploration signal, the same phase axis with different bending degree in seismic exploration record is detected. In order to extract the time-varying track line matching the same phase axis with different bending degree, finally resampling the seismic exploration record along the time-varying track line, and applying time-frequency peak filter to the signal after resampling. The accuracy of seismic signal recovery is improved. In order to verify the validity of the proposed algorithm, the proposed algorithm is applied to the processing of synthetic seismic exploration records and actual seismic exploration records. Compared with the results of traditional time-frequency peak filtering and radial channel time-frequency peak filtering, the experimental results show that the proposed algorithm is superior to the other two algorithms in signal-to-noise ratio (SNR) and signal peak value after filtering. It is shown that the proposed algorithm can recover the details of the effective signal and enhance the continuity of the cophase axis while suppressing the random noise.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P631.4;TN911.4

【参考文献】