【作者】 谢金娥；

【导师】 刘财； 郭全仕；

【作者基本信息】 吉林大学 ， 固体地球物理学， 2009， 博士

【摘要】 随着科技发展和社会的需要,人们对地震勘探资料品质的要求越来越高。本文所述三类方法通过提高地震资料的分辨率和信噪比,以改善地震资料质量。本文共分五个部分。第一章主要阐述了地震资料的分辨率和信噪比的研究现状以及文章的研究内容和创新点。其中主要研究方法为:基于自然梯度的盲反褶积算法、自适应波束压制高密点地震中面波和数学形态学压制面波。第二章研究了一种提取混合相位地震子波方法。算法的核心思想是自适应地迭代调整滤波器系数,以提取最优混合相位子波。为提高算法收敛速度,本文采用了基于自然梯度的迭代算法,根据地震信号的次高斯分布特性,提出概率密度函数,通过求解此函数,提取了混合相位子波。第三章分析了高密点地震数据的特性,提出自适应波束压制面波方法,通过设计代价函数,优化了自适应波束。本方法在理论模型和实际资料处理中取得了一定效果。第四章借鉴于数学形态学在图像处理领域的发展,创新性地将应用在聚类中数学形态学的思想应用于地震勘探中的面波压制中。通过基本运算膨胀、腐蚀、开运算和闭运算组成的算法对原始剖面中的面波进行压制,取得了一定效果。为了进一步改善剖面质量,对原始剖面进行中值滤波,再用本方法进行面波压制,效果明显。文章的结论部分对全文内容进行了总结和展望。更多还原

【Abstract】 The high signal-to-noise ratio (SNR) and resolution is the aim that people are pursuing for gaining good seismic profile and geologic interpretation. This paper tells us three methods in order to improve SNR and resolution of seismic data. Natural gradient blind deconvolution algorithm can extract true mixed-phase-wavelet. Adapt beamforming can suppressing single-sensor ground-roll, not only synthetic seismic data but also Yakenbei real seismic data, there is a obvious effect. Mathematical morphology also has a obvious effect on suppressing ground-roll. Natural gradient blind deconvolution algorithm extract true mixed-phase-wavelet, and improve the resolution.Estimating seismic wavelet is one of the most important problems of geophysics. True wavelet’s estimating is a important task in seismology, it is the base of seismic deconvolution, migration, feature extracting, geophysics interpretation and so on. Convention deconvolution assumes statistics under seismic wavelet and reflex coefficient are unknown, this will low the resolution and reliability of seismic data. Natural gradient blind deconvolution algorithm hasn’t need any assumes, using natural gradient estimate wavelet, gain true mixed-phase-wavelet, and use in deconvolution to improve resolution. First of all, we introduced mathematic foundation of original data’s can be separated, then we introduced mathematical principal of seismic blind deconvolution, and inducted compensate function, derived the seismic blind deconvolution’s iteration function, at last, we use this method to verify synthetic seismic data and real seismic data, the result shows than natural gradient blind deconvolution algorithm can estimate true mixed-phase-wavelet better, and it can estimate reflex coefficient better.Compare natural gradient blind deconvolution to convention deconvolution, it has many advantages: first, it hasn’t needed assuming seismic wavelet minimum phase; second, this algorithm used natural gradient method, and this algorithm seems stead, contraction fast, not sensible to noise, and those are very good to write software; third, this algorithm is adaptive, it can update filter coefficient; forth, in synthetic seismic data, though compare to convention deconvolution, it is obvious that natural gradient blind deconvolution can estimate seismic wavelet better. So, we uses this algorithm to real seismic data, we still have a good effect, it can improve seismic data’s resolution.At the same time, this paper also study on improving seismic data’s SNR. On single-sensor seismic section, the energy of ground-roll is very strong, reflect wave is hidden, and the SNR is very low. In convention seismic processing, we usually use High-pass filter to suppress ground-roll, it has problems, ground-roll still lie or many effect wave is filtered. F-K method and wavelet transform method also have the same disadvantages. So, we propose adaptive beamforming method to suppress ground-roll. This method uses the different apparent velocity of reflect wave and ground-roll, design cost function, make the filter coefficient better and better, suppress energy of ground-roll area. It is used in Xinjiang Yakenbei single-sensor data, and the effect is very good, SNR is improved.Adaptive beamforming mainly has those advantages. First, on single-sensor seismic section, the energy of ground-roll is very strong, and the SNR is very low. We transform the data to 2D Fourier area in order to make original data and ground-roll are in its self’s apparent velocity; Second, we need the beamforming can accept stand signals, at the same time, attenuate other noise, those noise and effect signals have same frequency in time area. We want to design useful filter to improve SNR; Third, when adaptive beamforming suppressing ground-roll, there exit space alias. Ground-roll is in low frequency area, effect wave and other noise are in wide frequency. Ground-roll and effect wave are mixed, especially when group interval is big, it is disadvantage to suppress noise. The group interval in single-sensor is 5m in most times, this decrease the space alias; Forth, the adaptive beamformging has contraction problems, it is up to step’s uncertainty. As adaptive beamforming is adaptive, it can be used in suppressing other linear noise; Fifth, in real seismic data processing, compare this method to convention band-pass filter, it is better than band-pass filter in seismic section and frequency section. So this method can be wide in seismic data processing, and has important theory and application value in improve SNR of seismic data.This paper also study mathematical morphology to suppress ground-roll. Mathematical morphology is stand on strict math theory, it’s basic idea and method make a important influence on image processing. In fact, mathematical morphology has compromise a new theory, and is applied in many areas. In this paper, we use this method to suppress ground-roll, though its basic calculation, such as dilation, erosion, opens, close. The result shows this method very well, and can perfectly suppress ground-roll. After midian filtering to original data, then suppressing ground-roll with this method, we can see the data became much clean than before, at the same time, the ground-roll has been suppressed and concordant axis became continuiously.更多还原