非平稳地震数据局部属性表征与应用研究

发布时间：2018-01-28 11:50

本文关键词： 非平稳局部属性反问题局部时频变换局部信噪比流式算法　出处：《吉林大学》2017年硕士论文　论文类型：学位论文

【摘要】：地震属性指的是地震数据经过数学变换得到的关于几何学、运动学、动力学或统计学特征。它作为地震资料属性特征的度量,能够有效反映地下介质的信息。因此,地震属性在地震数据处理和解释中起着重要作用。目前,地震属性种类较多,分类方法不一,考虑地震数据的全局特性和局部特性,可将地震属性分为全局属性与局部属性。全局属性指的是全部地震数据的几何学、运动学、动力学或统计学特征,反映的是地震资料的整体特性。局部属性指的是局部地震数据的几何学、运动学、动力学或统计学特征,反映的是地震资料的局部特性。一般情况下,局部属性特指的是单个地震数据点的属性特征。局部属性利用了地震数据中每个数据点及其邻域各点的局部信息,能够准确反映地震信号的局部信息特征,有利于地震数据更准确直观的处理和解释。局部属性种类较多,常见的有局部时频属性、局部相关性、局部信噪正交属性等。本文重点研究了两类局部属性,分别是局部时频属性和局部信噪比属性。其中,局部时频属性是将局部属性的数学原理应用到时频变换中,求得最小二乘意义下的时变傅里叶级数。局部时频谱能够准确反映地震数据的时频特征,可应用于基于时频特征的地震数据处理和解释中。本文在局部时频属性的基础上,研究了地震波衰减的补偿方法。通过频谱比值法求品质因子,并利用Kolsky衰减-频散模型在局部时频域中进行地震波的衰减补偿。本文研究的另一类局部属性是局部信噪比。针对目前已有的信噪比定义无法反映地震数据的局部质量问题,本文提出了局部信噪比的概念,定义了随时间空间变化的信噪比函数。自适应的信噪比函数使得求解问题成为数学欠定问题,本文利用每个数据点及其邻域各点的局部信息,使用正则化约束下的共轭梯度算法迭代求解局部信噪比的最小二乘解。时空变的信噪比函数能够反映地震数据中每一处数据点的质量好坏,为后续的处理和解释提供了直观准确的评判标准。求解实际地震资料的局部信噪比时,还需要知道含噪声地震数据所包含的信号部分和噪声部分。本文在过滤波和局部信噪正交属性的基础上,提出了级联信号估计方法来估计有效信号。该方法利用有效信号和噪声的相关性特征计算局部信噪比中的有效信号。局部信噪比估计方法利用了信号中每个数据点及其邻域各点的局部信息,避免了使用单个数据点而可能出现的信噪比不合理值,而且局部处理能够减少全局噪声对信噪比估算的影响,该方法可以更准确地表征地震资料信噪分布特征。另外,局部信噪比对去噪方法的评估也具有重要意义。局部属性的数学原理是利用正则化约束下的共轭梯度算法求解数学反问题。这种迭代式算法计算量大,在实际地震资料处理和解释中,其实用性受到影响。为此,本文研究了局部属性的流式算法。该算法以不断更新已得参数的方式来计算新的局部属性参数,这样一种不断更新的计算过程如同水流般沿数据点不断向前进行,因此称之为“流式”。流式算法用更新固定参数的方式代替自适应变化的参数,用代数相似关系代替正则化的局部平滑作用,用简单的代数运算代替复杂的迭代求解,所以,该算法极大的减小了内存占用量和计算量。将流式局部信噪正交属性与流式预测滤波的数学原理相结合,可以得到流式正交预测滤波方法。作为流式局部属性在地震数据去噪中的应用,该方法可以在有效去除随机噪声的同时,最大程度的保护有效信号。
[Abstract]:The seismic attribute is obtained by mathematical transform of seismic data on geometry, kinematics, dynamics or statistical characteristics. It is used as a measure of seismic attribute, which can effectively reflect the information of underground medium. Therefore, the seismic attribute plays an important role in seismic data processing and interpretation. At present, many kinds of seismic attributes, classification method a consideration of global characteristics and local characteristics of seismic data, seismic attribute can be divided into global and local property attributes. Global property refers to all the seismic data of the geometry, kinematics, dynamics and statistical characteristics, reflects the overall characteristics of seismic data. The local attribute refers to the local seismic data geometry. The kinematics, dynamics and statistical characteristics, reflects the local characteristics of seismic data. In general, local attribute refers to the single point of seismic data The characteristics of local properties using seismic data. The local information in each data point and its neighborhood of each point, can accurately reflect the local information of the characteristics of the seismic signal, is conducive to seismic data processing and interpretation more accurately and intuitively. Many local attribute types, common local time-frequency property, local correlation, local SNR orthogonal attributes etc.. This paper focuses on the two types of local properties, are the local time-frequency properties and local SNR properties. Among them, the local time-frequency property is the mathematical principle of using local properties of time-frequency transform, obtain the least squares sense time-varying Fourier series. Partial spectrum can reflect the time-frequency characteristics seismic data, seismic data processing and interpretation can be used in feature based on time-frequency. Based on the local time based on the properties of frequency, the compensation method of seismic wave attenuation by frequency spectrum ratio. Value method for the quality factor, and using the Kolsky attenuation dispersion model of seismic wave attenuation compensation in local time and frequency domain. A local attribute of this paper is the local SNR. According to the current SNR definition cannot reflect the local quality of seismic data, this paper puts forward the concept of local signal to noise the definition of change with time and space. The adaptive function of signal-to-noise ratio SNR function makes the solving problems become mathematics underdetermined problem, using the local information of each data point and its neighborhood of each point, using the least squares regularization constrained conjugate gradient iterative algorithm for solving local SNR of the temporal and spatial solution. The noise can reflect each data point on the quality of seismic data in function, provides intuitive and accurate criteria for subsequent processing and interpretation of seismic data. To solve the local signal-to-noise Than, also need to know the signal containing noise and noise of seismic data contained in part. Based on the filtering and local signal-to-noise orthogonal attributes, proposed signal estimation method to estimate the effective signal. Correlation between signal and noise characteristics using the method of calculating effective local signal SNR. Ratio estimation method using signal local information in each data point and its neighborhood of each point of the local signal to noise, avoid the use of a single data point may appear in the SNR is not reasonable, and local treatment can reduce the noise on the global SNR estimation, this method can characterize the signal-to-noise distribution characteristics exactly. In addition, the evaluation method of local denoising SNR is also important. The mathematical principle of local properties of the conjugate gradient algorithm is used to solve the mathematics question regularization constraints This kind of problem. The iterative algorithm of large amount of calculation, in the actual seismic data processing and interpretation, the practicability is affected. Therefore, this paper studies the flow properties of the local algorithm. The algorithm has to continuously update the parameters to calculate the local properties of the new parameters, such a constantly updated calculation process as water flows along like data points forward, so called "flow". Parameters of the streaming algorithm with fixed parameters instead of updating adaptive local smoothing function, instead of regularization using algebraic iterative solution, similar to that with simple algebraic operation instead of complex so the algorithm greatly reduced the memory and computation. The mathematical principle will flow noise and flow prediction filtering orthogonal attribute combination, can get the prediction filtering method of stream flow. As the orthogonal seismic attribute in number According to the application of de-noising, this method can effectively remove the random noise and protect the effective signal to the maximum extent.

【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P631.4

【相似文献】