当前位置: 首页>博士论文>资源详情
基于Ridgelet和Curvelet变换的地震资料噪声衰减与解释方法研究
中文摘要

石油天然气是人类社会赖以生存和发展的重要能源之一,且与国家安全及经济发展紧密相关。地震勘探是石油天然气勘探的重要手段。随着勘探领域的不断延伸,地震勘探难度越来越大,复杂的地表和地质条件对深部地震信号产生各种畸变和干扰。在地震记录上表现为信噪比低、噪声类型复杂、有效信号能量弱的特点。这严重制约着地震资料后续的高分辨率处理、目标准确成像及属性提取和反演。随着勘探的深入,储层预测成为勘探阶段必须考虑的重要因素,而储层的几何形态和空间展布是储层预测的一项主要内容,在岩性勘探开发日趋重要的今天,它对于提高预探布井和开发布井的成功率至关重要。Ridgelet变换和Curvelet变换是近年来在计算调和分析领域提出的两种新的信号表示方法。它们优异的信号表示功能,不仅引起了信号处理和图像处理工作者广泛关注,同时也受到了地球物理工作者的关注。 本论文针对地震资料噪声衰减问题,根据Ridgelet变换和Curvelet变换的多尺度、多方向特性,研究了基于Ridgelet变换和Curvelet变换的地震资料随机噪声和相干噪声衰减方法;根据Curvelet变换的奇性分析特点,研究了地震资料中储层的地层结构提取的Curvelet变换方法。本论文的研究成果主要有以下几个方面: (1)根据信号或图象的Curvelet系数之间的相互关联性,利用局部均值或中值滤波,提出了一种基于Curvelet变换的局部自适应萎缩阈值噪声衰减方法,并讨论了局部自适应滤波中邻域窗大小的选择原则。实验结果证实,与固定阈值方法相比,该方法能更好地提高信噪比。。 (2)根据Curvelet变换对信号的稀疏表示特性及多尺度、多方向性,提出了基于Curvelet变换的地震资料随机噪声和相干噪声衰减的阈值方法;针对VSP资料处理中上下行波场分离问题,根据上下行波经过Curvelet变换后主要能量分别集中在不同区域,提出了VSP上下行波场分离的Curvelet变换方法;针对强振幅下行波对上下行波场分离效果影响,提出了联合奇异值分解(SVD)和Curvelet变换的VSP上下行波场分离方法。将理论合成记录和实际资料利用文中方法处理结果与常用方法的处理结果比较,结果表明该方法不仅可以很好地对地震资料中噪声进行衰减,同时能更好地保持有效信号的动力学特性。 (3)根据Curvelet变换的奇性分析特征及地震资料瞬时相位在追踪地质层位的优势,提出了利用地震资料瞬时相位的地层结构提取的Curvelet变换方法。将提取的地层结构剖面和瞬时频率或其它与地层岩性相关属性联合显示,能更清晰、直观地解释有利储层的几何形态和空间展布特征。理论合成数据和实际资料处理验证了该方法的有效性。 (4)从Ridgelet变换与Radon变换的关系出发,建立了Ridgelet变换与地震信号处理中常用的τ-p变换的联系;根据地面地震炮集记录中面波的强能量、低视速度、低频率特点和Ridgelet变换所具有的平移、多尺度和多方向性,提出了基于Ridgelet变换的地震面波衰减方法。针对由于过强面波干扰在计算中所引起的强端点效应问题,提出了联合连续小波变换与Ridgelet变换的地震资料面波衰减方法。理论模型和实际资料处理效果证实了本文提出的方法不仅能较好地衰减地震记录中强面波干扰,同时可以保护有效信号的低频成分,为地震资料后续处理和分析提供更好的保证。 关键词:Curvelet变换;Ridgelet变换;噪声衰减;阈值估计;地层结构 论文类型:应用基础 本研究得到国家高技术研究发展计划“863”(编号:2006AA09A102-11)和国家自然科学基金重点项目(编号:40730424)和国家自然科学基金面上项目(编号:40674064)联合资助

英文摘要

The oil and gas is one of important energy sources on which human society rely for existing and developing, and it is close related with national security and economic development. Seismic exploration is an important means of oil and gas exploration. With the extension in the area of exploration, the seismic prospecting is becoming more difficult. The signal of deep part in siesmic data is distorted and disturbed under the complex surface and geological condition. So, the signal to noise ratio is low, the noise type is complicated, and the energy of effective signal is weak in seismic record. This has restricted some subsequent processing, such as high resolution processing, accurate imaging, attributes abstraction and inversion. Along with the exploration thorough, the reservoir prediction is becoming an important factor. The prediction of geometry and spatial distribution of reservoir is one of the primary task in reservior prediction. With the lithological exploration and development more and more important, it is very important to lay the exploration well and develeopment well. The Ridgelet transform and Curvelet transform are two new signal representation method in the field of computational harmonic analysis. They have not only caused the signal processing and the imagery processing worker widely attention, but also they have aroused geophysicist’s interesting. In view of the seismic data noise attenuation, according to the multi-scale, multi-directional properties of the Ridgelet transform and Curvelet transform, we has studied seismic data stochastic noise and the coherent noise weaken method based on the Ridgelet transform and Curvelet transform. According to the singular analysis characteristics of the Curvelet transform, we has studied the reservoir configuration extraction method in the seismic data baised on the Curvelet transform. The improtant research content of the article as follows: 1.According to the Curvelet coefficients of the signal or image are interrelated each other, a new local adaptive shrinkage threshold de-noising approach on Curvelet transform is presented, which uses local mean or median filter to the Curvelet coefficiences. In addition, the neighborhood window size selection principle to estimate threshold is discussed. The experimental results show that the proposed approach can improve the SNR in contrast to fixed threshold. 2.According to the sparse property of the Curvelet transform and the characteristics of multi-scale, multi-directional, a stochastic noise and the coherent noise in seismic data attenuation method is proposed based on the Curvelet transform. In allusion to the wavefield separation in Vertical Seismic Profiling (VSP) data, according to the up-going wave and down-going wave being located in different region of Curvelet domain, a method of VSP wave field separation is proposed. Focusing on the separation result to be affected by strong energy of down-going wave, a wavefield separation method is presented, which combines Singular Value Decomposition (SVD) with the Curvelet transform. The synthetic data and real data processing results show that the method can not only attenuate noise in the seismic data very well, but also can maintain dynamics characteristic of the effective signal. 3.According to the singular analysis characteristic of the Curvelet transform and the superiority of the seismic data instantaneous phase in the tracing geological horizon, a stratal configuration extraction method is proposed, which is based on the instantaneous phase and the Curvelet transform. The joint display of extracting stratal configuration section and the instantaneous frequency or other attributes section relating with the lithology can clearer and more intuitively interpretate the geometry shape and spatial distribution characteristic of superior reservoir. The synthetic data and real data processing results verify the effectiveness of the method. 4.From the relationship of the Ridgelet transform and the Radon transform, the realtion of the Ridgelet transform and the τ- p transform is established, and the forword transform and reconstruction method of the wave field is educed. According to the strong energy, low apparent velocity and low frequency characteristic of the ground-roll in seismic data, a ground-roll attenuation method based on the Ridgelet transform is proposed. Because the strong ground roll interference has caused the strong endpoint effect in the computation. A strong ground-roll attenuation method, which combines the continous wavelet transform and the Ridgelet transform, is proposed. Synthetical models and real data processing results confirmed the proposed method can not only better attenuate the strong ground-roll interference in seismic data, as well as protect the low-frequency components of effective signal, This provides a better guarantee for the seismic data following processing and the analysis. KEY WORDS: Curvelet transform; Ridgelet transform; Noise attenuation; Threshold estimation; Stratal configuration TYPE OF DISSERTATION: Application Fundamentals This work is supported by the National High Technology Research and Development Program of China (No.2006AA09A 102-11) and National Natural Science Foundation of China (No.47030424) and (No.40674064)

作者相关
主题相关
看过该书的人还在看哪些书