Steerable Pyramid分解域地震资料叠前去噪及断层检测

Random noise is unavoidable in seismic exploration. The existence of random noise in prestack seismic data not only lowers SN ratio of seismic data but also decreases the accuracy of dynamic and static corrections, harming final data quality. Many methods for attenuating random noise have received good effecting real application. They were developed by using poststack mathematic model, so that their effects in prestack application are restricted in some degree because of their own theoretical bases or some specific characteristics of prestack seismic data. At present, domestic oil exploration has transfered from structural hydrocarbon reservoir to subtle reservoir, and the requirement of exploratory technique for subtle reservoir is urgent. Some subtle reservoir such as fractured hydrocarbon reservoir of marine carbonate and lithologic hydrocarbon reservoir of clasolite expresses high anisotropic and directional in horizontal, so that the reservoir evaluation is difficult and conventional technology is very difficult to identify, which need some key seismic reservoir prediction technology to detect and characterize. And in the process of reservoir prediction, the recognition and detection of faults is extremely important. However, multi-scale and multi-direction detection technology for faults is in the stage of exploration at home and abroad..Consequently，the paper summarizes methods of pre-stack noise suppressing and fault detection of seismic data. It analyzes their merit and demerit in presenting and classifies the characteristics of seismic data. Based on the analyses, we analyze the fundamental principle of Steerable Pyramid decomposition and reconstruction in detail and improve it to make it more suit for pre-stack noise suppressing and fault detection of seismic data. We further our study on pre-stack noise suppressing , week signal recognition ,fault detection and program module design.

随机噪声是地震勘探中不可避免的一类干扰波,叠前记录中随机噪声的存在不但会引起资料信噪比的降低,而且会直接影响到动、静校正的精度,影响资料处理的质量。目前已有很多具有很好效果的衰减叠前随机噪声的方法，大多基于叠后数学模型开发,在叠前应用有时会因为其理论基础或叠前资料本身的某些特点而使应用效果受到某种程度的制约。同时，断层的检测和识别在隐蔽油气藏勘探中有着极其重要的作用，现有的断层检测算法大多没有针对断层的检测进行具体的设计，这些算法只是增强了断层的边缘特征，而不是增强断层本身，并且在远离断层的地方，仍然存在伪不连续性现象。本项目拟针对目前叠前地震资料随机噪声衰减及弱信号识别，断层检测相关研究方面存在的问题，在多尺度几何分析领域，通过Steerable Pyramid分解方法对叠前去噪和断层检测技术进行改进，以有效提高断层检测精度及地震资料的品质，丰富断层检测及改善资料品质相关理论技术。

项目对现行的断层检测及叠前地震随机噪声衰减及弱信号识别已有的方法进行了调研，在多尺度几何分析领域详细分析了2D、3D Steerable Pyramid分解算法。基于地震资料的特点，在Steerable Pyramid分解域围绕断层检测、叠前地震随机噪声衰减、叠前地震资料弱信号识别及算法实现等四个方面进行了较为深入的研究，开发了用于断层检测与成像及叠前强弱随机噪声衰减的算法及相应软件模块。.1、断层检测：详细分析了2D、3D Steerable Pyramid分解与重构的基本原理。输入的三维地震数据经过Steerable Pyramid分解后，得到了不同尺度不同方向的断层特征及走向。基于地震数据的特性对重构时方向估计算法进行了改进，获得稳定一致的估计方向，增加了跨尺度的平滑以增强方向性和线性性；针对二维地震断层加入了线性控制S函数以增强线性性，使断层检测能力大大加强。.2、叠前地震随机噪声衰减：在Steerable Pyramid分解域，采用阈值估计算法SoftLMAP和SoftLMMSE对叠前地震随机噪声进行衰减，结果证明Laplace分布模型比Gaussian分布模型更适合于描述地震数据的Steerable Pyramid分解系数。将SoftLMAP阈值去噪方法应用于实际地震资料，能比较彻底地去除随机噪声，去噪后的图像边缘保持较好，滤除噪声同时还保留了有效部分。.3、叠前地震弱信号识别：发展了降阈值硬阈值+均值滤波弱信号识别方法，可以从最大幅值为2倍弱信号强度的噪声中识别出弱信号，并且保证信噪比有足够的提高，原始地震数据中的强随机噪声得到有效的压制，弱同相轴得到保留，较好的保持振幅能量，保持频谱特征，保护断层边界。.4、软件工程化方面：在windows环境下使用vs2005和Qt工具，将2D、3D Steerable Pyramid分解断层检测及叠前地震资料随机噪声衰减及弱信号识别算法，形成了独立的程序模块，应用于实际资料的处理，取得了较好的效果。.理论模型和实际地震资料处理结果显示，将2D、3D Steerable Pyramid分解算法应用于地震资料断层的检测与成像及叠前随机噪声的衰减和弱信号的识别是成功的。本项目的完成，对今后地震数据处理及属性提取的研究有极大的意义。

内容获取失败，请点击重试