Texture attributes for detection of salt

Within areas of salt tectonics, seismic imaging requires extensive updating of the velocity model. This includes defining the boundaries of salt structures which are often characterized by changes in texture of the seismic signal rather than reflectivity. The main characteristics of texture inside salt structures are identified. Three groups of texture attributes are studied: gray-level co-occurrence matrix (GLCM) attributes, frequency-based attributes, and dip and similarity attributes. Various combinations of the selected attributes are tested in a supervised Bayesian classification method. Experimental results show that the classification performance improves by combining at least two texture attribute groups. The classifier computes an estimate of the pixelwise probability of salt. It can then be applied to compute the probability of salt on different seismic sections. Classification results were found more robust based on timeslices. The result from classification, the salt probability image, is then input to a segmentation algorithm that produces a smooth border delimiting the extent of the salt. The segmented salt contours corresponded fairly well to the contours provided by an interpreter.     

Salt diapirs without well defined boundaries – a feasibility study of semi‐automatic detection

Prestack depth imaging of seismic data in complex areas such as salt structures requires extensive velocity model updating. In many cases, salt boundaries can be difficult to identify due to lack of seismic reflectivity. Traditional amplitude based segmentation methods do not properly tackle this problem, resulting in extensive manual editing. This paper presents a selection of seismic attributes that can reveal texture differences between the salt diapirs and the surrounding geology as opposed to amplitude‐sensitive attributes that are used in case of well defined boundaries. The approach consists of first extracting selected texture attributes, then using these attributes to train a classifier to estimate the probability that each pixel in the data set belongs to one of the following classes: near‐horizontal layering, highly‐dipping areas and the inside of the salt that appears more like a low amplitude area with small variations in texture. To find the border between the inside of the salt and the highly‐dipping surroundings, the posterior probability of the class salt is input to a graph‐cut algorithm that produces a smooth, continuous border. An in‐line seismic section and a timeslice from a 3D North Sea data set were employed to test the proposed approach. Comparisons between the automatically segmented salt contours and the corresponding contours as provided by an experienced interpreter showed a high degree of similarity.     

The role of visual saliency in the automation of seismic interpretation

In this paper, we propose a workflow based on SalSi for the detection and delineation of geological structures such as salt domes. SalSi is a seismic attribute designed based on the modelling of human visual system that detects the salient features and captures the spatial correlation within seismic volumes for delineating seismic structures. Using this attribute we cannot only highlight the neighbouring regions of salt domes to assist a seismic interpreter but also delineate such structures using a region growing method and post‐processing. The proposed delineation workflow detects the salt‐dome boundary with very good precision and accuracy. Experimental results show the effectiveness of the proposed workflow on a real seismic dataset acquired from the North Sea, F3 block. For the subjective evaluation of the results of different salt‐dome delineation algorithms, we have used a reference salt‐dome boundary interpreted by a geophysicist. For the objective evaluation of results, we have used five different metrics based on pixels, shape, and curvedness to establish the effectiveness of the proposed workflow. The proposed workflow is not only fast but also yields better results as compared with other salt‐dome delineation algorithms and shows a promising potential in seismic interpretation.     

Prediction of effective porosity from seismic attributes using locally linear model tree algorithm

In this study, a locally linear model tree algorithm was used to optimize a neuro‐fuzzy model for prediction of effective porosity from seismic attributes in one of Iranian oil fields located southwest of Iran. Valid identification of effective porosity distribution in fractured carbonate reservoirs is extremely essential for reservoir characterization. These high‐accuracy predictions facilitate efficient exploration and management of oil and gas resources. The multi‐attribute stepwise linear regression method was used to select five out of 26 seismic attributes one by one. These attributes introduced into the neuro‐fuzzy model to predict effective porosity. The neuro‐fuzzy model with seven locally linear models resulted in the lowest validation error. Moreover, a blind test was carried out at the location of two wells that were used neither in training nor validation. The results obtained from the validation and blind test of the model confirmed the ability of the proposed algorithm in predicting the effective porosity. In the end, the performance of this neuro‐fuzzy model was compared with two regular neural networks of a multi‐layer perceptron and a radial basis function, and the results show that a locally linear neuro‐fuzzy model trained by a locally linear model tree algorithm resulted in more accurate porosity prediction than standard neural networks, particularly in the case where irregularities increase in the data set. The production data have been also used to verify the reliability of the porosity model. The porosity sections through the two wells demonstrate that the porosity model conforms to the production rate of wells. Comparison of the locally linear neuro‐fuzzy model performance on different wells indicates that there is a distinct discrepancy in the performance of this model compared with the other techniques. This discrepancy in the performance is a function of the correlation between the model inputs and output. In the case where the strength of the relationship between seismic attributes and effective porosity decreases, the neuro‐fuzzy model results in more accurate prediction than regular neural networks, whereas the neuro‐fuzzy model has a close performance to neural networks if there is a strong relationship between seismic attributes and effective porosity. The effective porosity map, presented as the output of the method, shows a high‐porosity area in the centre of zone 2 of the Ilam reservoir. Furthermore, there is an extensive high‐porosity area in zone 4 of Sarvak that extends from the centre to the east of the reservoir.     

基于克隆选择原理的核爆地震特征选择方法

为了解决核爆地震自动识别中最优特征子集的选择问题,根据克隆选择原理,提出了一种过滤与封装相结合的特征选择方法.该方法融合了封装式与过滤式特征选择方法的优点,利用局部化的类别可分性判据来处理核爆地震样本的多峰分布问题,通过设定独立的记忆抗体能够保证最终结果是搜索过的最佳特征组合,并且可以处理设定和不设定最优特征子集维数两种情况下的特征选择问题.首先通过UCI数据集中呈多峰分布的玻璃数据验证了该特征选择方法的有效性,进而将其应用到核爆地震特征选择中.核爆地震特征选择实验结果表明,该方法不仅有效地降低了特征空间的维数,而且使分类精度提高了2个百分点,与封装式特征选择方法相比,该方法的计算复杂度大为降低.     

基于地震属性约简的深度学习储层物性参数预测:以莺歌海盆地乐东区为例

储层物性参数作为描述储层特性、储层建模和流体模式的重要指标,其准确估算可以为储层预测提供有力参考依据,但传统储层物性参数反演方法无法兼顾反演精度及空间连续性。针对上述问题,本文引入地震属性作为深度学习算法输入,针对地震属性之间存在的信息冗余特征,利用随机森林-递归消除法对地震属性进行约简预处理,最终建立一种基于地震属性约简的储层物性参数预测方法。实际数据测试结果表明,地震属性约简的深度学习储层物性参数预测结果具有良好的精度及横向分辨率,证实本文方法的有效性。     

Common-reflection-surface stack — a real data example

The simulation of a zero-offset (ZO) stack section from multi-coverage reflection data is a standard imaging method in seismic processing. It significantly reduces the amount of data and increases the signal-to-noise ratio due to constructive interference of correlated events. Conventional imaging methods, e.g., normal moveout (NMO)/dip moveout (DMO)/stack or pre-stack migration, require a sufficiently accurate macro-velocity model to yield appropriate results, whereas the recently introduced common-reflection-surface stack does not depend on a macro-velocity model. For two-dimensional seismic acquisition, its stacking operator depends on three wavefield attributes and approximates the kinematic multi-coverage reflection response of curved interfaces in laterally inhomogeneous media. The common-reflection-surface stack moveout formula defines a stacking surface for each particular sample in the ZO section to be simulated. The stacking surfaces that fit best to actual events in the multi-coverage data set are determined by means of coherency analysis. In this way, we obtain a coherency section and a section of each of the three wavefield attributes defining the stacking operator. These wavefield attributes characterize the curved interfaces and, thus, can be used for a subsequent inversion. In this paper, we focus on an application to a real land data set acquired over a salt dome. We propose three separate one-parametric search and coherency analyses to determine initial common-reflection-surface stack parameters. Optionally, a subsequent optimization algorithm can be performed to refine these initial parameters. The simulated ZO section obtained by the common-reflection-surface stack is compared to the result of a conventional NMO/DMO/stack processing sequence. We observe an increased signal-to-noise ratio and an improved continuity along the events for our proposed method — without loss of lateral resolution.     

DELINEATION OF SHALLOW SALT DOMES AND SURFACE FAULTS BY TEMPERATURE MEASUREMENTS AT A DEPTH OF APPROXIMATELY 2 METRES*

The feasibility of using temperature measurements at a depth of about 2 m for locating and delineating salt domes and faults has been investigated both theoretically and in experimental field surveys. It is shown that measurable temperature anomalies in the soil are to be expected over shallow salt domes. In a field survey over a salt-dome area bordering the Groningen gas field, a large number of temperature measurements were made in small holes (2 m deep, 3 cm in diameter) within a relatively short time (some weeks). The results clearly indicate several temperature anomalies with differential temperatures of about 1°C. Comparison of our thermal contour map with interpretations of available seismic or gravity data, or with direct evidence from wells, showed an excellent correlation. Seismic data even support the shape of the thermal contours. Results in similar agreement with gravity or well data were obtained over salt ridges in a tropical area. Experiments showed that the technique worked as well in lakes and marshes as on dry land. In addition, some experimental evidence collected so far over shallow and surface faults is presented. In several cases, strong thermal anomalies coincided with known surface faults. A thermal model for a surface-fault zone is suggested which accounts satisfactorily for the observed thermal data. It suggests some diagnostic value for the fault's geometry. For shallow faults, however, lack of knowledge of subsurface detail prevented any unambiguous correlation with observed thermal anomalies. Accordingly any geological use of thermal analysis over shallow faults remains debatable. The field technique is simple, needs little correction and can, where useful, easily be included in routine gravity work to provide additional local information.     

基于混沌算法的地震信息网络入侵检测研究

为改善传统的基于机器学习的网络入侵检测方法只能检测已知入侵行为,对于未知入侵行为的检测存在误警率高、时效性差的不足,提出一种基于混沌算法的地震信息网络入侵检测方法。创建候选地震信息网络特征-混沌变量映射模型,实现变量之间的转化;采用混沌变量迭代演化算法进行地震信息网络特征选择;使用支持向量机对最优特征进行学习,为提高地震信息网络入侵检测精度,利用柯西蜂群算法对支持向量机参数进行寻优,建立网络入侵检测优化模型。仿真实验证明,基于混沌算法的地震信息网络入侵检测方法能有效实现高检测率、低误报率的入侵检测,具有很高的应用优势。     

Automatic signal detection based on support vector machine

Introduction The automatic processing of continuous seismic data is important for monitoring earthquake, in which real data recorded by field stations located in different regions is transmitted to data cen- tre through internet or satellite communication systems. Automatic processing will run firstly on data, afterwards these automatic processing results will be reviewed and modified. The load of interactive analysis would be increase if there were more false events or missed events after run…     

正则化偏移成像的全局优化快速算法

目前,偏移后的地震剖面往往只是一个地质构造图像,还不能为后续的岩性分析和油气储层属性的提取提供更精确的信息.为了得到高分辨率真振幅的图像,建议采用正则化偏移成像方法.针对本问题数据规模大和正演算子矩阵稀疏的特点,提出采用一种新的算法--无记忆拟牛顿-模拟退火法对偏移算子方程进行求解.该方法综合了无记忆拟牛顿法优良的局部...     

滨里海盆地东缘中区块盐下构造识别方法

滨里海盆地东缘中区块在下二叠统孔谷阶沉积了巨厚的盐岩层,由于盐层速度与围岩的速度存在很大差异,造成了下伏地层在时间剖面上存在上拉现象,形成了一些假构造圈闭或者使构造幅度发生变化,成为勘探陷阱.本文针对中区块盐下构造识别的难题,分析含盐盆地速度特征和盐下构造的影响因素,提出了正演模拟、基于地震叠加速度谱的变速成图和叠前深度偏移的识别方法.综合识别结果,并相互对比验证,最终消除盐丘造成的构造假象,有效识别了盐下构造,主要目的层的平均深度误差只有1%左右.上述方法对于解决含盐盆地地区以及速度横向变化剧烈区的构造问题具有借鉴意义.     

Seismic facies analysis through musical attributes

Seismic facies analysis is a well‐established technique in the workflow followed by seismic interpreters. Typically, huge volumes of seismic data are scanned to derive maps of interesting features and find particular patterns, correlating them with the subsurface lithology and the lateral changes in the reservoir. In this paper, we show how seismic facies analysis can be accomplished in an effective and complementary way to the usual one. Our idea is to translate the seismic data in the musical domain through a process called sonification, mainly based on a very accurate time–frequency analysis of the original seismic signals. From these sonified seismic data, we extract several original musical attributes for seismic facies analysis, and we show that they can capture and explain underlying stratigraphic and structural features. Moreover, we introduce a complete workflow for seismic facies analysis starting exclusively from musical attributes, based on state‐of‐the‐art machine learning computational techniques applied to the classification of the aforementioned musical attributes. We apply this workflow to two case studies: a sub‐salt two‐dimensional seismic section and a three‐dimensional seismic cube. Seismic facies analysis through musical attributes proves to be very useful in enhancing the interpretation of complicated structural features and in anticipating the presence of hydrocarbon‐bearing layers.     

A texture‐based region growing algorithm for volume extraction in seismic data

We present a novel approach to automated volume extraction in seismic data and apply it to the detection of allochthonous salt bodies. Using a genetic algorithm, we determine the optimal size of volume elements that statistically, according to the U‐test, best characterize the contrast between the textures inside and outside of the salt bodies through a principal component analysis approach. This information was used to implement a seeded region growing algorithm to directly extract the bodies from the cube of seismic amplitudes. We present the resulting three‐dimensional bodies and compare our final results to those of an interpreter, showing encouraging results.     

3D modeling of geological anomalies based on segmentation of multiattribute fusion

3D modeling of geological bodies based on 3D seismic data is used to define the shape and volume of the bodies, which then can be directly applied to reservoir prediction, reserve estimation, and exploration. However, multiattributes are not effectively used in 3D modeling. To solve this problem, we propose a novel method for building of 3D model of geological anomalies based on the segmentation of multiattribute fusion. First, we divide the seismic attributes into edge- and region-based seismic attributes. Then, the segmentation model incorporating the edge- and region-based models is constructed within the levelset-based framework. Finally, the marching cubes algorithm is adopted to extract the zero level set based on the segmentation results and build the 3D model of the geological anomaly. Combining the edge-and region-based attributes to build the segmentation model, we satisfy the independence requirement and avoid the problem of insufficient data of single seismic attribute in capturing the boundaries of geological anomalies. We apply the proposed method to seismic data from the Sichuan Basin in southwestern China and obtain 3D models of caves and channels. Compared with 3D models obtained based on single seismic attributes, the results are better agreement with reality.     

Common-offset common-reflection-surface attributes estimation with differential evolution

Common-reflection surface is a method to describe the shape of seismic events, typically the slopes (dip) and curvature portions (traveltime). The most systematic approach to estimate the common-reflection surface traveltime attributes is to employ a sequence of single-variable search procedures, inheriting the advantage of a low computational cost, but also the disadvantage of a poor estimation quality. A search strategy where the common-reflection surface attributes are globally estimated in a single stage may yield more accurate estimates. In this paper, we propose to use the bio-inspired global optimization algorithm differential evolution to estimate all the two-dimensional common-offset common-reflection surface attributes simultaneously. The differential evolution algorithm can provide accurate estimates for the common-reflection surface traveltime attributes, with the benefit of having a small set of input parameters to be configured. We apply the differential evolution algorithm to estimate the two-dimensional common-reflection surface attributes in the synthetic Marmousi data set, contaminated by noise, and in a land field data with a small fold. By analysing the stacked and coherence sections, we could see that the differential evolution based common-offset common-reflection surface approach presented significant signal-to-noise ratio enhancement.     

基于数据增广和CNN的地震随机噪声压制

卷积神经网络(Convolutional Neural Network,CNN)是一种基于数据驱动的学习算法,简化了传统从特征提取到分类的两阶段式处理任务,被广泛应用于计算机科学的各个领域.在标注数据不足的地震数据去噪领域,CNN的推广应用受到限制.针对这一问题,本文提出了一种基于数据生成和增广的地震数据CNN去噪框架.对于合成数据,本文对无噪地震数据添加不同方差的高斯噪声,增广后构成训练集,实现基于小样本的CNN训练.对于实际地震数据,由于无法获得真实的干净数据和噪声来生成训练样本集,本文提出一种直接从无标签实际有噪数据生成标签数据集的方法.在所提出的方法中,我们利用目前已有的去噪方法从实际地震数据中分别获得估计干净数据和估计噪声,前者与未知的干净数据具有相似纹理,后者与实际噪声具有相似的概率分布.人工合成数据和实际数据实验结果表明,相较于F-X反褶积,BM3D和自适应频域滤波算法,本文方法能更好地压制随机噪声和保护有效信号.最后,本文采用神经网络可视化方法对去噪CNN的机理进行了探索,一定程度上解释了网络每一层的学习内容.     

粗集理论在地震储层预测中的应用

在地震储层预测中,可采用的地震属性种类繁多,但太多地震属性常常会起到干扰作用,影响储层的预测精度,因此,为提高地震储层预测精度,把粗糙集理论融入到地震属性的优化中,利用粗糙集理论所具有的提取有用属性、简化信息处理的能力,优选出地震属性中的敏感属性是本文的研究目的,本文采用了一种基于属性方差的自组织神经网络量化方法,并运用基于区别矩阵的属性频率约简算法对地震属性进行优选,实例分析表明:该方法可行有效,可以最大限度地删除冗余地震属性,用优选出的敏感属性组合对多种储层参数进行预测均已取得了较好的效果.