Multisource least-squares reverse-time migration with structure-oriented filtering

The technology of simultaneous-source acquisition of seismic data excited by several sources can significantly improve the data collection efficiency. However, direct imaging of simultaneous-source data or blended data may introduce crosstalk noise and affect the imaging quality. To address this problem, we introduce a structure-oriented filtering operator as preconditioner into the multisource least-squares reverse-time migration (LSRTM). The structure-oriented filtering operator is a nonstationary filter along structural trends that suppresses crosstalk noise while maintaining structural information. The proposed method uses the conjugate-gradient method to minimize the mismatch between predicted and observed data, while effectively attenuating the interference noise caused by exciting several sources simultaneously. Numerical experiments using synthetic data suggest that the proposed method can suppress the crosstalk noise and produce highly accurate images.     

基于Contourlet域自适应Wiener阈值的同时震源波场分离

同时震源数据包含了多炮之间的串扰噪声,不能直接用于常规数据处理流程.因此,需要对混叠的波场进行分离得到常规采集的单炮记录.本文基于稀疏迭代反演分离,提出了一种具有尺度与空间自适应的Wiener阈值选取方法.该阈值选取方法能够根据不同迭代环境计算不同尺度下串扰噪声的方差和不同空间位置有效信号的方差,从而自适应调整阈值大小,最终通过对变换域系数进行收缩来达到去除串扰噪声的目的.理论模型数据和实际数据测试结果表明,本文方法能够快速有效地压制串扰噪声和保护弱有效信号,取得了比Contourlet域子带一致Wiener阈值方法和Curvelet域指数衰减阈值方法更好的分离效果.     

基于稀疏约束反演的三维混采数据分离

混合震源采集技术相对于传统的地震数据采集,在极大提高采集效率的同时引入了混叠噪声,很大程度上影响了成像结果的精度.二维混采数据中,我们通常利用混叠噪声在非共炮域呈非相干分布这一特点来压制混叠噪声,从而实现混合震源数据分离.相对于二维混采数据,三维混采数据具有数据量巨大,构建混合震源算子困难,混合度的增加引入了高强度混叠噪声的特点.针对上述问题,本文采用稀疏约束反演方法在Radon域实现混采数据分离,混叠噪声强度比较大的情况下,稀疏约束反演方法能够得到更高精度的分离结果;利用震源激发的GPS时间通过长记录的方式在共接收点道集对上一次迭代分离结果做混合、伪分离,实现了单个共接收点道集自身混合、伪分离,避免了对整个数据做运算,同时不需要构建混合震源算子.通过模拟数据和实际数据计算来验证上述方法的适用性.     

基于全变分原理的多震源混合数据直接偏移方法

多震源混合地震采集技术,即将多个震源以一定编码方式连续地激发,得到多炮混合的地震数据.该技术能减少地震采集时间,节约采集成本,但是混合数据的直接偏移会在成像剖面中引入严重的串扰噪声,影响成像效果.从数学上看,地震成像属于典型的数学物理反问题,可以采用线性反演方法求解一个正则化约束的最小二乘(LS)优化问题,获得更高质量的成像结果.全变分(TV)正则化方法是图像去噪和复原领域中广泛应用的热点技术,其能在去除噪声的过程中保留图像的边缘信息和不连续性.在对TV图像去噪复原方法原理分析的基础上,本文将多震源混合数据直接偏移成像问题转换成图像复原的极小化能量泛函问题,用TV正则化代替传统最小二乘偏移(LSM)中的L2范数正则化,提出基于全变分原理的混合数据直接偏移方法.该方法使用基于梯度的快速迭代收缩阈值与快速梯度投影组合算法——FISTA/FGP求解最优化问题,能有效压制串扰噪声,增强同相轴连续性,提高成像分辨率.理论模型测试结果表明:将本方法应用于混合数据,无论是去噪效果还是成像精度都得到显著改善.     

Acquisition design for direct reflectivity and velocity estimation from blended and irregularly sampled data

Blended acquisition along with efficient spatial sampling is capable of providing high-quality seismic data in a cost-effective and productive manner. While deblending and data reconstruction conventionally accompany this way of data acquisition, the recorded data can be processed directly to estimate subsurface properties. We establish a workflow to design survey parameters that account for the source blending as well as the spatial sampling of sources and detectors. The proposed method involves an iterative scheme to derive the survey design leading to optimum reflectivity and velocity estimation via joint migration inversion. In the workflow, we extend the standard implementation of joint migration inversion to cope with the data acquired in a blended fashion along with irregular detector and source geometries. This makes a direct estimation of reflectivity and velocity models feasible without the need of deblending or data reconstruction. During the iterations, the errors in reflectivity and velocity estimates are used to update the survey parameters by integrating a genetic algorithm and a convolutional neural network. Bio-inspired operators enable the simultaneous update of the blending and sampling operators. To relate the choice of survey parameters to the performance of joint migration inversion, we utilize a convolutional neural network. The applied network architecture discards suboptimal solutions among newly generated ones. Conversely, it carries optimal ones to the subsequent step, which improves the efficiency of the proposed approach. The resultant acquisition scenario yields a notable enhancement in both reflectivity and velocity estimation attributable to the choice of survey parameters.     

Research note: deblended-data reconstruction using generalized blending and deblending models

We introduce a concept of generalized blending and deblending, develop its models and accordingly establish a method of deblended-data reconstruction using these models. The generalized models can handle real situations by including random encoding into the generalized operators both in the space and time domain, and both at the source and receiver side. We consider an iterative optimization scheme using a closed-loop approach with the generalized blending and deblending models, in which the former works for the forward modelling and the latter for the inverse modelling in the closed loop. We applied our method to existing real data acquired in Abu Dhabi. The results show that our method succeeded to fully reconstruct deblended data even from the fully generalized, thus quite complicated blended data. We discuss the complexity of blending properties on the deblending performance. In addition, we discuss the applicability to time-lapse seismic monitoring as it ensures high repeatability of the surveys. Conclusively, we should acquire blended data and reconstruct deblended data without serious problems but with the benefit of blended acquisition.     

Separation method for multi-source blended seismic data

Multi-source seismic technology is an efficient seismic acquisition method that requires a group of blended seismic data to be separated into single-source seismic data for subsequent processing. The separation of blended seismic data is a linear inverse problem. According to the relationship between the shooting number and the simultaneous source number of the acquisition system, this separation of blended seismic data is divided into an easily determined or overdetermined linear inverse problem and an underdetermined linear inverse problem that is difficult to solve. For the latter, this paper presents an optimization method that imposes the sparsity constraint on wavefields to construct the object function of inversion, and the problem is solved by using the iterative thresholding method. For the most extremely underdetermined separation problem with single-shooting and multiple sources, this paper presents a method of pseudo-deblending with random noise filtering. In this method, approximate common shot gathers are received through the pseudo-deblending process, and the random noises that appear when the approximate common shot gathers are sorted into common receiver gathers are eliminated through filtering methods. The separation methods proposed in this paper are applied to three types of numerical simulation data, including pure data without noise, data with random noise, and data with linear regular noise to obtain satisfactory results. The noise suppression effects of these methods are sufficient, particularly with single-shooting blended seismic data, which verifies the effectiveness of the proposed methods.     

Simultaneous sources deblending via frequency-varying filter in cross-spread azimuth-offset domain

Simultaneous sources acquisition, also referred to as “blended acquisition”, involves recording two or more shots simultaneously. It allows for denser spatial sampling and can greatly speed up the field data acquisition. Thus, it has potential advantage to improve seismic data quality and reduce acquisition cost. In order to achieve the goal of blended acquisition, a deblending procedure is necessary. It attenuates the interference and thus improves the resolution of the pre-stack time migration image. In this paper, we propose an efficient deblending method, which applies frequency-varying median and mean filters to cross-spread azimuth-offset gathers (XSPR-AO). The method can be used with variable window sizes according to the characteristics of the interference. The effectiveness of the method is validated by a field data example.     

基于稀疏约束和多源激发的地震数据高效采集方法

地震勘探目标日趋复杂化和精细化,"两宽一高"等采集技术获得了广泛应用,从而导致当前地震数据采集周期越来越长、成本越来越高,如何解决日益增长的勘探成本问题成为当前地震采集领域的研究热点之一.针对上述问题,本文首先开展了基于稀疏性的地震数据高效采集方法理论研究,对地震数据稀疏性基本理论、稀疏约束下随机采样及其数据重建方法进行了深入探讨,提出使用改进的分段随机采样方法灵活地进行实际地震采集测网设计;详细阐述了多源地震激发方法,对多源地震数据分离方法开展了深入研究,提出了基于小窗口中值滤波与稀疏约束联合随机去噪的多源数据分离方法,并在数据分离处理中取得了较好的效果;将上述两种地震数据采集方案有机结合,提出了1)规则多源、随机检波点(DmsRg)、2)随机多源、规则检波点(RmsDg)和3)随机多源、随机检波点(RmsRg)等三种高效采集方案及相应的数据重建方案,满足了后续常规化数据处理的要求,并讨论了多源激发对数据成像的影响.基于Marmousi模型数据的数值试验表明,本文构建的基于稀疏约束和多源激发的高效采集方法理论对于提高地震数据采集效率、降低勘探成本具有重要的应用价值,建立的数据重建方法流程可以取得和常规数据接近的成像结果.本文方法虽然在数值试验中取得了较为理想的效果,但还需要得到野外实际数据采集的进一步检验.     

同时震源数据的加权结构增强最小二乘逆时偏移

同时震源（Simultaneous-source,SS）地震采集技术能有效地提高地震数据采集效率,但直接对SS混合数据偏移成像会在最后的成像剖面中引入很强的串扰噪声.将SS数据偏移成像看作一个反演问题,利用最小二乘（Least-squares,LS）求解是压制SS直接成像中串扰噪声的一种有效尝试.构造增强滤波（Structure-enhancing filter,SE）约束的最小二乘逆时偏移（LSRTM）方法可以有效地压制SS数据成像中的串扰噪声,但SE实质为低通滤波,会将成像中的陡倾角等细节信息平滑涂抹,降低成像分辨率.本文在利用SE对LSRTM约束的基础上,提出了基于加权构造增强约束的LSRTM方法（WSE-LSRTM）并应用于SS数据的反演成像中.该方法不仅能够有效地压制串扰噪声（cross-talk）、保留结构信息,而且可以保护成像中的陡倾角结构不被过度平滑而破坏.在对简单模型和复杂Marmousi模型的数值测试中,该方法都取得了良好的效果.     

Using surface multiples to estimate primaries by sparse inversion from blended data

For data acquired with conventional acquisition techniques, surface multiples are usually considered as noise events that obscure the primaries. However, in this paper we demonstrate that for the situation of blended acquisition, meaning that different sources are shooting in a time‐overlapping fashion, multiples can be used to ‘deblend’ the seismic measurements. We utilize the recently introduced estimation of primaries by sparse inversion (EPSI) methodology, in which the primary impulse responses are considered to be the unknowns in a large‐scale inversion process. With some modifications the estimation of primaries by sparse inversion method can be used for blended seismic data. As output this process gives unblended primary impulse responses with point sources and receivers at the surface, which can be used directly in traditional imaging schemes. It turns out that extra information is needed to improve on the deblending of events that do not have much associated multiple energy in the data, such as steep events at large offsets. We demonstrate that this information can be brought in during acquisition and during processing. The methodology is illustrated on 2D synthetic data.     

强环噪地区可控震源高效混叠采集应用效果

随着高密度以及深层地震勘探的发展和普及,采集数据量随之急剧增加,常规采集方法已经不能有效地适应这些大数据量的勘探项目需求,高效地震采集方法成为必然.近几年来,混叠采集是较新发展的一种高效采集方法.本文在调研目前常用的混合源和同时源方法的基础上,总结提出了混叠采集的新概念.根据概念建立起理论正演模型进行模拟,模拟混叠采集方法与常规采集方法的区别、不同参数对混叠效果的影响等,得出相关的结论.在华北东部廊固凹陷进行了验证性试验.廊固凹陷地处华北平原,交通发达,村庄密集,存在较大的空炮率和超强的环境噪声,对资料品质有较大的影响,野外试验与模拟试验结果大致相同,也有一定的差异.与以往规则型的混叠方式不同,本研究在试验中创新性引入了任意随机与不同激发信号的混叠方式方法,取得了一些新的认识.研究结果表明:混叠采集方法能显著地提高生产效率;混叠带来的噪声可以通过不同域来去除;混叠采集需要一定的合适的空间间隔;混叠采集数据品质略差于常规方法,但可以通过提高采集密度和生产效率来弥补;混叠参数选取要考虑平衡施工效率、噪声水平和资料品质.     

基于迭代去噪的多源地震混合采集数据分离

多源地震混合采集采用随机线性编码方式同时激发多个震源,检波器连续接收地震信号,获得波场混叠的炮记录.该采集技术能够显著提升采集效率和成像质量,其实现关键在于炮分离,即将相互混叠的多源波场数据彼此分离,获得传统采集的单炮记录.最小平方算法只能得到伪分离记录,不能去除混叠噪声.在高混合度的混叠数据中,混叠噪声的能量往往数倍于有效信号,炮分离难度倍增.但在伪分离记录中,该噪声在除共炮点道集以外的其它时域均为随机分布.本文研究提出了一种多时域组合迭代去噪的炮分离技术:通过运用多级中值滤波与Curvelet阈值迭代去噪算法,在不同时域根据混叠噪声特性采用相应的去噪手段,并设计迭代算法优化炮分离结果.实际资料处理结果证明:将本方法应用于高混合度的混叠数据,无论是分离质量还是计算效率,都有明显提升.     

Inversion-based deblending using migration operators

In this paper, we compare the denoising- and inversion-based deblending methods using Stolt migration operators. We use Stolt operator as a kernel to efficiently compute apex-shifted hyperbolic Radon transform. Sparsity promoting transforms, such as Radon transform, can focus seismic data into a sparse model to separate signals, remove noise or interpolate missing traces. Therefore, Radon transforms are a suitable tool for either the denoising- or the inversion-based deblending methods. The denoising-based deblending treats blending interferences as random noise by sorting the data into new gathers, such as common receiver gather. In these gathers, blending interferences exhibit random structures due to the randomization of the source firing times. Alternatively, the inversion-based deblending treats blending interferences as a signal, and the transform models this signal by incorporating the blending operator to formulate an inversion problem. We compare both methods using a robust inversion algorithm with sparse regularization. Results of synthetic and field data examples show that the inversion-based deblending can produce more accurate signal separation for highly blended data.     

基于深度卷积神经网络的地震数据随机噪声压制

本文以提高地震数据的成像质量为目标,提出一种智能的卷积神经网络降噪框架,从带有噪声的地震数据中自适应地学习地震信号。为了加速网络训练和避免训练时出现梯度消失现象,我们在网络中加入残差学习和批标准化的方法,并采用了ReLU激活函数和Adam优化算法优化网络。此外,Marmousi和F3数据集被用来对网络进行训练和测试,经过充分训练的网络不仅能在学习中保留地震数据特征,而且能去除随机噪声。首先充分地训练网络,从中提取出随机噪声,并保留学习到的地震数据特征,之后通过重建地震数据估算测试集中的波形特征。合成记录和实际数据的处理结果显示了深度卷积神经网络在随机噪声压制任务中的潜力,并通过实验验证表明了深度卷积神经网络框架有很好的去噪效果。      

基于脉冲检测的混合震源数据分离

混合震源采集技术相对于传统地震数据采集具有改善成像质量、提高采集效率的优势.减小混合炮中单炮之间的随机延时范围能够有效的提高采集效率,但这也给之后的混采数据分离带来了影响.混采数据经伪分离后非共炮域数据中的混叠噪声明显更加集中,不利于对混叠噪声进行压制.本文提出基于脉冲检测方法对混采数据进行分离,并且与迭代的多级中值滤波方法作对比,时间延时范围较大时,两种方法都能得到很好的分离结果;时间延时范围较小时,本文方法能更有效的去除混叠噪声,同时也能更好的保留细节信息.实际数据计算结果表明,本文方法一定程度上还能够有效压制其他随机噪声.     

Deghosting by echo‐deblending

A marine source generates both a direct wavefield and a ghost wavefield. This is caused by the strong surface reflectivity, resulting in a blended source array, the blending process being natural. The two unblended response wavefields correspond to the real source at the actual location below the water level and to the ghost source at the mirrored location above the water level. As a consequence, deghosting becomes deblending (‘echo‐deblending’) and can be carried out with a deblending algorithm. In this paper we present source deghosting by an iterative deblending algorithm that properly includes the angle dependence of the ghost: It represents a closed‐loop, non‐causal solution. The proposed echo‐deblending algorithm is also applied to the detector deghosting problem. The detector cable may be slanted, and shot records may be generated by blended source arrays, the blending being created by simultaneous sources. Similar to surface‐related multiple elimination the method is independent of the complexity of the subsurface; only what happens at and near the surface is relevant. This means that the actual sea state may cause the reflection coefficient to become frequency dependent, and the water velocity may not be constant due to temporal and lateral variations in the pressure, temperature, and salinity. As a consequence, we propose that estimation of the actual ghost model should be part of the echo‐deblending algorithm. This is particularly true for source deghosting, where interaction of the source wavefield with the surface may be far from linear. The echo‐deblending theory also shows how multi‐level source acquisition and multi‐level streamer acquisition can be numerically simulated from standard acquisition data. The simulated multi‐level measurements increase the performance of the echo‐deblending process. The output of the echo‐deblending algorithm on the source side consists of two ghost‐free records: one generated by the real source at the actual location below the water level and one generated by the ghost source at the mirrored location above the water level. If we apply our algorithm at the detector side as well, we end up with four ghost‐free shot records. All these records are input to migration. Finally, we demonstrate that the proposed echo‐deblending algorithm is robust for background noise.     

基于深度卷积神经网络的剪切波分裂质量检测

剪切波分裂是分析地震各向异性的一种重要手段,常规方法是利用网格搜索获取分裂参数,再通过不同方法的测量结果对比测量结果进行质量检测,这一过程会耗费大量计算时间。本文针对这一问题提出了一种利用深度卷积神经网络对剪切波分裂进行质量检测的新方法,对使用了Resnet残差结构的深度神经网络进行训练,直接对二分量剪切波波形数据的质量进行分类。整个过程为:神经网络通过卷积层提取波形特征,计算损失函数后反向传播训练模型参数,完成迭代训练后的模型对输入波形数据正向计算自动输出类型。本文利用川西台站接收到的实际数据以及随机生成的合成数据分别对该网络进行训练,均可以获得准确的分类结果。相比于通过多种剪切波分裂方法对比测量结果的质量检测方法,基于神经网络的方法可以省略网格搜索的计算过程直接判断质量类型,在运算速度上的优势明显,并可继续通过训练提高模型的精度,为提升剪切波分裂方法在数据处理过程中的操作效率提供帮助。     

Deblending of Simultaneous-source Seismic Data using Fast Iterative Shrinkage-thresholding Algorithm with Firm-thresholding

In this paper, an improved algorithm is proposed to separate blended seismic data. We formulate the deblending problem as a regularization problem in both common receiver domain and frequency domain. It is suitable for different kinds of coding methods such as random time delay discussed in this paper. Two basic approximation frames, which are iterative shrinkage-thresholding algorithm (ISTA) and fast iterative shrinkage-thresholding algorithm (FISTA), are compared. We also derive the Lipschitz constant used in approximation frames. In order to achieve a faster convergence and higher accuracy, we propose to use firm-thresholding function as the thresholding function in ISTA and FISTA. Two synthetic blended examples demonstrate that the performances of four kinds of algorithms (ISTA with soft- and firm-thresholding, FISTA with soft- and firm-thresholding) are all effective, and furthermore FISTA with a firm-thresholding operator exhibits the most robust behavior. Finally, we show one numerically blended field data example processed by FISTA with firm-thresholding function.     

Local outlier factor as part of a workflow for detecting and attenuating blending noise in simultaneously acquired data

A number of deblending methods and workflows have been reported in the past decades to eliminate the source interference noise recorded during a simultaneous shooting acquisition. It is common that denoising algorithms focusing on optimizing coherency and weighting down/ignoring outliers can be considered as deblending tools. Such algorithms are not only enforcing coherency but also handling outliers either explicitly or implicitly. In this paper, we present a novel approach based on detecting amplitude outliers and its application on deblending based on a local outlier factor that assigns an outlier-ness (i.e. a degree of being an outlier) to each sample of the data. A local outlier factor algorithm quantifies outlier-ness for an object in a data set based on the degree of isolation compared with its locally neighbouring objects. Assuming that the seismic pre-stack data acquired by simultaneous shooting are composed of a set of non-outliers and outliers, the local outlier factor algorithm evaluates the outlier-ness of each object. Therefore, we can separate the data set into blending noise (i.e. outlier) and signal (i.e. non-outlier) components. By applying a proper threshold, objects having high local outlier factors are labelled as outlier/blending noise, and the corresponding data sample could be replaced by zero or a statistically adequate value. Beginning with an explanation of parameter definitions and properties of local outlier factor, we investigate the feasibility of a local outlier factor application on seismic deblending by analysing the parameters of local outlier factor and suggesting specific deblending strategies. Field data examples recorded during simultaneous shooting acquisition show that the local outlier factor algorithm combined with a thresholding can detect and attenuate blending noise. Although the local outlier factor application on deblending shows a few shortcomings, it is consequently noted that the local outlier factor application in this paper obviously achieves benefits in terms of detecting and attenuating blending noise and paves the way for further geophysical applications.