Shot Domain Research Articles

Iterative Separation of Blended Seismic Data in Shot Domain Using Deep Learning

Accurate deblending techniques are essential for the successful application of blended seismic acquisition. Deep-learning-based deblending methods typically begin by performing a pseudo-deblending operation on blended data, followed by further processing in either the common-shot domain or a non-common-shot domain. In this study, we propose an iterative deblending framework based on deep learning, which directly addresses the blended data in the shot domain, eliminating the need for pseudo-deblending and domain transformation. This framework is built around a unique architecture, termed WNETR, which derives its name from its W-shaped network structure that combines U-Net and Transformer. During testing, the trained WNETR is incorporated into the iterative framework to extract useful signals iteratively. Tests on synthetic data validate the effectiveness of the proposed deblending iterative framework.

AdaTreeFormer: Few shot domain adaptation for tree counting from a single high-resolution image

The process of estimating and counting tree density using only a single aerial or satellite image is a difficult task in the fields of photogrammetry and remote sensing. However, it plays a crucial role in the management of forests. The huge variety of trees in varied topography severely hinders tree counting models to perform well. The purpose of this paper is to propose a framework that is learnt from the source domain with sufficient labeled trees and is adapted to the target domain with only a limited number of labeled trees. Our method, termed as AdaTreeFormer, contains one shared encoder with a hierarchical feature extraction scheme to extract robust features from the source and target domains. It also consists of three subnets: two for extracting self-domain attention maps from source and target domains respectively and one for extracting cross-domain attention maps. For the latter, an attention-to-adapt mechanism is introduced to distill relevant information from different domains while generating tree density maps; a hierarchical cross-domain feature alignment scheme is proposed that progressively aligns the features from the source and target domains. We also adopt adversarial learning into the framework to further reduce the gap between source and target domains. Our AdaTreeFormer is evaluated on six designed domain adaptation tasks using three tree counting datasets, i.e. Jiangsu, Yosemite, and London. Experimental results show that AdaTreeFormer significantly surpasses the state of the art, e.g. in the cross domain from the Yosemite to Jiangsu dataset, it achieves a reduction of 15.9 points in terms of the absolute counting errors and an increase of 10.8% in the accuracy of the detected trees’ locations. The codes and datasets are available at https://github.com/HAAClassic/AdaTreeFormer.

An novel SDA-CNN few shot domain adaptation framework for silent speech recognition

In BCI (brain-computer interface) applications, it is difficult to obtain enough well-labeled EEG data because of the expensive annotation and time-consuming data capture procedure. Conventional classification techniques that repurpose EEG data across domains and subjects lead to significant decreases in silent speech recognition classification accuracy. This research provides a supervised domain adaptation using Convolutional Neural Network framework (SDA-CNN) to tackle this problem. The objective is to provide a solution for the distribution divergence issue in the categorization of speech recognition across domains. The suggested framework involves taking raw EEG data and deriving deep features from it and the proposed feature selection method also retrieves the statistical features from the corresponding channels. Moreover, it attempts to minimize the distribution divergence caused by variations in people and settings by aligning the correlation of both the source and destination EEG characteristic dissemination. In order to obtain minimal feature distribution divergence and discriminative classification performance, the last stage entails simultaneously optimizing the loss of classification and adaption loss. The usefulness of the suggested strategy in reducing distributed divergence among the source and target Electroencephalography (EEG) data is demonstrated by extensive experiments carried out on KaraOne datasets. The suggested method achieves an average accuracy for classification of 87.4% for single-subject classification and a noteworthy average class accuracy of 88.6% for cross-subject situations, which shows that it surpasses existing cutting-edge techniques in thinking tasks. Regarding the speaking task, the model’s median classification accuracy for single-subject categorization is 86.8%, while its average classification accuracy for cross-subject classification is 87.8%. These results underscore the innovative approach of SDA-CNN to mitigating distribution discrepancies while optimizing classification performance, offering a promising avenue to enhance accuracy and adaptability in brain-computer interface applications.

Sparse seismic data regularization in shot and trace domains using a residual block autoencoder based on the fast Fourier transform

The increasing use of sparse acquisitions in seismic data acquisition offers advantages in cost and time savings. However, it results in irregularly sampled seismic data, adversely impacting the quality of the final images. In this paper, we develop the residual block fast Fourier transform-convolutional autoencoder (ResFFT-CAE) network, a convolutional neural network with residual blocks based on the Fourier transform. Incorporating residual blocks allows the network to extract high- and low-frequency features from seismic data. The high-frequency features capture detailed information, whereas the low-frequency features integrate the overall data structure, facilitating superior recovery of irregularly sampled seismic data in the trace and shot domains. We evaluate the performance of the ResFFT-CAE network on the synthetic and field data. On synthetic data, we compare the ResFFT-CAE network with the compressive sensing method using the curvelet transform. On field data, we conduct comparisons with other neural networks, such as the CAE and U-Net. The results demonstrate that the ResFFT-CAE network consistently outperforms other approaches in all scenarios. It produces images of superior quality, characterized by lower residuals and reduced distortions. Furthermore, when evaluating model generalization, tests using models trained on synthetic data also exhibit promising results. In conclusion, the ResFFT-CAE network indicates great promise as a highly efficient tool for regularizing irregularly sampled seismic data. Its excellent performance suggests potential applications in the preconditioning of seismic data analysis and processing flows.

Textual entailment classification using syntactic structures and semantic relations

In this paper, we propose a method exploiting syntactic structure, semantic relations and word embeddings for recognizing textual entailment. The sentence pairs are analyzed using their syntactic structure and categorization of sentences in active voice, sentences in passive voice and sentences holding copular relations. The main syntactic relations such as subject, verb and object are extracted and lemmatized using a lemmatization algorithm based on parts-of-speech. The subject-to-subject, verb-to-verb and object-to-object similarity is identified using enhanced Wordnet semantic relations. Further similarity is analyzed using modifier relation, number relation, nominal modifier relation, compound relation, conjunction relation and negative relation. The experimental evaluation of the method on Stanford Natural Language Inference dataset shows that the accuracy of the method is 1.4% more when compared to the state-of-the-art zero shot domain adaptation methods.

Seismic deghosting using convolutional neural networks

Ghost reflections deteriorate the quality of seismic data in towed-streamer acquisition, and various acquisition and processing solutions have been proposed to remove them from seismic data. A common issue with the proposed solutions is their limited ability to remove source-side ghosts because of the sparse source sampling. Satisfactory receiver-side deghosting solutions are facilitated by complementary measurements (e.g., particle motion data) for wavefield separation and also can be achieved using pressure data acquired at a single recording level only. We develop a solution based on convolutional neural networks (CNNs) to remove source- and receiver-side ghosts in the shot domain. The solution does not require complementary measurements, i.e., it can remove ghost reflections in conventional pressure data measured at a single recording level. Our method requires knowledge of the acquisition geometry to create training data that replicate the field acquisition geometry and require the ocean floor bathymetry to be known. A CNN learns to map ghost-contaminated gathers to corresponding ghost-free gathers through an iterative training process. We find that the CNN-based deghosting operator can remove ghost reflections from previously unseen data and demonstrate that the solution generalizes well when training is done on models unrelated to the actual field geology.

Seismic Data Enhancement Based on Common-Reflection-Surface-Based Local Slope and Trimmed Mean Filter

Seismic data often contain noise that can disturb or mask effective information. Noise elimination is an important and challenging task in seismic signal processing. Considering the high amplitude continuity of seismic events in the shot domain, this article proposes a structure-oriented denoising method that can enhance the effective events and suppress disturbing noise, including both incoherent and coherent noise. Based on the common-reflection-surface (CRS) travel time, the local slope of seismic events in the shot domain is deduced and estimated to provide structural information for plane-wave prediction. The proposed CRS-based slope depends on fewer parameters (two in 2D) than the conventional full CRS travel time (three in 2D), making it computationally efficient. Using the local slope, the third dimension is created using the plane-wave differential equation to predict the current trace from its neighbor traces and trimmed mean filtering (TMF) is applied in this dimension. The added dimension can be regarded as flattening the seismic events within a neighboring window and collapsing after application of the TMF. Synthetic and field datasets are employed to demonstrate the effectiveness of the proposed structure-oriented TMF. Compared with wavelet and plane-wave destruction methods, the proposed method can preserve more useful information with greater continuity in amplitude.

An Unsupervised Deep Learning Method for Direct Seismic Deblending in Shot Domain

An Unsupervised Deep Learning Method for Direct Seismic Deblending in Shot Domain

Improving signal fidelity for deep learning‐based seismic interference noise attenuation

AbstractDeep learning has shown a considerable potential to significantly improve processing efficiency but has not yet been widely deployed to production projects of seismic signal separation such as seismic interference attenuation. The main reasons are: First, the industry has high standards for signal fidelity, which are critical for the success of subsequent seismic imaging, and deep neural network methods have not yet matched the required level; second, the network's interpretability issue has affected many geophysicists and sponsors’ trust in the deep learning technique. To develop deep neural network methods towards the end of benefiting real‐world production, we first attempt to better understand their performance, especially in how they make use of local and global features of the data. A novel quantitative research of the overall network model behaviour on synthetic data is conducted. We simulate three types of coherent seismic data components in the shot domain, blend them together and then train a network to separate them. In this process, random noise, a component having only learnable local features, is selectively injected into the network's training pairs. Three network models sharing the same architecture are trained individually, and they show distinctive behaviours when applied to the same test data. Step‐by‐step analysis of each of them reveals that training the network with additional random noise injected into both the input and the output channel of the desired signal can lead to a decent prediction of the coherent noise based on good learning of the global features and, in the meantime, preserve almost all the data information from being lost. We propose this key lesson we learnt as a new method to improve the network's signal fidelity for shot‐domain seismic interference attenuation, which is essentially a signal separation task. Its effectiveness is demonstrated on field data from Africa with a comparison to a conventional physics‐based seismic interference attenuation method used in production.

3-D pre-stack diffraction separation by extending the PWD method with parametrized local slope

SUMMARY Diffractions comprise the seismic response of subsurface geological discontinuities and thus can provide detailed geological information during 3-D seismic exploration. The separation of weak diffractions from specular reflections is challenging, especially when the diffractions and reflections have similar kinematical characteristics in the 3-D pre-stack case. Conventional separation methods often estimate the local slope based on optimization or the Hilbert transform, which directly depends on the distribution behaviour of seismic events and may lead to the aliasing effects of the hyperbolic reflected and diffracted slopes in the shot domain. In this study, a different method is employed: local slopes are parametrized and constrained to the normal moveout velocity and ray parameter, which can distinguish reflections and diffractions in the shot domain and enhance the stability and accuracy of local slopes. Interestingly, when the receiver line and the geological edge are coplanar, the corresponding edge diffractions and reflections exhibit extremely similar behaviour, rendering them indistinguishable. Considering this phenomenon, a 3-D pre-stack diffraction separation strategy is proposed based on the estimation of the local slopes in two orthogonal directions. Thus, the accurate local slope can be used for plane-wave destruction when separating diffractions in the 3-D pre-stack domain. Synthetic and field data applications demonstrate that the proposed separation strategy is effective and can obtain high-quality diffraction wavefields for detecting the subsurface discontinuous structure.

Seismic de-multiple strategy in the submarine slope of Taiwan accretionary wedge

Reducing multiple contaminations in reflection seismic data remains one of the primary challenges in marine seismic data processing. Besides geological settings, its effectiveness is also dependent on the multiple removal methods. In this study, we undertook two legacy 2D multi-channel seismic data crossing the accretionary wedge off SW Taiwan to test the efficiency of various multiple-attenuation scenarios. The tectonic domain has resulted from the incipient arc-continent collision between the northern rifted margin of the South China Sea and the Luzon volcanic arc. The wedge extends from shallow water to deep water bathymetries, hence promoting both short-period and long-period multiples within the seismic records. A cascade of de-multiple methods was tested to attenuate multiple energy under various seafloor bathymetry and tectonic areas. The first step relies on the periodicity nature of multiples. Spatial dependent predictive deconvolution in the x-t domain was performed to attenuate reverberations and improve temporal resolution in the time domain. Wave-equation multiple attenuation (WEMA) was applied to suppress the water layer multiples based on a combination of numerical wave extrapolation in the shot domain through water layer and water bottom reflectivity. Surface-related multiple elimination (SRME) aimed to attenuate the residual water bottom multiple and peg-leg multiple by assuming surface-related multiples can be kinematically predicted via convolution of pre-stack seismic traces at possible surface multiple reflection locations. The second step exploits the spatial move-out difference behavior between primaries and multiples. Parabolic Radon transforms far-offset multiples by subtracting noise energy in the τ-p domain, whereas the frequency-wave number (F-K) filter aimed to eliminate any residual multiples energy in the F-K domain. Predictive deconvolution improved seismic resolution and suppressed sea-bottom reverberation energy in the continental and lower wedge slopes, but not in the upper wedge slope. WEMA, Radon filter, and F-K filter reduced multiples energy both at the continental slope and wedge slope; whereas SRME made minimal impact on both areas. Since the reflection seismic datasets stretch diverse tectonic environments and water depth, there was no single multiple attenuation method capable to suppress multiples in all tectonic environments and bathymetry.

Deep learning-based shot-domain seismic deblending

To streamline the fast-track processing of large data volumes, we have developed a deep learning approach to deblend seismic data in the shot domain based on a practical strategy for generating high-quality training data along with a list of data conditioning techniques to improve the performance of the data-driven model. We make use of unblended shot gathers acquired at the end of each sail line, to which the access requires no additional time or labor costs beyond the blended acquisition. By manually blending these data, we obtain training data that are fully adapted to the given survey, while having good control of their ground truth. Furthermore, we train a deep neural network using multichannel inputs that include adjacent blended shot gathers as additional channels. The prediction of the blending noise is added in as a related and auxiliary task with the main task of the network being the prediction of the primary-source events. Blending noise in the ground truth is scaled down during the training and validation process due to its excessively strong amplitudes. As part of the process, the to-be-deblended shot gathers are aligned by the blending noise. Implementation of field blended-by-acquisition data demonstrates that introducing the suggested data conditioning steps can considerably reduce the leakage of primary-source events in the deep part of the blended section. The complete proposed approach performs almost as well as a conventional algorithm in the shallow section and finds a great advantage in efficiency. It performs slightly worse for larger traveltimes, but it still removes the blending noise efficiently.

Prestack diffraction separation by parameterizing the reflection local slope

Diffracted waves provide an opportunity to detect small-scale subsurface structures because they give wide illumination direction of geologic discontinuities, such as faults, pinch outs, and collapsed columns. However, separating diffracted waves is challenging because diffracted waves have greater geometric amplitude losses and are generally weaker than reflections. To retain more diffracted waves, we have developed a prestack diffraction separation method based on the local slope pattern and plane-wave destruction (PWD) method. In general, it is difficult to distinguish between hyperbolic reflections and hyperbolic diffractions using data-driven local slope estimation in the shot domain. Therefore, we transfer slope estimation in the shot domain to velocity analysis in the common-midpoint domain and ray parameter calculation in the stack domain. The connection between local slope and the normal moveout velocity and the surface-ray parameter is known, which provides a novel approach for estimating the local slope of hyperbolic reflected waves in the shot domain. The estimated slope can provide an exact slope-based operator for the PWD method, thus allowing the PWD to separate diffracted waves from reflected waves in the shot domain. Synthetic and field data tests demonstrate the feasibility and effectiveness of our prestack diffraction separation method.

Modeling of the complex hydrocarbon traps by the shot domain acoustic finite difference method and data-processing

Numerical modeling studies have a widespread application in exploration seismology in order to understand the seismic reflection responses of hydrocarbon traps formed in relation to tectonic structure, lithological changes and unconformities in complex geological environments and to develop effective data processing strategies. In this study, the seismic modeling of two important hydrocarbon trap models (Granite Wash and Normal Fault Trap) was performed by the Finite Difference Method (FDM), which provides the solution of the acoustic wave equation. Seismic data models were carried out in the pre-stack shot environment, and the obtained shot data were passed through appropriate data-processing stages to obtain stack and migration (zero offset) sections. By converting the obtained migration sections to depth, the spatial location and dimensions of hydrocarbon traps on the section were determined and it has been observed that they are compatible by comparing with the initial geological models. Thus, the seismic responses of hydrocarbon trap structures were learned, the importance of data processing was understood, and zero offset cross-sections were obtained by processing of the generated synthetic shot records. Accordingly, it was observed that it is appropriate to make more and frequent shots in the investigation of granite wash type traps which are thin and short-width , whereas it is useful to make relatively less frequent shots in order to reduce the scattering intensity caused by the discontinuities of the fault type structures. Consequently, it is thought that before the field study for the hydrocarbon exploration, the modeling of the pre-stack shot instead of the post-stack modeling (zero offset) will contribute to the development of the data processing stages and the testing of the seismic section interpretation. In the future, such synthetic models and data processing will be developed for different complex trap structures and real data.

Case study: Improving the quality of the seismic reflection image for a Fujian mineral exploration data set with offset-domain common-image gathers

Seismic reflection is a proven and effective method commonly used during exploration of deep mineral deposits in Fujian, China. In seismic data processing, rugged depth migration based on wave-equation migration can play a key role in handling surface fluctuations and complex underground structures. Because wave-equation migration in the shot domain cannot output offset-domain common-image gathers in a straightforward way, the use of traditional tools for updating the velocity model and improving image quality can be quite challenging. To overcome this problem, we have used the attribute migration method. This works by sorting the migrated stack results for every single-shot gather into the offset gathers. The value of the offset that corresponds to each image point is obtained from the ratio of the original migration results to the offset-modulated shot-data migration results. A Gaussian function is proposed to map every image point to a certain range of offsets. This helps improve the signal-to-noise ratio, which is especially important in handling low-quality seismic data obtained during mineral exploration. Residual velocity analysis is applied to these gathers to update the velocity model and improve the image quality. The offset-domain common-image gathers are also used directly for real mineral exploration seismic data with rugged depth migration. After several iterations of migration and updating the velocity, our procedure achieves an image quality better than the one obtained with the initial velocity model. The results can help with the interpretation of thrust faults and deep deposit exploration.

Effect of sweep parameters on correlation wavelet

Vibroseis is an important seismic exploration method whose sweep signal parameters vary with rock layers deep in the earth. Controlling of sweep signal frequency band is extremely useful method in vibroseis data acquisition. The vibrator is a controlled energy source employed in seismic methods. Nowadays it is widely used and preferable seismic source for land seismic exploration. Suitable input signal is typically a sinusoid that changes frequency within a bandwidth. To obtain high-quality subsurface images, amplitude of the sweep signal and Klauder wavelet is important parameter of vibroseis. In this study, amplitude of Klauder wavelet variations with respect to different sweep parameters are investigated; tests have been made based on a variety of sweep numbers, sweep length, peak force ratio and sweep bandwidth. The results have been compared with theoretical values and the tested parameters are evaluated in the shot and processing domain.

Least-Squares Gaussian Beam Transform for Deblending Distance-Separated Simultaneous Sources

Simultaneous-source acquisition saves enormous acquisition costs and greatly enhances the quality of seismic data. However, it brings challenges to subsequent data processing due to the intense interferences of neighbor sources. Deblending is one of the methods to solve the problem of crosstalk noise. The deblending applied in the shot domain does not require the random scheduling that is used in the conventional simultaneous-source acquisition, so it has more flexibility. However, because the records of different sources show continuous traces in common shot gathers, they have the same characteristics and are difficult to separate directly. In this article, we use the least-squares Gaussian beam transform (LSGBT) to separate the useful seismic signals from crosstalk noise in the distance-separated simultaneous-sourcing (DSSS) survey and propose a novel deblending framework based on the LSGBT in the shot domain. Unlike most state-of-the-art Gaussian beam approaches that construct Gaussian beams in the frequency domain, the LSGBT constructs time-domain Gaussian beams, which can be considered as functions of amplitude, position, dip field, and so on. The essence of the proposed algorithm is that the blended data can be decomposed into the Gaussian beams that represent different dip-angle components. Thus, the single-source seismic records can be reconstructed in terms of a carefully selected combination of dip-oriented Gaussian beams. Two synthetic examples and one field data example show that the iterative deblending scheme based on LSGBT obtains better performance than the conventional frequency-wavenumber-based methods.

Cross-streamer wavefield reconstruction through wavelet domain learning

Seismic exploration in complex geologic settings and shallow geologic targets has led to a demand for higher spatial and temporal resolution in the final migrated image. Conventional marine seismic and wide-azimuth data acquisition lack near-offset coverage, which limits imaging in these settings. A new marine source-over-cable survey, with split-spread configuration, known as TopSeis, was introduced in 2017 to address the shallow-target problem. However, wavefield reconstruction in the near offsets is challenging in the shallow part of the seismic record due to the high temporal frequencies and coarse sampling that leads to severe spatial aliasing. We have investigated deep learning as a tool for the reconstruction problem, beyond spatial aliasing. Our method is based on a convolutional neural network (CNN) approach trained in the wavelet domain that is used to reconstruct the wavefield across the streamers. We determine the performance of the proposed method on broadband synthetic data and TopSeis field data from the Barents Sea. From our synthetic example, we find that the CNN can be learned in the inline direction and applied in the crossline direction, and that the approach preserves the characteristics of the geologic model in the migrated section. In addition, we compare our method to an industry-standard Fourier-based interpolation method, in which the CNN approach shows an improvement in the root-mean-square (rms) error close to a factor of two. In our field data example, we find that the approach reconstructs the wavefield across the streamers in the shot domain, and it displays promising characteristics of a reconstructed 3D wavefield.

Few shot domain adaptation for in situ macromolecule structural classification in cryoelectron tomograms.

Cryoelectron tomography (cryo-ET) visualizes structure and spatial organization of macromolecules and their interactions with other subcellular components inside single cells in the close-to-native state at submolecular resolution. Such information is critical for the accurate understanding of cellular processes. However, subtomogram classification remains one of the major challenges for the systematic recognition and recovery of the macromolecule structures in cryo-ET because of imaging limits and data quantity. Recently, deep learning has significantly improved the throughput and accuracy of large-scale subtomogram classification. However, often it is difficult to get enough high-quality annotated subtomogram data for supervised training due to the enormous expense of labeling. To tackle this problem, it is beneficial to utilize another already annotated dataset to assist the training process. However, due to the discrepancy of image intensity distribution between source domain and target domain, the model trained on subtomograms in source domain may perform poorly in predicting subtomogram classes in the target domain. In this article, we adapt a few shot domain adaptation method for deep learning-based cross-domain subtomogram classification. The essential idea of our method consists of two parts: (i) take full advantage of the distribution of plentiful unlabeled target domain data, and (ii) exploit the correlation between the whole source domain dataset and few labeled target domain data. Experiments conducted on simulated and real datasets show that our method achieves significant improvement on cross domain subtomogram classification compared with baseline methods. Software is available online https://github.com/xulabs/aitom. Supplementary data are available at Bioinformatics online.

Pre-stack Depth Migration based on VTI Media in Common Shot Domain

With the increasing complexity of seismic exploration targets, the basic imaging theory based on the basic assumption of simple isotropic media cannot meet the actual needs of oil and gas exploration at this stage. Therefore, the study of imaging methods in anisotropic media is of great significance to the exploration of various underground structures and mineral exploitation. In this paper, the SSF continuation operator for VTI anisotropic medium is derived and the one-way prestack depth migration with VTI media in common shot domain is realized. The VTI-HESS model is verified by calculation. Compared with prestack depth migration of various isotropic media, it can be seen that prestack depth migration of VTI media has better imaging effect because of considering anisotropic parameters, which is closer to the actual imaging effect of exploration data.