The popular deep learning-based karst caves recognition methods have higher precision than the traditional karst caves recognition approaches; however, they still face great challenges in processing complex seismic data and areas with strong reflection shielding. In this paper, to achieve high-efficiency and high-precision karst caves recognition results with the strong reflection shielding, we propose a wavefield separation-driven high-precision deep-learning karst caves recognition method (WCRM). WCRM is composed of the multitask two-stage seismic strong and weak reflection separation method (MTSM)-based high-precision karst caves recognition training data generation (MKCG), the high-precision deep-learning karst caves recognition network (HCRN), the loss function of WCRM, and the karst caves recognition result optimization (KCRO). MKCG aims to use the MTSM results to generate sufficient training data for HCRN; HCRN takes three-dimensional (3D) synthetic seismic data and the data augmentation results obtained by MKCG as its inputs, the corresponding karst caves recognition labels as its labels, and uses the 3D convolution kernel for high-precision karst caves recognition; The loss function of WCRM aims to calculate the loss function which focuses on the karst caves; KCRO aims to optimize the karst caves recognition results obtained by HCRN to obtain high-precision karst cave recognition results. One public synthetic 3D seismic dataset and one field 3D seismic dataset applications demonstrate that WCRM achieves high-precision karst caves recognition results.

The conventional seismic strong and weak reflection separation method (SRSM) encounters significant challenges due to the complexity of seismic data, the horizon time (or depth) accuracy of the target horizon, and the space-variant seismic wavelet, resulting in undesired seismic strong and weak reflection separation results. The popular seismic facies-guided trace-by-trace high-precision seismic SRSM can address the abovementioned issues and obtain high-precision seismic strong and weak reflection separation results; however, it still requires the horizon time of the target horizon, and its computational efficiency needs to be improved. In this article, in order to obtain high-efficiency high-precision seismic strong and weak reflection separation results without any horizon time, we propose a multitask two-stage deep learning seismic strong and weak reflection separation method (MTSM) based on the SRSM and the multitask deep learning network framework. MTSM consists of the SRSM-based seismic strong and weak reflection label automatic generation (SLG), the multitask two-stage seismic strong and weak reflection separation network (MTSN), and the energy balance loss function of MTSN. SLG aims to use SRSM and data augmentation to generate massive high-precision seismic strong and weak reflection labels, thereby providing sufficient high-precision training datasets for MTSN; MTSN aims to simultaneously output high-precision seismic strong and weak reflections, and the energy balance loss function of MTSN aims to address the imbalance between multiple loss functions resulting from the energy disparity between the seismic strong and weak reflections. An actual 3-D seismic dataset example demonstrates that MTSM has great potential as a technique for high-precision seismic strong and weak reflection separation.

The popular iterative Vertical seismic profiling (VSP) wavefield separation method has higher precision wavefield separation results than the conventional VSP scalar wavefield separation method, but it still faces challenges in achieving desirable high-precision wavefield separation results due to the time-variant wavelet, the complex wavefield, and the precision of wavefield flattening. In this paper, in order to minimize the impact of the factors mentioned above, we propose a trace-by-trace iterative VSP wavefield separation method (TISM) based on the gradually changing characteristics of VSP data in adjacent traces, the cross-correlation, and the iterative VSP wavefield separation method. TISM includes the flowchart of TISM, the target trace guided sub dataset automatic generation (TDG), the cross-correlation guided sub dataset optimization (CDO), and the cross-correlation guided wavefield separation result optimization (CWO). TDG aims to automatically generate the sub VSP dataset corresponding to the target trace and minimize the impact of the time-variant wavelet and complex wavefield. CDO aims to minimize the effect of wavefield flattening and form the high-precision wavefield flattened sub dataset for scalar wavefield separation. CWO aims to obtain the high-precision wavefield separation result. Synthetic and actual VSP data applications demonstrate that TISM can minimize the impact of the time-variant wavelet, the complex wavefield and wavefield flattening, thereby obtaining high-precision VSP wavefield separation results.

In recent years, Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) have become the mainstream segmentation methods for high-resolution remote sensing images (HRSIs). CNNs can quickly acquire the correlation between local neighboring pixels through convolutional operations, but it is difficult to establish global contextual relationships, resulting in limited segmentation accuracy. ViTs are able to establish reliable global semantic dependencies through the mechanism of self-attention, but the quadratic computational complexity of self-attention makes the ViTs present high accuracy but low efficiency. Therefore, in this letter, to balance the efficiency and accuracy of HRSIs segmentation, we combine the respective advantages of CNNs and ViTs to propose the FSegNet network. Specifically, we introduce FasterViT and utilize its efficient hierarchical attention to mitigate the surge in self-attention computation due to the high resolution of HRSIs. On this basis, we construct a lightweight decoder based on intensive computation, which achieves fast generation of segmentation results by reshaping and mapping multi-level features. Experiments on the ISPRS Potsdam and Vaihingen datasets show that the proposed FSegNet best balances performance and efficiency. The code is available at https://github.com/Rowan-L/FSegNet.

Multihorizon tracking deep learning methods have higher efficiency than single-horizon tracking deep learning methods, but they still have a great challenge to adapt to complex seismic data, resulting in undesired horizon tracking results. To achieve high-precision horizon tracking results, we propose a boundary-constrained multitask multihorizon tracking (BMTM) based on semantic segmentation, instance segmentation, and multitask learning (ML). The core idea of BMTM is to transform multihorizon tracking into strata recognition, using where the strata boundaries serve as the tracking results. BMTM comprises three components: strata label automatic generation, strata boundary-constrained multihorizon tracking network (BMTN), and strata boundary-constrained loss function. Strata label automatic generation automatically generates strata labels based on the input horizon labels. BMTN consists of a shared layer, an auxiliary task, and a main task. The auxiliary task takes the input horizon labels as its labels and employs semantic segmentation to directly output multihorizon tracking results. Main task takes the generated strata labels as its labels and employs instance segmentation with the outputs of auxiliary task to achieve high-precision strata (or horizon) tracking results. The strata boundary-constrained loss function aims to pay more attention to the strata boundary and ultimately improve the horizon tracking precision. One public 3-D synthetic seismic dataset study demonstrated the performance of BMTM, and one field 3-D seismic dataset application demonstrated that BMTM can be used for high-precision multihorizon tracking.

Intelligent seismic facies segmentation has recently gained significant attention, particularly with the application of deep learning technologies. However, existing methods face considerable challenges when processing complex seismic data, often due to their reliance on simplistic image mapping that fails to capture intricate geological structures and spatial relationships. Unlike natural image segmentation, seismic facies segmentation requires a global perspective that accounts for these complexities. To address these limitations, we propose the UMA-Net model, which integrates a U-shaped encoder-decoder structure with a mix-transformer (MiT) and an attention-enhanced convolution module (AECM) to enhance global feature extraction and seismic information acquisition. A key innovation of this study is the proximity constraint strategy (PCS), which shifts the focus from traditional mapping to predicting geological targets by analyzing variations in adjacent seismic data. This approach significantly improves phase boundary identification and reduces phase confusion, offering a novel solution to the challenges of seismic facies segmentation. Experimental results demonstrate that UMA-Net outperforms state-of-the-art networks in metrics such as pixel accuracy (PA) and intersection over union (IoU). Applied to seismic data from the F3 Netherlands work area, UMA-Net enhances segmentation accuracy while reducing reliance on labeled datasets, making it applicable to other regions facing similar geological challenges. This study not only advances the accuracy and efficiency of seismic interpretation but also opens new avenues for deep learning applications in seismic facies segmentation, particularly in improving model generalization across diverse geological settings.

Strong and weak reflection separation is crucial for seismic interpretation. The conventional strong and weak reflection separation method (CRSM) faces great challenges, due to the complex seismic data, the target horizon accuracy and the space-variant wavelet, resulting in undesired strong and weak reflection separation results. In this paper, in order to minimize the impact of the complex seismic data, the target horizon accuracy and the space-variant wavelet, we propose a seismic facies-guided trace-by-trace high-precision strong and weak reflection separation method (SRSM), which is based on the CRSM, the seismic facies and the concept of trace-by trace processing. SRSM includes the flowchart of SRSM, the seismic facies-guided target trace two-dimensional (2D) sub-seismic dataset automatic generation (SDG), the seismic facies-guided target trace 2D sub-seismic dataset optimization (SDO), and the strong and weak reflection separation result optimization (RSO). SDG aims to automatically generate the 2D sub-seismic dataset corresponding to the target trace to reduce the impact of the complex seismic data, the target horizon accuracy and space-variant wavelet based on the seismic facies, thereby providing high-consistency 2D sub-seismic dataset. SDO aims to use the correlation algorithm to automatically optimize the SDG result to minimize the impact of the complex seismic data, the target horizon accuracy and space-variant wavelet based on the seismic facies, ultimately providing high-consistency and high-continuity 2D sub-seismic dataset for wavefield separation. RSO aims to optimize the 2D strong and weak reflection datasets obtained by wavefield separation, ultimately providing 1D high-precision strong and weak reflection seismic data corresponding to the target trace. An actual 3D seismic dataset example demonstrates that SRSM has great potential as a technique for high-precision strong and weak reflection separation.

The popular geological anomaly (such as fault, river course, cave, and crack) identification methods, such as coherence cube, semblance, likelihood, and others, usually can achieve higher precision geological anomaly identification results when applied to the target horizon flattened seismic data, comparing to their counterparts using the target horizon-unflattened seismic data. However, these methods still face great challenges in achieving high-precision geological anomaly identification results, due to the complexity of the geological structure (or the seismic data) and the horizon tracking accuracy of the target horizon. To minimize the impact of the complexity of geological structure and the horizon tracking accuracy of the target horizon in geological anomaly identification, thereby obtaining high-precision geological anomaly identification results and providing precise labels for deep-learning-based geological anomaly identification methods, we propose a seismic facies-guided high-precision geological anomaly identification method (FHGI), basing on the concept of seismic facies and the cross-correlation algorithm. FHGI contains the flowchart of FHGI, and the seismic facies-guided trace-by-trace high-precision geological anomaly identification factor calculation (FTGC); in which FTGC consists of the target horizon-based seismic data flattening (THF), the seismic facies-guided target trace 2-D subseismic dataset generation (FTG), the cross-correlation algorithm-based target horizon further flattening (CFA), and the cross-correlation coefficient-based high-precision geological anomaly identification factor calculation (CGC). The THF aims to reduce the impact of the complexity of the geological structure and provide the input 3-D seismic data for the FTG. FTG aims to automatically generate the 2-D subseismic dataset corresponding to the target trace, thereby further reducing the impact of the complexity of the geological structure and providing the input 2-D subseismic dataset for CFA. CFA takes the target trace in the result of FTG as the reference for cross-correlation functions calculation and then uses them to further flatten the target horizon in the result of FTG, thereby minimizing the impact of the horizon tracking accuracy of the target horizon and providing the input 2-D subseismic dataset for CGC. CGC takes the target trace in the result of CFA as the reference for cross-correlation coefficient calculation and then uses them for high-precision geological anomaly identification factor calculation, thereby providing high-precision geological anomaly identification results. A public synthetic seismic dataset and actual 3-D seismic dataset examples demonstrate that FHGI has great potential as a technique for geological anomaly identification.

Accurate seismic velocity inversion is crucial for oil and gas exploration. The popular fixed-scale regular-grid adjoint-state first-arrival (or first-arrival travel-time) slope tomography seismic velocity inversion method (FFAST) (or adjoint-state first-arrival slope tomography seismic velocity inversion method with fixed-scale regular-grid) can obtain good seismic velocity inversion results, but it still faces great challenges in achieving desirable high-precision seismic velocity inversion results due to its fixed-scale regular-grid. In this article, we use the multiscale staggered grid to replace the fixed-scale regular-grid in FFAST for model parametrization and propose the multiscale staggered-grid adjoint-state first-arrival (or first-arrival travel-time) slope tomography seismic velocity inversion method (MFAST), thereby obtaining high-precision seismic velocity inversion result. The staggered-grid is composed of a finite set of fixed-scale regular-grids with spatially staggered (or overlapped) relationships, which aims to change the grid coordinate to fully sample the structure information in the velocity model space with multiple fixed-scale regular-grids. The multiscale staggered-grid is composed of multiple staggered-grids with different fixed scales, which aims to adapt to the different scale complex structures in the velocity model space; in which, the large-scale staggered-grid based MFAST aims to reconstruct the large-scale background structures, thereby providing the essential guidance (or prior) information for the small-scale staggered-grid based MFAST which aims to obtain the detailed structural information. The model parametrization with multiscale staggered-grid is achieved by performing the model parametrization with regular-grid multiple times; the mean or median value of the outputs of multiple model parametrizations with regular-grid is considered the output of MFAST in the current iteration, and used to iteratively update the velocity model obtained by MFAST in the previous iteration. The checkboard and Marmousi model testing validate the effectiveness of MFAST.

In the oil and gas exploration, delineating the tectonic structure of geological targets and predicting physical properties are paramount. As we all know, seismic velocity modeling is crucial for characterizing reservoirs. Enhancing seismic inversion accuracy to obtain quantitative parameters is vital in oil reservoir exploration. Current seismic waveform inversion utilizes the amplitude and timing of seismic records to reconstruct the velocity model, thereby refining the model for seismic migration. This article addresses the convergence and computational challenges in conventional inversion by proposing an adaptive iteration method for step updates. This method outperforms inexact linear searches and empirical formulas by updating the model effectively without additional seismic simulations. In the past, gradient optimization typically requires three seismic modeling in a single iteration, whereas adaptive iteration only requires two, offering practical cost savings. To address stability issues in complex scenarios, we combine parabolic search with adaptive iteration, which is the modified optimization technique, to enhance the convergence rate of objective function. Furthermore, by employing gradient preconditioning based on illumination and multigrid strategy, we achieve full strata velocity modeling from shallow to deep layers. Numerical tests on the theoretical anomaly model and BP model validate the effectiveness and precision of the adaptive iteration, significantly reducing the computational load of seismic inversion and accurately estimating deep strata information.

The popular deep-learning-based horizon tracking methods heavily relies on large volumes of well-labeled horizon data, which face significant challenges in achieving high-precision horizon tracking with limited label (or few sample), especially when encountering complex seismic data and geological structures with 1-D limited label. In this article, we propose a limited-label multiscale multihorizon tracking method (LMMT) based on the multimodal deep learning and (1-D limited label. In this method, the horizon is characterized in the seismic trace (1-D), the seismic profile (2-D), and the horizon slice (3-D). LMMT is comprised of the flowchart of LMMT, the 1-D convolution kernel single-modal multihorizon tracking method (OMT), the high-precision high-continuity horizon and strata optimization (HHO), and the 2-D (or 3-D) convolution kernel multimodal multihorizon tracking method (TMT). OMT takes the input 1-D limited horizon labels as its labels and utilizes a 1-D convolution kernel for strata division and multihorizon tracking. HHO aims to generate the 2-D (or 3-D) high-precision and high-continuity horizon and strata based on the 3-D horizon tracking results obtained by OMT or TMT, thereby providing high-precision high-continuity horizon labels and strata for TMT. TMT incorporates the 2-D (or 3-D) high-precision high-continuity horizon obtained by HHO as its labels, integrates the random masking result of the high-precision high-continuity strata obtained by HHO as the reference information, and utilizes a 2-D (or 3-D) convolution kernel for high-precision multihorizon tracking. Two 3-D seismic dataset applications demonstrate that LMMT achieves high-precision multihorizon tracking results with limited labels.

Due to the non-Gaussian distribution of erratic noise, conventional Gaussian denoising methods often encounter substantial challenges and pressures when suppressing this kind of noise. To overcome this challenge, several state-of-the-art (SOTA) schemes, for instance, robust low-rank approximation (LRA) and deep learning (DL) methods, have been designed and achieved promising results in the treatment of erratic noise. However, these SOTA denoising methods focus mainly on matrix-based modeling representations and fail to fully reflect the correlations associated with erratic noise and valid signals in the spatial dimension and thus may display suboptimal performance. As an alternative, a robust tensor DL (RTDL) denoising method for unsupervised 3-D seismic erratic noise suppression that involves the use of a reasonable combination of tensor sparse representation (SR) and a tensor neural network (tNN) is proposed in this study. The key to RTDL is to introduce a robust tensor sparse norm for erratic noise to exhaustively exploit the spatial tubular sparse distribution properties in 3-D space; notably, adding a tensor sparse norm to the tNN model yields a new data-driven model with 3-D erratic noise reduction capabilities. To find the optimized parameters of the new model, an efficiency tensor optimization method is established on the basis of alternating minimization (Alt), the aim of which is to alternately solve two subproblems involving tensor SR and a tNN. This paper presents well-designed experiments and satisfactory results compared with those of SOTA methods based on both synthetic and real field datasets.

Due to the poorly coupled geophones present in boreholes, vertical seismic profiling (VSP) data are known to suffer from intense coupling noise, which causes severe VSP image deterioration and significantly hinders subsequent processing. Thus, diverse denoising approaches are indispensable preprocessing steps for suppressing this kind of noise to achieve good results. Among them, robust principal component analysis (RPCA) is a common signal and noise separation model that is generally considered a highly promising method for removing intense coupling noise; however, handcrafted priors have limited denoising ability, especially for the low-rank assumption of useful signals. As an alternative, following the RPCA framework, this article proposes an unsupervised iterative robust deep convolutional autoencoder (IRDCAE) model to suppress intense VSP coupling noise without any assumptions regarding valuable signals. The key to the IRDCAE approach is the use of weighted column sparsity (WCS) to characterize the behavior of the intense coupling noise, where the weight prior is derived from the pure noise component before the first break. By adding a WCS regularization term to the conventional deep convolutional autoencoder (DCAE), our IRDCAE method transforms the model from an entirely data-driven model to a model+data driven approach. Thus, the IRDCAE approach has the advantages of both RPCA and DCAE, resulting in the ability to separate intense coupling noise from useful signals in an unsupervised manner by optimizing the IRDCAE model via an alternating minimization algorithm. The exceptional performance of the IRDCAE model is exhibited with synthetic and field VSP data.

Seismic facies analysis (SFA) plays a pivotal role in the interpretation of subsurface structures, with a pressing need to develop automated techniques for analyzing 4-D prestack seismic data. Various automated SFA methods, encompassing both supervised and unsupervised paradigms, have shown encouraging potential in fulfilling this demand. Nonetheless, supervised methods heavily hinge upon precious labeled seismic datasets of high caliber, and unsupervised methods handle all seismic samples indiscriminately during training, resulting in pronounced biases toward the majority classes of seismic data. As an alternative, this letter proposes class-imbalanced deep embedding clustering (CDEC), an unsupervised deep clustering methodology devised to analyze seismic data with class-imbalanced facies distributions. Within CDEC, a focal loss meticulously tailored to address class imbalance challenges is seamlessly integrated into the classic deep convolutional embedding clustering (DCEC). By balancing weights between the minority and majority seismic samples during network training, this approach adeptly attenuates biases toward the majority classes while concurrently bolstering the efficacy of SFA. Experimental evaluations conducted on synthetic and real field datasets compellingly underscore the effectiveness and utility of the proposed CDEC method.

Due to the precision of the first break, time-variant wavelet, and strata dip angle, the popular iterative vertical seismic profiling (VSP) wavefield separation method may not yield high-precision wavefield separation results. The single-stage multi-task U-Network VSP wavefield separation method can avoid the impact of the first break, time-variant wavelet, the strata dip angle, but it faces challenge in complex VSP wavefield due to its network performance. In this paper, based on the iterative VSP wavefield separation method, the U-Network and multi-task deep learning, we propose a two-stage multi-task U-Network VSP wavefield separation method. The two-stage multi-task U-Network VSP wavefield separation method comprises the two-stage multi-task U-Network, the loss function, and the synthetic VSP training data automatic generation. The two-stage multi-task U-Network aims to simultaneously output high-precision downgoing and upgoing wavefield, as well as the residual wavefield, while the synthetic VSP training data automatic generation aims to automatically generate numerous and various VSP training data. Applications of both synthetic and actual VSP data demonstrate that the two-stage multi-task U-Network VSP wavefield separation method can be widely used for high-precision VSP wavefield separation.

In the classic local window filtering seismic data denoising methods, the target sample is usually placed at the center of the given fixed-scale filter kernel. When the target sample is located on the structure edges, the filter kernel will cross the structure edges and leads to blurry structure edges. Multiscale adaptive right-angle side window filtering (MRSF) has better edge preservation capability. However, its 2-D filter kernel and eight right-angle side windows cannot better adapt to complex data, limiting its denoising capability. We extend the 2-D multiscale filter kernel in MRSF with the 3-D multiscale filter kernel. Meanwhile, we extend the eight 2-D right-angle side windows in MRSF with multiple 2-D and 3-D trapezoidal side windows. Finally, we propose the high-dimensional multiscale adaptive trapezoidal side window filtering (HMTSF). Synthetic and field 3-D seismic data examples demonstrate the good denoising capability of HMTSF.

Automatic extraction of high-precision roads from remote sensing images is crucial for path planning and road monitoring. However, there is room to improve the accuracy and generalization of existing methods in segmentation due to the challenges posed by ground object occlusion and complex backgrounds. Most existing methods rely on convolutional neural networks (CNNs), but the limitations of convolution prevent direct semantic interaction at a distance. In contrast, Mix-Transformer obtains long-term modeling capability through the self-attention mechanism, and inspired by it, we propose a multiscale self-adaptive network (UMiT-Net) based on the U-shaped structure. First, UMiT-Net extracts global features with the efficient Mix-Transformer backbone. Second, the dilated attention module (DAM) is used in the bottleneck of the network to fuse semantic features further to ensure the connectivity of the road. Third, in the decoder, to improve the accuracy of road segmentation, we construct the multiscale self-adaptive module (MSAM), which summarizes rich scene understanding from dense contexts with strip windows conforming to road morphology, and embed an edge enhancement module (EEM) to correct road edges. Finally, we design patch expanding (PE), which solves the problem of heavy computation of upsampling due to high resolution. The experimental results show that our UMiT-Net is substantially ahead of other state-of-the-art methods and has a significant improvement in generalization ability.

Seismic facies analysis is a crucial foundation for basin-fill studies and oil and gas exploration. With its rapid development, CNN-assisted interpretation is becoming increasingly popular. However, CNN models are often considered "black boxes" that lack transparency. To understand how CNN models classify seismic facies and visualize the contribution of each seismic attribute to the final predictive scoring, we have investigated class activation map (CAM) techniques and an explainable tool called Shapley additive explanations (SHAP) value. Based on real seismic data collected in the Sichuan basin, we compared the visualization performances of CAM and SHAP methods and found that the SHAP tool has better visualization capabilities than CAM methods, which only produce heat maps with positive values. Using SHAP values, we identified the importance of each seismic attribute and refined redundant attributes. This approach establishes a connection between seismic attributes and sedimentary environments and is a prime example of the capability of deep learning to discover knowledge beyond human experience. We applied the selected seismic attributes to generate a refined CNN model and compared it to the original CNN model, demonstrating the superiority of our proposed strategy. When we compared the predicted seismic facies using the refined CNN model based on SHAP features, the conventional K-means, SVM and Gaussian Naive Bayes methods, it is observed that our predicted map aligns well with geological knowledge with less prediction errors, demonstrating the effectiveness and feasibility of our developed strategy.

Conventional vertical seismic profiling (VSP) wavefield separation method (CSM) focuses on the wavefield separation result, not the residual. Usually, it cannot obtain desirable high-precision results. Iterative zero-offset VSP (ZVSP). However, IZSM only can be used for ZVSP data, which limits its application. The scalar wavefield separation algorithm used in IZSM is based on the SVD filtering, which limits its precision. The objective function threshold of IZSM is varied with the input ZVSP data, which limits its result precision and efficiency. To improve the wavefield separation precision and efficiency, we propose a high-precision iterative VSP wavefield separation method (HISM). HISM is suitable for both ZVSP and offset VSP (OVSP) data, and also applicable for any possible scalar wavefield separation algorithm The objective function threshold of HISM can be a fixed number. Therefore, HISM can obtain high precision and efficiency results. Actual ZVSP and OVSP data application demonstrates the performance of HISM.

Side window filtering (SWF) can effectively capture detailed image edges and is widely applied in image processing. However, its fixed-scale (or fixed-size) filter kernel cannot adapt to complex images, and the final output at the target pixel is only determined by the side window output with the minimum error functional, limiting its filtering capability. To further enhance the filtering capability of SWF, we first extend the traditional side windows with fixed-scale filter kernel to multiscale side windows by introducing the multiscale filter kernels, which leads to better complex image matching. Then, we further introduce an adaptively weighted parameter, which is inversely proportional to the error functional, to fully consider the contributions of all multiscale side windows to the final output. We finally propose the multiscale adaptive SWF (MASWF). Synthetic and field seismic data examples demonstrate that MASWF is a good potential technique for seismic data random noise attenuation and can be widely used in digital signal processing fields.

To reduce the impact of nontarget horizon regions and improve horizon tracking precision, we propose an attention based two-stage U-net horizon tracking method (ATUM). The ATUM consists of the horizon region label automatic generation and the attention module based two-stage U-Net (ATUN). Horizon region label automatic generation aims to automatically generate the target horizon region label of the target horizon label. In ATUN, the two stages (stages Ⅰ and Ⅱ) consist of the conventional encoder-decoder U-Net, and the two decoder parts are connected by the attention module. Stage Ⅰ treats horizon tracking as an objection detection problem. It takes the seismic data as its input and the automatically generated target horizon region label as its label, and finally obtains the target horizon region. Stage Ⅱ takes the results of stage Ⅰ with the corresponding seismic data as its input, and finally obtains the precise horizon. Two field three-dimensional seismic dataset studies demonstrated the performance of the ATUM for high-precision horizon tracking.

High-precision seismic fault detection and fault surface extraction (or grouping) are critical steps in reservoir characterization. In this paper, basing on the gradually changing characteristics of the target faults in the adjacent 2D seismic fault profiles, we propose an efficient and high-precision automatic fault surface grouping method (EHFG) for complex 3D seismic fault data; and it is realized using pairs of adjacent 2D seismic fault profiles without human intervention. EHFG comprises of high-precision fault separation (HFS) and high-precision fault labeling (or naming) (HFL); in which, HFS aims to separate the positive-slope and negative-slope faults in the seismic fault detection result, and ultimately obtain the high-precision positive-slope and negative-slope faults; HFL aims to label the separated positive-slope and negative-slope faults, and ultimately obtain the corresponding high-precision fault surface grouping results. An actual 3D seismic fault data example demonstrates that EHFG is a good potential technique for fault surface grouping.

Erratic seismic noise, following a (known or unknown) non-Gaussian distribution, poses a formidable challenge to conventional methods of random noise attenuation. Many erratic noise cancellation methods, for instance, robust reduced-rank and sparsity-promoting filtering, have been proven to achieve promising results in overcoming this challenge. Among them, deep learning (DL) methods require no assumptions about the underlying clear seismic image and are also more robust against erratic and random noise. However, the success of existing DL-based denoising methods strongly depends on supervised learning from a large number of ground-truth seismic images affected by erratic noise and their clean counterparts, which are typically unavailable in a real-world setting. As an alternative, this article presents an unsupervised DL method for erratic-plus-Gaussian noise removal based on a robust deep convolutional autoencoder (RDCAE). In the RDCAE, the mean squared error (mse) loss in a classic DCAE is replaced by the smooth Welsch function to exploit the concept of robust image denoising. In this way, the erratic noise is downweighted by means of a curbed weight defined in terms of the Welsch function. In contrast, the random noise is diluted by combining the mean square in the Welsch function and the total variation (TV). Subsequently, the training procedures required for solving the RDCAE are derived on the basis of the backpropagation (BP) algorithm for a neural network. Experiments conducted on both synthetic and real field datasets are reported to illustrate the efficacy of the proposed method.

The 3-D seismic data reconstruction can be understood as an underdetermined inverse problem, and thus, some additional constraints need to be provided to achieve reasonable results. A prevalent scheme in 3-D seismic data reconstruction is to compute the best low-rank approximation of a formulated Hankel matrix by rank-reduction methods with a rank constraint. However, the predefined Hankel structure is easily damaged by the low-rank approximation, which leads to harming its recovery performance. In this article, we present a structured tensor completion (STC) framework to simultaneously exploit both the Hankel structure and the low-tubal-rank constraint to further enhance the performance. Unfortunately, under the assumption of elementwise sampling used by existing methods, STC is intractable to be solved since Hankel constraints cannot be expressed as linear tensor equations. Instead, tubal sampling is proposed to describe the missing trace behavior more accurately and further build a bridge between tensor and matrix completion (MC) to overcome the solving issue in two aspects: through the bridge from tensor to MC, STC can be solved efficiently using MC from random samplings of each frontal slice in the Fourier domain. Through the bridge from matrix to tensor completion, various tensor models within the framework can be developed from noise-specific MC to meet the need for data reconstruction in changeable noise environments. Moreover, alternating-minimization and alternating-direction methods of multipliers are developed to solve the proposed STC. The superior performance of STC is demonstrated in both synthetic and field seismic data.

Wavefield separation is one critical step in the three-component (3C) vertical seismic profiling (VSP) data processing. The conventional 3C VSP wavefield separation method exhibits low precision polarization angle and cannot obtain desirable results. This letter proposed the right-hand rule 3C VSP wavefield separation method which has many advantages over the conventional 3C VSP wavefield separation method-particularly the polarization angle precision and wavefields continuity. Analysis of real 3C VSP data demonstrates that the right-hand rule 3C VSP wavefield separation method can enhance polarization angle precision and show a very desirable ability to produce high-continuity wavefields.

Synchrosqueezing transform (SST) has been used to address the problems in many diverse disciplines. However, the calculated instantaneous frequency (CIF) precision by SST decreases with the increase in time-sample interval and actual instantaneous frequency. This will lead to an incorrect time-frequency distribution. To eliminate the impact of the time-sample interval and to enhance the CIF precision and SST resolution, we propose an adaptive time-resampled high-resolution SST, in which the time-resampled interval of the time-frequency spectrum for instantaneous frequency calculation is adapted to the highest frequency in its effective frequency band and the desired CIF precision. A synthetic signal example shows that the adaptive time-resampled high-resolution SST can eliminate the impact of the time-sample interval and achieve correct and desired high-resolution time-frequency distribution. A real 3-D seismic data application demonstrates that the proposed approach is a good potential technique for high-precision signal processing and interpretation.