• Geophysics and Geophysical Exploration
• /
• v.25 no.2
• /
• pp.71-84
• /
• 2022

When acquiring seismic data, various types of simultaneously recorded seismic noise hinder accurate interpretation. Therefore, it is essential to attenuate this noise during the processing of seismic data and research on seismic noise attenuation. For this purpose, machine learning is extensively used. This study attempts to attenuate noise in prestack seismic data using unsupervised machine learning. Three unsupervised machine learning models, N2NUNET, PATCHUNET, and DDUL, are trained and applied to synthetic and field prestack seismic data to attenuate the noise and leave clean seismic data. The results are qualitatively and quantitatively analyzed and demonstrated that all three unsupervised learning models succeeded in removing seismic noise from both synthetic and field data. Of the three, the N2NUNET model performed the worst, and the PATCHUNET and DDUL models produced almost identical results, although the DDUL model performed slightly better.

• Economic and Environmental Geology
• /
• v.57 no.1
• /
• pp.51-71
• /
• 2024

Seismic exploration is one of the widely used geophysical exploration methods with various applications such as resource development, geotechnical investigation, and subsurface monitoring. It is essential for interpreting the geological characteristics of subsurface by providing accurate images of stratum structures. Typically, geological features are interpreted by visually analyzing seismic sections. However, recently, quantitative analysis of seismic data has been extensively researched to accurately extract and interpret target geological features. Seismic attribute analysis can provide quantitative information for geological interpretation based on seismic data. Therefore, it is widely used in various fields, including the analysis of oil and gas reservoirs, investigation of fault and fracture, and assessment of shallow gas distributions. However, seismic attribute analysis is sensitive to noise within the seismic data, thus additional noise attenuation is required to enhance the accuracy of the seismic attribute analysis. In this study, four kinds of seismic noise attenuation methods are applied and compared to mitigate random noise of poststack seismic data and enhance the attribute analysis results. FX deconvolution, DSMF, Noise2Noise, and DnCNN are applied to the Youngil Bay high-resolution seismic data to remove seismic random noise. Energy, sweetness, and similarity attributes are calculated from noise-removed seismic data. Subsequently, the characteristics of each noise attenuation method, noise removal results, and seismic attribute analysis results are qualitatively and quantitatively analyzed. Based on the advantages and disadvantages of each noise attenuation method and the characteristics of each seismic attribute analysis, we propose a suitable noise attenuation method to improve the result of seismic attribute analysis.

• Geophysics and Geophysical Exploration
• /
• v.27 no.2
• /
• pp.91-107
• /
• 2024

Single-channel seismic exploration has proven effective in delineating subsurface geological structures using small-scale survey systems. The seismic data acquired through zero- or near-offset methods directly capture subsurface features along the vertical axis, facilitating the construction of corresponding seismic sections. However, substantial noise in single-channel seismic data hampers precise interpretation because of the low signal-to-noise ratio. This study introduces a novel approach that integrate noise reduction and signal enhancement via matrix rank optimization to address this issue. Unlike conventional rank-reduction methods, which retain selected singular values to mitigate random noise, our method optimizes the entire singular value spectrum, thus effectively tackling both random and erratic noises commonly found in environments with low signal-to-noise ratio. Additionally, to enhance the horizontal continuity of seismic events and mitigate signal loss during noise reduction, we introduced an adaptive weighting factor computed from the eigenimage of the seismic section. To access the robustness of the proposed method, we conducted numerical experiments using single-channel Sparker seismic data from the Chukchi Plateau in the Arctic Ocean. The results demonstrated that the seismic sections had significantly improved signal-to-noise ratios and minimal signal loss. These advancements hold promise for enhancing single-channel and high-resolution seismic surveys and aiding in the identification of marine development and submarine geological hazards in domestic coastal areas.

• Geophysics and Geophysical Exploration
• /
• v.3 no.3
• /
• pp.83-87
• /
• 2000

The noise prediction filter using a local/remote reference was developed to obtain a high quality data from seismic surveys over the area where seismic transmission power is limited. The method used in the noise prediction filter is a 3-layer neural network whose algorithm is backpropagation. A NRF (Noise Reduction Factor) value of about 3.0 was obtained with appling training and test data to the trained noise prediction filter. However, the scaling technique generally used for minimizing EM noise from electric and electromagnetic data cannot reduce seismic noise, since the technique can allow only amplitude difference between two time series measured at the primary and reference sites.

• Journal of the Korean Geophysical Society
• /
• v.4 no.2
• /
• pp.113-120
• /
• 2001

The discrete wavelet transform(DWT) has potential as a tool for supplying discriminatory attributes with which to distinguish seismic events. The wavelet transform has the great advantage over the Fourier transform in being able to localize changes. In this study, a discrete wavelet transform is applied to seismic traces for identifying seismic events and picking of arrival times for first breaks and S-wave arrivals. The precise determination of arrival times can greatly improve the quality of a number of geophysical studies, such as velocity analysis, refraction seismic survey, seismic tomography, down-hole and cross-hole survey, and sonic logging, etc. provide precise determination of seismic velocities. Tests for picking of P- and S- wave arrival times with the wavelet transform method is conducted with synthetic seismic traces which have or do not have noises. The results show that this picking algorithm can be successfully applied to noisy traces. The first arrival can be precisely determined with the field data, too.

• Geophysics and Geophysical Exploration
• /
• v.15 no.4
• /
• pp.199-208
• /
• 2012

High-amplitude swell noises (HASN) are very difficult to eliminate from the marine seismic data. In this paper, we applied F-X filter and median filter in order to suppress HASN. Test data have been acquired on the northern offshore of the South Shetland Islands in December, 2010. Parts of data have been contaminated by HASN caused by bad weather during the cruise. We applied F-X filter and median filter to test data with HASN. After F-X filtering, most of non-coherent noises and small-amplitude swell noises are eliminated effectively but HASN are still remained significantly. With median filter, HASN was suppressed better than F-X filter, however some of non-coherent noises are still remains. We applied a cascade of two filters and results show HASN and non-coherent noises are suppressed effectively. After the cascade of two filtering, it is possible to define reflection layers clearly on the velocity spectrum and to produce better stacked section with a good signal-to-noise ratio.

• Geophysics and Geophysical Exploration
• /
• v.23 no.2
• /
• pp.72-88
• /
• 2020

Recent rapid advances in computer hardware performance have led to relatively low computational costs, increasing the number of applications of machine-learning techniques to geophysical problems. In particular, deep-learning techniques are gaining in popularity as the number of cases successfully solving complex and nonlinear problems has gradually increased. In this paper, applications of seismic data denoising methods using deep-learning techniques are introduced and investigated. Depending on the type of attenuated noise, these studies are grouped into denoising applications of coherent noise, random noise, and the combination of these two types of noise. Then, we investigate the deep-learning techniques used to remove the corresponding noise. Unlike conventional methods used to attenuate seismic noise, deep neural networks, a typical deep-learning technique, learn the characteristics of the noise independently and then automatically optimize the parameters. Therefore, such methods are less sensitive to generalized problems than conventional methods and can reduce labor costs. Several studies have also demonstrated that deep-learning techniques perform well in terms of computational cost and denoising performance. Based on the results of the applications covered in this paper, the pros and cons of the deep-learning techniques used to remove seismic noise are analyzed and discussed.

• The Journal of Engineering Geology
• /
• v.22 no.4
• /
• pp.459-466
• /
• 2012

A difficulty encountered in engineering seismic mapping is that reflection events from shallow discontinuities are commonly overlapped with coherent noise such as air wave, direct waves, head waves, and high-amplitude surface waves. Here, the radial trace transform, a simple geometric re-mapping of a trace gather (x-t domain) to another trace gather (v-t domain), is applied to investigate the rejection effect of coherent linear noises. Two different types of data sets were selected as a representative database: good-quality data for intermediate sounding (hundreds of meters) in a sedimentary basin and very noisy data for shallow (${\leq}50m$) mapping of the weathered zone and bedrock surface. Results obtained with cascaded application of the radial transform and low-cut filtering proved to be as good as, or better than, those produced using f-k filtering, and were especially effective for air wave and direct wave. This simple transform enables better understanding of the characteristics of various types of noise in the RT domain, and can be generally applied to overcoming diffractions and back-scatterings caused by joints, fractures, and faults commonly that are encountered in geotechnical problems.

• Geophysics and Geophysical Exploration
• /
• v.27 no.3
• /
• pp.171-180
• /
• 2024

The processing of seismic data involves analyzing earthquake wave data to understand the internal structure and characteristics of the Earth, which requires high computational power. Recently, machine learning (ML) techniques have been introduced to address these challenges and have been utilized in various tasks such as noise reduction and velocity model construction. However, most studies have focused on specific seismic data processing tasks, limiting the full utilization of similar features and structures inherent in the datasets. In this study, we compared the efficacy of using receiver-wise time-series data ("receiver array") and synchronized receiver signals ("time array") from shotgathers for pretraining a Bidirectional Encoder Representations from Transformers (BERT) model. To this end, shotgather data generated from a synthetic model containing faults was used to perform noise reduction, velocity prediction, and fault detection tasks. In the task of random noise reduction, both the receiver and time arrays showed good performance. However, for tasks requiring the identification of spatial distributions, such as velocity estimation and fault detection, the results from the time array were superior.

• Geophysics and Geophysical Exploration
• /
• v.21 no.1
• /
• pp.54-60
• /
• 2018

The data acquired from seismic exploration can be used to detect the existence of oil and gas resources through appropriate processing and interpretation. The seismic attributes indicating the existence of resources are extracted from amplitude information, where the Q-factor representing intrinsic attenuation plays an useful role of hydrocarbon indicator. So, the accuracy of Q-factor estimation is very important to investigate the existence of resources. In this study, we calculated the Q-factor and analyzed the error rate through a numerical example. To mimic real data, random noise was added to the synthetic data. With the noise-added data, the Q-factor was estimated and the error rate was analyzed by using the spectral ratio method (SRM) and peak frequency shift method (PFSM). Both methods provided a relatively accurate Q-factor when the signal-to-noise ratio was 90 dB. However, the peak frequency shift method (PFSM) produced better results than the spectral ratio method (SRM) as the level of random noise increased.