Attenuation of marine seismic interference noise employing a customized U‐Net

Sun, Jing; Slang, Sigmund; Elboth, Thomas; Greiner, Thomas Larsen; McDonald, S.; Gelius, Leiv‐J.

doi:10.1111/1365-2478.12893

Cited by 43 publications

(10 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…If any given layer in the network is built in such a way that important information might be lost during forward propagation, we may not be able to reach the desired result, despite having good training data and/or a large training data set. Some of the most popular architectures, the U-net [66] and the ResNet [32] (and their adaptions), have been proposed in multiple research papers for seismic reconstruction problems [40,43,75,87,88].…”

Section: Network Architecturementioning

confidence: 99%

Cross-streamer wavefield reconstruction through wavelet domain learning

et al. 2020

Self Cite

View full text Add to dashboard Cite

Seismic exploration in complex geological settings and shallow geological targets has led to a demand for higher spatial and temporal resolution in the final migrated image. Both 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 in order 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 investigate 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 in order to reconstruct the wavefield across the streamers. We demonstrate the performance of the proposed method on broadband synthetic data and TopSeis field data from the Barents Sea. From our synthetic example, we show 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 geological model in the migrated section. In addition, we compare our method to an industry-standard Fourier-based method, where the CNN approach shows an improvement in the RMS error close to a factor of two. In our field data example, we show that the approach manages to reconstruct the wavefield across the streamers in the shot domain, and displaying promising characteristics of a reconstructed 3D wavefield.

show abstract

Section: Network Architecturementioning

confidence: 99%

Cross-streamer wavefield reconstruction through wavelet domain learning

et al. 2020

Self Cite

View full text Add to dashboard Cite

show abstract

“…After the conventional contracting seen in other network architectures, the U‐Net then has an expansive path which combines both feature and spatial information which gives a more precise output than networks with just a contracting layer (Lian et al ., 2018). Additionally, the U‐Net has skip connections, allowing the network to capture high‐resolution detail and prevent the vanishing gradient problem (Sun et al ., 2020). This architecture allows the U‐Net to map features of the image which can then be used to recreate the image (Sankesara, 2019).…”

Section: Methodsmentioning

confidence: 99%

Pre‐migration diffraction separation using generative adversarial networks

Lowney

Lokmer

O’Brien

et al. 2021

Geophysical Prospecting

View full text Add to dashboard Cite

Diffraction imaging is the process of separating diffraction events from the seismic wavefield and imaging them independently, highlighting subsurface discontinuities. While there are many analytic‐based methods for diffraction imaging which use kinematic, dynamic or both, properties of the diffracted wavefield, they can be slow and require parameterization. Here, we propose an image‐to‐image generative adversarial network to automatically separate diffraction events on pre‐migrated seismic data in a fraction of the time of conventional methods. To train the generative adversarial network, plane‐wave destruction was applied to a range of synthetic and real images from field data to create training data. These training data were screened and any areas where the plane‐wave destruction did not perform well, such as synclines and areas of complex dip, were removed to prevent bias in the neural network. A total of 14,132 screened images were used to train the final generative adversarial network. The trained network has been applied across several geologically distinct field datasets, including a 3D example. Here, generative adversarial network separation is shown to be comparable to a benchmark separation created with plane‐wave destruction, and up to 12 times faster. This demonstrates the clear potential in generative adversarial networks for fast and accurate diffraction separation.

show abstract

“…The most popular auto-encoder (AE) network is the U-net (Ronneberger et al, 2015) with its long-range connections between the encoding and decoding levels that improve reconstruction accuracy. The U-net has been used in a wide variety of geophysical applications, such as salt segmentation (Shi et al, 2019), noise attenuation (Sun et al, 2019) and channel interpretation (Gao et al, 2021) and of course first break picking. We will compare a standard AE with the original U-net and vary its width and depth.…”

Section: E T H O D O L O G Y F O R F I R S T B R E a K P I C K I N G W I T H D E E P L E A R N I N G Implementation Detailsmentioning

confidence: 99%

First break picking with deep learning – evaluation of network architectures

Zwartjes¹,

Yoo²

2021

Geophysical Prospecting

View full text Add to dashboard Cite

In recent years, various convolutional neural network architectures have been proposed for first break picking. In this paper, we compare the standard auto-encoder and U-net architectures as well as versions enhanced with ResNet style skip connections. The U-net appears to have become the standard network for segmentation, judging from the number of published articles. Still, there is some variety in neural network architectural choices. In this paper, we assess the impact of neural network depth, width and input data size, as well as some small modifications for deep networks offered by the ResNet. In general, results improve as the networks get deeper, but with diminishing returns. The more complex the data, the more benefit the deeper networks bring. We use complete shot gathers, albeit rescaled for efficiency, to train the neural networks. For shot gathers with a simple piecewise linear moveout, this approach yields results with good accuracy when gathers are resampled to 128 × 128 samples. For shot gathers with more complex first break moveout, using our approach it is advised to stay close to the original dimension of each gather for best accuracy, at the expense of increased training times. A good trade-off between network depth, image size and training times is to use a nine-stage U-net with 256 sample images. Despite the advantages in other applications, the basic U-net outperforms a U-net with ResNet features. We show that changing the input data dimensions for trained networks does not work, despite the fact the fully convolutional networks are independent of image size. The U-net based first break picking is not sensitive to picking errors, as in many cases the neural network predictions are better than the training data where the training data have random mispicks. This suggests a practical application; namely, to train or re-train a pre-trained network on a single data set after conventional first break picking with the objective of improving conventionally picked first breaks.

show abstract

Attenuation of marine seismic interference noise employing a customized U‐Net

Cited by 43 publications

References 38 publications

Cross-streamer wavefield reconstruction through wavelet domain learning

Cross-streamer wavefield reconstruction through wavelet domain learning

Pre‐migration diffraction separation using generative adversarial networks

First break picking with deep learning – evaluation of network architectures

Contact Info

Product

Resources

About