Seismic data interpolation through convolutional autoencoder

Mandelli, Sara; Borra, Federico; Lipari, Vincenzo; Bestagini, Paolo; Sarti, Augusto; Tubaro, Stefano

doi:10.1190/segam2018-2995428.1

Cited by 85 publications

(37 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Mandelli et al . (, ) proposed to reconstruct missing seismic traces in the prestack domain by employing a convolutional autoencoder (AE). They applied this network to solve the joint problem of synthetic data interpolation and Gaussian‐noise attenuation.…”

Section: Introductionmentioning

confidence: 99%

“…Within the area of seismic data interpolation and reconstruction, Wang et al (2018a,b) introduced an eight-layer residual learning network (ResNet) based on CNNs to interpolate seismic data without aliasing. Mandelli et al (2018Mandelli et al ( , 2019 proposed to reconstruct missing seismic traces in the prestack domain by employing a convolutional autoencoder (AE). They applied this network to solve the joint problem of synthetic data interpolation and Gaussian-noise attenuation.…”

mentioning

confidence: 99%

See 1 more Smart Citation

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

Sun

Slang

Elboth

et al. 2020

Geophysical Prospecting

View full text Add to dashboard Cite

A B S T R A C TMarine seismic interference noise occurs when energy from nearby marine seismic source vessels is recorded during a seismic survey. Such noise tends to be well preserved over large distances and causes coherent artefacts in the recorded data. Over the years, the industry has developed various denoising techniques for seismic interference removal, but although well performing, they are still time-consuming in use. Machine-learning-based processing represents an alternative approach, which may significantly improve the computational efficiency. In the case of conventional images, autoencoders are frequently employed for denoising purposes. However, due to the special characteristics of seismic data as well as the noise, autoencoders failed in the case of marine seismic interference noise. We, therefore, propose the use of a customized U-Net design with element-wise summation as part of the skip-connection blocks to handle the vanishing gradient problem and to ensure information fusion between high-and low-level features. To secure a realistic study, only seismic field data were employed, including 25,000 training examples. The customized U-Net was found to perform well, leaving only minor residuals, except for the case when seismic interference noise comes from the side. We further demonstrate that such noise can be treated by slightly increasing the depth of our network. Although our customized U-Net does not outperform a standard commercial algorithm in quality, it can (after proper training) read and process one single shot gather in approximately 0.02 s. This is significantly faster than any existing industry denoising algorithm. In addition, the proposed network processes shot gathers in a sequential order, which is an advantage compared with industry algorithms that typically require a multi-shot input to break the coherency of the noise.

show abstract

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

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

Sun

Slang

Elboth

et al. 2020

Geophysical Prospecting

View full text Add to dashboard Cite

show abstract

“…Mandelli et al . (2019) used CNNs for interpolation and denoising. Recent studies using the generative adversarial networks (GANs) have shown promising results for seismic data processing and imaging applications.…”

Section: Introductionmentioning

confidence: 99%

Seismic data interpolation using deep learning with generative adversarial networks

Kaur

Pham

2020

Geophysical Prospecting

View full text Add to dashboard Cite

We propose an algorithm for seismic trace interpolation using generative adversarial networks, a type of deep neural network. The method extracts feature vectors from the training data using self-learning and does not require any pre-processing to create the training labels. The algorithm also does not make any prior explicit assumptions about linearity of seismic events or sparsity of the data, which are often required in the traditional interpolation methods. We create the training labels by removing traces from different receiver indices of the original datasets to simulate the effect of missing traces. We adopt the framework of the generative adversarial networks to train the network and add additional loss functions to regularize the model. Numerical examples using land and marine field datasets demonstrate the validity and effectiveness of the proposed approach. With minimal computational burden and proper training, the proposed method can be applied to three-dimensional seismic datasets to achieve accurate interpolation results.

show abstract

“…Recent successes in the application of machine learning algorithms have been reported in the areas of seismic interpretation (Huang et al, 2017;Di et al, 2018;Veillard et al, 2018), seismic data processing (Siahkoohi et al, 2018b;Sun and Demanet, 2018;Ovcharenko et al, 2018), seismic modeling and imaging Rizzuti et al, 2019;Moseley et al, 2018), and inversion (Mosser et al, 2018;Kothari et al, 2019). Several authors have addressed the seismic wavefield recovery (Siahkoohi et al, 2018a;Mandelli et al, 2018) as well. Mandelli et al (2018) trained a neural network to jointly denoise and interpolate 2D shot records by operating on extracted patches from shot records.…”

Section: Introductionmentioning

confidence: 99%

“…Several authors have addressed the seismic wavefield recovery (Siahkoohi et al, 2018a;Mandelli et al, 2018) as well. Mandelli et al (2018) trained a neural network to jointly denoise and interpolate 2D shot records by operating on extracted patches from shot records. Siahkoohi et al (2018a) recovered seismic wavefields by reconstructing subsampled frequency slices, by assuming that 5% of shot records are fully-sampled.…”

Section: Introductionmentioning

confidence: 99%

Deep-learning based ocean bottom seismic wavefield recovery

Siahkoohi

Kumar

Herrmann

2019

SEG Technical Program Expanded Abstracts 2019

View full text Add to dashboard Cite

Ocean bottom surveys usually suffer from having very sparse receivers. Assuming a desirable source sampling, achievable by existing methods such as (simultaneous-source) randomized marine acquisition, we propose a deep-learning based scheme to bring the receivers to the same spatial grid as sources using a convolutional neural network. By exploiting source-receiver reciprocity, we construct training pairs by artificially subsampling the fully-sampled single-receiver frequency slices using a random training mask and later, we deploy the trained neural network to fill-in the gaps in single-source frequency slices. Our experiments show that a random training mask is essential for successful wavefield recovery, even when receivers are on a periodic gird. No external training data is required and experiments on a 3D synthetic data set demonstrate that we are able to recover receivers for up to 90% missing receivers, missing either randomly or periodically, with a better recovery for random case, at low to midrange frequencies.

show abstract

Seismic data interpolation through convolutional autoencoder

Cited by 85 publications

References 20 publications

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

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

Seismic data interpolation using deep learning with generative adversarial networks

Deep-learning based ocean bottom seismic wavefield recovery

Contact Info

Product

Resources

About