Research output: Contribution to conference › Paper › peer-review
Accelerated deep learning-based estimation of wavefront dips and curvatures and their application to 3D prestack data enhancement. / Gadylshin, Kirill; Silvestrov, Ilya; Bakulin, Andrey.
2022. Paper presented at 2nd International Meeting for Applied Geoscience and Energy, United States.Research output: Contribution to conference › Paper › peer-review
}
TY - CONF
T1 - Accelerated deep learning-based estimation of wavefront dips and curvatures and their application to 3D prestack data enhancement
AU - Gadylshin, Kirill
AU - Silvestrov, Ilya
AU - Bakulin, Andrey
N1 - Conference code: 2
PY - 2022/8/15
Y1 - 2022/8/15
N2 - We present a novel workflow for the accelerated signal enhancement of massive 3D prestack seismic data utilizing a Local Wavefront Attributes Deep Neural Network. It is based on automatic local wavefront attributes estimation using a specially trained convolutional deep neural network. The general workflow is adaptive to a particular 3D prestack seismic volume. It requires performing a conventional semblance-based estimation of wavefront dips and curvatures for only about 1% of the whole amount of data. The verification of the proposed approach is done on challenging real datasets, both marine and land. Deep learning allows achieving a significant speed-up compared to the conventional method while preserving an acceptable quality of the results.
AB - We present a novel workflow for the accelerated signal enhancement of massive 3D prestack seismic data utilizing a Local Wavefront Attributes Deep Neural Network. It is based on automatic local wavefront attributes estimation using a specially trained convolutional deep neural network. The general workflow is adaptive to a particular 3D prestack seismic volume. It requires performing a conventional semblance-based estimation of wavefront dips and curvatures for only about 1% of the whole amount of data. The verification of the proposed approach is done on challenging real datasets, both marine and land. Deep learning allows achieving a significant speed-up compared to the conventional method while preserving an acceptable quality of the results.
UR - https://www.scopus.com/inward/record.url?eid=2-s2.0-85146652988&partnerID=40&md5=684fc6384a75acd47e8bb5aa79d602f8
UR - https://www.mendeley.com/catalogue/b232aac1-7865-3937-ae6d-0f582fc055e8/
U2 - 10.1190/image2022-3735882.1
DO - 10.1190/image2022-3735882.1
M3 - Paper
T2 - 2nd International Meeting for Applied Geoscience and Energy
Y2 - 28 August 2022 through 1 September 2022
ER -
ID: 45608098