4.7 Article

Convolutional Neural Network, Res-Unet plus plus , -Based Dispersion Curve Picking From Noise Cross-Correlations

Journal

JOURNAL OF GEOPHYSICAL RESEARCH-SOLID EARTH
Volume 126, Issue 11, Pages -

Publisher

AMER GEOPHYSICAL UNION
DOI: 10.1029/2021JB022027

Keywords

deep learning; surface wave; dispersion curves; ambient seismic noise; transfer learning; neural network

Funding

  1. Shenzhen Key Laboratory of Deep Offshore Oil and Gas Exploration Technology [ZDSYS20190902093007855]
  2. Key Special Project for Introduced Talents Team of Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou) [GML2019ZD0203]
  3. Shenzhen Science and Technology Program [KQTD20170810111725321]
  4. National Natural Science Foundation of China [U1901602, 41790465]
  5. leading talents of Guangdong province program [2016LJ06N652]

Ask authors/readers for more resources

The neural network Res-Unet++ can automatically and accurately extract fundamental dispersion curves and overtones from F-J dispersion spectra after training, showing high accuracies in synthetic and real data. The network's effectiveness in extracting dispersion curves and adaptability through transfer learning have been demonstrated, providing advantages in generating more effective dispersion points.
Ambient seismic noise cross-correlation has been widely applied in surface wave tomography at regional to global scales, including for seismic exploration of near-surface structures. Reliable seismic imaging requires the accurate selection of dispersion curves. However, manual picking has become cumbersome work with the increase in available correlation traces; it is even more difficult when the number of dispersion curves increases by using frequency-Bessel (F-J) transform. Here, we show that the neural network Res-Unet++ can automatically and accurately extract both fundamental dispersion curves and overtones from the F-J dispersion spectra after training the network. Results show that selected dispersion curves had high accuracies in the synthetic data (greater than 95%). The network could effectively extract both the fundamental and higher modes in real data, and transfer learning improved the adaptability of neural networks for different geological areas. The obtained dispersion curves from the real data agreed well with those acquired manually and were advantageous for generating more effective dispersion points.

Authors

I am an author on this paper
Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.7
Not enough ratings

Secondary Ratings

Novelty
-
Significance
-
Scientific rigor
-
Rate this paper

Recommended

No Data Available
No Data Available