What is it about?
This research is about the extraction of seismic diffraction events by using deep learning technique (convolutional U-net, transfer learning, t-SNE). We show that combination of simple events works well on really complicated seismic field data through our approach.
Featured Image
Photo by NASA on Unsplash
Why is it important?
Diffractions usually arise at small-scale heterogeneities such as fault and fractures, so we can get detailed subsurface images by extracting and utilizing them. By using DL approach, it can reduce the effort for parameter decision in conventional approaches and provide much faster results through transfer learning.
Perspectives
Our approach proposed machine learning-based systematic workflow for generation of synthetic training dataset and application to field data. However, as we discussed in the paper, 3D effect, far offset, late-time signal and field noise issues should be considered in the future research.
Sooyoon Kim
Hanyang University
Read the Original
This page is a summary of: Extraction of diffractions from seismic data using convolutional U-net and transfer learning, Geophysics, January 2022, Society of Exploration Geophysicists,
DOI: 10.1190/geo2020-0847.1.
You can read the full text:
Contributors
The following have contributed to this page
