Full text loading...
-
Deep Learning for Horizon Interpretation on 2D Seismic Data
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 82nd EAGE Annual Conference & Exhibition, Oct 2021, Volume 2021, p.1 - 5
Abstract
Horizon interpretation is one of the most important yet time consuming steps in a traditional seismic interpretation workflow. Traditionally, horizon interpretation is performed interactively. An interactive horizon tracker takes an interpreter’s sparse picks as constraints and fills in horizon segments in between the sparse picks, usually by comparing the similarity between two adjacent traces within a small temporal window. In some legacy oil fields, 2D seismic data are abundantly available. With the advancement of seismic processing and imaging techniques, reprocessing such legacy 2D seismic data may be the key to unlock additional value and bring new insights to the understanding of the subsurface. . In this study, we investigate using supervised deep learning to extract horizons on all the 2D seismic lines after training on a small subset of the available lines from the same region. Using a 2D shallow seismic dataset from offshore the Netherlands as an example, we are able to demonstrate that the proposed method is able to extract horizons consistently across multiple moderate quality 2D seismic lines.