An initial velocity model building method based on multimodal texture transfer network

View at publisher Summarise with VIRAT
Authors: Wang Lei, Liang Jiandong, Liang Shuaizhe, Zhu Jinping, Xin Kefeng, Zhaolei Wang
Cited by: 0
References: 5

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

We propose a joint well-seismic initial velocity model building method based on multimodal texture transfer network. This method predicts velocity model by building a texture transfer network to match the velocities and reflection layers, and introduces adaptive instance normalization by utilizing well logs to constrain velocity updating. In practical applications, an extra transfer learning strategy guided by well logs ensures the network’s generalization ability. This method has been applied in production in complex geological structure surveys in western China, achieving higher initial velocity model building accuracy compared to conventional constrained velocity inversion (CVI) method and structure-based interpolation method using well logs. The lateral variation of velocity model is more consistent with the geological features of the PSDM image and the PSDM CIP gathers are flatter than the conventional results.

Fields of Study & Subjects

Seismic Imaging and Inversion Techniques Seismic Waves and Analysis Geophysical and Geoelectrical Methods

Publication details

Journal / source
Second International Meeting for Applied Geoscience & Energy
Volume
1
Pages
2120–2124
Publication date
10 February 2026
Language
EN