Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Label-Free High-Precision Residual Moveout Picking Method for Travel Time Tomography based on Deep Learning

Hongtao Wang, Jiandong Liang, Lei Wang, Shuaizhe Liang, Jinping Zhu, Chunxia Zhang, Jiangshe Zhang

TL;DR

This paper tackles the challenge of accurate residual moveout (RMO) picking for travel-time tomography by introducing a cascade, deep-learning-based framework. It combines a multi-scale attention-based segmentation network (IFSN) with a slope-field–driven post-processing pipeline, including CG curvature splitting, DBSCAN-based merging, and robust Bayesian regression, all trained on synthetically generated common-image gathers (CIGs). Key contributions include a synthetic data generation method, new RMO-picking metrics, and a demonstrated gain in picking density and accuracy over semblance-based methods across model and field datasets. The approach promises faster, more reliable velocity-model updates and improved inversion performance in challenging seismic data settings. The work also provides ablation and hyperparameter analyses to guide practical deployment in varying data quality scenarios.

Abstract

Residual moveout (RMO) provides critical information for travel time tomography. The current industry-standard method for fitting RMO involves scanning high-order polynomial equations. However, this analytical approach does not accurately capture local saltation, leading to low iteration efficiency in tomographic inversion. Supervised learning-based image segmentation methods for picking can effectively capture local variations; however, they encounter challenges such as a scarcity of reliable training samples and the high complexity of post-processing. To address these issues, this study proposes a deep learning-based cascade picking method. It distinguishes accurate and robust RMOs using a segmentation network and a post-processing technique based on trend regression. Additionally, a data synthesis method is introduced, enabling the segmentation network to be trained on synthetic datasets for effective picking in field data. Furthermore, a set of metrics is proposed to quantify the quality of automatically picked RMOs. Experimental results based on both model and real data demonstrate that, compared to semblance-based methods, our approach achieves greater picking density and accuracy.

A Label-Free High-Precision Residual Moveout Picking Method for Travel Time Tomography based on Deep Learning

TL;DR

This paper tackles the challenge of accurate residual moveout (RMO) picking for travel-time tomography by introducing a cascade, deep-learning-based framework. It combines a multi-scale attention-based segmentation network (IFSN) with a slope-field–driven post-processing pipeline, including CG curvature splitting, DBSCAN-based merging, and robust Bayesian regression, all trained on synthetically generated common-image gathers (CIGs). Key contributions include a synthetic data generation method, new RMO-picking metrics, and a demonstrated gain in picking density and accuracy over semblance-based methods across model and field datasets. The approach promises faster, more reliable velocity-model updates and improved inversion performance in challenging seismic data settings. The work also provides ablation and hyperparameter analyses to guide practical deployment in varying data quality scenarios.

Abstract

Residual moveout (RMO) provides critical information for travel time tomography. The current industry-standard method for fitting RMO involves scanning high-order polynomial equations. However, this analytical approach does not accurately capture local saltation, leading to low iteration efficiency in tomographic inversion. Supervised learning-based image segmentation methods for picking can effectively capture local variations; however, they encounter challenges such as a scarcity of reliable training samples and the high complexity of post-processing. To address these issues, this study proposes a deep learning-based cascade picking method. It distinguishes accurate and robust RMOs using a segmentation network and a post-processing technique based on trend regression. Additionally, a data synthesis method is introduced, enabling the segmentation network to be trained on synthetic datasets for effective picking in field data. Furthermore, a set of metrics is proposed to quantify the quality of automatically picked RMOs. Experimental results based on both model and real data demonstrate that, compared to semblance-based methods, our approach achieves greater picking density and accuracy.

Paper Structure

This paper contains 26 sections, 28 equations, 9 figures, 5 tables.

Table of Contents

  1. Introduction
  2. A Cascade Picking Framework for RMO Picking
  3. Overview of Cascade Method
  4. Multi-Scale Attention Segmentation Network
  5. Multi-Scale Feature Extraction
  6. Information-Fusion Segmentation Network
  7. Trend-based Post-processing
  8. Computer Graphics-based Curvature Split Method
  9. New Distance Definition-based Clustering Method
  10. Robust Bayesian Regression Method
  11. Generation of Synthetic Samples
  12. Background Assumption
  13. CIG Generation
  14. New Metrics for RMO Picking
  15. Numerical Experiments
  16. ...and 11 more sections

Figures (9)

  • Figure 1: The four stages of our proposed cascade RMO picking framework.
  • Figure 2: Flowchart of IFSN (a) and Details of CBAM (b).
  • Figure 3: The influence of $z_0$, $\beta$ and $\gamma$ on the generated curvatures.
  • Figure 4: Classic CIG Image of Synthetic, BP, F-A, and F-B, respectively.
  • Figure 5: Detailed output of each step in inference process of the cascade method. (a) The CIG after AGC with the window of 15; (b)-(d) The output results of Step 1-4 of the cascade method.
  • ...and 4 more figures