Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MSSPN: Automatic First Arrival Picking using Multi-Stage Segmentation Picking Network

Hongtao Wang, Jiangshe Zhang, Xiaoli Wei, Chunxia Zhang, Zhenbo Guo, Li Long, Yicheng Wang

TL;DR

MSSPN addresses automatic FAT picking by decomposing the task into four coarse-to-fine segmentation stages that mimic manual processing. It integrates velocity-informed LMO corrections through RRVE and refines containment with RSN on cropped regions, followed by a robust post-processing step. Across seven field datasets, MSSPN achieves high accuracy on high/medium SNR data and substantial gains on low-SNR data, with explicit improvements via few-shot fine-tuning. The approach enhances cross-worksite generalization and robustness, offering a scalable solution for dense seismic data processing.

Abstract

Picking the first arrival times of prestack gathers is called First Arrival Time (FAT) picking, which is an indispensable step in seismic data processing, and is mainly solved manually in the past. With the current increasing density of seismic data collection, the efficiency of manual picking has been unable to meet the actual needs. Therefore, automatic picking methods have been greatly developed in recent decades, especially those based on deep learning. However, few of the current supervised deep learning-based method can avoid the dependence on labeled samples. Besides, since the gather data is a set of signals which are greatly different from the natural images, it is difficult for the current method to solve the FAT picking problem in case of a low Signal to Noise Ratio (SNR). In this paper, for hard rock seismic gather data, we propose a Multi-Stage Segmentation Pickup Network (MSSPN), which solves the generalization problem across worksites and the picking problem in the case of low SNR. In MSSPN, there are four sub-models to simulate the manually picking processing, which is assumed to four stages from coarse to fine. Experiments on seven field datasets with different qualities show that our MSSPN outperforms benchmarks by a large margin.Particularly, our method can achieve more than 90\% accurate picking across worksites in the case of medium and high SNRs, and even fine-tuned model can achieve 88\% accurate picking of the dataset with low SNR.

MSSPN: Automatic First Arrival Picking using Multi-Stage Segmentation Picking Network

TL;DR

MSSPN addresses automatic FAT picking by decomposing the task into four coarse-to-fine segmentation stages that mimic manual processing. It integrates velocity-informed LMO corrections through RRVE and refines containment with RSN on cropped regions, followed by a robust post-processing step. Across seven field datasets, MSSPN achieves high accuracy on high/medium SNR data and substantial gains on low-SNR data, with explicit improvements via few-shot fine-tuning. The approach enhances cross-worksite generalization and robustness, offering a scalable solution for dense seismic data processing.

Abstract

Picking the first arrival times of prestack gathers is called First Arrival Time (FAT) picking, which is an indispensable step in seismic data processing, and is mainly solved manually in the past. With the current increasing density of seismic data collection, the efficiency of manual picking has been unable to meet the actual needs. Therefore, automatic picking methods have been greatly developed in recent decades, especially those based on deep learning. However, few of the current supervised deep learning-based method can avoid the dependence on labeled samples. Besides, since the gather data is a set of signals which are greatly different from the natural images, it is difficult for the current method to solve the FAT picking problem in case of a low Signal to Noise Ratio (SNR). In this paper, for hard rock seismic gather data, we propose a Multi-Stage Segmentation Pickup Network (MSSPN), which solves the generalization problem across worksites and the picking problem in the case of low SNR. In MSSPN, there are four sub-models to simulate the manually picking processing, which is assumed to four stages from coarse to fine. Experiments on seven field datasets with different qualities show that our MSSPN outperforms benchmarks by a large margin.Particularly, our method can achieve more than 90\% accurate picking across worksites in the case of medium and high SNRs, and even fine-tuned model can achieve 88\% accurate picking of the dataset with low SNR.
Paper Structure (18 sections, 11 equations, 5 figures, 6 tables, 1 algorithm)

This paper contains 18 sections, 11 equations, 5 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Methodology
  3. Data Pre-processing
  4. Multi-Stage Segmentation Picking Network (MSSPN)
  5. Coarse-Segmentation Network (CSN)
  6. Regional Restriction based on Velocity Estimation (RRVE)
  7. Refine-Segmentation Network (RSN)
  8. Post-processing
  9. Quality Control
  10. Experiments
  11. Datasets
  12. Implementation Details
  13. Ablation Study
  14. Results
  15. Comparative Experiment
  16. ...and 3 more sections

Figures (5)

  • Figure 1: The flowchart of MSSPN
  • Figure 2: (a) Original gather grey image with manual first arrival label. (b) Downsampled gather grey image with interpolated label. (c) The ground-truth image of (b) in CSN.
  • Figure 3: (a) RANSACwVC regression processing. (b) Manually picked first arrival times and the reference first arrival times estimated by RANSACwVC. (c) The prediction map of CSN. (d) LMO gather with manually picked first arrival times (red points).
  • Figure 4: (a)-(f) are a picking results of a medium SNR gather in Halfmile. (a), (d) and (b), (e) are ground-truths and prediction maps of CSN and RSN, respectively. (c) is a gather after LMO marked FAT also named First Break (FB) times. (f) is a wiggle plot of a gather with final FB and the boundaries of RSN input. (g) and (h) are medium SNR gathers with FB of Lalor and DN-2, respectively.
  • Figure 5: A low SNR gather with FB of LS. (a) The picking result before fine-tuning. (b) The picking result after fine-tuning.