Scholarlink

This study conducts a preliminary investigation into methods for generating 3D models of the reactor interior at the Fukushima Daiichi Nuclear Power Plant using video footage acquired by underwater remotely operated vehicles (ROVs). The reconstruction methods evaluated included COLMAP, Reality Capture, DUSt3R, and MASt3R. While COLMAP and Reality Capture are classified as part of the computer vision (CV) generation, DUSt3R and MASt3R represent the artificial intelligence (AI) generation, and the underlying generation algorithms differ significantly. The overall workflow consists of three stages: (1) Preprocessing of video frames, (2) 3D point cloud generation, and (3) Real-time meshing and visualization. During the preprocessing stage, a custom image processing program was developed to remove floating objects and overlaid captions, aiming to improve subsequent reconstruction accuracy.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Basic Study on 3D Modeling of the Reactor Interior Using Underwater ROV Video Footage from Fukushima Daiichi Nuclear Power Plant

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Akio Doi,
  • Hiroki Takahashi,
  • Toru Kato,
  • Meguru Yamashita,
  • Gao Zhi Yi

摘要

This study conducts a preliminary investigation into methods for generating 3D models of the reactor interior at the Fukushima Daiichi Nuclear Power Plant using video footage acquired by underwater remotely operated vehicles (ROVs). The reconstruction methods evaluated included COLMAP, Reality Capture, DUSt3R, and MASt3R. While COLMAP and Reality Capture are classified as part of the computer vision (CV) generation, DUSt3R and MASt3R represent the artificial intelligence (AI) generation, and the underlying generation algorithms differ significantly. The overall workflow consists of three stages: (1) Preprocessing of video frames, (2) 3D point cloud generation, and (3) Real-time meshing and visualization. During the preprocessing stage, a custom image processing program was developed to remove floating objects and overlaid captions, aiming to improve subsequent reconstruction accuracy.