• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to previous page

Establishment of image recognition judgment method for dangerous actions in decommissioning work of Fukushima Daiichi Nuclear Power Station

Research Project

Project/Area Number 19K05324
Research Category

Grant-in-Aid for Scientific Research (C)

Allocation TypeMulti-year Fund
Section一般
Review Section Basic Section 31010:Nuclear engineering-related
Research InstitutionThe University of Tokyo

Principal Investigator

Demachi Kazuyuki  東京大学, 大学院工学系研究科(工学部), 准教授 (00292764)

Co-Investigator(Kenkyū-buntansha) 三木 大輔  地方独立行政法人東京都立産業技術研究センター, 開発本部開発第三部情報技術グループ, 副主任研究員 (70757343)
鈴木 俊一  東京大学, 大学院工学系研究科(工学部), 特任教授 (80767997)
Project Period (FY) 2019-04-01 – 2022-03-31
Project Status Completed (Fiscal Year 2021)
Budget Amount *help
¥4,420,000 (Direct Cost: ¥3,400,000、Indirect Cost: ¥1,020,000)
Fiscal Year 2021: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
Fiscal Year 2020: ¥1,300,000 (Direct Cost: ¥1,000,000、Indirect Cost: ¥300,000)
Fiscal Year 2019: ¥1,950,000 (Direct Cost: ¥1,500,000、Indirect Cost: ¥450,000)
Keywords作業安全 / 自然言語処理 / 動作認識 / 物体認識 / 階層型Scene Graph / 論理表現化 / 深層学習 / 位置推定 / 危険行為 / 画像認識 / アラーム
Outline of Research at the Start

「手」と「全身」の動作の画像認識技術を適用し、物体認識などの画像情報も組み合わせ、廃炉作業中の危険動作の自動判定及び表示手法を確立する。申請者が開発した手法では、5秒間の動画データを約1/70,000の画像ラベルおよび関節座標(=1.5kB)に圧縮しているため、市販のGPU計算機で十分にリアルタイムの画像認識計算が可能である。取得した画像情報に自然言語処理を適用することで、行為の危険度をリアルタイム判定するアルゴリズムを確立する。また、危険行為の定義のため、東京電力ホールディングスおよび協力会社に聞き取り調査を行い、想定される危険動作を分類しデータベースを構築する。

Outline of Final Research Achievements

In order to develop an automatic judgment method for dangerous actions during decommissioning work, a Scene Graph method was developed that expresses the situation in an image in a graph structure by fusing different deep learning models such as object recognition and posture estimation. In addition, a method has been developed to automatically convert the occupational safety rule, which is the basis of risk judgment by comparing with images, into a newly devised hierarchical Scene Graph using a natural language processing method and an ontology model. This research is the first in the world to integrate different deep learning models such as images (object recognition and posture estimation) and natural language processing. By comparing the two Scene Graphs, we have also succeeded in detecting work safety hazards with high accuracy in real time.

Academic Significance and Societal Importance of the Research Achievements

本研究の学術的意義は、画像(物体認識と姿勢認識と)と自然言語処理という異なる深層学習モデルを融合した手法としては世界初の研究であることである。一般的に深層学習を用いた研究は、画像は画像、言語は言語というように、各々の分野の中で閉じており異種深層学習同士を組みあわせることはない。しかし、人間の脳は視覚、聴覚、触覚、言語など多様な情報を同時に取り入れ比較することで情報を適切に処理している。本研究では、そのような異なる深層学習モデルを融合するための共通データフォームとしてグラフ構造が有効であることを提案し、かつその有効性を実証した。この成果は、異種深層学習モデル融合という新たあ分野の基礎を築いた。

Report

(4 results)
  • 2021 Annual Research Report   Final Research Report ( PDF )
  • 2020 Research-status Report
  • 2019 Research-status Report
  • Research Products

    (36 results)

All 2022 2021 2020 2019 Other

All Journal Article (7 results) (of which Int'l Joint Research: 3 results,  Peer Reviewed: 7 results,  Open Access: 3 results) Presentation (27 results) (of which Int'l Joint Research: 12 results,  Invited: 2 results) Remarks (2 results)

  • [Journal Article] Graph-based linguistic and visual information integration for on-site occupational hazards identification2022

    • Author(s)
      Chen Shi、Demachi Kazuyuki、Dong Feiyan
    • Journal Title

      Automation in Construction

      Volume: 137 Pages: 104191-104191

    • DOI

      10.1016/j.autcon.2022.104191

    • Related Report
      2021 Annual Research Report
    • Peer Reviewed
  • [Journal Article] Towards on-site hazards identification of improper use of personal protective equipment using deep learning-based geometric relationships and hierarchical scene graph2021

    • Author(s)
      Chen Shi、Demachi Kazuyuki
    • Journal Title

      Automation in Construction

      Volume: 125 Pages: 103619-103619

    • DOI

      10.1016/j.autcon.2021.103619

    • Related Report
      2021 Annual Research Report 2020 Research-status Report
    • Peer Reviewed
  • [Journal Article] owards on-site hazards identification of improper use of personal protective equipment using deep learning-based geometric relationships and hierarchical scene graph", Automation in Construction2021

    • Author(s)
      Shi Chen, Kazuyuki Demachi,
    • Journal Title

      Applied Sciences

      Volume: 125 Pages: 103619-103619

    • Related Report
      2021 Annual Research Report
    • Peer Reviewed
  • [Journal Article] A vision-based approach for ensuring proper use of Perosonal Protective Equipment (PPE) in decommissioning site of Fukushima Daiichi nuclear power station2020

    • Author(s)
      Shi Chen, Kazuyuki Demachi
    • Journal Title

      Applied Sciences

      Volume: 10.15 Issue: 15 Pages: 5129-5129

    • DOI

      10.3390/app10155129

    • Related Report
      2020 Research-status Report
    • Peer Reviewed
  • [Journal Article] Proposal of an insider sabotage detection method for nuclear security using deep learning2019

    • Author(s)
      Shi Chen, Kazuyuki Demachi
    • Journal Title

      Journal of Nuclear Science and Technology

      Volume: 56(7) Pages: 599-607

    • Related Report
      2019 Research-status Report
    • Peer Reviewed / Open Access / Int'l Joint Research
  • [Journal Article] Robust human pose estimation from distorted wide-angle images through iterative search of transformation parameters2019

    • Author(s)
      Daisuke Miki, Shinya Abe, Shi Chen, Kazuyuki Demachi
    • Journal Title

      Signal, Image and Video Processing

      Volume: 14(4) Pages: 693-700

    • Related Report
      2019 Research-status Report
    • Peer Reviewed / Open Access / Int'l Joint Research
  • [Journal Article] ROBUST HUMAN MOTION RECOGNITION FROM DISTORTED WIDE-ANGLE IMAGES FOR VIDEO SURVEILLANCE2019

    • Author(s)
      Miki Daisuke、Abe Shinya、Chen Shi、Demachi Kazuyuki
    • Journal Title

      The Proceedings of the International Conference on Nuclear Engineering (ICONE)

      Volume: 2019.27 Issue: 0 Pages: 1789

    • DOI

      10.1299/jsmeicone.2019.27.1789

    • NAID

      130007773690

    • ISSN
      2424-2934
    • Related Report
      2019 Research-status Report
    • Peer Reviewed / Open Access / Int'l Joint Research
  • [Presentation] 画像深層学習と自然言語処理を融合した検知手法の提案・開発・実装2022

    • Author(s)
      出町和之、陳実
    • Organizer
      日本原子力学会2022年春の年会
    • Related Report
      2021 Annual Research Report
  • [Presentation] A Dynamic Graph-based Time Series Analysis Framework for On-site Occupational Hazards Identification2021

    • Author(s)
      Shi Chen, Feiyan Dong, Kazuyuki Demachi
    • Organizer
      38th International Symposium on Automation and Robotics in Construction
    • Related Report
      2021 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Malicious Action Monitoring Technology By Coupling Deep Leanings Of Image Recognition And Natural Language Processing2021

    • Author(s)
      Kazuyuki Demachi
    • Organizer
      INMM/ESARDA 2021 Joint Annual Meeting
    • Related Report
      2021 Annual Research Report
    • Int'l Joint Research
  • [Presentation] A Novel Detection Approach to Preventing Theft of Nuclear Materials Using Deep Learning-based Object Detection and Human Pose Estimation2021

    • Author(s)
      Yuki Yokochi, Shi Chen, Kazuyuki Demachi
    • Organizer
      INMM/ESARDA 2021 Joint Annual Meeting
    • Related Report
      2021 Annual Research Report
    • Int'l Joint Research
  • [Presentation] A Graph-Based Scene Understanding Approach for Ensuring Proper Use of Personal Protective Equipment at the Decommissioning Site of Fukushima Daiichi Nuclear Power Station2021

    • Author(s)
      Shi Chen, Kazuyuki Demachi
    • Organizer
      28th International Conference on Nuclear Engineering (ICONE28)
    • Related Report
      2021 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Towards Malicious Action Detection for Nuclear Security via Integrated Deep Learning Based Image Recognition and Natural Language Processing2021

    • Author(s)
      Kazuyuki Demachi, Shi Chen, Masaki Sudo
    • Organizer
      28th International Conference on Nuclear Engineering (ICONE28)
    • Related Report
      2021 Annual Research Report
    • Int'l Joint Research
  • [Presentation] 核施設における通常の操作からの盗取の識別2021

    • Author(s)
      横地悠紀, 陳実, 出町和之
    • Organizer
      日本保全学会第17回学術講演会
    • Related Report
      2021 Annual Research Report
  • [Presentation] 画像認識と自然言語処理の連成による核セキュリティ悪意行為検知2021

    • Author(s)
      出町和之、陳実
    • Organizer
      日本保全学会第17回学術講演会
    • Related Report
      2021 Annual Research Report
  • [Presentation] 物体識別と姿勢推定を用いた核物質等盗取行為検知手法の提案2021

    • Author(s)
      横地悠紀, 陳実, 出町和之
    • Organizer
      2021年電子情報通信学会総合大会
    • Related Report
      2020 Research-status Report
  • [Presentation] 画像識別と自然言語処理の AIインターフェイス2021

    • Author(s)
      出町和之, 陳実, 須藤大揮
    • Organizer
      日本原子力学会2021年春の年会
    • Related Report
      2020 Research-status Report
  • [Presentation] A Real-Time Automated Approach for Ensuring Proper Use of Personal Protective Equipment (PPE) in Construction Site2020

    • Author(s)
      Shi Chen, Kazuyuki Demachi, Manabu Tsunokai
    • Organizer
      18th International Conference on Computing in Civil and Building Engineering
    • Related Report
      2020 Research-status Report
    • Int'l Joint Research
  • [Presentation] A Novel System for Automated Proper Use Identification of Personal Protective Equipment in Decommissioning Site of Fukushima Daiichi Nuclear Power Station2020

    • Author(s)
      Shi Chen, Kazuyuki Demachi
    • Organizer
      2020 International Conference on Nuclear Engineering
    • Related Report
      2020 Research-status Report
    • Int'l Joint Research
  • [Presentation] Weakly Supervised Deep Neural Network for Bearing Fault Diagnosis2020

    • Author(s)
      Daisuke Miki, Kazuyuki Demachi
    • Organizer
      2020 International Conference on Nuclear Engineering
    • Related Report
      2020 Research-status Report
    • Int'l Joint Research
  • [Presentation] Unnatural Human Motion Detection using Weakly Supervised Deep Neural Network2020

    • Author(s)
      Daisuke Miki, Shi Chen, Kazuyuki Demachi
    • Organizer
      2020 International Conference on Artificial Intelligence for Industries (AI4I)
    • Related Report
      2020 Research-status Report
    • Int'l Joint Research
  • [Presentation] 画像 AI と自然言語処理 AI とのインターフェイス2020

    • Author(s)
      出町和之、陳実
    • Organizer
      日本核物質管理学会(INMMJ)第41回年次大会
    • Related Report
      2020 Research-status Report
  • [Presentation] 自然言語処理を用いた作業安全判定2020

    • Author(s)
      須藤大揮、出町和之
    • Organizer
      日本核物質管理学会(INMMJ)第41回年次大会
    • Related Report
      2020 Research-status Report
  • [Presentation] eakly Supervised Graph Convolutional Neural Network for Human Action Localization2020

    • Author(s)
      Daisuke Miki, Shi Chen, Kazuyuki Demachi
    • Organizer
      EEE Winter Conference on Applications of Computer Vision (WACV), Colorado, United States
    • Related Report
      2019 Research-status Report
    • Int'l Joint Research
  • [Presentation] Skeleton based hand motion recognition using convolutional neural network. 2019 the Institute of Electronics, Information and Communication Engineers General Conference2020

    • Author(s)
      Shi Chen, Ryo Kubota, Kazuyuki Demachi
    • Organizer
      電子情報通信学会2019年総合大会
    • Related Report
      2019 Research-status Report
  • [Presentation] 弱教師付き学習による人物の不自然動作の検知2020

    • Author(s)
      三木 大輔,出町 和之
    • Organizer
      電子情報通信学会2019年総合大会
    • Related Report
      2019 Research-status Report
  • [Presentation] 弱教師付き学習による人物の不自然動作の検知2020

    • Author(s)
      有路翔太、三木大輔、出町和之
    • Organizer
      電子情報通信学会2019年総合大会
    • Related Report
      2019 Research-status Report
  • [Presentation] AI深層学習を用いた作業安全監視技術2020

    • Author(s)
      出町 和之
    • Organizer
      日本保全学会第20回保全セミナー
    • Related Report
      2019 Research-status Report
    • Invited
  • [Presentation] 作業安全監視のためのAIを活用した映像・信号解析技術2020

    • Author(s)
      三木 大輔
    • Organizer
      日本保全学会第20回保全セミナー
    • Related Report
      2019 Research-status Report
    • Invited
  • [Presentation] Insider sabotage detection for nuclear facilities using deep learning2019

    • Author(s)
      Shi Chen, Kazuyuki Demachi
    • Organizer
      The 27th International Conference on Nuclear Engineering. May 19-24, 2019
    • Related Report
      2019 Research-status Report
    • Int'l Joint Research
  • [Presentation] Robust Human Motion Recognition from Distorted Wide-Angle Images for Video Surveillance2019

    • Author(s)
      Daisuke Miki, Shinya Abe, Shi Chen, Kazuyuki Demachi
    • Organizer
      The 27th International Conference on Nuclear Engineering (ICONE27)
    • Related Report
      2019 Research-status Report
    • Int'l Joint Research
  • [Presentation] 広角映像の歪みに頑健な注目点検出手法の開発と人物動作解析への応用2019

    • Author(s)
      三木 大輔,阿部 真也,陳 実,出町 和之
    • Organizer
      日本保全学会第16回学術講演会
    • Related Report
      2019 Research-status Report
  • [Presentation] 深層学習による動画データからの手元動作認識2019

    • Author(s)
      出町和之、陳実
    • Organizer
      日本保全学会第16回学術講演会
    • Related Report
      2019 Research-status Report
  • [Presentation] 技術の現場と倫理の相互作用:AI技術を例に(3)保全分野におけるAI導入に向けた取り組み事例2019

    • Author(s)
      出町和之
    • Organizer
      日本原子力学会2019年秋の大会
    • Related Report
      2019 Research-status Report
  • [Remarks] 出町研究室 研究紹介

    • URL

      https://www.demachilab.org/research

    • Related Report
      2020 Research-status Report
  • [Remarks] 出町研究室 研究紹介

    • URL

      https://www.demachilab.org/research-in-detail-j

    • Related Report
      2019 Research-status Report

URL: 

Published: 2019-04-18   Modified: 2023-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi