A study on an analysis method for fire behavior using Deep Learning

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K04825
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 23020:Architectural environment and building equipment-related
• Research Institution: National Institute for Land and Infrastructure Management (2021-2022); Building Research Institute (2020)
• Principal Investigator: Iwami Tatsuya 国土技術政策総合研究所, 建築研究部, 防火基準研究室長 (20370744)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: AI / 火災検出 / 火災規模推定 / 発熱速度 / 燃焼実験 / ジオロケーション / 糸魚川火災

Outline of Final Research Achievements

Since many fires are expected to spread in the event of large earthquake, research and development has been carried out on technology to identify the location of fires and flames from aerial images using deep learning, and to accurately and quickly detect the burning conditions (heat release rate, flame shape, etc.).
A program was developed to detect and geolocate fires and estimate burning conditions based on video and camera parameters. And the effectiveness of the method was confirmed through verification using videos of a large-scale fire in Itoigawa City, Niigata Prefecture in 2016 and combustion experiments conducted in the past.

Free Research Field

広域火災

Academic Significance and Societal Importance of the Research Achievements

火災研究・火災対応の分野において深層学習手法を先進的に取り入れて活用することにより、これらの分野における今後の一層の有効活用に向けた研究の進展が期待できる。開発した技術は大規模地震時等における火災発生状況の正確・迅速な把握に資することが期待でき、災害対応の効率化による被害の軽減につながる技術である。併せて火災実験等で収集される映像データの自動解析にも活用可能であり、データ解析の効率化にもつながることが期待できる。