Development of Evaluation System of Driver's Cognitive Ability based on Driving Behavior

2023 Fiscal Year Final Research Report

• Project/Area Number: 20K20258
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 90150:Medical assistive technology-related
• Research Institution: Suwa University of Science
• Principal Investigator: Hashimoto Kohjiro 公立諏訪東京理科大学, 工学部, 講師 (00756588)
• Project Period (FY): 2020-04-01 – 2024-03-31
• Keywords: 自動車の運転場面検出 / 運転行動データ分析 / 深層学習

Outline of Final Research Achievements

The main cause of accidents among elderly drivers is a decline in cognitive abilities, and early detection of this decline is needed. Although assessment indices for cognitive abilities have been proposed based on existing research, they are designed for use with simulators and cannot be applied during actual driving. The reason for this is the scene detection problem. To calculate assessment indices, it is necessary to detect scenes from various situations encountered while driving that allow for the evaluation of cognitive abilities. Therefore, this study proposes a method for detecting scenes during driving that allow for the evaluation of cognitive abilities. This method adopts a supervised deep learning approach and confirms higher detection accuracy compared to existing scene detection methods. Additionally, domain adaptation techniques are applied, and a scene detection method using simulator data is also proposed.

Free Research Field

機械学習、画像処理、データマイニング

Academic Significance and Societal Importance of the Research Achievements

現在の自動車には車載カメラ、CANが搭載されており、運転中の走行映像及び操作情報の取得が可能である。本研究は、こうして得られる一連の運転行動データの中から高度な認知を要求する場面を検出する手法を提案したものである。これにより、既存の認知能力評価指標を用いることができ、実運転データに基づくドライバの認知能力評価が可能となる。これにより、高齢ドライバに限らず、運転能力低下の早期発見が可能となり、交通事故の未然防止が期待できる。