Development of a Depth-Camera-Based Foot Behavior Evaluation System with Automatic Hazard Scene Detection
危険場面の自動検知と深度カメラを用いた足行動評価システムの開発
- Delivery
- Available on the other site
- Click here to order.
- Publication code
- 20264469
- Paper/Info type
- JSAE Transaction
Vol.57 No.3
- Pages
- 679-685(Total 7 p)
- Date of publication
- May 2026
- Publisher
- JSAE
- Language
- Japanese
Detailed Information
| Category(J) | 論文 Translation |
|---|---|
| Category(E) | Paper |
| Author(J) | 1) 林 弘昭, 2) 洲﨑 一眞, 3) 菅野 重樹, 4) 一色 聡, 5) 亀﨑 允啓 |
| Author(E) | 1) Hiroaki Hayashi, 2) Kazuma Suzaki, 3) Shigeki Sugano, 4) Akira Isshiki, 5) Mitsuhiro Kamezaki |
| Affiliation(J) | 1) 早稲田大学, 2) 早稲田大学, 3) 早稲田大学, 4) NXキャッシュ・ロジスティクス株式会社, 5) 東京大学 |
| Abstract(J) | 事故予防には、危険に即座に対応する「予防動作」が重要である。本研究では、前方カメラから危険場面（例：路駐車通過、無信号交差点通過）を検出し、足元の深度カメラから足の構え行動を検出することで、場面に応じた足行動を自動的に評価するシステムを開発した。実験では、指導員の評価と90%以上一致した。 Translation |
| Abstract(E) | This study proposes a system that automatically evaluates drivers' foot behavior based on driver monitoring, with the aim of assessing daily driving habits. A forward-facing camera was used to capture the driving environment, and a pre-trained Faster R-CNN model was applied for object detection, classifying hazardous situations into five levels. In parallel, a depth camera installed near the pedals measured the distance to the pedals, enabling foot actions to be classified into the same five-level scale. A foot behavior was considered safe if its level was equal to or higher than the corresponding hazardous situation level. Experiments using eight days of driving data from seven professional drivers showed that the hazardous situation classification achieved over 90% ture positive accuracy in all scenarios, while the agreement rate between the system's foot behavior evaluation and instructor assessments exceeded 95%. |
