Scholarlink

In this paper, we present a method for detecting anomalous states of the driver inside a vehicle. The approach consists of three main parts. The first part includes 3D localization of the joints of the driver’s body. In the second step, appropriate features are selected for analysis. The third step involves selecting a suitable neural network that handles the final classification. Since the presented approach is primarily focused on the detection of anomalous behavior over time, the encoder-decoder architecture is chosen. On the basis of the experiments presented, it is shown that promising results can be achieved with the use of the proposed steps.

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

Driver Anomaly Detection Using 3D Human Pose Estimation

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Radovan Fusek,
  • Eduard Sojka,
  • Jan Gaura

摘要

In this paper, we present a method for detecting anomalous states of the driver inside a vehicle. The approach consists of three main parts. The first part includes 3D localization of the joints of the driver’s body. In the second step, appropriate features are selected for analysis. The third step involves selecting a suitable neural network that handles the final classification. Since the presented approach is primarily focused on the detection of anomalous behavior over time, the encoder-decoder architecture is chosen. On the basis of the experiments presented, it is shown that promising results can be achieved with the use of the proposed steps.