Scholarlink

<p>Healthcare workspaces can greatly benefit from the employment of robotic assistants in both clinical and non-clinical tasks. However, despite their advantages, a major shortcoming for the deployment of such robotic solutions, limiting their widespread market acceptance, is the fact that they were originally designed for large industrial and warehouse spaces. These are characterized by structured spaces and predictable environments, where robots move along predefined paths and interaction with humans is typically minimal. Herein, state-of-the-art computer vision methods are examined, which enable robots to detect the presence and identify the type of dynamic obstacles inside their visual field, so that they can adapt their navigation accordingly. For this purpose, the COCO128 dataset was augmented with an extra category consisting of nursing robots. Then, a new custom dataset was generated, comprising images of humans and a range of logistics/ nursing robots, captured within realistic hospital settings. Robotic vision systems were trained using contemporary deep learning methods (namely the YOLO—You Only Look Once—architecture and its variations) obtaining promising results in both human and robot detection. Ultimately, the goal of this study is to contribute towards safe robot navigation in healthcare spaces and the deployment of robotic fleets in less structured environments.</p>

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

Real-Time Object and Human Detection using YOLO for Autonomous Robot Navigation in Healthcare Workspaces

  • 期刊论文

  • 发表日期：

    2025年6月11日

查看全文
  • Eleftherios G. Vourkos,
  • Eftychios G. Christoforou,
  • Evropi Toulkeridou,
  • Nacim Ramdani,
  • Andreas S. Panayides

摘要

Healthcare workspaces can greatly benefit from the employment of robotic assistants in both clinical and non-clinical tasks. However, despite their advantages, a major shortcoming for the deployment of such robotic solutions, limiting their widespread market acceptance, is the fact that they were originally designed for large industrial and warehouse spaces. These are characterized by structured spaces and predictable environments, where robots move along predefined paths and interaction with humans is typically minimal. Herein, state-of-the-art computer vision methods are examined, which enable robots to detect the presence and identify the type of dynamic obstacles inside their visual field, so that they can adapt their navigation accordingly. For this purpose, the COCO128 dataset was augmented with an extra category consisting of nursing robots. Then, a new custom dataset was generated, comprising images of humans and a range of logistics/ nursing robots, captured within realistic hospital settings. Robotic vision systems were trained using contemporary deep learning methods (namely the YOLO—You Only Look Once—architecture and its variations) obtaining promising results in both human and robot detection. Ultimately, the goal of this study is to contribute towards safe robot navigation in healthcare spaces and the deployment of robotic fleets in less structured environments.