Scholarlink

This paper presents the design and evaluation of a four-wheeled mecanum mobile robot for real-time person tracking using computer vision. The system integrates YOLOv8n for object detection, Deep SORT for identity-aware tracking, and MediaPipe for gesture recognition. A custom PID controller regulates wheel motion based on image-derived tracking errors. Experiments conducted in indoor and outdoor environments with dynamic occlusions, clothing variations, and obstacles show that the robot maintains stable tracking, avoids static obstacles via ultrasonic sensing, and adapts to real-time distance and deviation changes. The system achieves high tracking accuracy with only minor performance degradation under intense lighting, demonstrating its effectiveness for real-world dynamic applications.

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

Development of a Vision-Based Mobile Robot for Real-Time Person Tracking Using YOLOv8n and Deep SORT

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Duong Dinh Tu,
  • Ho Sy Phuong,
  • Le Van Chuong,
  • Dinh Van Nam,
  • Phan Van Du,
  • Cao Nguyen Tien Dung

摘要

This paper presents the design and evaluation of a four-wheeled mecanum mobile robot for real-time person tracking using computer vision. The system integrates YOLOv8n for object detection, Deep SORT for identity-aware tracking, and MediaPipe for gesture recognition. A custom PID controller regulates wheel motion based on image-derived tracking errors. Experiments conducted in indoor and outdoor environments with dynamic occlusions, clothing variations, and obstacles show that the robot maintains stable tracking, avoids static obstacles via ultrasonic sensing, and adapts to real-time distance and deviation changes. The system achieves high tracking accuracy with only minor performance degradation under intense lighting, demonstrating its effectiveness for real-world dynamic applications.