Pupil Control Intelligent Wheelchair based on MediaPipe
Jingpei Zhou, Liankui Qiu, Yinggang Li, Zhengshuai Wang, Yixin Guo · 2023
In order to solve the problem that some quadriplegics cannot control the wheelchair autonomously, a vision tracking wheelchair control system which combines hardware and software is studied. The Raspberry Pi camera is mounted directly in front of the seat and is used to gather real-time information about the person's face. Using Google pre-trained MobileNetv2 based MediaPipe Face, as well as OpenCV image processing, to achieve the key point of face extraction. In consideration of People's Daily living habits, through the analysis of the direction of sight, eyes closed, mouth closed and other conditions output corresponding commands. The whole system runs on a Raspberry Pi 4B, and the output commands are sent to a motor controller to control the wheelchair. In addition to add automatic obstacle avoidance function to ensure the absolute safety of personnel. The system can operate in a variety of environments, such as background brightness and face occlusion, the system has high robustness.