Eye-tracking Control Wheelchair Method Based on Landmarks of Face

At the moment, the issue of fatigue and freedom restriction affects the intelligent wheelchair collection's control source. We looked into ways to enable quadriplegic people use wheelchairs by eye guidance. A manually annotated data set is used to address the issue that the face landmarks are overly numerous and the pupil landmarks are calculated. And the main components of heat map processing, which make it simple to increase the detection window. The lightweight convolution attention mechanism is integrated with the two-stage down and up sampling Unet network to reinforce the constraint between crucial locations and enhance the network's performance. In order to capture the most important details of the user's face in real time, we mounted the Raspberry Pi camera right in front of the wheelchair. We generate the appropriate command information for the wheelchair's motor drive by analyzing various photos. In order to ensure that users have a comfortable experience, additional security features include a specific face signal, speed restriction, and automatic obstacle avoidance. It can achieve 95% accurate real-time landmark detection through experimental verification. Additionally, it can control and track the position of wheelchairs for various users.