Research On An Auto-Following Method Of Wheelchair Based On Binocular Vision

Nowadays,the problem of social aging is prominent and the proportion of the disabled population is decreasing slowly.The huge population base of the elderly and the disabled has weakened the functions of social care for such populations.While,with the help of effective living assistance tools,some of them can take care of themselves.As important living assistance tools,wheelchairs are not only transportation vehicles,but also promote the elderly and the disabled to do some physical exercises and take part in social activities.Inspired by existing following products such as following shopping carts,following suitcases,it is hoped to realize the auto-following function of wheelchairs,which can automatically execute continuous and stable following actions after designating the target pedestrian.For users,it lightens the load,saves time and energy,and promotes communication with others.Therefore,this paper focuses on the research of the auto-following method of wheelchair based on binocular vision.In order to meet the requirements of realizing a friendly interactive environment between users and wheelchairs,the target pedestrian positioning and tracking,wheelchair following control and dynamic obstacle avoidance,a hybrid architecture,electronic hardware system scheme and software system functions of the auto-following wheelchair system are proposed.Aiming at the misrecognition and missing recognition of pedestrian detection and tracking algorithm based on monocular vision when the pedestrian changes posture or is occluded,the pedestrian detection and tracking method based on binocular vision is proposed.This method firstly detects pedestrians by applying human pose estimation technology and specifies the target pedestrian,then maps the key joints of the neck skeleton which represent pedestrians from the image coordinate system to the camera coordinate system,uses Kalmal filter to estimate the target pedestrian’s three-dimensional coordinates in the camera coordinate system,and finally tracks the target pedestrian based on the assumption of continuity of pedestrian space movement.Experimental results show that the pedestrian detection and tracking method based on binocular vision has good stability and real-time performance,which can locate and track the target pedestrian under various interferences.Taking into account normal indoor and outdoor scenes and characteristics of low-speed operation of wheelchairs,leader-follower formation model is used to maintain the position and posture relationship between the wheelchair and the target pedestrian while artificial potential field method is used for dynamic obstacle avoidance.Then proportional controller is adopted to output the linear velocity and angular velocity of the differential drive wheelchair for realizing the auto-following function of wheelchair.Experimental results show that wheelchair can follow the target pedestrian continuously and stably and avoid obstacles in indoor and outdoor scenes by using the following control method.The stereo vision tracking and following control methods are applied to ordinary electric wheelchairs.By adding embedded computers,touch screen,binocular camera,speech synthesis module,speakers and other electronic devices,the body structure of wheelchair is reformed,electronic hardware system is developed and software system based on ROS is realized.Functions such as automatic following,joystick control,human-computer interaction and voice broadcast are achieved.User testing confirms the effectiveness of the auto-following method of wheelchair based on binocular vision and the practicability of the auto-following wheelchair system.