Research On Intelligent Nursing Wheelchair Bed System Based On Machine Vision Navigation And Obstacle Avoidance Algorithms

Nowadays,the social aging is aggravating,meanwhile,science and technology are continuously developing.As the result,the multifunctional electric wheelchair bed with both nursing function and driving function gradually appear on the market.Such products can not only help users to carry out rehabilitation training,but also expand the range of activities of users.Aiming at the problem that the large size of the motorized wheelchair bed in wheelchair condition makes it difficult to drive,this research is proposed an intelligent wheelchair bed system based on machine vision navigation and obstacle avoidance.This system can reduce the driving difficulty of the wheelchair bed and improve the safety and applicability of the wheelchair bed.The hardware and software of the visual navigation and obstacle avoidance system is designed completely according to the market requirements.Deep image processing is applied to solve the semantic segmentation problem and detect the free space around the wheelchair bed.This research is proposed an indoor scene bilateral segmentation network which uses two-step deconvolution up-sampling instead of single-step interpolation up-sampling in bilateral segmentation network,and the model is trained on the indoor scene data set.By analyzing the camera's model and calibration algorithm,and comparing various monocular vision measurement methods,then the geometric relation method based on ground plane constraint is used to measure the position of obstacles.The path planning algorithm of the wheelchair bed is a combination of the Smooth A* algorithm and dynamic window method.The Smooth A* algorithm adds the attenuate cost of direction and can plan the global optimal path that is smoother and has fewer sharp turns.Dynamic window method performs the real-time local path planning according to the dynamic environment and considering the size and dynamic characteristics of wheelchair bed.To ensure that the wheelchair bed moves to the destination without collision.Finally,the visual navigation and obstacle avoidance system is deployed in the prototype of multifunctional nursing wheelchair-bed and the experiments are designed to verify everyalgorithm to test the whole system.The experimental results show that the improved bilateral segmentation network in this research achieves 90.03% CPA and 78.2% mean Io U,and the speed is 53 fps.The error of monocular measurement technology is less than 3% in the longitudinal distance and less than 5% in the transverse distance.The path planning algorithm combined with the Smooth A* algorithm and the dynamic window method is suitable for the navigation and obstacle avoidance task of the wheelchair bed,which can plan a smooth and convenient path.The whole system can meet the requirements of the tasks with a certain generalized performance.