Tracking Registration Technology Based On VSLAM Augmented Reality And Its Application In Hand Function Rehabilitation

Augmented reality developed from virtual reality technology has been applied in many fields in recent years,such as education,medical treatment,industrial maintenance,etc.Among them,the combination of augmented reality technology and hand function rehabilitation technology has gradually become a research hotspot in the field of rehabilitation.This paper designs a rehabilitation program for ADL(activities of daily living)by improving the registration algorithm of augmented reality:First of all,to address the problem that objects registered in VSLAM(visual simultaneous localization and mapping)cannot change posture information such as position and rotation,this paper proposes an augmented reality tracking algorithm based on template tracking and positioning.The method establishes a low-dimensional template for rapid positioning,and then uses an improved non-linear Kalman filter algorithm to predict motion during the tracking stage,reducing the matching area to achieve the goal of reducing time-consuming.Secondly,because most traditional VSLAM environments do not consider the impact of dynamic scenes on the augmented reality registration function,at the same time,there are a large number of dynamic targets in the hand function rehabilitation training scene proposed in this paper.Based on this,this paper proposes the VSLAM-AR algorithm for dynamic target tracking in dynamic scenarios.This method uses affine matrix and GMM(Gaussian mixture model)to establish candidate dynamic regions,assigns candidate dynamic regions at key frames through incremental models,and uses YOLOv3 and particle filter for continuous observation and tracking.The effectiveness of the algorithm proposed in this paper is verified by the TUM data set.Then,this paper combines the above two augmented reality tracking and registration algorithms,and proposes an AR algorithm based on dynamic target tracking under VSLAM.This algorithm can not only effectively eliminate the dynamic targets in the static scenarios,but also ensure that the virtual objects based on the dynamic template target tracking registration are still in the world coordinate system of VSLAM.In order to facilitate patients' hand function rehabilitation training,this paper,using OpenPose for two-dimensional gesture recognition,achieve interaction through collision detection after mapping the key points of the hand to the world coordinate system based on camera's pose information and depth information provided by VSLAM.At last,this paper,after integrating the above technologies,designs and develops a hand function rehabilitation system based on VSLAM augmented reality.Three different rehabilitation-training programs,i.e.,CIMT-AR,BAT-AR and TOT-AR,are designed by combining CIMT(constraint induced movement therapy),BAT(bilateral arm training)and TOT(task-oriented training)with augmented reality.This paper introduces the experimental process of each program and the principle of design,and analyzes the feasibility of the rehabilitation-training program through the experimental results.