Human Tracking Method For Mobile Robots Based On RGB-D Data

With the rapid development of robots and artificial intelligence,intelligent robots have become a research hotspot.Compared with traditional industrial robots,it has strong perception of the outside world and can complete dynamic,complex,flexible and precise operations in an unstructured environment.So it has great research and application value.But for the robot,the premise and basis of human-computer cooperation is human tracking.So the study of target tracking methods is of very significant meaning to intelligent robots.Based on the analysis of the research status of target tracking algorithm at home and abroad,this paper proposes a human tracking method for mobile robot based on RGB-D data and realize the human target tracking in complex environment.The main work of this paper are as follows:(1)A MeanShift target tracking method based on window adaptation is proposed to solve the problem that the traditional MeanShift algorithm causes the target to be biased or even lost due to the inability of the window to adapt adaptively.First,the tracking window of the target is updated by the distance information on the depth image.Then,the principal component analysis method and the Kalman filter are used to update the size of the target tracking window.(2)The depth information is introduced and a human target tracking method based on adaptive Kalman filtering is proposed.It solves the problem that the window adaptive MeanShift target tracking method causes the target lost due to target occlusion or background color is similar.This algorithm mainly uses the relationship between the mobile robot and the target to establish the mobile robot following model.Then,a selfadaptive factors is introduced in the hypothesis testing.Meanwhile,adaptive Kalman filter is designed to estimate the tracking state.(3)The human target tracking system of mobile robot is designed by ASUS Xtion Pro camera,PC and DFROBOT robot.Then,the algorithm is implemented and applied.Finally,the paper discusses the content of the full-text work,analyzes the advantages and disadvantages of the human target tracking algorithm,and forecasts its future improvement direction.