Vision-based Robot Personnel Tracking And Navigation Technology In Dynamic Environments

With the progress of technology and economic development,the national demand for quality of life is increasing,and mutually of the new spots of newest analysis,the event of intelligent robots has slowly shifted from industrial robots in factories accustomed improve productivity to service robots accustomed improve quality of life.The development of people's living commonplace and social productivity level has brought additional and additional robots into thousands of households,whether or not it's for private home scenes of sweeping robots,companion robots or for skilled scenes of messenger robots,building robots,etc.additional and additional service robots facilitate folks improve their quality of life.In service robots,auto-following function is a very general robot function,which has a great demand in many scenarios,and at present,there are also many companies making applications on supermarket shopping carts,travel boxes and other scenarios,but at present,most robots' auto-following functions need to wear special signs on the target,which greatly increases the limitation of auto-following function.In response to this,this paper will investigate the vision method to implement robot auto-following function,and focus on visual following and path planning,which are two crucial parts in the function implementation,and the main contents of this paper are arranged as follows:1.The requirements for the automatic following function of the service robot are analyzed,and the importance of visual following as well as path planning in the implementation of the automatic following robot is emphasized.The software architecture of the auto-following algorithm is presented,as well as the hardware components of the verification machine that implements the algorithms in this paper.2.The visual following algorithm in the automatic following robot algorithm is introduced,and the two important steps of classifier training and template matching and updating in the kernel correlation filtering(KCF)algorithm are described.Meanwhile,according to the flow of KCF algorithm,the shortcomings and defects of KCF algorithm are explained,and finally,according to its shortcomings,the improved KCF algorithm based on pedestrian detection and similarity judgment is proposed in this paper.3.The path planning algorithm in the automatic following algorithm is introduced.Firstly,the algorithm proposed in this paper to obtain the target spatial position based on the depth camera data is described,and an improved dynamic window algorithm(DWA)is proposed based on the obtained target spatial coordinates,the robot's own speed and other constraints,which improves the path planning capability of the DWA algorithm in short-distance dynamic scenes by adding the target position score and improving the orientation score and speed score on the basis of the scoring function of the traditional dynamic window algorithm.4.The feasibility of this auto-following algorithm is verified by simulation experiments and laboratory environment experiments.