Research Of Robot Dynamic Target Tracking Based On Vision

At this stage,more and more automated production line introduced the industrial robot technology.The conventional operation is to implement the process of repeated grab and placement in structured environment,which is less flexibility.In order to expand the application of industrial robots in various fields,the machine vision is introduced into the robot system to improve the adaptability to the working environment.In this paper,the current situation of the robot vision control system in the field of dynamic target tracking was analyzed,and some shortcomings of the current research were summarized,such as the poor portability of the image algorithm,less open of the control system,lack of intelligence of track system and so on.On this basis,a vision based three axis robot dynamic target tracking system was designed.The main research contents and achievement are as follows:The hardware platform of the vision control system based on PC was built on a three-axis linear arm.The vision subsystem was composed of two cameras,one use installation method with eye fixed in space to control the global vision,the other was fixed at the end of the robot to control the local visual field.The motion control subsystem using “Motion control card + Servo driver + Servo motor” structure mode was of high openness.The motion detection algorithm was the background estimation method,and the moving object tracking algorithm was based on SURF template matching.By combining the two methods,the dynamic target can be located quickly and accurately.The algorithm could be used to locate and track while the target color information and shape information were unknown.The algorithm was implemented by Halcon machine vision algorithm library.Global vision control system adopted the control mode based on the position.Zhang's calibration method was used to solve the inside and outside parameters in the MATLAB Calibration Toolbox;local visual control system adopted the control mode of uncalibrated based on the images.Kalman filter was used to online identification of the image Jacobian.Through the global and local control switch,it solved the problem that the global field of view was blocked by the robot arm,which made system unable to continue tracking and the target movement velocity was too fast to leave the local field of view.The whole process could be completely divorced from human intervention.The intelligent tracking of the dynamic target was realized.The human-computer interface software system was designed to verify the whole visual control system of the robot.Experiment test showed that the target tracking results achieve as expected,which proved the feasibility of the hardware and software scheme and the algorithm.Experience and deficiency was summed up,which laid the foundation for the further research and development of the whole vision control system in the future.