Research On Tracking And Grabbing Fast Moving Object Based On Real-time Vision

The first industrial robot was built in the late 1950 s.With the development of robot technology for nearly sixty years,human beings are replaced by robots in more and more fields.However,most of the industrial robots still use "teaching programming" or "off-line planning" to complete the task.Under this way,the flexibility of the production line is poor.The development of machine vision has gradually changed this situation.The ability of the robot to perceive the environment is improved,so that it has a higher flexibility,autonomy and intelligence under working.Robots can complete the tasks of detecting,tracking and grabbing on their own with the application of machine vision.In this paper,vision measurement is used to realize the real-time locating of the moving object,and industrial manipulator is used to realize the tracking and grabbing part.The main achievements of this paper are as follows:Firstly,two recognition algorithms are developed to realize the real-time locating of the moving bottles.Secondly,the state space model of the moving object is established and kalman filter with adaptive measurement covariance is used to track the movement of the object,which not only eliminates the recognition error,but also realizes the accurate prediction of the bottle in the future.Thirdly,a strategy of the estimation of grasping point is proposed.The grabbing time is calculated according to the motion state of the manipulator and the bottle and the tracking path of the manipulator is determined in the form of online trajectory planning.The effectiveness of the online trajectory planning strategy is verified by experiments.In the end,this paper designed a grab system.The manipulator achieves the task of grabbing the bottle by online trajectory planning method.