Positioning Error And Correction Of Visually Guided Grasping Manipulator On Operation Plane

Machine vision is widely used in the control and industrial inspection, intelligentrobots are widely used to grab the products on platform in industrial production line atpresent. Positioning of the products is necessary before grabing, and positioningaccuracy directly affects the result of grabing. Monocular vision and binocular visionare frequently-used visual positioning methods. Monocular vision positioning is analgorithm which uses image sequence taken by single camera to calculate thedirection and position of the objects to be measured. Compared to binocular vision,monocular vision is more and more widely used because of its convenient and simple.At present, Researches on monocular vision positioning and positioning error aremature, but there is seldom quantitative calculation for the location distribution of thepositioning error. Complicated nonlinear calibration is usually used to improve thepositioning accuracy. This paper uses a method to analyze the positioning error. First,builds a monocular vision positioning model, then, builds a monocular visionpositioning error model. After analyzing the error of model, the positioning error iscompensated and corrected by the monocular vision positioning error model thatestablished before.First, based on ideal pin-hole imaging model, the relationship between thehorizontal and vertical coordinates of the points on the plane in space and the imagecoordinates can be acquired to position the points on the plane. Then after the analysis,it is concluded that the key factor which affects the positioning error is radialdistortion of the imaging system. Positioning error model is established based on theradial distortion model. Finally, Through the experiment, calibrating the camera,locating the144corners of a planar checkerboard, fitting undetermined coefficients ofthe positioning error model. Experimental results indicate that the R-square of thefitting is up to over0.9. After compensating by the positioning error model, thepositioning error of the144corners of the planar checkerboard has a significantreduce, from2.15mm,1.94mm to0.62mm,0.29mm. It can be proved that thepositioning error model based on monocular vision raised by this paper is correct andradial distortion is the key factor which affects the positioning error.At last, based on the theoretical research, and the machine design of the system that provided by my tutor, the author finished the design of the modules of the systempreliminarily. Including the perfection attained design of the machine design, thehardware design and the software design.