Hand-eye Calibration Research For Welding Robot Based On Active Vision

Vision as one of the most important means in robot intellectualization, applied to weldedrobot can improve the efficiency and precision in significant measure. Where, the hand-eyeachieved the transition between the camera coordinate and the manipulator coordinate, andthe calibration result will have influence on the work precision of robot directly. Seeing thatthe hand-eye calibration is accomplished based on specific mechanical model so far. So, inthis study, a hand-eye calibration method which is widely used in industrial welding field ispresented.First, the framework of hardware and the necessary devices of the active vision systemare analyzed. And, according to the applied background the main technical data of the deviceswhich comprise the active vision system is studied, under this condition the suitable devices ischose, then a monocular structured light vision measurement system and cylindrical typewelding robot test platform is built successfully.Second, previous camera calibration methods are classified and studied, in which theZhang calibration method is specialized. The measurement principle of monocular structuredoptical system is analyzed and modeled. According to the idea of “vanish points”, animproved calibration method of inside parameters of active vision system is presented, whichuses the Least square method to optimize the calibration data effectively, and can get thecamera parameters and monocular structured optical system’s parameters at the same timeafter only one calibration. To verify the practicality of this method, a simulation is achievedon the cylindrical type welding robot test platform. The simulation result indicated that,compared with the previous method, this method reduce the calibration time of insideparameters, and the calibration accuracy has reached the industry standard at the same time.Third, intensive study the conventional hand-eye calibration methods, and conclude theprocedure of them. Then a improved hand-eye calibration method is proposed based on aconclusion of the previous calibration methods. The improved method calibrates the rotationmatrix and translation vector respectively based on the parameter analysis method. It need notthe specific calibration object, and can achieved the calibration of rotation matrix swiftly andeffectively utilizes the characteristic points to determine the relationship between the cameramoves fore and after. At the same time the translation vector can be calibrated according tothe relationship between camera and manipulator determined by virtual coordinate.Finally, to verify the proposed the hand-eye calibration method of active vision system,the tracked test is done. The result shows that this is a stable and accurate method.