Research On On-Line Detection Technology For Industrial Robot Based On Binocular Vision

Compared with the traditional machining method and the CNC machining, robot assisted machining has good flexibility, large operating space, high efficiency, low cost, easy to be integrated with other advanced technology, which has been used for casting grinding, deburring, chamfering, rapid prototyping, carving, etc. With advantages such as improving work efficiency of robot, easy for complex trajectory planning, collision and interference inspection, convenient to show programming result, robot off-line programming is widely used in the production practice. However, because of the uncertainty of the processing environment such as the robot position error, the fixture positioning error, uncertainty of components, the application of robot off-line programming is restricted. Using multiple information fusion such as visual information and force sensing information to improve robot perception of the environment, thus to provide real-time quality testing of parts or processing parameters become the inevitable choice.In this paper, the online detection system based on binocular stereo vision is introduced:First, outlined the study status of robot binocular stereo vision technology. Second, using d-h method to establish the Motoman UP6robot kinematics model, proposed a simple inverse solution algorithm, and based on this an offline programming system access machining path from DXF file is completed, which can provide the source of machining information. And then analyzes mathematical model of the on-line detection system based on binocular vision, including the camera imaging model and the principle of binocular stereo imaging, using Bouguet’s stereo calibration algorithm to obtain the ideal binocular stereo imaging model, and realizes the3D reconstruction based on Harris corner points matching and the3D reconstruction based on block matching algorithm. Again, this paper put forward a fast hand-eye calibration algorithm, which is simple, fast, and of high precision. Finally, this paper implements the stereo vision robot control system based on Kalman filter, which can effectively locate moving targets on-line, and complete machining task at the same time.The development of the software system is based on QT, Open CV and PCL (Point Cloud Library), using C++to complete each experiment. And the results show that the positioning accuracy of binocular vision system meets the requirement.