Research On Binocular Vision Based Heterogeneous Parts Identification,location And Grasping Technology

Aiming at the problem that the polishing process of heterogeneous faucet needs manual teaching and positioning grasping,which can't achieve full automation,a binocular vision detection scheme is designed and a binocular vision guided robot recognition,positioning grasping system is developed.The main research contents are as follows:1.Overall system design.According to the precision requirement of the system,the models of industrial camera and lens are analyzed and determined.The fixture is designed according to the characteristics of the faucet.The hardware control system is designed according to the fetching process.2.System calibration research.According to the imaging principle of the camera,the radial distortion and tangential distortion are analyzed,the mathematical model of the binocular camera is established,the calibration experiment is carried out,the internal and external parameters of the camera are obtained,and the coordinate system calibration of the tool and the workpiece are completed.By using eye-to-hand hand-eye method,the matrix relationship between the camera and the robot is obtained,and the pose data obtained by the visual system is converted to the robot coordinate system.3.Research on image preprocessing.Median filter,Gaussian filter and other algorithms are studied to remove the influence of image noise and ensure the accuracy of recognition and positioning.The method of Gamma correction is used to enhance the image and reduce the influence of image preprocessing and illumination change on the image characteristics.4.3D coordinate calculation of objects.The method and constraint conditions of stereo matching technology are analyzed,the binocular imaging model is studied,and the types of template matching are described.The grayscale stereo matching technology based on NCC was used to match the images collected by the binocular camera,and the 3D coordinates is calculated according to the binocular imaging model and system calibration results.5.Object attitude calculation.The three-dimensional coordinate can only locate the object,but the deflection Angle of the faucet needs to be obtained in order to realize the locating grasp.According to the Angle difference returned by stereo matching,the object's deflection Angle was calculated,and then the attitude was converted from Euler Angle to the description form of four elements.Finally,the three-dimensional coordinates and deflection Angle of the object were written into the robot control cabinet to realize positioning and grasping.6.System software development.Based on VS2015 development environment,C# and Halcon are used to develop system software,including image acquisition module,binocular calibration module,hand-eye calibration module,communication module,automatic operation system and other modules.7.System experiment.Experiments show that the success rate of grasping the faucet with different postures is about 94.4%.The positioning accuracy of XYZ in three directions was ±0.25 mm,0.07 mm and ±0.2mm by using a three-axis optical motion platform.The comprehensive positioning accuracy of the system is 0.538 mm,which meets the requirement of grasping.