Research On Robot Vision Guided Cutting System In Complex Working Conditions

With the increasing development of the manufacturing industry,traditional bevel cutting methods with manual cutting and semi-automatic cutting have become increasingly difficult to meet industrial needs.At the same time,manual cutting and semi-automatic cutting are difficult to guarantee cutting precision and cutting efficiency.Therefore,it is urgent to study the complex vision robot cutting vision guidance which can realize panoramic image acquisition,automatic workpiece recognition,automatic and accurate workpiece cutting path and cutting in the factory environment.The system realizes the automation of the bevel cutting.The paper starts from the actual factory cutting environment,builds an experimental platform,simulates the real cutting environment,uses a larger working platform,and selects cutting workpieces with different thicknesses in the experiment.The external axis is mounted on the KUKA robot as an external actuator,and the camera is fixed to the robot in an eyein-hand manner.Combining the advantages of 2D images and 3D point clouds,using 2D camera for image stitching and template matching,the workpiece recognition and centroid position acquisition are completed;the 3D camera is used for point cloud stitching to obtain the workpiece point cloud contour in the working platform,and according to 2D The workpiece recognition result matches the target workpiece point cloud,and the target workpiece robot cutting path is extracted to realize the visual guidance cutting of the complex working condition robot.The main work of the thesis is as follows:(1)In the aspect of two-dimensional image processing,the image stitching algorithm for the experimental environment described in the paper is mainly studied.According to the characteristics of the experimental environment studied in this paper,the image mosaic algorithm based on the ORB feature of ROI region is selected for image mosaic.The feature points of the Fast corner point are used to identify the feature points,and the corresponding feature matching is realized by using the Brief feature description and the Hamming distance.The matching result is purified by the random sampling consistency algorithm,and the image registration is realized according to the purified feature points,and finally the weighted fusion is implemented.Two-two stitching of the image.And this algorithm is used to splicing multiple working platforms to obtain a panoramic image of the working platform that meets the accuracy and time requirements.After the splicing is completed,the contour extraction based on HSV color segmentation and template matching are used for target recognition,and the centroid position of the workpiece is obtained to prepare for the three-dimensional point cloud processing.(2)In the aspect of 3D vision,the point cloud stitching algorithm for this experimental environment is mainly studied.Based on the experimental environment of the paper,the point cloud collected by the AT camera is subjected to point cloud filtering and point cloud denoising,and then the point cloud is segmented to obtain the point cloud surface of the workpiece.Then,the point cloud coarse splicing method based on covariance matrix and the ICP point cloud splicing algorithm based on kDtree data structure of ROI region are studied.Based on the original ICP algorithm,image segmentation based on ROI region is used to reduce Iterate the number of point clouds,use the kd-tree data structure to accelerate the ICP algorithm,and finally achieve precise stitching of the point cloud.After experimental verification,the image stitching algorithm described in the paper has a stitching time of less than 2.5s and a stitching error of less than 0.1mm,which meets the requirements of subsequent processing.(3)It is proposed to realize the robot vision guidance by combining the two-dimensional camera and the three-dimensional camera.Using their respective advantages,using 2D vision for image stitching and target recognition,the centroid position of the target workpiece is obtained.The three-dimensional camera is used for point cloud splicing,and the target workpiece cutting path is extracted according to the two-dimensional image recognition result,thereby realizing visual guidance cutting of complex working conditions.The final visual guidance error of the system is less than 2mm,which meets the cutting accuracy requirements of the workpiece groove.Through the combination of two-dimensional vision and threedimensional vision,the paper realizes the robot vision guiding groove cutting path acquisition in complex working conditions that meets the precision requirements.It has certain feasibility and practical value in industrial production.