Preliminary Study On Mechanical Arm Unloading Device Based On Binocular Vision Positioning

With the continuous improvement of the human cost,enterprises began to seek a new logistics mode to replace the traditional way which relying on manpower and was inefficiency.Intelligent logistics was the ultimate solution,and it was an effective way to reduce logistics cost and improve logistics efficiency by using vision robot technology.Aiming at the existing container logistics mode,the paper put forward an unloading method which based on mechanical arm and binocular visual positioning.Taking regular cartons and irregular tire stack in containers with upper opening for application object,a mechanical arm unloading device based on binocular vision positioning was studied and developed.The design,analysis and experiment of the device were studied in great detail which lead to provide the theory and basis for the development of this product.Finally,on the basis of the study,a mechanical arm device for unloading tyre and carton based on binocular vision positioning was successfully developed.First of all,according to the structural characteristics of the tire,the pneumatic tire gripper with error adaptability was designed.The key components of the gripper(right angle retractor)were studied,the structure of the right angle retractor was improved by static analysis.The kinematics of the manipulator was studied,and the binocular vision system was designed on the third axis of the manipulator.The parallel binocular vision measurement model and its calibration technology were studied.The binocular vision sensor system was established through single camera calibration,stereo calibration and eye-in-hand calibration.Secondly,through template matching based on the SURF feature point,the algorithm of carton location was designed.A fast nearest neighbor matching algorithm for SURF feature points was studied;the RANSAC algorithm was used to estimate the fundamental matrix to realize the epipolar constraint filtering and to find the target that matches the template;the RANSAC algorithm was also used to estimate homography matrix to realize the homography transformation between planar images to locate marker points;finally,the marking point of carton was positioned through the binocular vision measurement model.The relative positioning accuracy of the algorithm was 1.43%—2.94%.Thirdly,a binocular vision localization algorithm for tyre marking line was designed.Binocular images of the tire stack were preprocessed by threshold segmentation,removal of small area and two value refinement.The matching points of the labeled lines are extracted by the line scanning binocular image,and the stereo matching is realized by using the sequence constraints and the disparity range constraints;the RANSAC algorithm was used to filter the error and noise points in the reconstruct point cloud of the mark line,the high precision fitting of the mark line and tire positioning were realized.The relative error of tire marking line radius was 1.72%.Finally,the grasping experiment based on binocular vision positioning was designed.The carton grab experiment was realized by using the pressure sensing vacuum sucker and the binocular vision system;through several experiments on the visual positioning of tire,it was verified that the positioning accuracy of the vision system could meet the requirements of the system.