Research On Key Technology Of Sorting System Based On Binocular Vision

With the improvement of industrial automation,the traditional manual operation method of sorting out different kinds of workpieces does not meet the increasingly accelerated production line work requirements,especially in the radiation work environment,it is urgent to realize automatic sorting.The combination of machine vision and industrial robots will improve the efficiency of sorting and the accuracy of identification,thus driving the industry's transition to automation.Therefore,how to integrate existing visual recognition technology with robot control technology is very important.The goal of this paper is to scatter the occluded workpieces,and to study a size and pose detection technology based on binocular parallel vision.Through the mechanical arm combined with the results of machine vision recognition,the sorting operation of different types of workpieces is completed.Technical support is provided for picking.The main research contents are as follows:According to the task requirements of sorting,the imaging principle of intersecting and parallel camera installation is analyzed.The two parallel camera installation schemes are established in this paper,and the specific installation parameters are given.According to the installation characteristics of the camera,binoculars are designed.The calibration method of the visual and hand-eye system,and the calibration of the parameters in the experiment,determined the overall research plan for image detection of the target workpiece.The image enhancement method in preprocessing is studied.A hybrid enhancement algorithm is proposed for the characteristics of image acquisition.The necessity of image polar line correction is analyzed.The influence of each parameter change on the matching result in stereo matching is analyzed,and the selection method of specific matching parameters for different situations is determined.The principle of 3D reconstruction and the method of obtaining 3D pose are analyzed.The image segmentation method of target workpiece is proposed for the working conditions of this paper.The method of feature extraction,contour fitting,size and pose detection for target workpiece is given.The recognition of the pose and size of the target workpiece to be grasped in the stacked workpiece is realized.Based on the camera SDK and MFC framework,the modules of image acquisition and image processing of the binocular vision sorting system softwareplatform were developed.Based on the V-REP platform,the system simulation of the sorting process was completed.The experimental system platform was built and the detection accuracy of size and pose was analyzed.The results show that the whole sorting experiment meets the requirements in practical work.