Research On Binocular Vision Accurate Positioning Technology For Material Sorting

Stereo vision guidance technology has been widel y used in the material sorting industry,using stereo vision technology to reconstruct the three-dimensional geometric in formation of objects,to realize the recognition and positioning of objects in three-dimensional space.Binocular stereo vision technology can realize the accurate identification and accurate positioning of targets in a single scene,but in complex environm ents,the positioning accuracy is reduced due to material stacking and shielding.In this paper,the precise positioning technology of binocular vision for material sorting is studied.The specific research contents are as follows:(1)Research on adaptive threshold matching algorithm based on SIFT.According to sparse matching point accuracy of stereo matching problem,to match sift algorithm to perform anal ysis of the threshold and the matching point,obtain the graph of the matching point sift algorithm threshold and,according to the matching rate using an adaptive threshold for optimization,get a more reasonable matching effect,but onl y b y the threshold filter can not effectivel y correct matching points,so the introduction of RANSAC iterative algorith m of false matching points out further improved,and verified the boxes to get higher matching accuracy under various scenarios.(2)Research on dense surface reconstruction algorithm based on SFM.SFM algorithm based on sift algorithm,in view of the boxes SFM point cloud sparse clutter problem,on the basis of SFM,CMVS/PMVS algorithm combining with the selected point set,will feature points and feature points around the form,change the point cloud constraints between the rest will be another integration into dense point cloud,make boxes of dense point cloud outline with details is more clear,poisson reconstruction algorithm is introduced to repair the hole between the dense point cloud,get more abundant details more stereo box box model,implement a variet y of scenarios boxes of 3 d reconstruction,according to the standard size and actual size compa red to verify the accuracy of 3 d reconstruction,boxes.(3)Construction and experimental verification of binocular vision robot sorting platform.This pap er guide the robot to complete the scraping of the five boxes,with actual grab as standard coordinates,the coordinates of then to rebuild after coordinate transformation theory of the five boxes,boxes theoretical coordinates and attitude,information an d the actual contrast standard stance information,anal ysis of error sources,according to the robot grab indicators,ensure accuracy of three axis is less than 5 mm,attitude accuracy of less than 5 °,the experimental results show that the overall system results meet the requirements of fetching with boxes and verify the accuracy of the whole of the binocular vision guidance s ystem.