Research On Key Technologies Of Recognition And Grabbing Based On Binocular Vision

With the development of intelligent manufacturing,machine vision technology has become a research hotspot,especially binocular stereo vision technology.At present,most of China's automated production,manufacturing,and carring processes are completed by manual control or robot teaching mode,which has poor flexibility and wastes human resources.Therefore,combining vision technology with robots to automatically complete the grabbing work in the manufacturing process can effectively increase the degree of automation of industrial production and expand the application field of robots.This topic mainly carried out systematic theoretical research on the key technologies of recognition and capture based on binocular vision.The main research contents are as follows:The imaging principle of the camera is analyzed;the camera calibration method is studied,the two different calibration methods are compared under Matlab environment and Halcon environment,and the Halcon calibration method with higher calibration accuracy is selected for binocular vision calibration research;The hand-eye calibration model is established and analyzed,and a fixed-point hand-eye calibration method based on Halcon is proposed,which accurately obtains the interrelation between camera coordinates,robot's base coordinates,and end grip coordinates with full consideration of camera distortion,then verifies the feasibility of the calibration method proposed in this paper.Image gray processing and smoothing methods are researched,different smoothing filtering methods are analyzed;different feature detection methods are studied,which includes corner features,edge features and texture features.The detection of feature points by Moravec operator,Harris operator and SUSAN operator is mainly analyzed,and an improved Harris corner detection operator is proposed,which can effectively improve the accuracy of corners by automatically adjusting the response threshold.The image stereo matching algorithm and commonly used constraints are studied.In recent years,the SIFT algorithm,which is a research hotspot of image matching,is analyzed.An improved SIFT stereo matching algorithm combining Harris operator and epipolar constraint is proposed.The feature points extracted by Harris operator are used to replace the detection of extreme points in SIFT algorithm.The feature points have been reduced by more than 86%,the improved algorithm has shortened the time by more than half,and the matching accuracy has been increased by an average of 7.8%,effectively improving the real-time and accuracy of image matching.The principle of template matching is used to identify the target object and verify the recognition effect under different conditions.The principle of three-dimensional reconstruction is analyzed,and three-dimensional reconstruction is performed according to the coordinates of the position of the target object;an experimental platform is set up to study the recognition and grab system of target object,analyze the composition and selection of each module of the grabbing system,establish the communication between various modules,and design grabbing software for experiments.The experimental results show that the method and system involved in this subject can meet the actual needs of production,and the average grabbing success rate can reach more than 95%.