Research On Moving Workpiece Recognition And Robot Grasping System Based On Binocular Vision

Automatic grabbing of workpieces is very important of the modern industrial assembly line.The binocular vision technology can recover the spatial information of the target object by using the principle of parallax.And combining the robot with it can solve the disadvantages of the traditional production line for the robot to grasp through the offline teaching mode,which is of great significance for improving production efficiency and automation.Based on the requirement of robots to accurately and efficiently grasp the workpieces in the field of intelligent manufacturing,this paper proposes a moving workpiece recognition and robotic grabbing system based on binocular vision,which improves the automation of the crawling process.In this paper,the key technologies involved in the system were introduced in detail,the required software and hardware were selected,three different binocular vision models were analyzed,and a complete system workflow architecture was designed.The image is enhanced and smoothed by histogram equalization and median filtering through comparing the characteristics of different image preprocessing algorithm.The feature extraction and matching algorithms are studied in deeply,through the analysis of the characteristics of traditional image feature extraction algorithms,based on the advantages of accelerated segmentation test feature(FAST)and accelerated robust feature(SURF),as well as the dimensionality reduction characteristics of principal component analysis algorithm(PCA),an improved SURF algorithm is proposed to conduct feature extraction and use the distance measure as a similarity measure technique to perform coarse matching of the workpiece image.The experimental results show that the running time of the algorithm is 3 times higher than that of the traditional SURF algorithm,and the matching correct rate is 13%higher than the traditional SIFT algorithm.The target workpiece recognition and spatial localization methods are studied,and the effectiveness of the improved SURF algorithm for image recognition is verified by comparing the recognition effect of traditional image recognition algorithm and improved SURF algorithm in different environments such as illumination,zoom and rotation.The random sampling consensus algorithm(RANSAC)is used to eliminate the wrong matching point pairs to achieve accurate recognition of the target workpiece image.The plane position information of the target workpiece image is determined by calculating the centroid of the workpiece using the target edge method.And the Zhang Zhengyou calibration experiment is carried out,which results were conbined to complete the three-dimensional reconstruction of the centroid position at a certain moment.Through the calculation of the horizontal distance of the belt moved in the period from the reset state to the realization of robot arm,the three-dimensional space coordinates of the gripping point were determined.Then,the kinematics of the robot arm was analyzed,the hand-eye calibration experiment was carried out,the software system integrating binocular vision recognition and robot capture was developed,and the vision subsystem and the capture subsystem was combined on the experimental platform to realize the specific work applications of the binocular vision technology.The results show that the moving workpiece recognition and robotic grabbing system based on binocular vision can meet the requirements of accurate recognition of target workpiece images and real-time robot capture requirements with good results.