Research Ou Multi Type Workpiece Recoguitiou And Positioning Technology Based On Visual Guidance

In the industrial production line,the loading,unloading and gripping of workpieces are mainly carried out by manual control or by using pre-displayed industrial robots,which have obvious defects in production cost and poor flexibility.Combining machine vision technology with industrial robot technology to classify and grasp workpieces can improve the automation and intellectualization of grasping effectively.This paper analyzes and optimizes the key technologies such as image preprocessing,recognition and location,and robot grasping by setting up the identification and positioning system of binocular vision robot,which provides a theoretical basis for industrial visualization and localization,as well as for the research of robot intelligent gripping.Firstly,a mathematical model of camera imaging is established,the distortion model is analyzed,binocular calibration is completed,and the structural parameters between the two cameras are determined.Based on the principle and process of Zhang’ s calibration method,the camera calibration is accomplished by using Halcon internal function library programming and Matlab calibration toolbox,and the distortion coefficient of camera lens is obtained.The calibration parameters are optimized by eliminating the left and right lens distortion based on the Halcon camera.The error and precision of two calibration methods are calculated and compared.Finally,the principle of hand eye calibration is studied,and the parameters of hand eye matrix are obtained through experiments.Then,after analyzing the basic principle of image preprocessing method,combined with image filtering and image enhancement,the noise in the workpiece image is eliminated,and the influence on image brightness is solved.After analyzing and optimizing the template matching recognition method based on shape feature,Zernike moment method is used to extract sub-pixel contour lines of workpieces for pattern matching,which improves the precision of the workpiece recognition.By setting the termination search threshold of similarity measure,the speed of template matching is improved,and false matching is prevented.The Gaussian pyramid principle of image is studied.The optimal number of layers is determined by experiments,and the speed of workpiece identification is improved.The recognition precision and speed of the optimal template matching recognition algorithm are verified by identifying many kinds of workpieces in different scenes.Secondly,the paper studies the spatial location technology of the target workpiece.Based on Bouguet algorithm to complete binocular stereo correction,the commonly used stereo matching methods are described.Before stereo matching,histogram matching is used to optimize the workpiece image to improve the accuracy of stereo matching algorithm.Then the integral image theory is introduced into stereo matching to reduce the complexity and amount of computation of matching cost function,so as to complete the stereo matching of workpiece image efficiently and accurately.Then the 3-D reconstruction matching of work-piece image is accomplished by region matching,parallax is extracted and its value is calculated.Combining with the principle of 3D reconstruction and camera calibration parameters,the target workpiece is positioned in space.Finally,the hardware platform of the system is constructed by selecting the key parts,and the software platform developed by Halcon and VC + + is combined to complete the whole software of binocular vision robot localization system.This paper analyzes the strategy of robot grasping multi type stacked workpiece,designs and carries out the robot grasping experiment.The experimental results show that the robot vision grasping system can successfully complete the task of multi type workpiece recognition and positioning and robot grasping,which verifies the effectiveness of the recognition and positioning algorithm and the feasibility of the grasping system.