| With the rapid development of industrial automation control technology and computer technology,industrial automation has penetrated into the production of various industries.Automatic dispensers are widely used in the production of integrated circuits,printed circuit boards,electronic components,automotive parts,handbags,and packaging boxes.Machine vision is one of the important technologies of today's dispensing machines.Its appearance has revolutionized the dispensing industry.With the rapid development of industry,the accuracy of dispensing has been continuously improved.The calibration and image processing of dispensing systems based on machine vision are playing an increasingly important role in the dispensing production line.In the actual dispensing production line,because the position of each workpiece is not fixed,the position error correction must be performed by visual positioning technology to obtain the actual precise dispensing position,so the accuracy of visual positioning plays a vital role.Based on the monocular vision technology and taking PCB board as the target,the real-time position and system calibration of the dispensing point have been studied in depth.The main work is as follows:(1)Aiming at the problem of PCB board positioning,an improved feature point matching algorithm of BRISK algorithm is given.This algorithm solves the problem of uneven distribution of feature points caused by the fixed detection threshold of BRISK operator.By setting the feature point re-projection error threshold,the wrong matching point pairs are eliminated to solve the problem of BRISK algorithm with many mismatches.Based on the improved BRISK algorithm,the feature point matching relationship is obtained.The projection matrix between the real-time scene image and the standard image is calculated using the RANSAC algorithm to complete the Mark point matching of the PCB board.(2)In accordance with the requirements of system detection accuracy and speed,an improved edge detection algorithm of Canny operator is given.Guided filtering is used instead of Gaussian filtering.The gradient amplitude calculation in two directions is added to solve the problem of edge blurring and improve the edge.positioning accuracy.The edge contour is obtained based on the improved Canny operator's edge detection algorithm,and then a straight line detection algorithm that combines Hough detection and least squares method is used to fit the contour of the PCB board,and then the Mark points for positioning are obtained.(3)A dispensing vision system model was established for the dispensing experiment platform,and the camera imaging model and the lens distortion model were analyzed.The mapping relationship between the plane of the dispensing platform and the imaging plane is obtained by solving the transformation relationship between the four coordinate systems,and the parameters of the vision system model for dispensing are solved using Zhang Zhengyou's calibration method.According to the actual situation,a system calibration method is used to obtain the scale factor between the pixel and the real object.Aiming at the location of the glue position,the sampled image is grayed out,image denoising filtering and binary segmentation of the image are laid,which lays a good foundation for accurate positioning.In this paper,a dispensing vision system model is established,and the improved BRISK feature matching algorithm is used for PCB recognition.The edge detection algorithm using the improved Canny operator exists for the traditional algorithm.The problems of fuzzy edges and susceptible to noise are improved,and the effectiveness and robustness of the algorithm are verified through experiments.Compared with the traditional algorithm,the matching accuracy is improved,and the average positioning error is reduced to less than 0.1118 mm.Based on the research of the PCB vision machine positioning system,a positioning software system is developed based on the MATLAB platform.Easy to operate. |
PDF Full Text Request