Research On Workpiece Size Measurement Based On Matlab Binocular Vision

In this paper,the binocular vision measurement and the accuracy of its measurement methods and algorithms are improved as research target.Then,design a binocular vision measurement platform on Matlab by using mathematical model of binocular stereo vision and related image processing method.And improve the methods and algorithms for important steps.Its content mainly includes calibration of binocular camera,image preprocessing,feature extraction,stereo matching.The specific research content is as follows:(1)The characteristics between traditional calibration method,self-calibration method,active visual calibration method,Zhang Zhengyou calibration method are introduced.Then,calculate the calibration parameters with higher precision,by using positioning advantage of the center point and the positioning method of this calibration plate.Provides a computing foundation for the entire binocular vision measurement process.(2)Before feature extraction,Analyze the problem in the original image of the camera where the feature needs to be extracted,and Choose the appropriate pretreatment method.For corner features,analyze the problem of traditional Harris corner detection methods in the current environment.A method for obtaining the coordinates of corner points by using the straight line detected by Hough transform is proposed.For circular features,analysis of the causes of the presence of gray areas around circular features in the current environment,then,using the actual situation of the circular feature itself and its gray histogram,the appropriate threshold is selected to process the original image,and measuring the processed circular feature information.(3)For the corner characteristics detected in this paper,the analysis shows that the commonly used stereo matching method does not apply to the stereo matching of the corner points detected in this paper.Then combined with the mathematical principle of normalized cross-correlation method,a "point-window-point" matching method is proposed.The experimental results show that this method has a more accurate matching rate for the corner points detected in this paper,and this method is also suitable for matching the circular features.(4)Combining the detection and matching methods proposed in this paper,the design of binocular vision measurement software is completed on Matlab.and the threedimensional space coordinates of the feature points in the world coordinate system can be calculated by the software.Finally,according to the actual value of the workpiece to be tested,it is proved that the result obtained by the software operation are more accurate and efficient than the traditional methods.