Research On Machine Vision Technology For Measuring Shaft Diameters

Machine vision technology uses computer to simulate visual function, and determines dimension information of objects to be measured through acquiring and processing their images. Using machine vision technology to measure element dimensions in mechanical industry will have outstanding advantages of non-contact, high-speed and high-degree of automation, so this technology has obtained more and more attention.Aiming at limitations of visual measurement methods at home and abroad, this article considered geometrical characteristics and optical imaging process of shaft elements, and proposed a measurement method of shaft diameters based on machine vision.Firstly, this article researched and analyzed camera calibration technique, to eliminate influence of lens distortions on measurement accuracy for shaft diameters. On assumption of pinhole camera model, linear and nonlinear imaging models of camera were established; To reduce cost of calibration and make actual operation convenient, a two-step calibration method based on plane was studied in detail; For improving precision of camera calibration and facilitating dimension measurement, several improvements were made for this method, including:(1) Modifying projection direction of distortion model, which could avoid solving high-time equation by numerical methods during solving world coordinates of feature points in reverse; (2) Adding two tangential distortions in distortion model besides two radial distortions, which could correct lens distortions better, meanwhile not affect optimization stability of parameters; (3) Solving initial values of camera interior parameters by using feature points in central area of images, which could improve precision of initial values of camera interior parameters during optimization; Finally, experiments of camera calibration showed that, using more images during calibration could make results of parameters calibrated more stable, moreover, considering tangential distortions, choosing appropriate constraint for interior parameters, and using feature points in central area of images could improve calibration precision effectively. To improve resolution of image measurement, it is necessary to acquire subpixel location of edges from images. Several common subpixel edge detection methods were researched, including gray moment method, spatial moment method, gradient interpolation method, relevant interpolation method and gaussian fitting method; Considering deficiencies of traditional evaluation methods (method using edges simulated, repeated experiments using practical edge), a evaluation method was put forward, combining camera calibration and edge detection, and taking actual measurement error as criterion of accuracy evaluation, which conforms to requirements of actual measurement. Experiment of evaluating edge detection accuracy showed that, gradient interpolation method had the highest detecting precision and best stability.When calibrating exterior parameters of objects to be measured using the two-step calibration method based on plane, dimentions to be measured should be lied on the calibration plane, while it is difficult to realize for mechanical elements. Therefore, based on geometric relationships of shaft imaging, a world coordinate system and a mathematical model for measuring shaft diameters were established. Exterior parameters which showed shaft gestures in this model could be calibrated by shafts with known diameters.Finally, accuracy of the measurement method of shaft diameters proposed in this article was tested by measurement experiments. Firstly, accuracy of the method was tested in static condition, and results showed that, position of camera installation and number of shafts with known diameters used to calibrate exterior parameters had no obvious influence on measurement precision; Secondly, in dynamic conditions, shaft diameters after several cuttings were measured, and experimental data indicated that, using more images was helpful for improving accuracy and stability of measurement, and rotating speed of shaft had no effect on measurement results.The work and research in this article has certain significance for applying machine vision measurement technology in engineering.