Research On The Visual Measurement Technology Of Workpiece Size Based On Sub-Pixel Positioning

The measurement on workpiece size is needed indispensably in machinery manufacturing,because whether the measurement result is accurate or not makes directly influence on the quality of the final workpiece.The traditional workpiece size measurement is mainly completed by using a universal measuring tool or operating a special high-precision measuring instrument.But there exist issues of high cost,low efficiency and high labor intensity in this manual measurement method,which is inconsistent with the development requirements of mass production in modern industry.In recent years,the rapid development of machine vision technology and precision manufacturing technology gives the promotion to the visual measurement technology in the development,whose emerging measurement is characteristics of fast speed,high accuracy,and non-contact.The edge detection is one of the key links in the visual measurement technology,and the detection accuracy of its algorithm directly determines the accuracy of the measurement.However,the accuracy in traditional edge detection algorithms has only one pixel quantitatively,i.e.,the edge can only be judged to be within a certain pixel,but not to a more precise position in that pixel.It is confirmed that the detection accuracy is less than one pixel in the sub-pixel edge positioning algorithm.This paper,in order to further improve the measurement accuracy,takes gears as the research object and makes deeply the research on the sub-pixel edge positioning and the related visual measurement technology.Firstly,it is concerned with the design of the vision measurement system.The performance indicators of light sources,industrial cameras,optical lenses and other hardware are analyzed in detail in the research,the appropriate hardware equipment is selected and the visual measurement platform is built under laboratory conditions.Due to the fact that there is a certain degree of non-linear distortion of the camera objective lens,the Open CV calibration tool is used to complete the camera position calibration and the original image of the gear is corrected for the reduce of measurement error according to the distortion calibration coefficient.Secondly,research is related to the image preprocessing and edge detection algorithms,including the relevant theories about the image preprocessing(the image filtering and threshold segmentation included),the traditional whole pixel-level edge detection algorithms,and the comparison of the effects of processing gear images in the simulation experiments.Thirdly,the research is connected with the work how to improve Zernike moment sub-pixel edge positioning algorithm.Three existed types of sub-pixel edge positioning algorithms are studied for the improvement on the measurement accuracy.A new edge judgment is put forward and the optimal threshold is calculated by the maximum between-class variance for the resolution of the existed issues of detecting thicker lines of edges and time-consuming and laborious manual adjustment of thresholds.And on this basis,the detection accuracy of Canny operator,spatial moment,gray moment,Ghosal algorithm are compared and analyzed,and the detection accuracy of Zernike moment is thus improved.It is shown in the experimental simulation that when the improved Zernike moment sub-pixel edge positioning algorithm is detected by the ideal square edge,the measurement result deviates from the true value by about 0.05 pixels,which is significantly better than other detection algorithms for the verification of algorithm effectiveness and superiority.And fourthly,the research in on the measurement experiment and analysis.The least squares circle fitting algorithm is used to fit the top circle and the root circle.The other parameters of the gear are deduced from the gear parameter calculation formula.According to the pixel equivalent obtained by the camera size calibration,the pixel value is converted to the actual size and the experimental results are analyzed for the realization of the gear size measurement.It is indicated in the experimental simulation that in the comparison with Canny operator detection and manual measurement,the improved Zernike moment measurement is closer to the true value with higher detection accuracy.In addition,the visual measurement fluctuates smoothly,which proves that the measurement system designed in this paper is stable in the performance with the undeniable significance in engineering application.