Research On Key Techniques Of Visual Inspection System For The O Ring Geometry Dimension

Machine visual detection technology has the advantages of rapidity, high precisionand stability. With the increasing renewal of machine visual technology, visualmeasurement is booming in modern industrial production. As the commonly usedsealing element, O ring obtains wider and wider application, but its measurementworkload is quite heavy. Limited by the staff’s mental state and work experience, themeasuring accuracy of traditional manual measurement method is often not so high.Besides, the working efficiency is low in the large-scale testing. To solve theseproblems, this paper presents an accurate and speedy measuring method for the O ringdimension based on machine vision.Based on two-dimensional visual measurement and the research on active visualtechnology, this paper constructed the hardware platform of the O ring dimensionmeasuring system, including motion control parts such as three-dimensional servomoving mechanism, sensors, motion control card and the visual positioning parts suchas lighting system, image acquisition system, imaging system.On the basis of some common auto-focus methods, sharpness function value wasused to evaluate the clarity of the imaging system. By comparing the differentdefinitions of the O ring image, Brenner function was proved more satisfactory, and canbasically achieve the correct focus. In the process of peak searching, the curve fittingwith high efficiency and the Mountain Climbing Servo with simplicity are used incoarse focusing and fine focusing, with5mm initial step size and0.5mm small step sizerespectively. During the coarse focusing, the second-order polynomial based on theleast square method was adopted to predict theoretical peak position, and then searchthe best focus position in the range of theoretical peak position±2.5mm based onMountain Climbing Servo.Then, this paper studies the image pre-processing method and the image trackingmethod of the O ring. During the course of image pre-processing,3×3template wasselected in Gaussian filter. The results showed that this kind of filter could eliminate theimage noise better, without debasing the quality of the picture. After that, the O ringimage was transformed into binary image with the segmentation threshold of200. In theend, the control system obtains the complete O ring image information by tracking theedge. Finally, by selecting Canny algorithm as the ultimate edge detection operator andcalibrating the camera, this visual measuring system could compute the real size of theO ring with the method combining image segmentation based on morphologyprocessing and circle fitting based on Least Square..