Research About The Detection Method Of Arc’s Dimension

With the development of science and technology, vision detection technology has been widely used in modern industrial production. It is efficient and labor saving to use vision detection technology on machine parts measurement. Moreover, compared with traditional detection technology, it is more precise and has the advantage of realtime and non-contact. Thus, there are obviously scientific significance and practical value to do a deep research on the vision detection technology on the geometry and dimension of workpiece.The thesis focuses on the detection of curve edges, doing a research on the measurement method based on machine vision. First of all, a research on several filtering method has been carried on through simulation experiments to compare the noise removing effects. Combing the real environment of workpiece production, we choose the median filter method as the filtering method to reduce the impaction resulting from the image transmission decoding process.The study on the completion and the clarity of the detected image, the advantages and disadvantages of several pixel detection operators is carried afterwards. And based on these, the sub-pixel detection methods is discussed, including gray moment method, interpolation method, fitting method and bidimension gray moment method basing on 9×9 template. Combining the detecting object in the experiment and the contour edge gray distribution of the object, the bidimension gray moment method is chosen for the best sub-pixel edge detection method in the chapter.According to the character of edge curvature variation, a method of contour recognition and segmentation is developed and compared to the traditional manual detection method. The detecting precision and recognition time of both is discussed.At the final, a typical workpiece is chosen to test the efficiency of the methods used in the thesis. The test uses a developed camera calibration algorithm to correct the distortion caused by the camera lens. And the calibration parameters are used to convert the sub-pixel coordinates to the corresponding feature points in world coordinates. After fitting the world coordinates, the curve diameters are detected, and the factors effecting the detecting accuracy are analyzed.