Research On Key Technologies In Dimensional Inspection System Of Thin Sheet Parts Based On Machine Vision

Machine vision inspection technique is a technique which can obtain image of the object to be inspected by means of machine vision, and then compares the image with given criterion in advance to ascertain quality of the object. Machine vision inspection technique is a non-contact inspection method which is integrated machine vision theory, image processing, precision measurement, pattern recognition, artificial intelligence, and other techniques. Machine vision inspection technique takes image of the object to be inspected as the means and the carrier of inspection and information transformation. Line scanning step adaptive optimization, subpixel edge detection, primitive recognition of planar contour and the development of dimensional inspection system of thin sheet parts based on machine vision are studied systematically in this dissertation.CAD information-based line scanning step adaptive optimization method is proposed with the help of the thought of scanning step adaptive optimization in the free-form surface reverse engineering. The proposed method resolves the problem that the conventional methods need to scan the object twice in order to adaptively optimize the scanning step, and can provide the dimensional values and their tolerances for the successive dimensional automatic inspection. A line scanning step adaptive optimization algorithm of arc primitive is developed by curve fitting allowance. A line scanning step adaptive optimization algorithm of straight line primitive is developed by equal line segment length. A region line scanning step adaptive selection criterion is concluded based on topological structure between primitives.The CAD information-based line scanning step adaptive optimization is realized in the dimensional inspection system of thin sheet parts based on machine vision. The inspection accuracy and scanning data amount of the method are analyzed by the dimensional inspection experiments of the thin sheet part of the HGA (Head Gimbal Assembly) of the hard disk and the precision standard parts respectively.An image calibration algorithm is discussed on the basis of the imaging property of line scanning. The maximum between-cluster variance method (Ostu method) is selected as the gray scale threshold segmentation method of the inspection system. The blob area threshold method is studied to eliminate noises of the image according to actual state of the line scanning image. The mathematical morphological edge detection method is selected as the pixel level edge detection method of inspection system. Aiming at the improvement of the orientation accuracy and inspection speed of subpixle edge detection, a rectangle lens subpixel edge detection method based on cubic spline interpolation is proposed after further studies of the existed subpixle edge detection methods. Taking the subpixel edge detection of line scanning image of the precision standard parts for instance, the inspection accuracy of proposed subpixel edge detection method is analyzed. The inspection accuracy and inspection speed of proposed subpixel edge detection method are compared and analyzed by contrast experiments between the proposed subpixel edge detection method and some other subpixle edge detection methods.According to the two indices of inspection accuracy and inspection speed, a planar contour primitive recognition method based on curvature and HOUGH transform is proposed after deep studies of the existed dominant point detection methods. A contour point classification algorithm based on neighborhood values is developed, and a curvature threshold method is selected to filter the contour points, and a projection height method is selected to distinguish the property of the primitive and classify the contour points, and the straight line primitive and arc primitive segmentation and merging algorithms are constructed respectively by HOUGH transform. The inspection accuracy and inspection speed of the proposed planar contour dominant point detection method are compared and analyzed by contrast experiments between the proposed method and some classical dominant point detection methods according to the edge information of image. The dominant point detection ability of the proposed planar contour dominant point detection method is tested by an emulational planar contour which includes all kinds of dominant points.A dimensional inspection system of thin sheet parts based on machine vision is developed by the research fruits mentioned above. The calculation methods of the dimension and dimensional error of the primitive are studied, and a least square straight line method is proposed to calculate the straightness error of the straight line primitive, and a least square circle method is proposed to calculate the roundness error of the circle primitive. The dimensions of the thin sheet part of the HGA of the hard disk and the precision standard parts are inspected respectively by the dimensional inspection system of thin sheet parts based on machine vision, and the inspection time is analyzed, and the inspection accuracy of the inspection system is analyzed by the inspection results of typical dimension of the thin sheet part and the diameter dimension of the precision standard parts. The experimental results indicate that the inspection accuracy of the dimensional inspection system of thin sheet parts based on machine vision can reach toμm level, the inspection time can satisfy the requirements of on-line real time inspection, and moreover, the inspection system can recognize the dimensions which need to be inspected automatically and realize the non-contact, automatic inspection of the dimensions by acquiring the CAD information of the part to be inspected. The system and the inspection algorithm are confirmed and agreed by the corporative enterprise.