Research On Composite Analysis And System Development Of Parts Manufacturing Based On Industrial CT Images

With the increasing complexity of the current workpiece structure,it is very important to precisely evaluate the errors in the manufacturing process.Due to the influence of measurement tools and the operation of surveyors,the traditional measurement methods are often inefficient and easy to cause irreversible damage to the workpiece.The industrial CT technology can obtain the internal structure and exact size of the workpiece under the premise of not damaging the structure of the measuring workpiece.Moreover,it has the advantages of high efficiency and high resolution.In view of the above problems,this paper studies the component analysis of parts manufacturing based on industrial CT images,and realizes the two-dimensional comparison software.The main research contents are as follows:(1)In order to obtain the test model for image registration,the test model needs to be pretreated.In this paper,an image segmentation method based on the combination of the distribution estimation algorithm(EDA)and the two-dimensional Otsu method(two-dimensional OTSU)is studied.The method searches the spatial sampling and statistical learning to predict the best region of the search,so as to obtain the optimal threshold,so that the foreground color and the background color of the test model can be separated as much as possible.Through the verification,the algorithm is fast and stable,and the image segmentation is accurate.(2)For industrial CT image and two-dimensional CAD image registration,it is necessary to extract the edge of the image.In this paper,the method of spatial moment subpixel edge detection based on error compensation is studied.First of all,by combining the estimation of distribution algorithm and two-dimensional Otsu method for image segmentation of industrial CT image with noise;and then use the Canny algorithm for edge detection of pixel level image after segmentation,the model of industrial CT image pixel points;finally,the detection method of spatial moment subpixel edge error compensation based on the point set model for sub-pixel edge detection.(3)The initial registration technique of point set model and two-dimensional CAD model based on minimum enclosing rectangle and primitive feature is studied.First,determine the minimum bounding rectangle point set model by sorting point to get the edge point set and the set of convex polygons;then,the minimum bounding rectangle to determine the centroid point set model to get the translation parameters at the beginning of registration;finally,through the line,arc and circle of primitive features initial registration and rotation matrix inversion matrix to achieve initial registration.Fourth,this paper studied the technology of fine registration of model which combined singular value decomposition and iterative closest point(SVD-ICP).Then,this paper realized precise registration of the two-model by SVD-ICP based on the rough registration handling.(4)Study and improve the fine registration technique.The traditional ICP algorithm is improved by searching the nearest point,removing the mismatching points and constructing the error metric function.The experimental results show that the ICP algorithm can effectively improve the computing speed and the registration accuracy.The contents of this paper and the research results have been achieved in the 2D alignment system,completed the initial registration and fine registration function of industrial CT image and 2D CAD images,and the development of the error model between calculation and color cloud picture display.