Research And Implement Of Manufacturing Error Analysis Method Based On Indurstrial CT Test

As the requirements of manufacturing precision of mechanical product is higher and higher, objective and accurate evaluation of manufacturing error of components is the urgent needs. As a kind of nondestructive testing technology, the industrial CT technology can obtain the slice images of components under the premise of not destroying the structure of existing parts which provides new ideas and methods for the quantization of manufacture errors for components. In order to improve the analysis precision of components manufacturing errors, this paper studies the analysis method of components manufacturing errors based on Industrial CT test, and this paper also developed the Error Analysis-ICT manufacturing errors analysis software. The main content of this paper includes industrial CT image segmentation, pixel and sub-pixel edge detection, the initial registration based on the minimum bounding rectangle and primitive feature, the accurate registration based on SVD-ICP, the calculation and display of errors.First,we propose an image segmentation method combined with distribution-estimation algorithm and two-dimensional Otsu method(2D OTSU). This method takes the two-dimensional Otsu formula of image for adaptation degree function, searches the optimal threshold through the evolutionary approach of building and sampling the probabilistic graphical model. The convergence speed is fast and stable and the image segmentation is accurate.Second,this paper studied the method of pixel and sub-pixel edge detection. For pixel level edge detection method, mainly introduces advantages, disadvantages and the applicable scope of some typical edge detection algorithm, then this paper showed the detection results of threshold segmentation of industrial CT images with algorithm and verified the Canny algorithm’s feasibility as a practical pixel level edge detection algorithm; As for the sub-pixel edge detection methods, studied of the spatial moment subpixel edge detection method based on error compensation. This paper introduced the principle of the algorithm and took the nut of the CT images as experimental materials and verified the effectiveness of the sub-pixel edge detection method.Third, this paper studied the rough registration of point cloud model and 2D CAD model based on the minimum bounding rectangle and primitive feature. This paper also described the minimum bounding rectangle with two-dimensional fixed-point polygon ring and determined the translation parameters according to the centroid coordinate between the minimum bounding rectangles of the two models.Then,this paper determined the rotation parameters of the two models through selecting primitive characteristics of lines, circles or arcs and realized the rough registration of them.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.Fifth, the development and implementation of the system. On the basis of RTV which is a existing CT image processing software,this paper developed and realized the functions of aforementioned image segmentation, pixel level edge detection, sub-pixel edge detection, two-model coarse registration and two-model accurate registration. Finally,this paper realized the error calculation between models after registration and color cloud picture display, after that,this paper output interesting error information. The test results showed that the system had reached the expected requirements in accuracy, availability, ease of use and other aspects and verified the feasibility of scheme and correctness of method of this paper.