Crack Detection Algorithm Of Industrial CT Images Based On C-V Model

Image segmentation is a foundation of image measurement, recognition and analysis. It divide image into different parts, find the region of interest and to facilitate further processing. Based on industrial CT image, crack region is split from the image using crack detection method for the sequent crack measurement and analysis. The accuracy of image segmentation affects the effectiveness of the following task, so it is very important.C-V model which based on the theory of partial differential equations is a novel and effective method in image segmentation. It makes the evolving curve of zero level set function stop on the boundary of target regions by use of internal energy depended upon the curve of zero level set function and external energy depended on image data. Compared with traditional image segmentation methods, C-V model has some special advantages: It can accommodate itself to the changes of the object's Geometric topology and obtain continuous boundary. Facet model is an image segmentation method, which has an advantage of high precision and noise suppression. But it sometimes can not obtain continuous boundary.In this paper, the following three parts are mainly studied: (1) The C-V model and the facet model are studied and applied to the crack detection of two-dimensional CT images. (2) A new method which integrate C-V model with facet model is researched. The steps of this method are as follows: Firstly, the crack edge is detected by C-V model in the whole image; Secondly, a new crack edge is detected by facet model based on the result of C-V detection; Thirdly, the final crack edge is attained by combining the two eges. (3) Three-dimensional C-V model which can be used to detect three-dimensional crack directly are studied. As an extension of two-dimensional C-V model, Three-dimensional C-V model avoids dimensional reduction. The experiment results show that this method can detect satisfactory the three-dimensional cracks and lay a foundation for the sequent stress analysis.