Research On Edge Detection Based On Fuzzy Theory For Inner Defect Slice-Images

In the process of industry, the information of products and parts always obtain from a series of slice-images with CT scan. However, as to the complexity of the inner defect and the inherent limitation of X radial, the slice-images usually be fuzzy or contain illusive edges. In order to extract the exact edges of the inner defect, it is necessary to detect the accurate edges of the inner defect. This paper makes a series of research on edge detection for inner defect slice-images based on fuzzy theory.Firstly, the paper expounds the important position of edge detection in three-dimensional reconstruction, and analyses the advantage and disadvantage of traditional edge detection algorithms, then concludes the weakness that using traditional edge detection methods on the inner defect CT images, finally, proposes the edge detection method based on fuzzy theory for inner defect CT images.Secondly, an improved algorithm is proposed after analyzing the classical fuzzy edge detection method-Pal algorithm. Considering the deficiency about the threshold of Pal algorithm, the improved method separates the objective and background, then fuzzies them respectively, thereby establishing a new membership function. And also, smoothes the space image to reduce the noise, finally, an edge-connection step is added. The improved method avoids the defect of Pal algorithm that all of the pixels do not distinguished to process. Then the hole-defect CT image edge detection experiment carries on, results show that the improved algorithm is an efficient edge extraction algorithm for inner defect CT images.Finally, a new distance measure based on intuitionistic fuzzy set theory, called intuitionistic fuzzy divergence, has been proposed to be used in edge detection for the CT images with irregular inner defect. Intuitionistic fuzzy set takes into account the membership, non-membership and hesitation degree, which is more in line with the nature of fuzzy objects in the objective world. Also the result for edge detection is completely dependant on the selection of hesitation constant and thereby be the hesitation degree. It is opening up new ideas to fuzzy edges detection.