Research On Edge Detection Technology Of Inner Defect Based On Cross-Image

Edge detection based on CT image has become a hot research at present at home and abroad.The main defects of structure are crack and holes.The defect of crack takes on a character of tree -geometry,and the material density of holes is different form surrounding material,so gray value of ICT is also different.Th- erefore,geometry structure is the character parameters of crack detect,and gray value is the character parameters of hole detect.For different types of detect,we also should choose different edge detection method.The previous edge detecti- on algorithm was not detecting to the type of detect.This paper takes this as a background to carry on edge detection of inner crack and hole defect in cutive- slices systematically.Firstly,aiming at the character of the defect of crack which is often very thin and long,and takes on a tree -geometry,a new edge detection method is proposed on the basis of gray-scale morphology of top-hat transform and curvature evaluation,that is,first using top-hat transform based on gray-scale morphology to extract the edge of the pre-crack,then using curvature evaluation algorithm to eliminate the edge which is similar to the geometry of crack,in this to obtain the crack edge accurately.Secondly,aiming at the character of the area of holes where material density is lower than the surrounding material, so the black area is showed in CT image,a edge detection method is proposed on the basis of entropy on Interval-valued fuzzy sets and curvature evaluation, that is, first using entropy on Interval-valued fuzzy sets of all the pixes in image to extract the edge of holes,then binating the original image and to have a regional labeling in the binary image;at last ,in order to obtain the last edge we need to have a operation between the the processed image and the image which after the edge extraction.Finally,the method for two detection proposed above is analysed,and the accuracy of the new algorithm was verified.