Research On X-ray Image Crack Detection

With the development of industrial technology,the X-ray detection technology has become one of the important methods of NDT technology because of its wide adaptability and low costing.However,the traditional X-ray detection image has low contrast and high noise,which makes the low defect detection accuracy.Improving the contrast and segmentation efficiency of X-ray images has become a current research hotspot in image detection.In this paper,the crack defect detection in X-ray image is studied based on three aspects: weld area extraction,weld image enhancement and crack defect segmentation,and defect classification:(1)X-ray image is composed of background area and weld area.The defect of the sample mainly exists in the weld area.Extracting the weld area can solve the problem of long detection time due to data redundancy.First,the image binarization method were used to segment the image automatically.Then,determined the weld area and reduced the detection range according to the geometric and generation features of the image.(2)Aiming at the problems that it is difficult to solve the low contrast of weld image and difficulty in segmentation,two weld image enhancement methods were proposed.The first one is local equalization and local variance processing,which solves the problems of image dim and edge blur.It not only can enhance the contrast of image effectively,but also can improve the visual effect of the image,makes the output image have better visibility,and convenient to directly use the OTSU algorithm for image segmentation.The second one is the OTSU threshold optimization algorithm based on gray level transformation of the image.By introducing gray transformation function,the relationship between gray transformation(image enhancement)and maximum between-class variance was established.The gray level transformation function which makes the variance between classes reach the maximum is found to solve the problem of image segmentation.In order to accurately segment crack defects,morphological corrosion method is utilized to fill in narrow intervals and eliminate irrelevant details;(3)According to the features of defects and relevant evaluation standards,the classification of defects is realized.First,the defects were marked by the connected area marking method.Then,the geometric and gray level features of crack defects and other defects were extracted and calculated.Finally,based on the extracted features,defects are classified and finally realized the automatic detection of crack defects in X-ray images.In order to verified the feasibility of the method,15 defects were detected respectively.The results show that the method can classify 15 defects effectively.