Research On Automated Location And Peeling Display For Industrial CT Image Defects Based On Fractal Dimension

As one of the most important processing steps in digital image analysis, auto-location and peeling display of defects are of great significance to improve the detection efficiency, to reduce the labor intensity and to ensure the safety of products. As a kind of advanced non-destruction detection technologies, namely detecting the inner structure of products without destruction, Industrial Computed Tomography (ICT) has been widely used in many fields, such as aviation, oil, steel, manufacturing, rail transport and so on. The purpose of auto-location of defects is to detect the defects from CT images automatically and to obtain the information about those defects as precise as possible. Based on the present research development of defects recognition for ICT images, this paper introduces fractal dimension into auto-location of defects and then peels off the defects from those located regions by the improved moment invariant threshold segmentation algorithm.Fractal was proposed in 1975 by the French mathematician, B.B.Manderbrot, on the basis of studying the self-similarities of vast irregular structures. As an important feature of fractal, fractal dimension breaks through the restriction that the dimension in Euclid space must be integral and has been applied in many scientific fields. The essence of the fractal dimension is to measure the extent of irregularity, and in terms of image processing, it is extraordinarily sensitive to gray change on the image gray surface. Therefore, fractal dimension is appropriately used to detect and locate defects of images. This paper studies the applications of fractal dimension in automatic location and peeling display of defects in ICT images. The main work was stated as follows.Firstly, the defects of ICT images are located automatically based on the improved method with fractal dimension. To avoid the influence on detection of noise in the images, a pre-processing algorithm, which filters out the noises and enhances the images simultaneously, is studied. The applied self-adaptive threshold strategy can efficiently improve the automatic level on condition of ensuring the precision of location. The proposed algorithm can self-adaptively locate defects along their true edges and has strong robustness against the random noises. Because the ratio of the located region is small, if any operation in the located regions is needed, the suggested algorithm can reduce the computations and increase the detection efficiency. Then, on the basis of the regions automatically located by fractal dimension, a method on condition of keeping the intensity moment invariable is applied into multi-goal peeling display of ICT image defects, and some feature parameters about defects are also calculated. The moment invariant threshold segmentation algorithm combines the local grads information and conducts the segmentation independently within each located region to avoid the effects of thresholds on the results when segmented as a whole. Compared with edge detection by direct Facet model in the whole image, the method used Facet model in those located areas is faster. Furthermore, calculations about the feature parameters of defects can provide accurate location and size of defects for repair of work piece and so on. Experiments also verify that the proposed algorithm can peel off the defects of images effectively.