| The internal of industrial castings often have some defects due to the influence ofthe subjective and objective factors in its production process. Defect detection isnecessary before the use of these industrial castings because serious defects will resultin safety hazards. DR (Digital Radiograph) system has a broad application prospectswith its good image quality and high detection efficiency. The digital grayscale imagesare obtained by the dynamic range stretching and rounding of the floating-point data,which is obtained through the X-ray scanning of industrial castings. This paper differsfrom the normal directly defect detection on the grayscale images, but the defectdetection is after the synthesis of color image which is synthesized by the grayscaleimage. Image segmentation methods based on CV model and region growth arerespectively used on the defect detection of the grayscale images and color image,experimental results show that the color image has a better segmentation results.Generally, after floating point data is obtained by using the DR system scanningthe industrial castings, all the data is directly stretched to get grayscale images, thismethod lose a lot of data and the grayscale images’ details lose seriously, and the imagedefects displayed result is not satisfactory. Or only a piece of data is stretched to getgrayscale image, this method only show part of the defects because defects of thedifferent thickness of industrial castings parts corresponds to different data segments. Ordifferent data segments are stretched to get grayscale images which display the defectsof the different thickness of the industrial castings parts. In this paper, according to thecharacteristics of the DR images, auto interception of different data segments is capableof displaying grayscale images of the defects of the industrial castings’ differentthickness parts, and then synthesized into a color image. First of all, doing so canenhance the visual effect, because the human eye is sensitive to color, it can identifythousands of colors but only can distinguish dozens of gray levels; Secondly, the pseudocolor image contains our concern defects of different thickness of the industrial castingsparts; Finally, compared with the approach of pieces of grayscale images to show thedefects of different thickness site, obviously using a color image saves time and effortand intuitive.The C-V model is an image segmentation method which is proposed by Chan andVese based on simplify the M-S (Munford-Shan) model. It has advantages of M-S model and the reduction of the model computational complexity.The C-V model isextended to the case of N-dimensional value in the vector images by Chan and Vese, soit can be used to segment grayscale images and color images. In this paper, the CVmodel is respectively used on railway truck side frame DR gray scale images andsynthetic color images for defect detection, experimental result shows that the lattereffect is better. The segmentation accuracy of the C-V model is high, but if there isstrong edge in the image, such as the castings external borders, the castings dividing lineof thick and thin walled region and etc, C-V model can only segment the local imageand to the whole image segmentation is not perfect. However, based on region growingsegmentation method can partition the entire image. As a common image segmentationmethod, region growing uses an image of the space character, but automatically selectedseed point is difficult. Meanwhile, the region growing is respectively used on railwaytruck side frame DR gray scale images and synthetic color images for defect detection,experimental result shows that the latter effect is better.Finally, the application software which according to MFC (Microsoft FoundationClasses) has the following functions: the original floating point data stretch to grayscaleimages, grayscale images synthesize color image, C-V model segment grayscale imageand color image, region growing method segment grayscale image and color image, etc. |
PDF Full Text Request