Image Processing Algorithms And Their Applications For Flatness Detecting

Image processing for defect examinations of industrial products is spreading with the development of computer technology. The main contents of this dissertation are basic image processing algorithms and their applications for flatness detecting of cold-roll steel sheets.The main topics of this dissertation which are shown on the stage of flatness detecting project with image segmentations of Rice as their clue are three kinds of basic algorithms, histogram transform, image binarization and image processing based on morphology, which are deeply studied. In the part of histogram transform, a histogram equalization algorithm based on Hall theorem for histogram equalization is presented and the properties of histogram equalization are analysed in theoretical terms. Then, the drawbacks of histogram equalization are shown. Furthermore, a compression method of gray levels is presented in this part. In the part of image binarization, a binarization method based on means of classes which is proved later that this algorithm is equivalent to Otsu's algorithm is presented, and two algorithms, histogram equalization algorithm based on image binarization and image binarization algorithm based on histogram equalization are put forward. In the part of image processing based on morphology, firstly, a scheme for structuring elements selection is presented. Secondly, an image denoising algorithm based on binary morphology and an image enhancement algorithm based on gray-level morphology associated with proper structuring elements is fulfilled. At last, on the basis of these three kinds of basic algorithms put forward previously, image segmentations of Rice are quite successful.Three practical flatness detecting methods employing these basic algorithms are developed. The flatness detecting algorithm based on top-hat transform and the flatness detecting algorithm based on comparison(I) is put forward at first. Next, the nature of cold-roll steel sheets images, that all the images from cold-roll steel sheets with defects differ from one another, but two images without defects from two neighboring cold-roll steel sheets respectively are almost the same, is discovered by analysis. Finally, based on the nature, the flatness detecting algorithm based on comparison(II) is developed. Practice shows that these three flatness detecting algorithms are practical, however, flatness detecting algorithm based on comparison(II) whose detecting accuracy is up to 100% is better than others with preferable comprehensive performance.