PROCESSING OF DIGITAL IMAGES OF INDUSTRIAL OBJECT SURFACES DURING NON-DESTRUCTIVE TESTING

Abstract

The paper presents modern approaches to processing of images obtained with the help of industrial equipment. Usage of pixel modification in small neighborhoods, application of uniform image processing while changing brightness level, possibilities for combination of several images, threshold image processing have been described in the paper. While processing a number of images on a metal structure containing micro-cracks and being under strain difference between two such images have been determined in the paper. The metal structure represents a contour specifying the difference in images. An analysis of the contour makes it possible to determine initial direction of crack propagation in the metal. A threshold binarization value has been determined while processing the image having a field of medium intensity which are disappearing in the process of simple binarization and merging with the background due to rather small drop between the edges. In this regard an algorithm of a balanced threshold histogram clipping has been selected and it is based on the following approach: two different histogram fractions are “weighed” and if one of the fractions “outweighs” then last column of the histogram fraction is removed and the procedure is repeated again. When there is rather high threshold value a contour break (disappearance of informative pixels) may occur, and when there is a low threshold value – a noise (non-informative pixels) may appear. The paper shows implementation of an algorithm for location of contact pads on image of semiconductor crystal. Algorithms for morphological processing of production prototype images have been obtained in the paper and these algorithms permit to detect defects on the surface of semiconductors, to carry out filtration, threshold binarization that presupposes application of an algorithm of a balanced threshold histogram clipping. The developed approaches can be used to highlight contours on the surface images of mechanical engineering products and prepare them as production prototype images in accordance with enterprise documentation. Such approach makes it possible to remove noise on radiographs without introduction of additional distortions in the processed image; highlight defects of welded joints on the images.