Use of image processing in iron and steelmaking industry

Image analysis differs from image processing as used in image correction, e nhancement or computer graphics, because of a final step of quantification. It is this which has enable image analysis to become an investigation meth od used for more than thirty years in French iron and steelmaking industry. The ability to objectively quantify the content of images, then to provide data which can be handled by computers and finally to obtain accurate stat istical measures is one of the reasons behind ifs development. Another reas on is the ability to help a human operator (sometimes even replace him) in tiresome and repetitive tasks. At this time quantitative metallography rema ins an important activity. Further the increase in computing power and info rmation processing have led to an extension of the field. In laboratories a nd in production lines, Visual inspection remains the final judge for under standing, interpreting or controlling a process. The examples presented her e do not provide an exhaustive coverage or image analysis but rather the di versity of its scope. The requirements for strength and toughness of pipelines steels are becomin g more and more critical. Many studies have shown the importance of second phase islands (martensite-austenite islands, carbides,...) in fracture prop erties. The quantitative analysis of the morphology distributions of these components is thus an important step for microstructure characterization. R esults of these analyse are used in the models which predict elongation to failure. The small size of martensite-austenite (MA) islands requires that images be acquired using scanning electronic microscope. The metallographic attack u sed (Villela) does not allow distinguish MA islands from carbides, We have developed some software with the aim of detecting and measuring the morphol ogical characteristics of MA islands (area, elongation, Feret diameters) an d minimal distances between neighbours. Roughness transfer from rolling cylinders to plates is a critical process l imit. An understanding of mechanisms involved drives ifs improvement, Furth er, gains require knowledge of the contact ratio during rolling. An estimat ion of this important parameter can be made from three dimensional maps of both surfaces in contact. In steel making, using scrap as raw material requires the elimination of no n ferrous materials in order to supply clean metal suitable for low alloyed steel. Copper is the main unwanted metal. In practice, a large proportion of copper in scrap is recovered since it is mechanically bound or alloyed t o steel. For this manual sorting is needed. Automation of this repetitive a nd inefficient step is possible by the use of image analysis. The system us es colour image decomposition into three basic components (lightness, hue a nd saturation) then their analysis. The aim is to detect objects with conta ining the colour copper. The surface aspect of plates is one of the main quality criteria for custom ers. Consequently control this parameter and quick correction in case of a problem are very important. Moreover, some surface defects can cause plan f ailure in downstream operations. The faster the line speed is, the more dif ficult becomes the detection of defects. Automatic Area Inspection Systems become necessary for lines with a speed of many hundred of meters per minut e. Now, human operators who used to manually inspect the plates all the day prefer to rely on the screen and to confirm or reject its diagnostic. These examples show the variety of image analysis uses. Classical metallogr aphic measurements allow the accurate characterization needed for the elabo ration of high quality products as there become more and more sophisticated . This method is not only used in laboratory measurements but has also beco me a tool for industrial control which contributes to increase productivity and quality.