| At present, sinter is the main burden for blast furnace production. Sinter quality has a close relationship with its composition, property and structure. Therefore, it is very important to measure and analyze composition and microscopic feature of sinter.Sinter can be defined as a kind of complex containing various minerals, the composition of which are iron-containing minerals, gangue minerals and binding phase formed by them. Its microstructures were observed under microscope, mainly including granular crystal structure, porphyritic crystal structure, melting crystal structure, eutectic crystal structure etc. Mineral compositions of sinter are complicated and various, presenting inter-wrapping and complicated shape feature. And mineral composition and microstructure of sinter vary by raw materials and sintering process.In this paper, characteristics of sinter mineralogy micrograph are summarized based on micro-images of sinter. Selection principle of image segmentation and recognition algorithms for sinter mineralogy micrograph processing is explored. Hence, region growing algorithm, fuzzy c-means algorithm and watershed algorithm are chosen. Region growing algorithm has faster processing speed and higher accuracy, compared with the other algorithms by image segmentation and recognition. Therefore, it is more applicable to recognition of sinter mineralogy micrograph.Image recognize system for iron ore sinter mineralogy was developed by VC++ and Openvc. The system consists of functions such as image reading, recognition of sinter mineral composition by region growing algorithm, image filtering, expert subsystem, and results evaluation, etc. The relative error of sinter composition recognized by the system was less than 5%, comparing with that tested under optical microscope.Content of main minerals of sinter can be acquainted by this system which has high speed and accuracy, meeting the requirement of sinter mineralogy processing during actual operation.
|
PDF Full Text Request