A new algorithm for segmenting overlapped green pellets to improve automatic pelletizing process

In common terms, pellets means balls which produces from iron ore and other additions and after baking they will be used to produce sponge iron. One of the steel industries sections is pelletizing which need to have required granulation. Mobarakeh steel complex want to use image processing to calculate pellets size with accuracy to mechanize size measurement tools and decrease human faults. In this research, we assume pellets as circles and by finding the centers and their boundary and using methods we can estimate their radius. The most important problem we are trying to solve it, is pellets sticking, overlapping and underlapping on each other. To find the range of pellets center, by not considering small pellets we have 95% accuracy and to find range of area where the pellets are in there by Otsu binarization the error is not significant. By using this method, number of pellets which are stick to a pellet and their position can be estimated. The false positive error of our method is about 4% and the true negative error is about 12%.