Iron ore green pellet diameter measurement by using of image processing techniques

Abstract

Automatic Quality control is a vital process in many manufacturing process such as steel industry. Pellet size monitoring and control is a critical process which is done in steel making to improve quality of products. According to the technical reports, pellet size should be fall in the range of 9-16 mm in diameter. Larger or smaller pellets could degrade the final products and impose extra overheads to industry. In this paper, a new method is proposed for measuring the pellet size using practical Image Processing algorithms. In this method, active contour with Chen-Vese method is used to eliminate the images backgrounds and achieving a distinguishable plot of the objects. After detecting distinct elements existing in the image, the number of pellets in each object is determined and each object is classified as singular, double, triple or more pellets using an SVM classifier. Finally, morphological methods are used to estimate the real size of pellets and the pellets size histogram is presented. This practical method was applied in Mobarakeh Steel Complex, where the method was tested on about 1000 prototypes. Results showed that we have 95.1% of accuracy for detection of one pellet elements and in classification by SVM 95.6% of elements were classified correctly.