A conditional watershed technique for mineral particle characterization from images

The Tuscaloosa Research Center, Bureau of Mines, is currently studying automated and efficient measurement of the distribution of mineral particles in the feed, concentrates, and tailings during minerals beneficiation to better understand the physical and chemical factors that affect the efficiency of these operations. Conclusions derived from varying these factors as part of experimental analyses requires timely and accurate characterization of particles. The application of gray-level mathematical morphology using the watershed transformation on acquired-digital images combined with color separation allows a robust particle characterization. In images where the gray level changes slightly between touching particles, the ability to draw a one-pixel-wide line between these particles facilitates separation. Recent advances in gray-level morphology theory and practice have permitted these techniques to be workable on a microcomputer. This paper summarizes the implementation of the conditional watershed transformation using a microcomputer and the use of color classification for identification of minerals of different types.<
>