Title :
Automatic ore image segmentation using mean shift and watershed transform
Author :
Amankwah, Anthony ; Aldrich, Chris
Author_Institution :
Process Eng. Dept., Univ. of Stellenbosch, Stellenbosch, South Africa
Abstract :
In this paper, we present a novel method for segmenting ore images specifically for estimating the size distribution of ore material on conveyer belt. The segmentation system uses the mean shift and watershed algorithm. The mean shift algorithm is used to identify pixel clusters of particular modes of the probability density function of the image data. The pixel clusters are then used to generate markers for the watershed transform and shadow areas in ore image. Experimental results show that the proposed algorithm is not only faster than the standard methods but also more robust.
Keywords :
conveyors; image segmentation; mineral processing; minerals; particle size; probability; transforms; automatic ore image segmentation; conveyer belt; image data; mean shift algorithm; ore material; pixel clusters; probability density function; segmentation system; size distribution; standard methods; watershed transform; Algorithm design and analysis; Clustering algorithms; Estimation; Image segmentation; Pixel; Software algorithms; Transforms; Mean Shift; Ore size distribution estimation; Watershed Transform;
Conference_Titel :
Radioelektronika (RADIOELEKTRONIKA), 2011 21st International Conference
Conference_Location :
Brno
Print_ISBN :
978-1-61284-325-4
DOI :
10.1109/RADIOELEK.2011.5936391
