The objective of the work was to determine the particle size of powder
s displayed in bulk by using image analysis. As image data treatment,
two basic operations of mathematical morphology, grey-level erosion an
d dilation, were studied for the characterization of the particle size
. The procedure was tested on two products: steel marbles taken as sph
erical model particles and ground pea kernels. Five fractions (A, B, C
, D and E) of particle size varying from 0.300 to 1 mm were collected
and mixtures were made from the fractions A, C and E. Monochrome image
s of a thin layer of the particles in bulk were acquired. Successive s
teps of erosion and dilation were applied to the images. The variation
s of grey level observed in the images at each step were measured and
granulometric curves were built. The curves were compared by principal
component analysis. In the case of marbles, the finest fractions were
overestimated because projected surfaces were visualized while the mi
xtures were made by weighting the fractions. In the case of ground pea
s, some demixing of the particles occurred that led to an underestimat
ion of the finest fraction. Despite this, the correlation coefficient
between the median diameter measured by laser diffraction and the firs
t principal component was -0.95.