Grain size characterization of aluminum alloys can be correlated with
thermomechanical processing properties [1]. To predict the processing
characteristics of these alloys under certain combinations of strain,
deformation, and temperature, the metallographic measure of the grain
size can be used. Most of the techniques that have been proposed so fa
r do not provide reliable and reproducible quantitative metallographic
measurements of the grain size, due to human error [2, 3]. Considerin
g that this manual task is also tedious to perform, a general color im
age analysis algorithm is proposed to automate the characterization pr
ocess using an optical microscope with polarized light. This algorithm
was tested on several ingots and on rolled-aluminum samples. The resu
lts show robustness in several conditions, even when the grains can ba
rely be seen by a human operator. Time constraints specific to industr
ial settings were taken into account when implementing the algorithm.
A complete characterization of the entire surface of an ingot can be o
btained within a reasonable time limit. The same algorithm was tested
with other types of color material. The proposed framework for handlin
g segmentation of color grains fulfills industrial requirements for th
e characterization of materials seen under a microscope with polarized
light. (C) Elsevier Science Inc., 1996.