非表示:
キーワード:
-
要旨:
This paper presents an information system that supports a materialographic
laboratory in classifying material samples based on microstructure images. The
system uses database and Web technologies to manage its information and make it
accessible to Internet users. Its core is a classifier, based on support vector
machines, that provides an automatic diagnosis of the material class of a given
sample. The classifier uses texture features from an underlying image analysis,
the so-called Haralick parameters, and stereologic features such as fractal
dimension, Euler parameter, etc. In addition to the classifier, the system
provides a sensitivity analysis that allows the user to understand which
features are most influential for certain classification decisions. The system
is fully operational and can be used on the Web.