2008
EMSS 2008 Proceeding
Artificial neural network-based classification of vector sets for surface inspection
Authors: Michael Gyimesi, Felix Breitenecker, Wolfgang Heidl, Christian Eitzinger
Abstract
In applications of pattern recognition a set of objects - usually represented by feature vectors - is extracted from an image and needs to be classified as a whole set of objects, meaning that some properties of the set of objects are an aggregation of the single feature vectors ? and the classification of the set of objects may depend on exactly these properties. If the number of objects, respectively the number of features, is not known or limited a priori standard classification algorithms such as support vector machines or linear classifiers cannot applied in a straightforward way due to the fixed size of the number of features in these methods. Therefore the set of object?s ?structure? may not be implemented properly. In this paper we will discuss some issues of these problems and propose recurrent neural networks (RNN) as a promising method to use for such problems.
I3M Scientific Sponsors
I3M Industrial Sponsors
I3M Media Sponsors
