Object Recognition from 2D Images
using Kohonen Self-Organized Feature Maps

H. M. Lakany*, E. G. Schukat-Talamazzini**, H. Niemann***, and M. A. R. Ghonaimy****

* Institute of Statistical Studies and Research, Cairo University, Egypt; University of Edinburgh, Scotland, U.K.;
e-mail: hebaal@dai.ed.ac.uk, tel.: (+44)-131-650-3080; fax: (+44)-131-650-6899

** Institut für Informatik, Friedrich-Schiller-Universität, Jena, Germany;
e-mail: schukat@informatik.uni-jena.de; tel.: (+49)-3641-6-38-703; fax: (+49)-3641-6-38-726

*** Friedrich-Alexander Universität, Erlangen, Germany;
e-mail: niemann@informatik.uni-erlangen.de; tel.: (+49)-9131-85-7774; fax: (+49)-9131-303811

**** Ain Shams University, Cairo, Egypt; e-mail: adeeb@frcu.eun.eg

Abstract—In this paper, we propose an algorithm for recognition of objects from 2D images. The algorithm is
based on neural networks. It uses Kohonen self-organized feature maps (SOFM) as a vector quantizer whose
performance is compared to that of EM technique. A codebook is designed using SOFM and a simple labelling
procedure is used to label the feature vectors, then the labelled training data set of different objects are recog
nized by the resilient propagation network. The algorithm is applied to 2D images of industrial objects.

Pleiades Publishing home page | journal home page | top

If you have any problems with this server, contact webmaster.