|
Vol: 51(65) No: 2 / June 2006 Object Identification in Color Images by Neural Network Classifiers Octavian Pastravanu Department of Automatic Control and Applied Informatics, Faculty of Automatic Control and Computer Science, Technical University "Gh. Asachi" of Iasi, Blvd. Mangeron 53A, 700050 Iasi, Romania, phone: +40-232-230751, e-mail: opastrav@delta.ac.tuiasi.ro Mihaela-Hanako Matcovschi Department of Automatic Control and Applied Informatics, Faculty of Automatic Control and Computer Science, Technical University "Gh. Asachi" of Iasi, Blvd. Mangeron 53A, 700050 Iasi, Romania, phone: +40-232-230751, e-mail: mhanako@delta.ac.tuiasi.ro Keywords: neural networks, classification, color images, region separation, teaching aids. Abstract Neural-network-based classification is a classical topic for the training programs in various areas of applied informatics and computer engineering. Therefore the organization of laboratory experiments placing emphasis on the intuitive support of classification represents a challenge for many educators working in the aforementioned fields. This paper reports our experience in guiding students to master neural network classifiers by object identification in color images. The visual information plays a key role in getting a deeper insight into the proposed classification problems. Thus, the color clusters created in the RGB (red-green-blue) cube allows separating regions with different signification in a color image. We use color images from metallurgy such that the region identification responds to concrete investigations requested by material science (alloy structure and composition). This framework founded on an image processing background attracts students to develop numerous experiments, as well as stimulates their initiative in defining the color classes and assessing the quality of the classifiers. Our approach is illustrated by a relevant case study which compares the classification results provided by three types of neural networks with supervised learning, namely MLP (multi-layer perceptron), LVQ (learning vector quantization) and PNN (probabilistic neural networks). References [1] S. Haykin, Neural Networks: A Comprehensive Foundation, 2nd Edition, Prentice Hall, 1998. [2] R.M. Haralick and L.G. Shapiro, "Image segmentation techniques", Computer Vision, Graphics and Image Processing, vol. 29, pp. 100-132, 1985. [3] S. Makrogiannis, G. Economou, S. Fotopoulos and N.G. Bourbakis, "Segmentation of color images using multiscale clustering and graph theoretic region synthesis", IEEE Trans. on Systems, Man and Cybernetics, Part A, vol. 35, Issue 2, pp. 224-238, March 2005. [4] Guo Dong and Ming Xie, "Color clustering and learning for image segmentation based on neural networks", IEEE Trans. on Neural Networks, vol. 16, Issue 4, pp. 925-936, July 2005. [5] *, Microstructures of Copper and Copper Alloys, http://64.90.169.191/resources/properties/microstructure/. [6] H. Demuth and M. Beale, Neural Network Toolbox User's Guide, The MathWorks, Inc., Natick, MA, USA, 2005. [7] *, Image Processing Toolbox User’s Guide, The MathWorks, Inc., Natick, MA, USA, 2005. [8] *, Image Analysis Cookbook 6.0, http://www.reindeergraphics.com, 2005. [9] D.F. Specht, "Probabilistic Neural Networks", Neural Networks, vol. 3, pp. 109-118, 1990. |