Image Processing Techniques for Cork Tiles Classification

An intelligent, automated visual inspection system is investigated in this paper. It is used for pattern recognition and classification of four different types of cork tiles. The process includes image acquisition with a CCD camera, texture feature extraction, statistical processing of the feature vectors, and cork tiles classification with feed-forward Neural Networks (NN) employing a hybrid global optimization technique called GLP_¿S. We use co-occurrence method and the Laws filter masks to generate image texture characteristics. Several different NN topologies, reflecting variety of texture features are simulated, evaluated and their generalization abilities discussed and assessed. Reported test results show very encouraging recognition and classification rate of up to 95%.