Development of recognition system for billet identification

This paper presents off-line character recognition that is applied to billet identification system in steel-making industry. The identification characters are so corrupted in this application that we need noise-proof and robust character segmentation and extraction algorithms. We propose a new local adaptive thresholding and character extraction methods. For classification of characters we use subspace classifier using KLT. The methods are tested on real industrial application and proved that they are successful applied to such highly noisy conditions.