Robust image binarization method for billet identification in steelmaking process

This paper presents a vision system technique for the real-time identification of billet characters in the steel production line. Normally, billets are mixed in the yard, thus, making the process for automatic identification of billet codes very difficult to conduct. In steelmaking industry, the environment around billets has a serious effect on the accuracy of character recognition. The qualities of captured images are very poor because the surfaces of billets are uneven and marking tasks are done under very hot and hostile environment. Therefore, it is required to make special preprocessing algorithm with high robustness. This paper proposes an image binarization method based on the edge information. Compared with the classical binarization methods, the proposed method is effective for the binarization of images with noise, low contrast and non-uniform illumination. It is applied to real industry application in Pohang Iron and Steel Company (POSCO), and the obtained results are compared with results from the existing techniques.