Abstract
In this paper, we propose a new recognition procedure for a degraded character string printed on a steel plate for factory automation. The character string consists of alphanumerals and '-'. The procedure includes the image emphasis with Laplacian Gaussian filter, the extraction of the character string image, the segmentation-recognition with the dynamic programming, and the high accuracy character recognition. We evaluated the accuracy for 1036 images (8, 806 characters) scanned by the monochrome video camera at an actual steel producting factory line. The average recognition rate was 99.2% for the character recognition, and 91.6% for the character string recognition.
