Abstract:

To meet the need of realtime shoot and identification of numerical codes printed on goods labels, due to the small size of numerical codes and the small, fuzzy, complex images taken by iPhones from long distance, we propose a series of image processing to improve the recognition accuracy. First, we choose numerical regions manually, and then separate numerical codes from the background, thus obtaining a numerical image. Then the numerical image is transformed to a whiteblack one by graying and binaryzation. Each numerical code image is segmented by projection method, and then they are normalized, sharpened, and thinned. We finally adopt the nearest neighbor method to recognize the numerical codes based on the extracted statistical features. Experimental results prove the high recognition accuracy.

Key words: numerical code label;image processing;numeral recognition;iPhone