The Anti-fake And Classification Technology Research On Ultraviolet Images Of RMB

In recent years, RMB anti-counterfeiting classification recognition technology is the more popular topic in the field of image processing and pattern recognition, which has high value on theoretical research. It plays a considerable strategic importance role in national financial security, so it has broad application prospects.This research gets the salient features of RMB ultraviolet image which the ordinary visible light image do not has by analyzing and understanding RMB ultraviolet image. RMB can be classified quickly and accurately by using these features and high imitation counterfeit money can be detect which common algorithm can not. The algorithm that this research proposed will be useful for secure circulation of RMB.Based on RMB ultraviolet image anti-counterfeiting Classification recognition technology includes four processes: the RMB ultraviolet image acquisition, the image prepossessing,the denomination and facing recognition and authenticity recognition.In the process of the image prepossessing, removing noise, equalization of histogram, edge detection and tilt correction are studied. Radon convert is used to line detection which removes interference of noise with respect to point detection.In the process of the denomination and facing recognition,size characteristics is used for denomination recognition. Facade M code area match method is put forward which contribute to denomination and authenticity recognition. Finally, facing recognition is studied by the means of template matching. Above algorithms achieved relatively good results.In the process of the RMB anti-counterfeiting discriminating, the correlation coefficient of facade M code area and the texture of reverse side extracted by Gabor filter are considered helpful to anti-counterfeiting discriminating. The mean number of pixels whose Gabor value exceeds the threshold and the correlation coefficient are treat as ultimate feature which are adaptive for lightness and a small amount of contamination. Finally, supervised training method is introduced and genuine and fake currency classifier based on Fisher core is designed.Algorithm proposed by this article are verified on RMB, Proving that banknote recognition technology used in this paper have certain advance. More satisfactory results have been achieved.