Abstract:The traditional algorithm and system designed by using the color sensitive sensors and the mechanism of changing color of the color ink is facing the complicated hardware design and constrains about the space inside the banknote counting machine in the practical aspects. In view of this, an algorithm of using the color information of the color ink image based on a single ordinary camera to effectively achieve authentication is proposed. Under the condition of the same color temperature, the color aberration between the color of the color ink themselves and the background color is treated as the principal component of features, and the other color features are combined, it can achieve whether the currency is true or false by the nearest neighbor classifier in the color ink image which is acquired by the CMOS image sensor—MT9T001. Experimental results show that the false accept rate(FAR) is high and the false alert rate is low. Compared with the existing algorithms, it did not only simplify the hardware design and space requirements, it also has a good performance of grasping the false. Now it has been applied to an embedded system.