Fake banknote detection using multispectral images

With advancement of sensor technologies, it is now possible to manufacture cost-effective multispectral sensors for ATM (automatic teller machine). Using multispectral images, one can better cope with counterfeit banknote problems. In this paper, we propose a counterfeit banknote detection using multispectral images in visual and infrared spectrum. In the proposed method, we divided a banknote into a number of blocks and extracted features from the blocks. To reduce processing time for real-time applications, we applied block selection algorithms. Since ATMs have a limited computing power, we used linear and quadratic classifiers. Experimental results show promising results.