Image Processing Based Detection of Counterfeit Indian Bank Notes

Currency counterfeiting is a significant offense which has a profound impact on the assets and capital of the citizens thus having an adverse effect on the nation's finances. The schemes currently existing to combat the falsification of banknotes are complex, hardware-based and inaccessible to the common people. In this paper, a unique authentication system is proposed which is compact, mobile and devoid of any hardware components. Certain security features, such as security thread and latent image, embedded on the note are utilized to help ascertain its legitimacy. The methodology involves the extraction and encoding of these security features. Given the prominence of the security thread in certain image planes, a clustering algorithm, k-means is applied for classification. The latent image, segmented via template matching was encoded using HOG descriptor and classified with an SVM model. The result is illustrated with the aid of performance parameters and overall accuracy.