Offline Grayscale Handwritten Signature Verification

With the popularity of credit card, the attendant security issues are increasingly prominent,traditional password authentication method can not meet the needs of the community. Usingbiological characteristics to verify personal identity can effectively solve the security issues ofauthentication. Signature characteristic is one of biological characteristics. It is widely used incontracts, certificates, agreements, documents and other instruments. As some signatures wereforged, maybe cause harm to society. So the researching of offline signature system is veryimportant, and it has practical significance and theoretical value.This paper considers SIFT features to most the sole basis of the signature verification, whilethe traditional method use the static features and pseudo-dynamic features. The handwrittensignature verification of the status quo at home and abroad, as well as the developments areintroduced first. Then put forward a viable identification algorithm based on gray-scale signatureimage: in the preprocessing,"Quartet" is used to determine signature image boundary. Then thesignature image should be normalized. Subsequent the extraction of the SIFT feature points inthe grayscale image are used to the follow-up signature verification work. Matching SIFT featurepoints using RANSAC algorithm to eliminate the misallocation.Each piece of signature image may have a number of SIFT features and the location of theseSIFT features will be different, you can not use the matching way of traditional staticcharacteristics and pseudo-dynamic characteristics. This article extract multiple signatures incommon SIFT features as universal SIFT features of the signature and deposite them to thesignature SIFT feature library. In the final decision-making stage, using the features in the SIFTfeature library match with the signature which need to be identify. If the number of successfulmatch points is over a certain amount that the signature is real.The final experimental results show that the system accuracy rate reaches88.5%., theaccuracy of the identification of high-level signature also reaches80%, higher accuracy in asimilar signature verification system.