Entire Information Compare Method In Fingerprint Recognition Processing

Automated Fingerprint Identification Technology is one of the most widelyused technologies in the field of biometric identification at present. It has superbadvantages over other techniques. Its key steps include fingerprint image pre-processing (enhancement, sharpness, binarization and simplification), featureextraction and feature matching. Up to now, many similar works exist in theliterature on this topic. However, there are many unfavorable factors such as sweatstains, finger shed skin bubble and the fingerprint natural break. Especially, onlylow-quality fingerprint images can be captured at the criminal spot. As a result, thetraditional techniques of automatic fingerprint recognition can only achieve lowratio of correct recognition, and its application is constrained.The accurate recognition of low-quality fingerprint image is an issue to beresolved in the field of automatic fingerprint recognition. This thesis starts from thispoint and attempts to solve this problem. We found that some fingerprint informationis lost during the process of feature extraction, which leads to less informationinvolved in fingerprint comparison. Apparently, it is harmful to the automaticrecognition of low-quality fingerprint image. Therefore, we propose a complete-information fingerprint comparison scheme. In other words, all the fingerprintinformation is involved in the fingerprint comparison. Correspondently, a softwaresimulation system is designed and developed to verify the proposed scheme. Themain works of this thesis are summarized as follows.First, the state-of-art of fingerprint recognition is briefly overviewed.Especially, the technical challenges for low-quality fingerprint are analyzed. Then,we point out that complete fingerprint information should be involved in the processof fingerprint recognition. That is the main idea behind complete-informationcomparison for fingerprint recognition.Second, a complete-information comparison scheme for fingerprint recognitionis proposed, which can improve the discrimination capability of human eyes. Itborrows some featuring methods in the human seal verification such as angularcomparison and color enhancement, into fingerprint recognition. Different fromtraditional fingerprint recognition based on global or local features, the proposed approach realizes the accurate recognition of low-quality fingerprint by human-machine interaction during the process of fingerprint comparison. Moreover, someimage enhancement techniques are utilized into the process of grayscale fingerprintimage, and the recognition ratio is improved significantly.Third, an experimental software is designed and developed for complete-information fingerprint comparison. The experimental results on several low-qualityfingerprint image show that the proposed approach can achieve better accuracy. It issuitable for the identification verification by low-quality fingerprint image,especially in the circumstances of band ATM machines and criminal spots.