Research Of Fast Fingerprint Recognition Algorithm Based On Large-scale Fingerprint Database

With the rapid development of computer science and technology, the automaticfingerprint recognition technology has made great progress since the1980s. Currently,many countries, including China, have specialized companies or institutions engaged inthe research of automatic fingerprint identification. While the continuous expansion of thefingerprint database capacity, fingerprint recognition system will have problems such asrate drop, match slower etc. At home and abroad, fingerprint recognition technology for alarge-scale fingerprint database has become a hot and difficult research. In this paper, wedo the following work and try to solve the issue about quickly identify in a large-scalefingerprint database:Based on the fuzzy mathematical theory, a quality evaluation system about fingerprintimage was proposed. Through its judgment, you can know whether fingerprint imagecomply with the identification requirements or not. Selected assessment indicators includegray distribution features, effective area of the fingerprint image, center offset of thefingerprint image, spectrum structure. According to the fuzzy weights matrix and fuzzyrelationship matrix, we can calculate the comprehensive evaluation index of thefingerprint image, and judge the quality of the fingerprint image in accordance with thecomprehensive evaluation index.For a large-scale fingerprint database fingerprint recognition, we need classifyfingerprint to narrow the range of linear matching. For this problem, the paper proposed amultidimensional characteristic of fingerprint classification, in order to improve the speedand efficiency of fingerprint identification. The classification features including thenumber of singular point, the direction of the ridge, the ridges between the core and thedelta, are organized into the index of kd-tree. Through the kd-tree, we find the appropriatecategory to narrow the matching range.Sometimes, there are still many fingerprint images after fingerprint classification, which fail the requirements of real-time query. Learning from perception hash similarityqueries, this paper further narrow the scope of the linear matching. The experimental dateshow that, this method greatly improve the speed of fingerprint identification at the sametime not greatly affect the erroneous recognition rate while select an appropriate thresholdvalue.The experimental results show, the new methods put forward in this paper can largelyimprove the efficiency of large-scale fingerprint database fingerprint identification.