Fingerprint Image Retrieval And Matching Algorithm Research

With the advent of electronic banking, e-commerce and smart cards, people pay more attention to the privacy and security of information stored in various databases. In many automatic personal identification technologies, the fingerprint identification is becoming one of the most popular, convenient and reliable personal authentication technologies. Many fingerprint products have been offered nowadays, however, most of the key technologies have not been publicized on account of commercial interest, while the better performance is required. So more reliable and accurate automated fingerprint identification systems (APIS) have been of great interest to researchers. On the base of existing works, the key technologies of fingerprint identification are studied in the thesis in which the main contents are as follows.(1) A fingerprint image identification system is built, including preprocessing, feature extraction, post-processing, indexing and matching of fingerprint images.(2) In order to implement the fast fingerprint identification in face of large database, this paper presents a fingerprint identification approach by integrating the indexing and matching algorithms. Two new indexing algorithms are proposed by modifying the triangle minutiae feature vector. The first indexing algorithm is based on irregular array and the second one is based on regular array. The hashing and distribution array are used to seek the potential triangles quickly for decreasing the indexing time.(3) A new matching algorithm is presented based on the indexing information that is introduced into the minutiae matching process in a simple and effective way, which solves the problem, with low computational cost, of selecting reference point pairs. The matching process is composed by two different stages. In the first stage, an unvaried bounding box is used to search the first batch of matching minutiae pairs. In the second stage, a varied bounding box is used to search the matching minutiae pairs omitted in the first stage in order to tolerate the non-linear distortion of the images. The matching algorithm can correctly recognize the fingerprints from the same fingers while distinguishing the fingerprints from the different fingers quickly.All the proposed algorithms have been implemented in a practical fingerprint image processing system in the paper. The experimental results show that the false acceptance ratio (FAR) is approximately 0% and the false rejection ratio (FRR) is less than 1.001%, and its main performance criterion reach the requirement of design. The proposed indexing and matching algorithm performs well both in matching speed and in toleration of the non-linear deformation, besides its shift and rotation invariant property.