SIFT-Based Low-Quality Fingerprint LSH Retrieval and Recognition Method

Abstract

Most of the existing fingerprint retrieval systems are based on the overall characteristics and detailed features of fingerprints, and their performance is poor in the cases of low-quality fingerprint images, such as incomplete fingerprint images. In order to improve the recognition speed, accuracy, and robustness of automated fingerprint recognition systems based on large-scale fingerprint databases, in this paper, we propose a fast fingerprint classification retrieval and identification method based on Scale Invariant Feature Transform (SIFT) and Local Sensitive Hash (LSH) algorithms. A method based on scale space theory extracts SIFT feature point descriptors of relatively high-quality fingerprint images in accordance with the principle of a greater matching contribution rate, uses multi-template image feature fusion technology to build a stable fingerprint feature template database, achieves the storage and retrieval of highdimensional SIFT features using the LSH algorithm, and carries out matching progressively by level on the basis of SIFT’s matching principle of close neighboring priority scale. Experimental results show that the proposed method has strong penetration, high retrieval efficiency, good recognition accuracy, and strong robustness, thereby providing a fast and efficient retrieval and matching mechanism for the automated recognition of the large-scale fingerprint database, with strong practicality.