Multi-level fingerprint continuous classification for large-scale fingerprint database using fractal analysis

Abstract

A three-level classification method using fractal analysis was proposed to improve the speed, accuracy, and robustness of an automated recognition system for a large-scale fingerprint database. Low-quality fingerprints were first eliminated via an assessment algorithm with a multi-level progressive discriminant factor. Next, three-level classification was done for fingerprints with acceptable quality. The fingerprints were sorted into six categories according to fingerprint types. Classification was made based on the number of ridge lines between the singular points of each fingerprint. Categorisation was done in terms of the fractal dimensions of the stable-quality region of each fingerprint image. With the second and third levels of classification, continuous classification and redundancy retrieval could be achieved. The experimental results using the NIST-4 fingerprints database established that the proposed method has various advantages, including fast retrieval speeds, strong adaptability, and great robustness, making it particularly suitable for automated classification and recognition matching for large-scale databases.