A New Method for Multi-level Fingerprint Classification based on Inflexion-like Eigenvector
Wang Zhi-Jian,Ning Xin-Bao,Yang Xiao-Dong
DOI: https://doi.org/10.3321/j.issn:0469-5097.2007.01.006
2007-01-01
Abstract:Automatic fingerprint identification technique (AFIT) has realized fingerprint identification through fingerprint acquisition, minutiae extraction and the comparison with template stored in the database, which is based on the uniqueness of fingerprints. In the past few years, people have attached much importance to AFIT and it has become a hotspot of science and industry throughout the world. Now it has graduately been applied in access control, time management, finance, public security and e-commerce. It is believed to bring a revolution in the identification area. This paper for the first time successfully extracted a new parameter suggesting the characteristics and regularity of variation in the orientation field from the image of a single fingerprint, which was defined as inflexion-like eigenvector. Using these new eigenvectors and core points obtained from inflexion-like as the classification feature, as well as the classification policy of multi-level tree structure, we could acquire the maximum inter-class distance and the minimum intro-class variance, making all kinds of fingerprints in diverse levels and different feature spaces classified in linear manner. Under the testing condition of high dimensional ratio, the accuracy rate of classification was achieved to above 99 %. Moreover, this classification method also had a good robustness to those fingerprint images of lower quality. Therefore, a real-time classification was realized authentically, which could arrive at high accuracy and consistency. This method has strong application in practice and also some values in popularizing.