Fingerprint classification based on learned features

In this paper, we present a fingerprint classification approach based on a novel feature-learning algorithm. Unlike current research for fingerprint classification that generally uses well defined meaningful features, our approach is based on Genetic Programming (GP), which learns to discover composite operators and features that are evolved from combinations of primitive image processing operations. Our experimental results show that our approach can find good composite operators to effectively extract useful features. Using a Bayesian classifier, without rejecting any fingerprints from the NIST-4 database, the correct rates for 4- and 5-class classification are 93.3% and 91.6%, respectively, which compare favorably with other published research and are one of the best results published to date.

I. Introduction

The Henry system is a systematic method for classifying fingerprints into five classes: Right Loop (R), Left Loop (L), Whorl (W), Arch (A), and Tented Arch (T). Fig. 1 shows an example of each class. This system of fingerprint classification is commonly used by most of the developers and users, although the scheme adopted by the FBI defines eight classes [1]. The most widely used approaches for fingerprint classification are based on the number and relations of the singular points (SPs), which are defined as the points where a fingerprint's orientation field is discontinuous. Using SPs as reference points, Karu and Jain [2] present a classification approach based on the structural information around SPs. Most other research uses a similar method: first, find the SPs and then use a classification algorithm to find the difference in areas, which are around the SPs for different classes. Several representations based on principal components analysis (PCA) [3], a self-organizing map (SOM) [4], and Gabor filters [5] are used. The problems with these approaches are a)

it is not easy to detect the SPs and some fingerprints do not have SPs;

b)

the uncertainty about the location of SPs is large, which has great effect on the classification performance since the features around the SPs are used.