5. Conclusion
In this paper, a novel fingerprint classification algorithm is
presented. There are mainly three contributions in this paper. Firstly,
a regularized orientation model is proposed to improve fingerprint
orientation extraction. The experimental results over FVC-onGoing
orientation extraction benchmark demonstrate its effectiveness. Secondly,
orientation image, complex filter responses as well as ridge line
flows are combined to represent a fingerprint, which characterizes
both ridge flow and singularities. Thirdly, a heuristic classifier is
proposed for the fingerprint classification, which is robust to large
intra-class variabilities and small inter-class variabilities. The experimental
results over NIST SD 4 demonstrate the effectiveness of the
proposed classification algorithm.
In summary, the proposed algorithm is able to make use of
orientation image and complex filter response to achieve a better
accuracy. Most of misclassified fingerprints are with severe noise
because it is very difficult to extract accurate orientation. Fig. 10 shows
some examples. In addition, some Left or Right fingerprints are
mistaken as Tented Arch fingerprints and some Tented Arch fingerprints
are mistaken as Left or Right fingerprints. Fig. 11 illustrates such
cases. In future, elaborate classification strategy is needed to distinguish
Tented Arch fingerprints from Left and Right fingerprints.