A robust damaged fingerprint identification algorithm based on deep learning
Yani Wang, Zhendong Wu, Zhang Jianwu, Chen Hongli · 2016
With the development of science and the improvement of social information, Biological Recognition Technology (BIT) is becoming increasingly important. Among them, the fingerprint identification technology has become the hot spot because of its feasibility and reliability. The traditional fingerprint identification method relies on matching feature points to get the similarity. Undoubtedly, this method needs a long time to find the feature points, and with the rotation, scaling, damage and other problems of the fingerprint, the robustness is decreased seriously. Aiming at these problems, we propose a robust damaged fingerprint recognition algorithm, which is based on Convolution Neural Network (CNN) of deep learning. It not only has a high resistance to abnormal degeneration, and the recognition process is also simpler than the feature points matching algorithm. In the end of the essay, the recognition rate based on deep learning is compared with the fingerprint identification algorithm based on Kernel Principal Component Analysis (KPCA). Experiments' results show that fingerprint recognition based on deep learning has a higher robustness.