基于卷积神经网络的低质量指纹识别研究

Low-Quality-Fingerprint Identification based on Convolutional Neural Network

当今社会,各种不稳定因素的增加,使得人们越来越关注识别技术的可靠性。随着科学技术的发展,生物识别技术被广泛应用到各个方面,也使其成为身份鉴定的主要手段之一。其中,指纹识别更因其方便性和可靠性受到研究者们的推崇。传统的指纹识别方法依赖于对特征点的一一比对来得到指纹之间的相似性,无疑是较为成熟且识别率较高的一种方式。但是,这种方法需要花费大量的时间来寻找特征点,且指纹图像的质量对最终的识别率具有关键性影响。因此,针对这些问题提出了基于卷积神经网络的多分块低质量指纹识别算法。它将指纹细化图做分块处理,然后将分块后的指纹图像和指纹原始细化图均输入到卷积神经网络中进行分类识别。实验结果表明,所提算法有效解决了低质量指纹识别率低的情况。

In modern society, the reliabilty of identity technology attracted much attention from people as the potential threat becomes even more popular, and biological identification technology begins to act as one of the main solutions in the field of authentication, and thus is widely applied in many aspects of the industries. Of the biological identifications, the fingerprint identification, for its convenience and stability, is highly praised by the researchers. The traditional fingerprint identification method depends on the comparison of similarities of between the two fingerprints, and proves to be a fairly mature solution with relatively high identification ratio. However, this method is quite time-consuming in locating the feature points of the fingerprints, and the quality of fingerprint image sample has the essential influence on the final identification rates. Foe this reason, the multi-block low-quality-fingerprint identification algorithm based on convolutional neural network is proposed. This method firstly separates the fingerprint thinning image into several blocks, then transfers the separated image blocks and the entire original image together to the convolutional neural network for classification and indentification. The experiment results indicate that the proposed algorithm could effectively solve the problem of low identification rate when dealing with the low-quality-fingerprint images.