计算机工程与应用 ›› 2020, Vol. 56 ›› Issue (10): 134-140.DOI: 10.3778/j.issn.1002-8331.1901-0162

• 模式识别与人工智能 • 上一篇    下一篇

基于密集连接卷积网络的小面积指纹识别方法

陈文燕,范文博,杨钧宇   

  1. 西安科技大学 电气与控制工程学院,西安 710054
  • 出版日期:2020-05-15 发布日期:2020-05-13

Small-Size Fingerprint Recognition Method Based on Densely Connected Convolutional Network

CHEN Wenyan, FAN Wenbo, YANG Junyu   

  1. School of Electrical and Control Engineering, Xi’an University of Science and Technology, Xi’an 710054, China
  • Online:2020-05-15 Published:2020-05-13

摘要:

针对基于细节特征点的传统指纹识别方法在小面积指纹识别时识别率明显下降的问题,提出一种基于密集连接卷积网络的小面积指纹识别方法。对指纹原图进行图像增强处理,充分利用密集连接卷积网络特征复用的优点构建提取指纹特征的深度学习模型,并将二值特征引进训练模型,依据指纹图像的二值特征和特征向量实现小面积指纹的注册和识别。实验结果表明,所提出的方法在自建数据集中正确识别率达到98.57%,高于基于细节特征点的传统指纹识别方法,基本满足智能移动端的应用要求。

关键词: 指纹识别, 密集连接卷积网络, 二值特征, 深度学习

Abstract:

Focused on the issue that the accuracy rate of the traditional fingerprint methods based on minutiae reduces significantly when dealing with the small-size fingerprint, a small-size fingerprint matching method based on densely connected convolutional networks is presented. Firstly, the image enhancement of original fingerprint images is performed. Additional, a deep learning model for extracting fingerprint features is built with the advantages of the feature reuse of densely connected convolutional networks, and the binary features is introduced. Finally, the small-size fingerprint registration and recognition are accomplished according to the binary features and feature vectors of fingerprint images. The experimental results show that the proposed method achieves a correct recognition rate of 98.57% in the self-built database, which is much higher than the traditional fingerprint matching methods based on minutiae and meets the application requirements of the intelligent mobile terminal.

Key words: fingerprint recognition, densely connected convolution network, binary features, deep learning