• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>International Journal of Innovative Computing Information and Control >CURVATURE GRAY FEATURE DECOMPOSITION BASED FINGER VEIN RECOGNITION WITH AN IMPROVED CONVOLUTIONAL NEURAL NETWORK
【24h】

CURVATURE GRAY FEATURE DECOMPOSITION BASED FINGER VEIN RECOGNITION WITH AN IMPROVED CONVOLUTIONAL NEURAL NETWORK

机译：改进的卷积神经网络的基于曲线灰色特征分解的手指静脉识别

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Finger vein recognition (FVR) is a technique for identity authentication based on finger vein images (FVIs) that are acquired using a specific device, which has become one of hot spots in the field of biometrics. The main idea of traditional FVR schemes is to directly extract features from FVIs or finger vein patterns (FVPs) and then compare features among FVIs to find the best match. However, the features extracted from FVIs contain much redundant data, while the features extracted from FVPs are greatly influenced by image segmentation methods. Recently, in order to improve the recognition rate and release the high complexity of image preprocessing, a finger vein recognition method based on deep belief network (DBN) with the features extracted from curvature gray images (CGI) has been proposed. However, the training process of DBN is somewhat time-consuming, and the background information of CGI may affect the recognition rate. In order to further improve the accuracy and speed up the training process, a new FVR algorithm based on the improved convolutional neural network (CNN) and curvature gray feature decomposition (CGFD) is proposed in this paper. First, we calculate the curvature of an FVI using a two-dimensional Gaussian template. Then we extract two gray images from the FVI with different scales and add these two images to obtain a CGI. Unlike the previous method, we further decompose this image into two components named vein curvature gray feature image (VCGFI) and background curvature gray feature image (BCGFI). Finally, using VCGFIs as input, an improved CNN is trained and used to recognize the identity of the input FVI. Experimental results show that our scheme is effective and better than traditional schemes and the previous DBN based method.
机译：手指静脉识别（FVR）是一种基于手指静脉图像（FVI）的身份验证技术，该手指静脉图像是使用特定设备获取的，该特定设备已成为生物统计领域的热点之一。传统FVR方案的主要思想是直接从FVI或指静脉图案（FVP）中提取特征，然后比较FVI之间的特征以找到最佳匹配。但是，从FVI提取的特征包含大量冗余数据，而从FVP提取的特征受图像分割方法的影响很大。近来，为了提高识别率并释放图像预处理的高复杂度，提出了一种基于深度信念网络（DBN）的具有从曲率灰度图像（CGI）提取的特征的指静脉识别方法。但是，DBN的训练过程比较耗时，并且CGI的背景信息可能会影响识别率。为了进一步提高精度，加快训练过程，提出了一种基于改进的卷积神经网络（CNN）和曲率灰度特征分解（CGFD）的FVR算法。首先，我们使用二维高斯模板计算FVI的曲率。然后，我们从FVI中提取不同比例的两个灰度图像，并将这两个图像相加以获得CGI。与以前的方法不同，我们将该图像进一步分解为两个分量，分别称为静脉曲率灰度特征图像（VCGFI）和背景曲率灰度特征图像（BCGFI）。最后，使用VCGFI作为输入，训练了一种改进的CNN，并将其用于识别输入FVI的身份。实验结果表明，该方案是有效的，并且优于传统方案和以前的基于DBN的方法。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号