基于卷积神经网络的手指静脉采集识别系统

Finger-vein Acquisition and Recognition System Based on Convolutional Neural Network

手指静脉采集识别技术是一种新的生物特征识别技术,因其活体性、高稳定性、高安全性等优点,具有广阔的市场应用前景,但现今市场已有的手指静脉采集识别设备存在着识别准确率不高、采集效果不佳等问题,无法达到快速准确识别的要求。本文提出了一种基于卷积神经网络的手指静脉识别系统,系统分为采集和识别两大模块。基于血红蛋白对近红外光的吸收特性制作了新型的手指静脉图像采集装置,预处理采用Canny算子提取手指静脉图像边缘,并采用中线拟合矫正方式修正了图像旋转问题;基于U-Net网络优化设计手指静脉识别技术,提出基于AlexNet网络的深度学习方法,对采集模块获取的自建图片数据集进行识别分析,实验结果表明,本文方法的识别准确率可达到96.65%,相比U-Net网络提升了2.17%,同时软硬件一体化设计使得系统的稳定性得到提升。

Finger vein acquisition and recognition technology is a new biometric recognition technology. It has broad market application prospects because of its advantages of liveness, high stability and high security. However,the existing finger vein acquisition and recognition equipment in the market has some problems, such as low recognition accuracy and poor acquisition effect, which can not meet the requirements of fast and accurate recognition. This paper presents a finger vein recognition system based on convolutional neural network. The system is divided into two modules: acquisition and recognition. Based on the absorption characteristics of hemoglobin to near-infrared light, a new type of finger vein image acquisition device is made. The edge of finger vein image is extracted by Canny operator in preprocessing, and the image rotation problem is corrected by midline fitting correction;Based on the optimal design of finger vein recognition technology based on u-net network, a deep learning method based on Alex net network is proposed to recognize and analyze the self built picture data set obtained by the acquisition module. The experimental results show that the recognition accuracy of this method can reach 96.65%, which is 2.17% higher than that of u-net network. At the same time, the integrated design of software and hardware improves the stability of the system.