一种基于改进多任务学习的手背静脉识别方法与系统
A Method and System of Dorsal Hand Vein Recognition Based on Improved Multi-task Learning
郑音飞 1刘高凯 2罗泽熠 1段会龙 1徐正国3
作者信息
- 1. 浙江大学
- 2. 浙江大学湖州研究院
- 3. 浙江大学;浙江大学湖州研究院
- 折叠
摘要
针对当前手背静脉识别产品较为缺乏的问题,为了促进深度学习技术在手背静脉识别领域的落地应用,提出一种基于改进多任务学习的手背静脉识别方法,同时开发一套基于软硬件协同的手背静脉识别系统.来自采集手背静脉数据集上的实验结果显示,改进算法在嵌入式设备上能够取得99.59%的准确率、0.437%的等误率、小于1 s的识别时间,足以满足大多数常见应用场景对识别性能的需求,为手背静脉识别方法的落地应用提供了一种有效的解决方案.
Abstract
In order to promote the application of deep learning technology in the field of dorsal hand vein recognition,this paper proposes a dorsal hand vein recognition method based on improved multi-task learning,and develops a dorsal hand vein recognition system based on hardware and software cooperation.The experimental results on the self-collected dataset of dorsal hand vein show that the improved algorithm can achieve 99.59%accuracy,0.437%equal error rate and less than 1s recognition time on embedded devices,which is enough to meet the requirements of recognition performance in most application scenarios.The system developed above provides an effective solution for the application of the recognition method of dorsal hand vein.
关键词
手背静脉识别/多任务学习/基础模型/改进模型/Jetson/NANO/系统开发Key words
dorsal hand vein recognition/multi-task learning/basic mode/improved model/Jetson NANO/system development引用本文复制引用
出版年
2024
NETL
NSTL
万方数据