摘要
针对假币的特征未知以及样本数量不平衡的局限性问题,提出基于半监督辅助分类生成对抗网络的纸币红外特征鉴伪算法。辅助分类生成对抗模型可以扩充样本的数据集,经过半监督的方式训练得到分类器进行分类,实现对纸币红外特征的鉴伪。实验结果表明,该算法能提高假币鉴伪的准确率以及泛化能力。
Focusing on the limitation problems of the unknown characteristics and the unbalanced sample size of counterfeit paper currency,an infrared characteristic identification algorithm is proposed based on adversarial networks generated by semi-supervised auxiliary classification.This model can expand sample data set,which is trained in a semi-supervised manner to obtain a classifier with the ability to identify the infrared characteristics of paper currency.The experimental results show that the algorithm can improve accuracy rate and generalization ability in identifying counterfeit paper currency.
作者
陈小静
曹语含
张学东
CHEN Xiaojing;CAOYuhan;ZHANG Xuedong(School of Computer Science and Software Engineering,Universityof Science and Technology Liaoning,Anshan 114051,China)
出处
《辽宁科技大学学报》
CAS
2021年第1期50-55,80,共7页
Journal of University of Science and Technology Liaoning
关键词
红外纸币鉴伪
辅助分类生成对抗网络
半监督
ACGAN
infrared characteristics identification of paper currency
adversarial network generated by auxiliary classification
semi-supervision
