摘要
为提高银行卡号识别正确率,可利用基于深度学习的视觉识别技术来解决目前存在的问题,拓展现有光学识别技术。首先使用数据增强手段对采集的图像集进行预处理,制作数据集,然后对制作好的数据集采用深度学习网络,使用CTPN网络算法实现文本定位,最后将处理后的数据集输入到DenseNet网络中进行模型训练,分析训练前后数据,表明基于深度学习的视觉识别技术可有效提升银行卡号识别的正确率,并保证了识别效率。
In view of the deficiency of the recognition accuracy of bank card number,this paper proposes the use of visual recognition technology based on deep learning to solve the existing problems and expand the existing optical identification technology.Firstly,data enhancement is used to preprocess the collected image set to make the data set.Then,the deep learning network is used to locate the text in the finished data set.Finally,the processed data set is input,the model is trained in DenseNet network,and the data before and after training are analyzed.The results show that the visual recognition technology based on deep learning can effectively improve the accuracy of bank card number recognition and ensure the recognition efficiency.
作者
刘成
李正辉
高基豪
LIU Cheng;LI Zheng-hui;GAO Ji-hao(Zhengzhou Railway Polytechnic,Zhengzhou,Henan,China 450000)
出处
《湖南邮电职业技术学院学报》
2020年第4期35-38,共4页
Journal of Hunan Post and Telecommunication College
基金
2020-2023年国家自然科学基金“复杂环境下铁路轨道在线故障诊断的迁移学习方法研究”(课题编号:62003313)
2019-2021年国家自然科学基金“一类复杂系统精细建模与多级多粒度故障诊断方法研究”(课题编号:61803086)。
