摘要
针对图书馆馆藏书籍的管理和查找带来的大量繁琐手工工作,提出基于书脊文本自动识别和定位的智能图书管理方法。该方法采用神经网络技术搭建文本识别模型对海量书架上的书脊文本进行自动识别,并基于自动识别出的图书信息搭建馆藏图书的数据库,为实现对馆藏图书的智能检索提供可靠的数据支持。所设计的文本识别模型,在考虑包括图像失真和各种光照等条件的基础上,利用监督学习的训练模型来加速神经网络的训练,并在几个基准数据集上达到最先进的性能。通过实验表明,所提出图书智能管理方法具有较好的文本识别准确率和检索正确率,能够帮助图书馆大大减少管理书籍库存所需的体力劳动。
Aiming at a large number of tedious manual work brought by management and search of collection books,an intelligent library management method based on convolutional neural network model detection is proposed.This method uses deep neural network technology to design a text recognition model,and realizes the text recognition of the spine on the bookshelf in the library through automatic text recognition to realize the construction and automatic retrieval of library inventory.The designed text recognition model,based on the considerations of image distortion and various lighting conditions,utilizes a supervised learning training model to accelerate the training of neural networks and achieve the most advanced performance on several benchmark data sets.Experiments show that the proposed book intelligent management method has better text recognition accuracy and retrieval accuracy,and can help the library greatly reduce the manual labor required to manage book inventory.
作者
杨敏
YANG Min(Shaanxi Xueqian Normal University,Xi'an 710061 China)
出处
《自动化技术与应用》
2018年第12期145-150,共6页
Techniques of Automation and Applications
关键词
文本识别
卷积神经网络
递归神经网络
CTC训练模型
图书管理
text recognition
convolutional neural network
recurrent neural network
CTC training model
library management
