摘要
当今档案电子化过程中遇到的OCR识别倾斜和横向布局字体困难、印章遮挡导致OCR识别错误、OCR技术难以进行表格格式解析和表格内容识别等难题。为了解决这些问题,使用MobileNetV2网络实现了横向布局的档案扫描件角度转正,并应用U-Net网络和目标检测技术实现了印章去除技术,同时运用U2-Net网络实现了表格格式解析技术。此外,还引入了图片倾斜矫正技术和OCR文字识别技术,构建了全面的档案电子化系统。经实验验证,该系统能够高效地将纸质文档扫描成PDF,并对PDF中的文字进行解析、横向布局的页面矫正及对表格格式和内容进行解析。该系统结合高性能服务器和全天候队列数据处理,大大提高了海量档案电子化的效率,为企事业单位的档案管理带来了很大的便捷和效率优势。
In the current process of digitizing archives,there are difficulties in OCR recognition such as skewed and horizontally arranged fonts,seal occlusion leading to OCR recognition errors,and difficulties in table format parsing and table content recognition using OCR technology.In order to solve these problems,MobileNetV2 network is used to realize the Angle correction of transverse layout file scanning,and U-Net network and target detection technology are used to achieve seal removal technology,and U2-Net network is used to achieve table format analysis technology.In addition,the technology of image tilt correction and OCR character recognition are introduced,and a comprehensive electronic file system is built.The experimental results show that the system can scan paper documents into PDF efficiently,parse the text in PDF,correct the horizontal layout of the page,and parse the form format and content.The system combines high-performance server and all-weather queue data processing,greatly improves the efficiency of electronic mass archives,and brings great convenience and efficiency advantages for the archives management of enterprises and institutions.
作者
肖雪丽
冷颖雄
谢洁芳
邓茵
周彦吉
Xiao Xueli;Leng Yingxiong;Xie Jiefang;Deng Yin;Zhou Yanji(Dongguan Power Supply Bureau of Guangdong Power Grid Co.,Ltd.,Dongguan,Guangdong 523109,China)
出处
《机电工程技术》
2024年第1期176-179,共4页
Mechanical & Electrical Engineering Technology
基金
南方电网公司科技项目资助(031900KK52220044)。
关键词
图像分割
图像分类
档案电子化
OCR
image segmentation
image classification
electronic records
OCR
