基于OCR识别技术的碎片化时空信息库异常文件检测方法  

Intelligent Detection Method of Abnormal Files in Fragmented Spatiotemporal Information Based on Intelligent OCR Recognition Technology

作　　者：杨飞 宋吉星 王宜春 杨伟迪 赵璟 YANG Fei;SONG Jixing;WANG Yichun;YANG Weidi;ZHAO Jing(Shenshuo Railway Company,CHN Energy Baoshen Railway Group Co.,Ltd.,Yulin 719316,China)

机构地区：[1]国能包神铁路集团神朔铁路公司,陕西榆林719316

出　　处：《武汉理工大学学报（信息与管理工程版）》2023年第6期967-971,共5页Journal of Wuhan University of Technology：Information & Management Engineering

基　　金：包神铁路神朔公司书面合同扫描识别比对开发项目(YXHT202112096822).

摘　　要：为优化碎片化时空信息库中包含的异常文件的智能检测过程,降低工作人员的疲劳感,提高检测的精度与效率,提出基于OCR识别技术的碎片化时空信息库异常文件检测方法。利用图像二值化和旋转矫正等技术对扫描图像进行预处理,基于连接预选框网络的文本检测网络和端到端文本识别网络来自适应提取预处理图像中包含的文字信息。将识别出的文字信息描述为定长字节序列,得出该文件的统计特征,并对比信息库异常文件的标志特征,输出异常文件检测结果。结果表明,该异常文件检测方法的F-Score达到0.9以上,证明该方法具有良好的性能,能够在碎片化时空信息库中对异常文件进行高效准确的智能检测。In order to optimize the intelligent detection process of abnormal files contained in fragmented spatiotemporal information databases,reduce staff fatigue,and improve detection accuracy and efficiency,an intelligent detection method for abnormal files in fragmented spatiotemporal information databases based on intelligent OCR recognition technology was proposed.By utilizing techniques such as image binarization and rotation correction to preprocess scanned images,a groundbreaking combination of text detection networks based on connected pre-selection boxes and end-to-end text recognition networks was utilized to adaptively extract the text information contained in the preprocessed images.The recognized text information was described as a fixed length byte sequence,and the statistical characteristics of the file were obtained,and the sign characteristics of the abnormal file in the information database were compared,and the intelligent detection results of the abnormal file were output.The results show that the F-Score value of the proposed intelligent detection method for abnormal files reaches more than 0.9,which proves that the method has good performance.It can efficiently and accurately detect abnormal files in a fragmented spatiotemporal information database,reduce staff fatigue,and improve detection accuracy and efficiency.

关 键 词：OCR识别技术 碎片化 信息库 异常文件 智能检测 统计特征 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]