基于DBnet和众筹策略的气象纸质表格快速数字化方法及系统

A Rapid Digitization Method of Meteorological Paper Forms Based on DBnet and Crowdfunding Strategies

气象纸质表格历史资料有很高的科学研究价值和很重要的历史保留意义。数字档案通过数字化的方式将纸质原始资料进行收集、保存和提供各类数字信息资源。本文提出了一种“纸质表格扫描成像”+“图像碎片化处理”+“众筹识别”的快速数字化方法及系统,利用气象大数据资源,应用DBnet模型、DSCC算法等技术对纸质气象表格扫描图像进行图像碎片处理;通过用户的登录行为实现“众筹录入”,完成气象纸质表格资料的数字化档案工作。经验证,快速数字化系统的一录正确率约为99.7%,高于传统数字化人工键入的一录正确率(95.6%);录入时效比传统数字化提高了22.2%。该系统保证了数据正确性、提高了工作效率,实现了气象纸质表格资料快速形成数字档案的目标,也为数字档案工作提供了新的思路;在新冠疫情导致传统数字化人工聚集工作模式面临较大风险的背景下,具有一定的现实意义。

The historical data of meteorological paper forms have high scientific research value and important historical preservation significance. Digital archives are the collection, preservation and provision of various digital information resources of paper-based original materials through digital methods. This research proposes a rapid digitization method and system of “paper form scanning imaging”+“image fragmentation processing”+“crowdfunding recognition, ” using meteorological big data resources, DBnet model, DSCC algorithm and other technologies to process the scanned images of paper weather forms for image fragmentation. “Crowdfunding Entry” is realized through user login behaviour, and the digital file work of weather paper table data is completed. It has been verified that the accuracy rate of one-recording of the fast digital system is about 99.7%, which is higher than the accuracy rate of one-recording of traditional digital manual input(95.6%);the input time efficiency is improved by 22.2% compared with traditional digitalization. The system ensures data accuracy, improves work efficiency, achieves the rapid formation of digital archives of meteorological paper forms and data, and provides new ideas for digital archives. The COVID19 epidemic has caused difficulties in traditional digital manual aggregation work models. Under the background of big risks, it has certain practical significance.