基于图像的试验数据识别与管理系统

发布时间：2019-05-22 20:46

【摘要】：随着科研技术的快速发展,企业测量、测试能力不断提高,对产品的测试任务也不断增多,产生了大量的试验数据。近年来企业尝试采用建立数据管理系统完成对试验数据的采集和管理,但已完成的试验数据仍存储在纸质表格中,需要大量人力劳动将数据录入管理系统中。为了减少人力投入,更有效的提高工作效率,本课题研究了表格内容识别技术,设计了一种针对多种样式表格进行内容提取的算法;完成了基于图像的试验数据识别软件的设计,实现了试验数据的快速定位和准确识别;将试验数据识别软件作为试验数据采集终端,搭建了试验数据管理系统,对企业的流程管理与数据管理等方面做了研究。本文的主要工作如下:(1)研究了表格内容识别技术。表格内容识别需要排除横向、纵向表格线对于文本内容的干扰,实现字线分离,再通过定位单元格,提取其中内容并完成文字识别。(2)基于表格内容识别算法设计了一整套试验数据自动识别软件。对扫描得到的表格图片完成预处理,提高图像质量,完成版面分析,突出有用信息;提取出表格框线,定位所有单元格区域,切割出文字内容,完成字符识别;训练手写数字库,进一步提高手写字体的识别效率;人工后向纠错,完全修正识别结果。经过多次测试,该软件的自动识别正确率在90%以上,经过少量人工干预,完全能够满足企业级应用要求,有效的解决了自动录入的难题。(3)在PC终端搭建了实验数据管理系统,完成了安全性维护管理、试验任务管理、试验流程管理和数据库管理四个模块的设计与实现。研究了系统的体系架构和基于SOA的中间件技术,合理规划了试验管理流程,搭建了数据库模型,提高了管理试验数据的效率。
[Abstract]:With the rapid development of scientific research technology, the measurement and testing ability of enterprises is improving, and the testing tasks of products are also increasing, resulting in a large number of test data. In recent years, enterprises have tried to use the establishment of data management system to complete the collection and management of test data, but the completed test data are still stored in paper forms, which requires a lot of manpower to input the data into the management system. In order to reduce the manpower investment and improve the work efficiency more effectively, this paper studies the table content recognition technology, and designs a content extraction algorithm for a variety of style tables. The design of the test data recognition software based on image is completed, and the fast positioning and accurate recognition of the test data are realized. The test data recognition software is used as the test data acquisition terminal, and the test data management system is built, and the process management and data management of the enterprise are studied. The main work of this paper is as follows: (1) the technology of table content recognition is studied. Table content recognition needs to eliminate the interference of horizontal and longitudinal table lines to text content, realize the separation of word lines, and then through the positioning cell, Extract the content and complete the text recognition. (2) based on the table content recognition algorithm, a set of experimental data automatic recognition software is designed. The scanned table images are preprocessed, the image quality is improved, the layout analysis is completed, and the useful information is highlighted. The table frame lines are extracted, all the cell areas are located, the text content is cut out, and the character recognition is completed. The handwritten digital library is trained to further improve the recognition efficiency of handwritten fonts, and the manual backward error correction is carried out to completely correct the recognition results. After many tests, the automatic recognition accuracy of the software is more than 90%. After a small amount of manual intervention, the software can fully meet the requirements of enterprise applications. The problem of automatic input is solved effectively. (3) the experimental data management system is built in PC terminal, and the design and implementation of four modules: security maintenance management, test task management, test flow management and database management are completed. The system architecture and middleware technology based on SOA are studied, the test management flow is planned reasonably, the database model is built, and the efficiency of managing test data is improved.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41;TP315

【参考文献】