基于数据颜色特殊性的加分表格识别系统研究

发布时间：2018-04-26 01:10

本文选题：表格识别 + 单元格　；参考：《昆明理工大学》2017年硕士论文

【摘要】：目前,许多的手写表格信息更需要录入、存储在计算机中。在教育系统中,处理试卷加分表格过程中,存在着以下问题:(1)不恰当的使用人力资源,造成人力资源的浪费。教育系统内的工作人员很大部分是高学历、高素质的尖端科研人员,却在做着这种单调而重复的工作,无疑对社会及国家造成了大量的人力资源的浪费。(2)大量的小题加分计算会造成人员的疲劳、计算的失误、效率低下。长时间的批改试卷已经产生了精神上和生理上的疲劳,这种情况下再进行加分计算容易产生失误且效率较低。本文考虑应用表格图像识别技术来解决教育工作者在试卷表格的加分过程中存在的问题。为了解决以上存在的问题,本文考虑从两个方面进行表格图像的识别。一方面是表格线单元格的检测,一方面是表格内数据识别。首先,查阅大量的相关文献,了解了表格识别相关技术,分析了表格识别相关技术的国内外的研究现状,并进行了评述。其次,本文以试卷加分表格为识别对象,针对该表格进行了特征分析,确定了根据数据颜色的特殊性,进行表格识别的基本思路。第三,本文应用预处理对图像进行预处理,应用直线检测的方法进行表格线的检测,获得表格单元格位置情况。第四,根据颜色特殊性直接对图像中表格内数据进行提取,实现数据图片的切分并获得的单个数字图片。由于数据提取的特殊性,数据间的边界模糊、区分不明确。本文提出了数字区域扩散边界界定法,实现每个单元格数据的切分。利用滴水算法将两位数数据字符切分,实现单个数字图片的获得。应用OCR技术对单数字进行识别并计算单元格数据。最后,根据单元格位置情况,提出识别数据与单元格排序关联算法对识别的表格数据按顺存储。本文利用Microsoft Visual Studio开发平台,利用EmguCV图片处理库,实现试卷加分表格识别系统原型的开发,并计算出总成绩。验证出本文提出的针对具有特殊颜色表格识别的相关方法具有一定的可行性;本系统的开发具有一定的使用价值。
[Abstract]:At present, many handwritten form information needs to be input, stored in the computer. In the educational system, in the process of processing the test papers and adding the score forms, there are the following problems: 1) improper use of human resources, which results in the waste of human resources. Most of the staff in the education system are highly educated and highly qualified cutting-edge researchers, but they are doing this monotonous and repetitive work. There is no doubt that a large amount of human resources is wasted to the society and the country. (2) A large number of small problems and points will cause fatigue, error in calculation and inefficiency. A long period of marking has resulted in mental and physical fatigue. In this case, the calculation of bonus points is prone to errors and low efficiency. In this paper, we consider the application of table image recognition technology to solve the problems that educators have in the process of adding scores to test papers. In order to solve the above problems, this paper considers two aspects of table image recognition. On the one hand, it is the detection of table line cells, and on the other hand, it is the data recognition in the table. Firstly, we refer to a large number of relevant literature, understand the related technology of table recognition, and analyze the current research situation of the related technology at home and abroad, and make a review. Secondly, this paper takes the test paper plus score table as the recognition object, carries on the characteristic analysis to the table, and determines the basic idea of the table recognition according to the particularity of the data color. Thirdly, the preprocessing method is used to preprocess the image, and the method of line detection is used to detect the table line, and the position of the table cell is obtained. Fourthly, the data in the table is extracted directly according to the color particularity, the segmentation of the data picture is realized and a single digital picture is obtained. Because of the particularity of data extraction, the boundary between data is fuzzy and the distinction is not clear. In this paper, a digital region diffusion boundary demarcation method is proposed to segment each cell data. Two-digit data characters are segmented by drip algorithm to obtain a single digital picture. OCR technology is used to identify single digits and calculate cell data. Finally, according to the location of the cells, the algorithm of associating the recognition data with the sorting of the cells is proposed to store the identified table data in sequence. This paper uses Microsoft Visual Studio development platform and EmguCV image processing library to realize the prototype of test paper plus score table recognition system, and calculate the total results. It is verified that the method proposed in this paper is feasible for the recognition of special color tables and the development of the system has some practical value.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】