[1]孟彩霞,魏荣娟.移动端体检报告影像识别及数据分析应用研究[J].计算机技术与发展,2018,28(03):187-191.[doi:10.3969/ j. issn.1673-629X.2018.03.040]
　MENG Cai-xia,WEI Rong-juan.Research on Application of Image Recognition and Data Analysis of Physical Examination Reports in Mobile Terminal[J].,2018,28(03):187-191.[doi:10.3969/ j. issn.1673-629X.2018.03.040]
点击复制

移动端体检报告影像识别及数据分析应用研究()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
28
期数:
2018年03期
页码:
187-191
栏目:
应用开发研究
出版日期:
2018-03-10

文章信息/Info

Title:
Research on Application of Image Recognition and Data Analysis of Physical Examination Reports in Mobile Terminal
文章编号:
1673-629X(2018)03-0187-05
作者:
孟彩霞; 魏荣娟
西安邮电大学 计算机学院,陕西 西安 710000
Author(s):
MENG Cai-xia; WEI Rong-juan
School of Computer Science,Xi’an University of Posts and Telecommunications,Xi’an 710000,China
关键词:
移动应用开发; 影像文字识别; 字符分割; 数据可视化
Keywords:
mobile application development; image character recognition; character segmentation; data visualization
分类号:
TP301
DOI:
10.3969/ j. issn.1673-629X.2018.03.040
文献标志码:
A
摘要:
提出了移动端体检报告影像识别及数据分析方法。 该方法通过移动端设备拍照的方式获取用户的体检报告图像,采用图像预处理、字符切分和识别等智能图像处理技术以及医学符号分析、数据库和数据可视化等技术,设计与实现了基于移动端的个人医疗健康管理原型系统。 在对移动端体检报告影像识别方法的研究中,采用平均法进行彩色图像灰度化,利用中值滤波法去噪以及最大类间方差法实现二值化,改进了表格线检测及消除算法;字符切分法在传统字符切割前增加了字段切割,利用投影法加边界阈值判断法实现字段切割,对字符段再次利用投影法并结合基于识别的切分方法实现单字符分割;采用三次卷积插值算法进行归一化,最后利用模板匹配法完成识别。 在对体检数据分析及可视化方法的研究中,受限于实验数据资源的不足,主要对识别结果的个体历史体检数据进行了简单对比分析,画出可视化图,最后结合实验结果给出健康建议。初步的实验结果证实该方法可以有效地提取体检结果文字信息,并将体检信息以图形化的方式呈现给用户,便于用户查看自己的健康数据变化,及时了解自身健康状况。
Abstract:
A method of image identification and data analysis for physical examination report on mobile terminal is proposed,which acquires the physical examination report image of users by taking pictures of mobile terminal device,and uses the intelligent image processing technologies like image preprocessing,character segmentation and recognition,as well as others like medical symbol analysis,database and data visualization for design and implementation of a prototype system for personal health management based on mobile terminal. In the study of the image identification methods for physical examination reports,we improve a table line detection and elimination algorithm through the mean method to transform the color image to gray one,denoising by the median filtering and binaryzation according to Otsu algorithm. Before traditional character segmentation,the character fields segmentation implemented by methods of projection and boundary threshold is added in character segmentation,and the single character segmentation is achieved by projection algorithm again in combination with segmentation al-
gorithm based on recognition. We adopt the three convolution interpolation algorithm for normalization. Finally,the template matching is used to complete the recognition. In the study of physical data analysis and visualization,due to the limitation of finite data resources,we mainly analyze the individual historical physical examination data of the identified results and draw their visual maps. Finally the health advice is given by experiment. The preliminary results show that the proposed method can effectively extract the examination results of text in formation,and will present medical information graphically to the users,which is convenient for them to understand their own health data changes and know their own health status in time.
更新日期/Last Update: 2018-05-15