多模版匹配的特殊符号识别定位算法研究  

A SPECIAL SYMBOL RECOGNITION AND LOCATION ALGORITHM BASED ON MULTI-TEMPLATE MATCHING

作　　者：曹长玉 郑佳春[1] 黄一琦 Cao Changyu;Zheng Jiachun;Huang Yiqi(Xiamen Key Laboratory of Marine Intelligent IOT Terminal Development and Application,Jimei University,Xiamen 361021,Fujian,China)

机构地区：[1]集美大学厦门市海洋智能终端研发及应用重点实验室,福建厦门361021

出　　处：《计算机应用与软件》2021年第3期175-180,209,共7页Computer Applications and Software

基　　金：福建省科技计划重点项目(2017H0028);福建省自然科学基金项目(2013J01203,2015J01265)。

摘　　要：针对试卷智能批阅场景模式,由于Tesseract-OCR缺少特殊符号包,直接定位符号存在较多漏检等问题,提出具有覆盖保留机制的多模板匹配方法。通过OCR定位空白试卷中的符号分别建立多类型元素的方块、圆圈、括号模板集;而对于试卷中的直线,通过筛选查找轮廓的方法建立多类型元素的直线模板集,综合多模板匹配技术提高试卷中符号的识别性能及定位准确率。经实际试卷测试结果表明:该算法符号定位准确率、精确度和召回率均高于94%;查找轮廓法定位直线准确率达96%,模板匹配直线定位准确率、精确度和召回率高于87%;将空白试卷符号坐标应用于学生作答试卷,能较完美地定位手写答案。In the intelligent review scene mode of test papers,because Tesseract-OCR lacks the special symbol package,there are many missed detection problems such as missing symbols.A multi-template matching method with overlay retention mechanism is proposed.The square,circle and bracket template sets of multi-type elements were established by OCR locating the coincidence of blank test papers.For the straight line in the test paper,the line template set of multi-type elements was established by method of filtering the contour search,and then the integrated multi-template matching technology improved the recognition performance and positioning accuracy of the symbols in the test paper.The actual test paper test results show that the symbolic positioning accuracy,precision and recall rate are higher than 94%;the accuracy rate of line positioning by searching contour method is 96%,and the template matching line positioning accuracy,precision and recall rate are higher than 87%;the blank test paper symbol coordinates are applied to students answer paper,which can perfectly locate the handwritten answer.

关 键 词：智能批阅 模板匹配 符号 查找轮廓 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]