{"title":"DEEP LEARNING BASED RACING BIB NUMBER DETECTION AND RECOGNITION","authors":"Y. Wong","doi":"10.5455/jjcit.71-1562747728","DOIUrl":null,"url":null,"abstract":"","PeriodicalId":36757,"journal":{"name":"Jordanian Journal of Computers and Information Technology","volume":"1 1","pages":""},"PeriodicalIF":0.9000,"publicationDate":"2019-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"8","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Jordanian Journal of Computers and Information Technology","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.5455/jjcit.71-1562747728","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q4","JCRName":"COMPUTER SCIENCE, INFORMATION SYSTEMS","Score":null,"Total":0}
健康的生活方式趋势在世界范围内变得越来越突出。举办了无数的马拉松赛跑活动,激发了不同年龄、性别和国家的人们的兴趣。这种多样化的真相增加了理解大量马拉松式图像的难度,因为这一过程通常是人工完成的。为此,本文提出了一种基于深度学习的马拉松比赛号码布定位与识别新方法。以前,所有的RBN应用系统都是通过使用图像处理技术开发的,这限制了性能的实现。在该系统中有两个阶段:阶段1:RBN检测和阶段2:RBN识别。在阶段1中,You Only Look Once version 3 (YOLOv3)由单个卷积网络组成,用于通过多个边界框和这些框的分类概率来预测跑步者和RBN。YOLOv3是一种新的分类器网络,性能优于其他最先进的网络。在第二阶段,使用卷积递归神经网络(CRNN)对每个输入图像生成一个标签序列,然后选择概率最高的标签序列。CRNN可以直接从序列标签(如单词)中训练,而不需要任何字符注释。因此,CRNN对检测到的RBN内容进行识别。对基于平均精度和编辑距离的实验结果进行了分析。所开发的系统适用于马拉松或长跑赛事,实现了选手的自动定位和识别,从而提高了赛事控制和监控以及赛事数据后处理的效率。