欢迎来到天天文库
浏览记录
ID:36742421
大小:1.22 MB
页数:6页
时间:2019-05-14
《局部特征脸型分类方法》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、第12卷第1期智能系统学报Vol.12№.12017年2月CAAITransactionsonIntelligentSystemsFeb.2017DOI:10.11992/tis.201605021网络出版地址:http://kns.cnki.net/kcms/detail/23.1538.TP.20170227.1805.016.html局部特征脸型分类方法孙劲光,邓智硕(辽宁工程技术大学电子与信息工程学院,辽宁葫芦岛125105)摘要:本文针对传统脸型分类算法特征点定位不准和过度依赖轮廓曲线的问题,提出了一种人脸轮廓圆形邻域局部特征表达方式和脸型分类模型。首先,初步定位脸型轮廓特征点;然后
2、,在特征点周围选取三重八连通圆形邻域,通过计算一级邻域、拓展邻域与中心区域间的纹理变化,生成二进制编码序列,构造脸型局部特征向量;最后,设计OVO⁃RBF⁃SVM多分类模型,实现脸型分类。本文方法在CAS⁃PEAL人脸库上进行脸型类型判别,获得了94?28%的准确率;在相同情况下,分别与基于主动形状模型和基于下颌曲线模型的脸型类型判别方法进行对比,准确率分别提高了6?64%和6?58%。本文所研究的方法在一定程度上解决了特征点定位相对不准确导致误差增加的问题,同时尽可能多利用图片原始信息,保证轮廓特征提取的准确率,具有较强的鲁棒性。通过实验证明本文方法适用于脸型分类。关键词:脸型分类;圆形邻域
3、;特征编码;局部特征表达;多分类支持向量机中图分类号:TP391.41文献标志码:A文章编号:1673-4785(2017)01-0104-06中文引用格式:孙劲光,邓智硕.局部特征脸型分类方法[J].智能系统学报,2017,12(1):104-109.2英文引用格式:SUNJinguang,DENGZhishuo.Localfeaturefacialclassificationmethod[J].CAAItransactionsonintelligentsystems,2017,12(1):104-109.Localfeaturefacialclassificationmethod12SUN
4、Jinguang,DENGZhishuo(SchoolofElectronicandInformationEngineering,LiaoningTechnicalUniversity,Huludao125105,China)Abstract:Consideringtheproblemswherethefeaturepointsoftraditionalfacialclassificationalgorithmsarenotlocatedinthepositionoftheactualfeaturepointsandareheavilydependentuponthecontourcurve,
5、afacialcontourcircularneighborhoodlocalfeatureexpressionandafacialclassificationmodelwereproposed.First,thepreliminaryfacialcontourfeaturepointswerelocatedandthenaroundthefeaturepoints,thetripleeightconnectedround⁃neighborhoodwasselected.Bycalculatinganeighborhoodlevelandexpandingtheneighborhoodwith
6、thecentralareabetweenthetexturechanges,thebinarycodesequencewasgeneratedandthetectonicfaciallocalfeaturevectorscanbecreated.Then,thefaceswereclassifiedbydesigningtheOVO⁃RBF⁃SVMclassificationmodel.TheexperimentwasconductedontheCAS⁃PEALfacelibraryforfacialcontourfeaturediscrimination,achieving94.28%ac
7、curacyrate;underthesamecircumstances,theface⁃typediscriminationmethodswhicharebasedontheactiveshapemodelandjawcurvemodelwerecompared,andtheaccuracyrateraised6.64%and6?58%,respectively.Toacertainextent
此文档下载收益归作者所有