欢迎来到天天文库
浏览记录
ID:55974473
大小:2.66 MB
页数:8页
时间:2020-03-24
《动态光照下驾驶人面部特征识别算法与试验研究.pdf》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31卷第10期2014年10月公路交通科技JournalofHighwayandTransportationResearchandDevelopmentV01.31NO.100ct.2014doi:10.3969/j.issn.1002—0268.2014.10.016动态光照下驾驶人面部特征识别算法与试验研究程文冬1’2,付锐1,张文霞1(1.长安大学汽车运输安全保障技术交通行业重点实验室,陕西西安710064;2.西安工业大学机电工程学院,陕西西安710032)摘要:基于机器视觉的驾驶人面部特征识别受光照的影响很大
2、。为克服由于动态光照引起的背景干扰,面部特征弱化的问题,采用一种基于KalmanFiltering的光照自适应AKF算法,通过高斯概率密度函数建立G,(i。7)算子,实现驾驶室背景的分割;在HSI色彩空间中通过阈值分割算法提取面部肤色区域,最终建立了眼鼻坐标搜索模型;进行了不同的照度与头部姿态下的AKF—HSI算法试验,测试统计前景分割率kfront一小肤色分割率k幽。与眼鼻识别率6,在2×104—10X104lx的照度下,眼鼻的平均识别率占达到82%一92%。结果表明AKF—HSI融合算法对动态光照下眼鼻-/7,另'1
3、具有较好的鲁棒性.照度E、头部姿态与硬件设备AGC是眼鼻识别的最主要影响因素。关键词:交通工程;眼鼻特征;机器视觉;AKF—HSI融合算法;照度中图分类号:U491.2+54文献标识码:A文章编号:1002—0268(2014)10—0097—07AlgorithmandExperimentalResearchofDriver’SFaceFeatureRecognitionunderDynamicIlluminationCHENGWen.don91”,FURuil,ZHANGWen.xial(1.KeyLaborator
4、yofAutomotiveTransportationSafetyTechnologyofMinistryofTransport,Chang'anUniversity,Xi’anShaanxi710064,China;2.SchoolofMechatronicEngineering,Xi’anTechnologicalUniversity,Xi’anShaanxi710032,China)Abstract:Therecognitionofdriver'sfacialfeaturesbasedonmachinevision
5、isinfluencedgreatlybyillumination.Inordertoovercomedynamicilluminationinducedbackgroundinterferenceandweakeningoffacialfeature,aself-adaptiveAKFalgorithmbasedonKalmanfilteringisadopted.Real—timesegmentationofdrivingcabbackgroundisrealizedbyGf(i,J)operatorestablis
6、hedbyGaussianprobabilitydensityfunction.ByclusteringperformanceofskincolorandfacialfeatureinHSIcolorspace,skinregionandfacialfeaturearerecognizedandlocatedbythethresholdsegmentationalgorithm.Finally,aneye—nostrilsearchmodelissetupbytheircoordinatevaluesequence.Th
7、eAKF—HSIalgorithmexperimentisthencarriedoutunderdifferentilluminanciesandheadposes.Frontgroundsegmentationratekf。。t£。。。d,skinsegmentationratek。k.nandeye-nostrilrecognitionrate6ofdifferentpersonsareobtained.whichshowsthattheaveragerecognitionrate6reachesupto82%一92
8、%withtheoutdoorilluminationof2×104—10X104Ix.Theresultshowsthattheanti·-interferenceperformanceofAKF--HSIfusionalgorithmisrobusttoeye—-nostrilrecognitionunderdy
此文档下载收益归作者所有