视频序列中的笑脸识别技术分析-analysis of smiley face recognition technology in video sequences.docxVIP

  • 6
  • 0
  • 约4.45万字
  • 约 64页
  • 2018-08-01 发布于上海
  • 举报

视频序列中的笑脸识别技术分析-analysis of smiley face recognition technology in video sequences.docx

视频序列中的笑脸识别技术分析-analysis of smiley face recognition technology in video sequences

摘要随着模式识别、计算机视觉技术的发展,以及移动多媒体商业中应用需求的推动,笑脸识别技术得到了越来越多研究人员的重视。在笑脸识别的整个流程中,对笑脸特征的提取至关重要,它是后续进行分类识别的唯一依据。目前的特征提取技术从应用场合的角度,主要分为针对单张图像和针对视频序列两种。针对单张图像的特征提取虽然算法众多,但是单张图像本身的信息量是有限的;而针对视频序列的算法,目前只是致力于获取序列中图像之间的变化情况。所以,将序列中的变化信息与单张图片的特征结合起来,获取更加丰富且有效的笑脸特征,是一个非常有意义的研究方向。本文重点针对视频环境下笑脸识别系统中的特征提取技术进行研究,力图通过特征的提取来提高笑脸识别系统的准确率。本文首先分析了现有的笑脸识别相关的特征提取和分类识别技术理论;其次,在光流法和PHOG算法的基础上,提出了Optical_PHOG特征提取技术,该方法将人脸区域的光流特征和嘴巴区域的PHOG特征,使用串接系数进行合理串接,从而形成Optical_PHOG特征向量;同时引入随机森林分类识别算法,对识别结果进行统计,验证了Optical_PHOG特征能够更好地区分笑脸与非笑脸特征之间的差别;最后,结合本文研究的算法,搭建了一个视频序列环境下的笑脸识别系统,通过摄像头采集视频序列,实现了对视频序列中的人脸图像进行实时的笑脸识别。关键词:光流金字塔梯度方向直方图笑脸识别视频环境AbstractWiththerapidadvancementofPatternRecognitionandComputerVisionaswellasthepromotionbroughtbythemobilemultimediacommercial,SmileRecognitionearnmoreattentionbyanincreasingnumberofresearchers.Theextractionofsmilefeature,whichistheonlyevidenceofsubsequentclassificationwork,playsacrucialroleinthewholeprocessofSmileRecognition.Fromthepointofviewofapplications,thecurrentfeatureextractiontechnologyismainlydividedintotwoparts:featureextractionforsingleimageandforthevideosequence.Althoughthealgorithmsforsingleimagearenumerous,thereexistslimitationintheinformationcontentofsingleimage.Thealgorithmsforvideosequencesnowareonlyusedforgettingthechangesbetweentheimagesinasequence.Therefore,combiningthevariableinformationinthevideosequenceandthefeaturesinsingleimage,togetmoreabundantandeffectivesmilefeaturebecomesasignificantresearchdirection.ThisessayfocusesonthefeatureextractiontechnologyofSmileRecognitionSystemforvideosequences,tryingtoimprovetherecognitionaccuracybyfeatureextraction.Firstly,Theexistingfeatureextractiontechnologyandrecognitiontheoryrelatedtosmilerecognitionareanalyzed.Then,afeatureextractionalgorithmcalledOptical_PHOGisproposedwhichisbasedonOpticalFlowandPHOGalgorithm.ThisalgorithmexecutereasonableconcatenationsofopticalflowfeatureinfaceareaandPHOGfeaturearoundmouthtoformaOptical_PHOGfeaturevector;Meanwhile,itbringsinRandomForestalgorithm,provingthatOptical_PHOGfeaturec

您可能关注的文档

文档评论(0)

1亿VIP精品文档

相关文档