计算机视觉+人体姿态识别+双目视觉.docx
- 文档编号:1952048
- 上传时间:2022-10-25
- 格式:DOCX
- 页数:15
- 大小:768.15KB
计算机视觉+人体姿态识别+双目视觉.docx
《计算机视觉+人体姿态识别+双目视觉.docx》由会员分享,可在线阅读,更多相关《计算机视觉+人体姿态识别+双目视觉.docx(15页珍藏版)》请在冰豆网上搜索。
计算机视觉+人体姿态识别+双目视觉
Computervisionapplication
院(系)电子与信息工程学院
专业集成电路工程
学生吕广兴14S158054
Computervisionapplication
Thedirectory
Report:
Computervisionapplication2
1.Theobjectoftheproject2
2.Themethodandtheprincipleappliedtotheproject2
2.1Platform2
2.2TheprincipleoftransformtheRGBimagetothegrayimage2
2.3Theprincipleofimageenhancement2
2.4Theprincipleofthresholding3
2.5Theprincipleofclassifier3
3.Thecontentandtheresultoftheproject4
3.1Themainstepsintheproject4
3.2Abouthumanbodyposturerecognition4
Aboutthreekindsofmethodsaremostcommon:
4
3.3.Stereovision11
4.Reference17
Report:
Computervisionapplication
1.Theobjectoftheproject
TheobjectoftheprojectisGesturerecognitionandlocationintheinteriorofpeople.
2.Themethodandtheprincipleappliedtotheproject
2.1Platform
TheplatformisbasedonVisualStudio2012andOpenCV2.4.10.
2.2TheprincipleoftransformtheRGBimagetothegrayimage
TherearethreemajormethodstotransformtheRGBimagetothegrayimage.
ThefirstoneiscalledthemaximumvaluethatissetthevalueofR,G,andBtothemaximumofthesethree.
Gray=R=G=B=max(R,G,B)
ThesecondoneiscalledmeanvaluewhichissetthevalueofR,G,andBtothemeanvalueofthesethree.
Gray=R=G=B=(R+G+B)/3
ThethirdoneiscalledweightedaveragethatisgivingdifferentweightstotheR,GandBaccordingtotheimportanceorotherindicators,andthenaddingthethreepartstogether.Infact,human’seyeisveryhighsensitivetogreen,thenred,lastblue.Gray=0.30R+0.59G+0.11B
2.3Theprincipleofimageenhancement
Imageenhancementistheprocessofmakingimagesmoreuseful.Therearetwobroadcategoriesofimageenhancementtechniques.Thefirstoneisspatialdomaintechnique,anditisadirectmanipulationofimagepixelsthatincludespointprocessingandneighborhoodoperations.Thesecondoneisfrequencydomaintechnique,anditisamanipulationofFouriertransformorwavelettransformofanimage.
Theprincipleofthemedianfilteristoreplacethevalueofapixelbythemedianofthegraylevelsintheneighborhoodofthatpixel(theoriginalvalueofthepixelisincludedinthecomputationofthemedian).Itforcesthepointswithdistinctgraylevelstobemoreliketheirneighbors.
Inaddition,wealsoapplythemorphologicalimageprocessingaftersmoothing.Morphologicalimageprocessing(ormorphology)describesarangeofimageprocessingtechniquesthatdealwiththeshape(ormorphology)offeaturesinanimage.ThebasicidealofMorphologyistouseaspecialstructuringelementtomeasureorextractthecorrespondingshapeorcharacteristicsintheinputimagesforfurtherimageanalysisandobjectrecognition.Themathematicalfoundationofmorphologyisthesettheory.Therearetwobasicmorphologicaloperations:
erosionanddilation.
2.4Theprincipleofthresholding
Thresholdingisparticularlyusefulforsegmentationinwhichwewanttoisolateanobjectofinterestfromabackground.Atthesametime,thresholdingsegmentationisusuallythefirststepinanysegmentationapproach.Theblowformulaisthebasicprincipleofimagesegmentation.Whenthegraylevelisnobiggerthanthethreshold,wewillsetthepixelvaluezero(black).Incontrast,whenthegraylevelisbiggerthanthethreshold,wewillsetthepixelvalue255(white).
Whenitcomestothethreshold,wegetthevaluethroughtheimagehistograms
2.5Theprincipleofclassifier
Theclassifierisaalgorithmordevicethatseparatesobjectsintodifferentclasses.Usually,aclassifierconsistsofthreeparts.Firstoneisthesensor,forinstance,imagingdevice,fingerprintreader,etc...Secondoneisfeatureextractor,forexample,edgedetectororpropertydescriptor.ThirdoneisclassifierwhichusestheextractedfeaturesfordecisionmakingorEuclidiandistance,orothermethods.
Featuresshouldcanberegardedasthedescriptorsweintroducedbefore.Andthefeatureshouldberepresentativeandusefulforclassification.
Whenitcomestothefeaturespace,thesetofallpossiblepatternsformthefeaturevector.Eachfeaturevectorisapointintheso-calledfeaturespace.Similarobjectsyieldsimilarmeasurementresults.Nearbypointsinfeaturespacecorrespondtosimilarobjects.Distanceinfeaturespaceisrelatedtosimilarity.Pointsthatbelongtothesameclassformacloudinfeaturespace.
Dividethedatasetintoatrainingsetandatestset.Theperformanceofaclassifiershouldbeassessedbytheclassificationerroronanindependenttestset.Thissetshouldnotcontainobjectsthatareincludedinthetrainingset.Determineadecisionboundarybyminimizingtheclassificationerrorofthetrainingset.Determinetheclassifierperformancebycomputingtheclassificationerrorofthet
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- 计算机 视觉 人体 姿态 识别 双目