特征描述算子评测.doc

特征描述算子评测.doc

ID:57648715

大小:336.00 KB

页数:5页

时间:2020-08-30

特征描述算子评测.doc_第1页
特征描述算子评测.doc_第2页
特征描述算子评测.doc_第3页
特征描述算子评测.doc_第4页
特征描述算子评测.doc_第5页
资源描述:

《特征描述算子评测.doc》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库

1、FeaturedescriptorcomparisonreportSharingmyresearchworkofbehaviorofseveraltypesoffeaturedescriptors.Thisarticleisanupdateofold"Comparisonoffeaturedescriptors"post.I'veaddedabrandnewORBfeaturedescriptortothetestsuite,alsoSIFTdescriptorincludedaswell.Andanewversiono

2、fLAZYdescriptorpresentinthistesttoo.Forthistestihavewrittenspecialtestframework,whichallowsmetoeasilyaddthenewkindofdescriptorsandtestcasesandgeneratereportdatainCSV-likeformat.ThaniuploaditinGoogledocsandcreatethisawesomecharts.Fivequalityandoneperformancetestwa

3、sdoneforeachkindofdescriptor.oRotationtest- thistestshowshowthefeaturedescriptordependsonfeatureorientation.oScalingtest- thistestshowshowthefeaturedescriptordependsonfeaturesize.oBlurtest- thistestshowshowthefeaturedescriptorisrobustagainstblur.oLightingtest- th

4、istestshowshowthefeaturedescriptorisrobustagainstlighting.oPatterndetectiontest-thistestperformsdetectionofplanar平面object(image)ontherealvideo.Incontrasttothesynthetic综合tests,thistestgivesarealpictureoftheoverallstabilityoftheparticulardescriptor.oPerformancetest

5、isameasurementofdescriptionextractiontime.Allqualitytestsworksinsimilarway.Usingagivensourceimagewegenerateasynthetictestdata:transformedimagescorrespondingfeaturepoints.Thetransformationalgorithmdependsontheparticulartest.Fortherotationtestcase,it'stherotationof

6、thesourceimagearoundit'scenterfor360degrees,forscaling-it'sresizingofimagefrom0.25Xto2xsizeoforiginal.Blurtestusesgaussianblurwithseveralstepsandthelightingtestchangestheoverallpicturebrightness.Thepatterndetectiontestdeservesaspecialattention.Thistestisdoneonver

7、ycomplexandnoisy videosequence.Soit'schallengingtaskforanyfeaturedescriptoralgorithmtodemonstrateagoodresultsinthistest.Themetricforallqualitytestsisthepercentofcorrectmatchesbetweenthesourceimageandthetransformedone.Sinceweuseplanarobject,wecaneasilyselecttheinl

8、iersfromallmatchesusingthehomographyestimation.IuseOpenCV'sfunctioncvFindHomographyforthis.Thismetricgivesverygoodandstableresults.Idonooutlierdetectionofmatch

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。