毕业论文外文翻译-基于视觉的矿井救援机器人场景识别

《毕业论文外文翻译-基于视觉的矿井救援机器人场景识别》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、附录英文原文ScenerecognitionforminerescuerobotlocalizationbasedonvisionAbstract：AnewscenerecognitionsystemwaspresentedbasedonfuzzylogicandhiddenMarkovmodel(HMM)thatcanbeappliedinminerescuerobotlocalizationduringemergencies.Thesystemusesmonocularcameratoacquireomni-dir

2、ectionalimagesofthemineenvironmentwheretherobotlocates.Byadoptingcenter-surrounddifferencemethod,thesalientlocalimageregionsareextractedfromtheimagesasnaturallandmarks.TheselandmarksareorganizedbyusingHMMtorepresentthescenewheretherobotis,andfuzzylogicstrategyis

3、usedtomatchthesceneandlandmark.Bythisway,thelocalizationproblem,whichisthescenerecognitionprobleminthesystem,canbeconvertedintotheevaluationproblemofHMM.Thecontributionsoftheseskillsmakethesystemhavetheabilitytodealwithchangesinscale,2Drotationandviewpoint.There

4、sultsofexperimentsalsoprovethatthesystemhashigherratioofrecognitionandlocalizationinbothstaticanddynamicmineenvironments.Keywords:robotlocation;scenerecognition;salientimage;matchingstrategy;fuzzylogic;hiddenMarkovmodel1IntroductionSearchandrescueindisasterareai

5、nthedomainofrobotisaburgeoningandchallengingsubject[1].Minerescuerobotwasdevelopedtoenterminesduringemergenciestolocatepossibleescaperoutesforthosetrappedinsideanddeterminewhetheritissafeforhumantoenterornot.Localizationisafundamentalprobleminthisfield.Localizat

6、ionmethodsbasedoncameracanbemainlyclassifiedintogeometric,topologicalorhybridones[2].Withitsfeasibilityandeffectiveness,scenerecognitionbecomesoneoftheimportanttechnologiesoftopologicallocalization.Currentlymostscenerecognitionmethodsarebasedonglobalimagefeature

7、sandhavetwodistinctstages:trainingofflineandmatchingonline.Duringthetrainingstage,robotcollectstheimagesoftheenvironmentwhereitworksandprocessestheimagestoextractglobalfeaturesthatrepresentthescene.Someapproacheswereusedtoanalyzethedata-setofimagedirectlyandsome

8、primaryfeatureswerefound,suchasthePCAmethod[3].However,thePCAmethodisnoteffectiveindistinguishingtheclassesoffeatures.Anothertypeofapproachusesappearancefeaturesinclu