人體檢測中若干關(guān)鍵技術(shù)研究.pdf

1、電子科技大學(xué)碩士學(xué)位論文人體檢測中若干關(guān)鍵技術(shù)研究姓名：張睿申請學(xué)位級別：碩士專業(yè)：通信與信息系統(tǒng)指導(dǎo)教師：皮亦鳴20060520ABSTRACTABSTRACTObjectdetectionisoneofthehoaesttopicsincomputervisionareasItisanobviousdenotationforcomputerintelligenceaswellThisdissertationfocusestheres

2、earchonhumandetection，whichisanextraexampleofobjectdetectionHumandetectionisthekeypointforautomaticrecognitionofhumanactivitiesandimageunderstandingIthasawideuseinsurvenillancesystemandhumancomputerinteractionHumanbodyis

3、non—rigidbodyItischallengingduetotheappearancevariation(colortexture)，thepresenceofposeambiguity，andthedifferentviewsofhumanbodyinanimageThereforeitisarguablythemostdifficultyproblemincomputervisionThisdissertationstudie

4、sthecuttingedgemethodsforhumandetectionComparedwithourresearch，theprobabilityoptimizationmethodsareexploredandimproved，whichisthenappliedtothehumandetectionproblemindissertationTheresultsshowthatthismethodispromisingThem

5、ainworksofthisdissertationareasfollow：1ThebasicmethodsofobjectdetectionarestudiedEspeciallyfeatureextractionandstatisticallearningareincludedThekeyproblemisalsoexplored2Thestate—ofarthumanbodydetectionmethodsarestudiedTh

6、efocusisthechallengeduetothehumannon—rigidbodyaswellascomplicatedappearanceandgesture3BasedontheupperresearchandourspecificproblemAccordingtothevariousappreance，gestureandviewangleofhumanbody，thepracticabilityofuppermeth

7、odsisanalyedandtheirdrawbacksarepointedoutAndtheprobabilityoptimizationmethodsareintroducedintoourdetectionprocess，weinterplayedbetweenlowlevelfeaturesandhighlevelmode，constructedfeaturespaceandhumanmodel，andfinallyuseMC