欢迎来到天天文库
浏览记录
ID:37348751
大小:4.85 MB
页数:67页
时间:2019-05-22
《图像融合在地面目标检测及特征提取中的应用研究》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、国防科学技术大学硕士学位论文图像融合在地面目标检测及特征提取中的应用研究姓名:罗有敢申请学位级别:硕士专业:信息与通信工程指导教师:庄钊文20091101国防科学技术大学研究生院硕十学位论文ABSTRACTWiththerapiddevelopmentofsensortechnologyanditswideuseinmilitaryaffairs,remotesensingandindustry,themultisensordatafusionhasbecometheresearch’Shottopic.Asanimportan
2、tbranchofdatafusion,thepurposeofimagefusionistoreduceuncertaintyandtominimizeredundancyintheoutputwhilemaximizingusefulinformation.Thisdissertationfocusesongroundtargetdetectionandfeatureextractionbasedonimagefusion.Themajorworkandconclusionsareasfollows:1.Thecompari
3、sonaboutdifferentimagefusionalgorithmsarestudied.Thesealgorithmsareappliedinmulti-focusimagefusion,multi-spectralimagefusion,aswellasIRandvisibleimagefusion.Experimentalresultshowsthatmulti-resolutionanalysisismoreuniversalandreliableinimagefusiontoimproveimagedefini
4、tion,entropyandSNR.Finally,itisconcludedthatthewavelettransformwasthekeydirectiontobeinvestigated.2.Smallgroundtargetdetectionbasedonimagefusionisresearched.Itisdimculttodetectsmallgroundmovingtargetbysingleframe’imageduetoitslowSNRandbeinglackoftexturefeature.Subseq
5、uently,anapplicationsystembasedondifferentialimagefusionispresented,whichWasusedindualbandinfraredsmallgroundtargetdetectionundercomplexbackground.Thismethoddecomposedthedifferentialimageintomultistagelevelbywavelettransform.Thenthedetectionresultswereacquiredbyrecon
6、structeddifferentialimages.ExperimentalresultshowsthatthemethodcanbemoreefficienttodetectlowSNRmovingsmalltargetsundercomplexbackgroundandreducefalsealarm.3.Featureextractionbasedonregionimagefusionisresearched.Toke印thespatialresolutionandtexturedetailinformationofvi
7、sibleimagesandsensitivityofIRimage,astationarywaveletmethodbasedonROIisproposed.Forhi.曲-frequencypartinwavelettransformationdomain,weadoptthefusionruleofchoosingthemaxcoefficientvalue,whileforlow—frequencypart,adoptthefusionruleoflocalenergybasedonneighborhoodpixelgr
8、ay.ThenthefeatureextractionresultsareacquiredinthereconstructedROIimage.Experimentalresultshowsthatit’Smoreefficienttogettheobjectf
此文档下载收益归作者所有