高光谱成像.docx
- 文档编号:25639825
- 上传时间:2023-06-11
- 格式:DOCX
- 页数:19
- 大小:626.82KB
高光谱成像.docx
《高光谱成像.docx》由会员分享,可在线阅读,更多相关《高光谱成像.docx(19页珍藏版)》请在冰豆网上搜索。
高光谱成像
Hyperspectralimaging
FromWikipedia,thefreeencyclopedia
Two-dimensionalprojectionofahyperspectralcube
Hyperspectralimaging,likeotherspectraljmaging_,collectsandprocessesinformation
fromacrosstheelectromagneticspectrum_.Thegoalofhyperspectralimagingistoobtainthespectrumforeachpixelintheimageofascene,withthepurposeoffindingobjects,identifyingmaterials,ordetectingprocesses.皿
Thehuman_eye_seescolorofvisible_light_inmostlythree_bands_(red,green,andblue),spectral_imaging_dividesthespectrumintomanymorebands.Thistechniqueofdividingimagesintobandscanbeextendedbeyondthevisible.Inhyperspectralimaging,therecordedspectrahavefinewavelengthresolutionandcoverawiderangeofwavelengths.
Engineersbuildhyperspectralsensorsandprocessingsystemsforapplicationsinastronomy,agriculture,biomedicalimaging,geosciences,physics,andsurveillanee.Hyperspectralsensorslookatobjectsusingavastportionoftheelectromagneticspectrum.Certainobjectsleaveunique'fingerprints'intheelectromagneticspectrum.Knownasspectralsignatures,these'fingerprints'enableidentificationofthematerialsthatmakeupascannedobject.Forexample,aspectral_signature_foroilhelpsgeologistsfind
newoilfields単
Contents
[hide]
・1Hyperspectralimagesensors
«2Technologiesforhyperspectraldataacquisition
oooo
2.1Spatialscanning
2.2Spectralscanning
2.3Non-scanning
2.4Spatiospectralscanning
«3Distinguishinghyperspectralfrommultispectralimaging
«4Applications
o
4.1Agriculture
o
4.2Eyecare
o
4.3Foodprocessing
o
4.4Mineralogy
o
4.5Surveillanee
o
4.6Physics
o
4.7Astronomy
o
4.8Chemicalimaging
o
4.9Environment
*5Advantagesanddisadvantages
*6Softwareresources
*7Seealso
*8References
*9Externallinks
Hyperspectralimagesensors[edit]
Figurativelyspeaking,hyperspectralsensorscollectinformationasasetof'images'.Eachimagerepresentsanarrowwavelengthrangeoftheelectromagneticspectrum,alsoknownasaspectralband.These'images'arecombinedtoformathree-dimensional(x,y,Nhyperspectraldata_cube_forprocessingandanalysis,wherexandyrepresenttwospatialdimensionsofthescene,and入representsthespectraldimension(comprisingarangeofwavelengths).已
Technicallyspeaking,therearefourwaysforsensorstosamplethehyperspectralcube:
Spatialscanning,spectralscanning,snapshotimaging,andspatio-spectralscanning.
HyperspectralcubesaregeneratedfromairbornesensorsliketheNASA'sAirborne_
Visible/lnfraredImagingSpectrometer(AVIRIS),orfromsatelliteslikeNASA'sEO-1with
itshyperspectralinstrumentHyperion.也However,formanydevelopmentandvalidation
studies,handheldsensorsareused.困
Theprecisionofthesesensorsistypicallymeasuredinspectralresolution,whichisthewidthofeachbandofthespectrumthatiscaptured.Ifthescannerdetectsalargenumberoffairlynarrowfrequencybands,itispossibletoidentifyobjectseveniftheyareonlycapturedinahandfulofpixels.However,spatial_resolution_isafactorinadditiontospectral
resolution.Ifthepixelsaretoolarge,thenmultipleobjectsarecapturedinthesamepixelandbecomedifficulttoidentify.Ifthepixelsaretoosmall,thentheenergycapturedbyeachsensorcellislow,andthedecreasedsignal-to-noise_ratio_reducesthereliabilityof
measuredfeatures.
Theacquisitionandprocessingofhyperspectralimagesisalsoreferredtoasimaging_
spectroscopy_or,withreferencetothehyperspectralcube,as3Dspectroscopy.
Technologiesforhyperspectraldataacquisition[edit]
Spatialscanning
HSI
scanning
Non-scanning
431
Spatio-spectralscanning
Photosillustratingindividualsensoroutputsforthefourhyperspectralimagingtechniques.
right:
Slitspectrum;monochromaticspatialmap;'perspectiveprojection'ofhyperspectralcube;
Fromleftto
wavelength-codedspatialmap.
Non-scanningSpaliQ-Spcctral
(Snapshot|Scanning
Acquisitiontechniquesforhyperspectralimaging,visualizedassectionsofthehyperspectraldatacube
withitstwospatialdimensions(x,y)andonespectraldimension(lambda).
Therearefourbasictechniquesforacquiringthethree-dimensional(x,y,?
)datasetofa
hyperspectralcube.Thechoiceoftechniquedependsonthespecificapplication,seeingthateachtechniquehascontext-dependentadvantagesanddisadvantages.
Spatialseanning[edit]
Inspatialseanning,eachtwo-dimensional(2-D)sensoroutputrepresentsafullslitspectrum(x,0Hyperspectralimaging(HSI)devicesforspatialscanningobtainslitspectrabyprojectingastripofthesceneontoaslitanddispersingtheslitimagewithaprismoragrating.Thesesystemshavethedrawbackofhavingtheimageanalyzedperlines(withapush_broom_scanner_)andalsohavingsomemechanicalpartsintegratedintotheopticaltrain.Withtheseline-scansystems,thespatialdimensioniscollectedthrough
platformmovementorscanning.Thisrequiresstabilizedmountsoraccuratepointinginformationto,reconstruct?
theimage.Nonetheless,line-scansystemsareparticularly
commoninremotesensing,whereitissensibletousemobileplatforms.Line-scansystemsarealsousedtoscanmaterialsmovingbyonaconveyorbelt.Aspecialcaseoflinescanningispointscanning(withawhisk_broom_scanner_),whereapoint-like
apertureisusedinsteadofaslit,andthesensorisessentiallyone-dimensionalinsteadof2_d[5][9][10]
Spectralscanning[edit]
Inspectralscanning,each2-Dsensoroutputrepresentsamonochromatic
('single-colored'),spatial(x,y)mapofthescene.HSIdevicesforspectralscanningaretypicallybasedonopticalband-passfilters(eithertuneableorfixed).Thesceneisspectrallyscannedbyexchangingonefilterafteranotherwhiletheplatformmustbestationary.Insuch'staring',wavelengthscanningsystems,spectralsmearingcanoccurifthereismovementwithinthescene,invalidatingspectralcorrelation/detection.
Nonetheless,thereistheadvantageofbeingabletopickandchoosespectralbands,andhavingadirectrepresentationofthetwospatialdimensionsofthescene.
Non-scanning[edit]
Mainarticle:
Snapshot_hyperspectral_imaging
Innon-scanning,asingle2-Dsensoroutputcontainsallspatial(x,y)andspectral(0data.
HSIdevicesfornon-scanningyieldthefulldatacubeatonce,withoutanyscanning.
Figurativelyspeaking,asinglesnapshotrepresentsaperspectiveprojectionofthedatacube,fromwhichitsthree-dimensionalstructurecanbereconstructed.[5][11]Themost
prominentbenefitsofthesesnapshot_hyperspectral_imagingsystemsarethesnapshot
advantage(higherlightthroughput)andshorteracquisitiontime.Anumberofsystemshavebeendesigned,includingcomputedtomographicimagingspectrometry(CTIS),fiber-reformattingimagingspectrometry(FRIS),integralfieldspectroscopywithlensletarrays(IFS-L),multi-apertureintegralfieldspectrometer(HyperpixelArray),integralfieldspectroscopywithimageslicingmirrors(IFS-S),image-replicatingimagingspectrometry(IRIS),filterstackspectraldecomposition(FSSD),codedaperturesnapshotspectralimaging(CASSI),imagemappingspectrometry(IMS),andmultispectralSagnacinterferometry(MSI).凹However,computationaleffortandmanufacturingcostsarehigh.
Spatiospectralseanning[edit]
Mainarticle:
Spatiospectralseanning
Inspatiospectralseanning,each2-Dsensoroutputrepresentsawavelength-coded('rainbow-colored',入=Xy)),spatial(x,y)mapofthescene.Aprototypeforthistechnique,
introducedin2014,consistsofacameraatsomenon-zerodistaneebehindabasicslitspectroscope(slit+dispersiveelement).[6[13]Advaneedspatiospectralscanningsystems
canbeobtainedbyplacingadispersiveelementbeforeaspatialscanningsystem.
Scanningcanbeachievedbymovingthewholesystemrelativetothescene,bymovingthecameraalone,orbymovingtheslitalone.Spatiospectralscanningunitessomeadvantagesofspatialandspectralscanning,therebyalleviatingsomeoftheirdisadvantages.—
Distinguishinghyperspectralfrommultispectral
imaging[edit]
Multispeotral/
檳y口它「旦p聖gtnalCompa打呂on
HyperspectralandMultispectralDifferences.
Hyperspectralimagingispartofaclassoftechniquescommonlyreferredtoasspectral_
imaging_orspectral_analysis_.Hyperspectralimagingisrelatedtomultispectral_imaging_.Thedistinctionbetweenhyper-andmulti-spectralissometimesbasedonanarbitrary"numberofbands"oronthetypeofmeasurement,dependingonwhatisappropriatetothepurpose.
Multispectralimagingdealswithseveralimagesatdiscreteandsomewhatnarrowbands.Being"discreteandsomewhatnarrow"iswhatdistinguishesmultispectralinthevisiblefromcolor_photography_.AmultispectralsensormayhavemanybandscoveringthespectrumfromthevisibletotheIongwaveinfrared.Multispectralimagesdonotproducethe"spectrum"ofanobject.Landsat_isanexcellentexampleofmultispectralimaging.
Hyperspectraldealswithimagingnarrowspectralbandsoveracontinuousspectralrange,andproducethespectraofallpixelsintheseene.Soasensorwithonly20bandscanalsobehyperspectralwhenitcoverstherangefrom500to700nmwith20bandseach10nmwide.(Whileasensorwith20discretebandscoveringtheVIS,NIR,SWIR,MWIR,andLWIRwouldbeconsideredmultispectral.)
'Ultraspectral'couldbere
- 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:
- 光谱 成像
![提示](https://static.bdocx.com/images/bang_tan.gif)