資源描述：
《面向高分一號遙感影像的自動(dòng)幾何配準(zhǔn)算法對比》由會員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在工程資料-天天文庫。
1、2015年12月農(nóng)業(yè)機(jī)械學(xué)報(bào)第46卷增刊doi:10．6041/j．issn．1000-1298．2015．S0．042*面向高分一號遙感影像的自動(dòng)幾何配準(zhǔn)算法對比112121王媛葉思菁岳彥利劉帝佑熊全朱德海(1．農(nóng)業(yè)部農(nóng)業(yè)信息獲取技術(shù)重點(diǎn)實(shí)驗(yàn)室，北京100083;2．中國農(nóng)業(yè)大學(xué)國土資源部農(nóng)用地質(zhì)量與監(jiān)控重點(diǎn)實(shí)驗(yàn)室，北京100083)摘要:遙感影像的幾何配準(zhǔn)是影像后續(xù)處理的重要前提和遙感農(nóng)情監(jiān)測等應(yīng)用的重要保障。不同的自動(dòng)幾何配準(zhǔn)算法在配準(zhǔn)效果上存在差異，單一配準(zhǔn)算法難以滿足所有類型數(shù)據(jù)的配準(zhǔn)要求。根據(jù)不同地形特征和不同時(shí)相
2、特征，選擇了平原和山地、夏季和冬季4個(gè)實(shí)驗(yàn)區(qū)，以現(xiàn)有的基于區(qū)域的互相關(guān)法、互信息法和基于特征的SIFT算法為基礎(chǔ)，分別對上述4個(gè)實(shí)驗(yàn)區(qū)的高分一號影像數(shù)據(jù)進(jìn)行自動(dòng)配準(zhǔn)實(shí)驗(yàn)，對比3種算法的配準(zhǔn)精度、配準(zhǔn)效率和穩(wěn)－5定性。實(shí)驗(yàn)結(jié)果表明:應(yīng)用SIFT算法進(jìn)行配準(zhǔn)，4組實(shí)驗(yàn)結(jié)果均目視接邊效果良好且均方根誤差達(dá)到10數(shù)量級，滿足精度要求。該方法簡單、高效，可以應(yīng)用于農(nóng)情遙感監(jiān)測等日常業(yè)務(wù)。關(guān)鍵詞:遙感影像自動(dòng)幾何配準(zhǔn)精度評價(jià)高分一號中圖分類號:TP751文獻(xiàn)標(biāo)識碼:A文章編號:1000-1298(2015)S0-0260-07Contra
3、stofAutomaticGeometricＲegistrationAlgorithmsforGF-1ＲemoteSensingImage112121WangYuanYeSijingYueYanliLiuDiyouXiongQuanZhuDehai(1．KeyLaboratoryofAgriculturalInformationAcquisitionTechnology，MinistryofAgriculture，Beijing100083，China2．KeyLaboratoryofAgriculturalLandQuali
4、ty(Beijing)，MinistryofLandandＲesources，ChinaAgriculturalUniversity，Beijing100083，China)Abstract:Thegeometricalregistrationofremotesensingimageisanimportantpremiseforthesubsequentprocessingofimage．Andit'salsoanimportantsecurityfortheapplication，suchasagriculturalcond
5、itionmonitoring．Differentalgorithmsofautomaticgeometryregistrationleadtovariousregistrationeffects．It'shardtomeettheregistrationrequirementsofallimages．Fourtestingtypesofplains，mountains，summerandwinterwereselectedbasedthefeaturesofterrainandtime．Themainthreeregistr
6、ationmethodswere:crosscorrelationalgorithmbasedonregiongray，mutualinformationalgorithmbasedonregiongrayandSIFTalgorithmbasedonfeatures．SIFTfeatureisthepartialfeatureoftheimage，whichcankeeptheinvarianceinrotating，scale-zoomingandbrightnesschanging．Thentheautomaticgeo
7、metricregistrationwasmadeforfourclassesofGF-1remotesensingimageusingtheabovethreealgorithms．TwokindsofexperimentswereconductedforGF-1remotesensingimageundervariousconditionssuchasdifferentterrainsanddifferentimagingtime．Thecomparisonofdifferentgeometricregistrationa
8、lgorithmsweremadeintheaspectsofaccuracy，efficiencyandstability．TheresultsshowthattheSIFTalgorithmisthemostappropriateone．Thevisualedgeeffe
