Abstract
Realtime,accurateandreliableagriculturalinformationisanimportantprerequisitefor
promotingthedevelopmentofagriculturalmodernization,andplaysanimportantguidingrole
informulatingnationalagriculturalpolicies.Remotesensinghastheadvantagesofwidearea
andshortcycle.Currently,withremotesensinggraduallybecomingapracticaltechnology,it
hasbeenappliedtomanyfields,andagricultureisoneofthemostwidelyusedfieldsof
remotesensing.Therefore,theprecisemonitoringofcropshasimportantresearchsignificance
foragriculturalinformatization.However,thecurrentsituationoffarmerscultivationinChina
hasresultedinthelong-termfragmentationandmiscellaneouschangesofmostruralareas
cultivatedland.Somecultivatedlandplotsareirregular,anddifferentcropsintersect
seriously.Cropplantingtypeschangeirregularly,andhigh-precisioncropextractionis
difficult,howtoachievehigh-precisioncropextractionbasedonthecharacteristicsof
scatteredandscatteredincropisahotissueincropremotesensingrecognition.
Combinedwiththeaboveanalysis,thispapertakesBeijing,TianjinandHebeiprovinces
astheresearchareas,takestheGF-6remotesensingimagedatain2020asthedatasource,
andstudiesandanalyzestheimpactofcropslandscapetypesoncropremotesensing
recognitionandclassificationalgorithmsbasedoncropspatialdistributioncroplandscape
model,combinedwithmachinelearninganddeeplearning.Andrealizesthecropextraction
accordingtothecharacteristicsofdifferentfarmlanddivisiontypes,andfinallyrealizesthe
high-precisionextractionofwheatinBeijing,TianjinandHebeiregion.Thispapermainly
completesthe