文档介绍：第 卷第期 河南科技学院学报渊自然科学版冤 年 月
50 atgrainrecognitionandclassificationisanimportant
accuratelyextractthecharacteristicsofwheat,basedontheexistingdatasetconstructionmethods,thispaperuses
multiplepicturesofthesamewheatgrain,addssubangleprocessing,andconstructsasubanglewheatgrainvariety
picturelibrarybasedonthethreeanglesofwheatgrainbellyditchup,bellyditchdownandbellyditchforward.
Firstly,1000grainsof6wheatvarietieswithlargeplantingareainHuangHuaiwheatareawerecollected,andeach
collection,,therecognitioneffectis
,theperformanceofventralgrooveforwardangleisbetterthantheother
,%,whichisbetterthanthatofanglemixeddataset.
Theexperimentalresultsshowthatthemulti-anglewheatgrainclassificationdatabaseconstructedinthispaper
helpstheclassificationmodeltoextractwheatgrainfeaturesmoreaccurately,avoid