资源描述：
《基于图像的水稻灯诱害虫自动识别技术的研究》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、中国水稻科学(ChinJRiceSci),２０１５,２９(３):２９９－３０４http://www．ricesci．cnDOI:１０．３９６９/ji．ssn．１００１Ｇ７２１６．２０１５．０３．００９２９９基于图像的水稻灯诱害虫自动识别技术的研究冼鼎翔１姚青１,∗杨保军２罗举２谭畅１张超１徐一成２,∗(１浙江理工大学信息学院,杭州３１００１８;２中国水稻研究所,杭州３１０００６;∗通讯联系人,EＧmail:qＧyao＠zstu．edu．cn;xuyicheng＠caas．cn)AutomaticIdentificati
2、onofRiceLightＧtrappedPestsBasedonImagesXIANDingＧxiang１,YAOQing１,∗,YANGBaoＧjun２,LUOJu２,TANChang１,ZHANGChao１,XUYiＧcheng２,∗(１CollegeofInformation,ZhejiangSciＧTechUniversity,Hangzhou３１００１８,China;２ChinaNationalRiceResearchInstitute,Hangzhou３１０００６,China;∗Corresponding
3、author,EＧmail:qＧyao＠zstue．duc．n;xuyicheng＠caasc．n)XIANDingxiang,YAOQing,YANGBaojun,etal．AutomaticidentificationofricelightＧtrappedpestsbasedonimages．ChinJRiceSci,２０１５,２９(３):２９９Ｇ３０４．Abstract:AutomaticidentificationandcountofricelightＧtrappedpestsisacommonandimpor
4、tantpestforecastingmethodinpaddyfields．However,mostofthelightＧtrappedpestsareunnecessarytobemonitoredandmustberemoved．ThismanualmethodistimeＧconsumingandfatiguingwithalowaccuracyrate．WedevelopedanautomaticmethodforidentifyingricelightＧtrappedpestsbasedontheimage
5、s．Firstly,wedividedtheimagesintoseveralgroupsaccordingtotheirmorphologicalfeatures．Eachgrouphasthreeclassifications:thebackＧsideimageofapest,theabdomenＧsideimageofthispest,andnonＧforecastingpestimagesimilartothispest．Then,thirtyＧonefeaturesincludingthecolor,shap
6、eandtexturefeatureswereextractedfromeachinsectimage．Finally,threesupportvectormachineclassifierswithposteriorprobabilitieswereusedtotrainandtestthethreegroupsofinsects,respectively．Intheresults,thebackＧsideimageandtheabdomenＧsideimageofapestwereconsideredasthesa
7、mespecies．Weachieveda９７．４％accuracyrateinthethreespeciesofricelightＧtrappedpests．Keywords:lightＧtrappedpests;imagefeatures;supportvectormachineclassifier;automaticidentification;nonＧforecastingpestsrejection;forecasting;rice冼鼎翔,姚青,杨保军,等．基于图像的水稻灯诱害虫自动识别技术的研究．中国水稻科
8、学,２０１５,２９(３):２９９Ｇ３０４．摘要:利用灯光诱杀稻田害虫,并识别与计数这些害虫是水稻害虫的一种常规但非常重要的测报方法.在灯光诱杀的昆虫中,大多数昆虫是不需要测报的,因此,在人工识别灯诱测报害虫时,需要排除这些昆虫.这种人工识别与计数害虫的方法效率低、任务重、识别准确率差.我们提出了一种基于图像的水稻灯诱害虫自动识别方
