欢迎来到天天文库
浏览记录
ID:53570131
大小:693.69 KB
页数:5页
时间:2020-04-18
《基于能量信息的毛豆豆荚螟高光谱图像检测-论文.pdf》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、分轿检测,.35,No.14,2014基于能量信息的毛豆豆荚螟高光谱图像检测马亚楠’,黄敏,李艳华。,张憨,步培银。(1.江南大学轻工过程先进控制教育部重点实验室,江苏无锡214122;2.食品科学与技术国家重点实验室,江苏无锡214122;3.无锡出入境检验检疫局,江苏无锡214101)摘要:为了寻求快速有效的毛豆内部豆荚螟的检测方法,将高光谱图像技术应用于毛豆内部的豆荚螟无损检测。以225个样本为研究对象,首先采用平均灰度值的方法自动获取毛豆感兴趣区域,然后提取400~lO00nm波长范围内共94个波段的能量信息作为特征参数,最后结合支持
2、向量数据描述分类器建立豆荚螟的分类检测模型。研究结果显示,在自动提取的感兴趣区域验证集中,正常样本的分类精度为100%,有虫样本分类精度为75%,验证集的总体分类精度为95.6%,可有效识别出含豆荚螟的毛豆样本。关键词:毛豆,感兴趣区域,能量,支持向量数据描述Detectionofinsect-damagededamamebasedonimagepowerusinghyperspectralimagingtechniqueMAYa—nan,HUANGMin,LIYan-hua3,ZHANGMin。,BUPei-yin(1.KeyLaborat
3、oryofAdvancedProcessControlforLightIndustry,MinistryofEducation,JiangnanUniversity,Wnxi214122,China;2.StateKeyLaboratoryofFoodScienceandTechnology,JiangnanUniversity,Wuxi214122,China:3.WuxiEntry-exitInspectionandQuarantineBureau,Wuxi214101,China)Abstract:Inordertoseekaquick
4、andeficientdetectionmethodofedamame,hyperspectralimagingtechniquewasappliedtothenOndeslructivedetectionofinsect—damagededamameinthisstudyltwaswelIknownthattheROIofthevegetablesoybeanpodsisthepositionofthebeans,AROIselectionapproachbasedonthemeangrayvaluesinthehorizontalcoor
5、dinateandvertica1coordinatewasproposed.Inthisexperiment,hyperspectraltransmissionimageswereacquiredfrOmnormalandinsect-damagedvegetablesoybeansf225beans),ThesebeanswereusedastheresearchsamplesFirst,aregionofinterest(ROI)ofedamamewasextractedautomaticallyusingthemeangrayvalu
6、emethodfrOmhyperspectralimages.Then.theimagepowerOfROJwasextractedasclassificationfeature.whichthespectralregioncovered400N1000rimandcontained94wavelengths.Atlast,supportvectordatadescription(SVDD)wasusedtodeveloptheclassificationmodelsfortheinsect—damagededamame.Inthevalid
7、ationset,theresultsindicatedtheautomaticextractingRQImethodbasedOnthemeangrayvalueachieved100%accuracyforthenormalsamples,75%accuracyforthefnsect—damagedsamples.and95.6%overaliclassificationaccuracy.whichcoulddiscriminatefnsect—damagededamame.Keywords:edamame;aregionofinter
8、est;power;supportvectordatadescription中图分类号:TS207.3文献标识码:A文章编号:1002—0306(2014)14—0
此文档下载收益归作者所有