基于支持向量机的大枣内部虫害无损检测

大枣内部虫害的光谱检测是利用大枣本身的光特性,获取与大枣内部虫害有关的物理化学信息,并利用NIR光谱与化学计量学方法建立定量模型来准确测定物质某些成分的含量。文章对160个大枣样品的近红外光谱测量数据进行二阶导数处理,找出测量波长范围内具有最大样本识别能力的有效波长,再用主成分分析进行降维处理,最后通过支持向量机算法对预测集大枣样本有无虫害进行判别,平均判别正确率为93.5%,并且算法比较稳定。综上,所测样品保持完整,不被破坏;仅通过对样品的一次NIR光谱的简单测量,就能同时测定物质的多种成分数据;可对复杂体系进行多组分同时测定,在短时间内获得分析结果,有利于工业化生产的实时、在线检测,自动化...

Full description

Saved in:
Bibliographic Details
Published in东北农业大学学报 Vol. 45; no. 2; pp. 94 - 102
Main Author 陈红光 王健 中野和弘 敖长林
Format Journal Article
LanguageChinese
Published 东北农业大学水利与建筑学院,哈尔滨,150030%中国食品发酵工业研究院,北京,100015%日本新潟大学农学部,日本新潟950-2181%东北农业大学理学院,哈尔滨,150030 2014
Subjects
Online AccessGet full text
ISSN1005-9369

Cover

More Information
Summary:大枣内部虫害的光谱检测是利用大枣本身的光特性,获取与大枣内部虫害有关的物理化学信息,并利用NIR光谱与化学计量学方法建立定量模型来准确测定物质某些成分的含量。文章对160个大枣样品的近红外光谱测量数据进行二阶导数处理,找出测量波长范围内具有最大样本识别能力的有效波长,再用主成分分析进行降维处理,最后通过支持向量机算法对预测集大枣样本有无虫害进行判别,平均判别正确率为93.5%,并且算法比较稳定。综上,所测样品保持完整,不被破坏;仅通过对样品的一次NIR光谱的简单测量,就能同时测定物质的多种成分数据;可对复杂体系进行多组分同时测定,在短时间内获得分析结果,有利于工业化生产的实时、在线检测,自动化分级。
Bibliography:Spectrum detection of jujube internal pests is used optical character of jujube, to obtain the physical and chemical information of jujube internal pests, and to establish the quantitative model by using NIR spectroscopy and chemometrics method to accurately determine the content of the material ingredients. The paper carded out second derivative to original sample data of near infrared spectrum measurement of 160 jujube samples, and selected the effective wavelengths that had the big identification capability among the wavelength range, using primary constituent analytical to reduce dimension processing. The last, theaverage right forecasting rate of identifying the intact and infested jujubes was about 93.5 % for the predicting set by using the algorithm of SVM, and the algorithm were proved stable, Summing up the above, the test sample could be intact, not destroyed; could determined the variety of material composition data based on simple measurement of NIR spectra for the sample at the same time; could a
ISSN:1005-9369