摘要
采用正常玉米样品40个,发霉玉米样品41个,建立了电子鼻对霉变与正常样品的识别模型,优化了10个传感器的组合,并对32个未知样品进行判别,其中霉变样品15个,正常样品17个。结果表明传感器优化前后主成分分数分别为86.34%和97.54%,优化后提高了11.2%。采用Euclid、Malahanobis、Correla-tion以及DFA四种算法对检验集未知样品进行判定,优化前总判别率分别为Euclid:68.75%,Malahanobis:75%,Correlation:84.38%,DFA:81.25%;优化后总判别率分别为Euclid:68.75%,Malahanobis:75%,Correla-tion:90.63%,DFA:87.5%。优化后Correlation和DFA法的判别率比优化前提高,其中Correlation法达90.63%。在对霉变和正常玉米判别时,霉变样品的判别率要远高于正常样品的判别率。
40 normal samples and 41 moldy samples were collected to domesticate the electronic nose.10 sensor arrays were optimized,and 32 unknown samples were tested by the electronic nose,including 15 moldy samples and 17 normal ones.The principal component percentages before and after being optimized were 86.34% and 97.54% respectively,and the optimized one increased by 11.2% compared with the original one.Four different pattern recognition algorithms were applied to identify the unknown samples.Before the optimization,the total discrimination rates were Euclid:68.75%,Malahanobis:75%,Correlation:84.38%,DFA:81.25%.After optimization,the total discrimination rates were Euclid:68.75%,Malahanobis:75%,Correlation:90.63%,and DFA:87.5%.The discrimination rate of the algorithm correlation and DFA was improved after being optimized,and the correlation reached 90.63%.With regard to the discrimination between moldy and normal corns,the discrimination ratio of moldy sample was far higher than the discrimination rate of normal ones.
出处
《中国粮油学报》
EI
CAS
CSCD
北大核心
2011年第10期103-107,共5页
Journal of the Chinese Cereals and Oils Association
基金
国家科技支撑计划(2009BADA0B00-5)
关键词
电子鼻
玉米
霉变
快速检测
electronic nose
corn
moldy
rapid detection