CAJ | 학술논문

快捷方便检测粮粒中是否有储粮害虫及其种类，对储粮害虫虫情调查和监测，实现粮库中储粮害虫的准确检测和口岸害虫快速检测检疫具有重要实践意义。以最主要的初期性害虫6米象为例，对6米象、谷壳、米粉样品进行了太赫兹时域光谱测试，获得了样品在0.2～1.6 THz 波段的折射率和吸收光谱，分析了这些样品的特征吸收谱，并利用 PLS -DA 方法对含有6米象和不含6米象的谷粒样品的太赫兹光谱进行鉴别分析，结果表明用0.2～1.6 THz 范围内的 THz 吸收光谱结合 PLS -DA 方法对校正样本建立判别模型，其校正和验证结果与实际分类变量的相关性高，交叉验证均方根误差（RMSECV）和预测均方根误差（RMSEP）都小于0.150，建立的 PLS -DA 分类模型对检测样本的判别准确率为100％，为检测粮粒中是否有储粮害虫提供快速方便的鉴别分析方法。
쾌첩방편검측량립중시부유저량해충급기충류，대저량해충충정조사화감측，실현량고중저량해충적준학검측화구안해충쾌속검측검역구유중요실천의의。이최주요적초기성해충6미상위례，대6미상、곡각、미분양품진행료태혁자시역광보측시，획득료양품재0.2～1.6 THz 파단적절사솔화흡수광보，분석료저사양품적특정흡수보，병이용 PLS -DA 방법대함유6미상화불함6미상적곡립양품적태혁자광보진행감별분석，결과표명용0.2～1.6 THz 범위내적 THz 흡수광보결합 PLS -DA 방법대교정양본건립판별모형，기교정화험증결과여실제분류변량적상관성고，교차험증균방근오차（RMSECV）화예측균방근오차（RMSEP）도소우0.150，건립적 PLS -DA 분류모형대검측양본적판별준학솔위100％，위검측량립중시부유저량해충제공쾌속방편적감별분석방법。
Detecting the stored -product insects in the grain kernels and determining the species of these in-sects quickly and accurately are very significant to survey and monitor the damage by the pest.It also has an impor-tant practical significance in the accurate detection of pests in food warehouse and the rapid detection of pests in im-port and export ports.The Sitophilus zeamais (S.zeamais),a common stored grain insect associated with food -pro-cessing facilities worldwide was used as the test insect in this study.The study involved identification of 0.2 ~1.6 THz absorption characteristics and refractive indices of S.zeamais,chaff flour,rice flour and mixtures of them by ter-ahertz time -domain spectroscopy (THz -TDS).The absorption characteristics of these samples (S.zeamais,chaff flour and rice flour)were analyzed.Partial least -squares discriminate analysis (PLS -DA)was applied to classify the healthy grain powder and the grain powder mixed with different concentration S.zeamais into two groups.The re-sults demonstrated that the calibration and prediction results were highly correlated to the real classification variables by using of the discrimination model,which was built by PLS -DA method and the absorption spectrum in the region of 0.2 ~1.6 THz.The root mean square error of calibration (RESECV)and root mean square error of cross -valida-tion (RESEP)were both less than 0.150.The correct classifications were 100% by building PLS -DA discrimination model.This study provided a rapid and convenient method to detect the insect -damaged grain kernels.