激光生物学报Issue(1)：38-45,8.DOI:10.3969/j.issn.1007-7146.2015.01.006
奇异数据筛选法在玉米籽粒蛋白质近红外光谱检测中的应用
Outlier Detection for Measurement of Protein Content in Maize Kernels Based on Nearinfrared Reflectance Spectroscopy
摘要
Abstract
As the classical chemical analysis of protein content in maize kernel was slow and destructive,and the exist- <br> ence of the outliers in the near infrared (NIR)spectra would affect the accuracy and stability of the NIR models,we ap-plied outlier detection methods for measuring protein content in maize kernel based on near infrared spectroscopy.3 out-lier screening methods,leverage method,resampling by half-mean method (RHM),leverage method,and monte-carlo sampling method (MCS),were compared to detect outliers in the protein spectra and the least squares support vector machine (LS-SVM)models were built with using partial least squares regression (PLSR)method to extract the optimal component scores and using niche ant colony algorithm (NACA)to optimize the parameters (γand σ2 )of the LS-SVM model.The results showed that the performances of the LS-SVM models with those samples removed the outliers were better than the LS-SVM model with all samples.The prediction results of the validation set also showed that the MCS method was optimal for detecting outliers in the spectra of the protein of the whole maize kernel based on NIRS.关键词
玉米籽粒蛋白质/奇异数据筛选法/偏最小二乘支持向量机 (LS-SVM)/小生境蚁群算法 (NACA)Key words
protein content in maize kernel/outlier screening methods/the least squares support vector machine (LS-SVM)/niche ant colony algorithm (NACA)分类
化学化工引用本文复制引用
梁秀英,李小昱,杨万能..奇异数据筛选法在玉米籽粒蛋白质近红外光谱检测中的应用[J].激光生物学报,2015,(1):38-45,8.基金项目
国家高技术研究发展计划（863计划，2013AA102403）；华中农业大学博士启动基金（项目编号2662014BQ038）；Fundamental Research Funds for the Central Universities （）
