摘要
文章采用近红外光谱分析方法验证不同条件下氨化和碱化处理玉米秸秆粗蛋白含量测定效果。选择不同超声条件对玉米秸秆样品前处理,获取54种样品,粗蛋白含量为2.535 6%~6.854 7%,依据X-Y残差法剔除29、30、38、51号异常样本,交互验证决定系数R2C由0.679升至0.840,将剩余50个样品划分为校正集(40samples)及验证集(10 samples),选用OSC方法对光谱去噪处理,对比平滑处理(windowsize 15),R2C由0.827升至0.865,选取波段9 781~1 093 cm-1作为特征波段,对比SVRM、PCR及PLS 3种粗蛋白定量分析模型,选取SVRM(C=0.01,Gamma=100)为最佳模型,校正集决定系数R2C为0.833,RMSEC为0.389,验证集决定系数R2P为0.914,RMSEP为0.296。结果表明,近红外光谱分析方法测定玉米秸秆氨化、碱化处理后粗蛋白含量可行。
A rapid detection method based on near infrared spectroscopy （NIR） was used toquickly verify the crude protein of corn straw in ammoniation and alkalization under different conditionsin this paper. Different ultrasonic conditions were selected to pretreat the corn straw. A total of 54samples were obtained. The crude protein content was 2.5356%-6.8547%. Four abnormal samples, 29,30, 38 and 51, were excluded. The cross validation coefficient of determination （R2C） of the remaining50 samples rose from 0.679 to 0.84. All the remaining 50 samples were divided into calibration set （40samples） and test set （10 samples）. After denoising the spectrum by way of OSC and contrasting smoothing（windowsize 15）, R2C increased from 0.827 to 0.865. The 9 781- 1 093 cm- 1 band wasselected as the characteristic band to compare three crude protein quantitative analysis models, SVRM,PCR and PLS. Finally, SVRM （C=0.01, Gamma=100） was chosen as the best model. It＇s correctiondecision coefficient R2C was 0.833, RMSEC was 0.389, validation set decision coefficient R2P was0.914, and RMSEP is 0.296. The results showed that it is feasible to use near infrared spectroscopyanalysis method to determine the crude protein contents of corn straw under ammoniation andalkalization conditions.
出处
《东北农业大学学报》
CAS
CSCD
北大核心
2017年第12期68-79,共12页
Journal of Northeast Agricultural University
基金
国家重点研发计划项目(2016YFD0700204-02)
作者简介
沈维政（1977-）,男,教授,博士,博士生导师,研究方向为农业信息化.E-mail：wzshen@neau.edu.cn
