【摘要】 研究旨在探讨利用全株紫花苜蓿(Medicagosativa L.)样品的近红外漫反射光谱信息,建立能够预测其茎组分营养价值的校正模型的可行性。将66份不同年份、品种、茬次和生育期的紫花苜蓿全株样品徒手分离茎叶后,按一定的茎叶比重新混合成198份实验样品(建模样品138份,检验样品60份)。采用傅里叶变换近红外漫反射光谱技术(FT-NIRS),结合偏最小二乘法(PLS),建立了茎组分粗蛋白(CP)、中性洗涤纤维(NDF)、酸性洗涤纤维(ADF)、粗灰分(CA)和体外可消化干物质(IVDDM)含量的预测模型。除NDF含量的预测模型外,其他4个指标预测模型的建模效果和实际预测均较好,交叉检验相关系数(rCV)为0.8523~0.9007,交叉检验标准误差(RMSECV)为0.72%~3.96%,检验样品的预测值与化学值的相关系数(r)为0.9255~0.9512。而NDF含量预测模型的RCV,RMSECV,r分别为0.8214,3.70%和0.9020,模型只可用作粗略估测。更多还原

【Abstract】 The present research was attempted to predict the qualities of stem of alfalfa (Medicago sativa L.) without separation from the whole plant by near infrared reflectance spectroscopy and discussed the feasibility of using the near infrared reflectance spectra information of the whole object to predict the qualities of a certain part. Sixty six whole alfalfa hay samples of separated stems from leaves were collected and they were distinguishing by years, cultivars, cuts and growing periods. There were 138 calibration samples and 60 validation samplers. Fourier transform-near infrared reflectance spectroscopy(FT-NIRS) and partial least square(PLS) were used to set up the calibration models of stem’s crude protein (CP), neutral detergent fiber(NDF), acid detergent fiber (ADF), crude ash (CA) and in vitro digestible dry matter (IVDDM) contents. All models showed great calibration and prediction performances except the one of stem’s NDF content. The correlation coefficients of cross-validation (rCV) were between 0.852 3 and 0.900 7, the root mean square errors of cross-validation (RMSECV) were between 0.72% and 3.96% and the correlation coefficients of NIRS values and chemical values (r) were between 0.925 5 and 0.951 2. However, rCV, RMSECV and r of the model of stem’s NDF content were 0.821 4, 3.70% and 0.902 0, respectively. It wasn’t exact enough and would be used for rough predicting only. All of the results showed that near infrared reflectance spectra information of whole alfalfa hay could be used to predict some components of its stem exactly. It was the maiden attempt of using near infrared reflectance spectra information of the whole objects to evaluated the qualities of a certain part.更多还原