Abstract: Moisture content affects the quality of pasture. Leymus chinensis is rich in vital vitamins,protein,neutral detergent fiber,acid detergent fiber,fat and other essential nutrients for livestock. The quality of Leymus Chinensis is highly susceptible to water,soil and gas during the harvesting and storage process which greatly influence the nutrient contents.In order to effectively reduce redundant non-information variables,improve the prediction accuracy and stability of Leymus chinensis moisture content using near-infrared model,this study uses four spectral feature interval selection methods,including interval partial square (Interval Partial Least -Squares Regression,iPLS),Backward interval PLS (BiPLS),Synergy interval PLS (SiPLS),and Successive projections algorithm (SPA) to establish a predictive model for moisture content of Leymus chinensis. The results showed that the SiPLS method is most suitable for the screening of the characteristic wavelength of Leymus chinensis,followed by the BiPLS method. The worst method is iPLS,and RPD=2.648>2.50. This indicates that the near-infrared spectroscopy model of SiPLS is completely feasible in predicting the moisture content of Leymus chinensis,and the prediction accuracy is above 96.13%.

Key words: Near infrared spectral characteristics, Moisture content of Leymus Chinensis, Characteristic wavelength

摘要： 水分是牧草最为重要的品质属性，水分含量的多少直接影响牧草品质的变化。羊草（Leymus chinensis）因富含重要的维生素、蛋白质、中性洗涤纤维、酸性洗涤纤维、脂肪等家畜必需营养成分，在收获和储藏过程中极易受到生长地的水、土、气等的影响而发生营养成分损失或变质，因此为了有效降低冗余无信息变量，提高羊草水分含量近红外模型的预测精度和稳定性，本研究采用4种光谱特征区间选择方法，包括间隔偏二乘法（Interval partial least-squares regression，iPLS）、向后区间偏最小二乘法（Backward interval PLS，BiPLS）、联合区间偏最小二乘法（Synergy interval PLS，SiPLS）、和连续投影算（Successive projections algorithm，SPA）建立羊草水分含量的预测模型。结果表明：SiPLS方法最适合用于羊草水分含量特征波长的筛选，其次为BiPLS方法，最差的方法为iPLS，同时，相对分析误差（Residual predictive deviation，RPD）=2.648>2.50。这表明SiPLS的近红外光谱模型在预测羊草水分含量的应用上完全可行，预测精度在96.13%以上。

关键词: 近红外光谱, 羊草水分, 光谱特征波长