Wang Tao, Wang Li, Guo Tao, Shi Yanli, Liu Baocang, Li Fei
State Key Laboratory of Grassland Agroecosystems, Key Laboratory of Grassland Livestock Industry Innovation, Ministry of Agriculture and Rural Affairs, Engineering Research Center of Grassland Industry, Ministry of Education, College of Pastoral Agriculture Science and Technology, Lanzhou University, Lanzhou, China.
Xinjiang Taikun Group Chang Feed Co., Ltd., Changji Hui Autonomous Prefecture, China.
Anim Biosci. 2025 Sep;38(9):1921-1933. doi: 10.5713/ab.24.0872. Epub 2025 Apr 11.
This study aims to establish an accurate and reliable near-infrared spectroscopy to enable rapid, efficient, and non-destructive evaluation of the nutritional quality of common vetches across different regions and varieties.
A total of 190 samples from various regions and varieties were selected for this study, which were divided into a calibration set (4:1 ratio) and validation set. The original spectrum of the calibration set is subjected to 10 different pretreatment techniques in combination with first and second derivatives, while the prediction model was established by combining the measured value of common vetch with the modified partial least squares method.
The results indicate that the calibration root-mean-square error and cross-validation root-mean-square error values range from 0.01 to 1.25 and 0.01 to 1.37, respectively. The determination coefficients of cross-validation (R2 CV) for phosphorus (P), potassium (K), magnesium (Mg), and iron (Fe) are relatively low at 0.82, 0.86, 0.82, and 0.74, respectively; however, all other indicators have R2 CV values above 0.90. The predicted root means square errors (RMSEP) for common vetch indexes range from 0.01 to 1.87, with RMSEP values higher than 1.0 observed for crude protein, neutral detergent fiber, acid detergent fiber, and ash indices, whereas RMSEP values lower than or equal to 1.0 were obtained for other indicators. The measured coefficient of determination (R2p) demonstrates that the R2p values for each nutrient element and mineral element vary from 0.70 to 0.96. The residual prediction deviation (RPD) values for Mg exhibit relatively low levels, while the RPD values for other indicators exceed 2.0.
These findings suggest that this study provides a viable approach to evaluate the nutritional composition and mineral element content of different varieties and regions of common vetch.
本研究旨在建立一种准确可靠的近红外光谱法，以实现对不同地区和品种的普通野豌豆营养品质进行快速、高效和无损评估。
本研究共选取了来自不同地区和品种的190个样本，将其按4:1的比例分为校正集和验证集。校正集的原始光谱结合一阶和二阶导数进行10种不同的预处理技术处理，同时采用改进的偏最小二乘法结合普通野豌豆的测量值建立预测模型。
结果表明，校正均方根误差和交叉验证均方根误差值分别在0.01至1.25和0.01至1.37之间。磷（P）、钾（K）、镁（Mg）和铁（Fe）的交叉验证决定系数（R2 CV）相对较低，分别为0.82、0.86、0.82和0.74；然而，所有其他指标的R2 CV值均高于0.90。普通野豌豆指标的预测均方根误差（RMSEP）在0.01至1.87之间，粗蛋白、中性洗涤纤维、酸性洗涤纤维和灰分指标的RMSEP值高于1.0，而其他指标的RMSEP值低于或等于1.0。测定决定系数（R2p）表明，各营养元素和矿质元素的R2p值在0.70至0.96之间。镁的残留预测偏差（RPD）值相对较低，而其他指标的RPD值超过2.0。
这些结果表明，本研究为评估不同品种和地区普通野豌豆的营养成分和矿质元素含量提供了一种可行的方法。
