影响花生秧近红外光谱预测准确性因素的分析

Analysis of Factors Affecting Near-infrared Spectroscopy Prediction Accuracy of Peanut Seedlings

采集来自河南省不同地区和不同品种的花生秧样品,分别采用湿化学方法和近红外扫描技术(NIRS)进行主要营养成分含量的测定和预测,从样品数量、测定指标、粉碎粒度和计量方法等方面分析了影响花生秧近红外光谱预测准确性的因素。结果表明:预测准确性随着花生秧样品数量的增加而提高;对干物质、酸性洗涤纤维、中性洗涤纤维、粗蛋白质含量的预测效果较好,而粗脂肪、钙和磷的含量低,不能进行准确预测;花生秧的最佳粒度为40目;建立预测模型的最佳离群值为2.5~5.0;花生秧不同测定指标预测模型的最佳计量方法各不相同,但均宜采用标准正态变换处理(SNV)方法。

Peanut seedling samples from different regions and varieties of Henan Province were collected to analyze various factors affecting the accuracy of peanut seedling with near-infrared spectroscopy prediction from the number of samples,measurement index,crushing granularity and measurement method,and wet chemical method and near-infrared scanning technology(NIRS)were used to determine and predict the main nutrients.The results showed that the prediction accuracy increased with the increase of peanut seedling samples.The prediction effect of DM,ADF,NDF and CP was good.EE,Ca and P were not accurately predicted because of the low content.The optimal granularity was 40 orders.The best outliers were 2.5~5.0.At the same time,the best measurement methods for prediction models of different measuring indexes of peanut seedlings are different,but the standard normal transformation(SNV)method was suitable for all of them.