论文部分内容阅读
采用近红外光谱分析技术,建立番茄中游离氨基酸总量预测模型。从一至三穗成熟果实中共采集番茄108个,其中84个做校正集,24个做验证集。从数据归一化、数据格式、数据平滑几个方面选择不同的光谱预处理方法,确定最佳方法为:“Mean Centering”+“Second derivative”+“Norris derivative filter”。将全波数范围(40000~11000)cm~(-1)划分为70个区间,得到最佳建模区间组合为9,10,17,56,57,61。利用偏最小二乘法建立预测模型,得到相关评价指标R、RMSEC、RMSEP及模型预测准确率分别为0.936、6.72μg/100g、7.15μg/100g和92.5%。评价指标及对验证集的预测结果表明,所建模型用来实现对番茄中游离氨基酸总量进行无损、快速预测是可行的。
Near infrared spectroscopy was used to establish the prediction model of total free amino acids in tomato. A total of 108 tomato plants were collected from one to three spike fruits, of which 84 were calibration sets and 24 were validation sets. From the aspects of data normalization, data format and data smoothing, we choose different spectral preprocessing methods to determine the best method: “Mean Centering ” + “Second derivative ” + “Norris derivative filter ”. The full wave number range (40000 ~ 11000) cm ~ (-1) is divided into 70 intervals, the best modeling interval combination is 9,10,17,56,57,61. The partial least squares method was used to establish the prediction model. The related evaluation indexes R, RMSEC, RMSEP and model prediction accuracy were 0.936, 6.72μg / 100g, 7.15μg / 100g and 92.5% respectively. The evaluation index and the forecasting result of the validation set show that it is feasible that the model can be used to predict the total amount of free amino acids in tomato without loss and rapidness.