【摘要】 研究了近红外光谱定量分析模型对于样品温度的适应性。以 4 2个不同品种的大豆为实验材料 ,用 2台光谱仪分别独立测定了样品在 5种温度下的近红外光谱。对于 2台光谱仪测定的光谱 ,均依据光谱信息选择部分光谱 ,采用PLS回归法对大豆样品的粗蛋白质和粗脂肪含量分别建立了近红外光谱定量分析模型 ,并以剩余样品对模型进行预测检验。 4个模型的预测结果均表明 :超过 94 %的检验样品的预测相对误差小于 5 % ,说明了预测样品处于 5～ 4 0℃时 ,模型都有较好的预测效果。更多还原

【Abstract】 This paper was about the temperature adaptability to quantitative analysis model of near infrared spectrum. 42 soybean samples were used and measured the near infrared spectrum independently and separately with two spectrometers at 5 different temperatures. The spectra were then divided to calibration set and validation set according to the information of spectrum. Two quantitative analysis models with the calibration set were established and used to estimate the contents of raw protein and raw fat. The same operations were made for the remained spectra. The results showed that the relative residuals of the 94% tested samples were within a range of ±5%. The models had a high adaptation to temperature during predicting the contents of raw protein and fat in soybeans.更多还原