探讨了基于不同数据预处理方法的正交信号校正在秸杆饲料近红外光谱模型传递中的应用.以141个秸杆青贮饲料样品为研究对象,以其粗蛋白含量为目标参数,研究了基于无处理、局部中心化、全局中心化和Z-score标准化预处理方法的正交信号校正,在源仪器(SPECTRUM ONE NTS)和目标仪器1(ANTA-RIS)与目标仪器2(FOSS 6500)之间的模型传递效果.实验表明:对于两台傅里叶变换型近红外光谱仪,采用局部中心化、全局中心化和Z-score标准化预处理方法的正交信号校正均可成功实现模型传递,其中局部中心化和全局中心化法的作用效果基本一致,且优于Z-score标准化法.对于傅立叶变换和光栅型近红外光谱仪,全局中心化的作用效果明显优于其它3组处理效果,且只有全局中心化预处理的正交信号校正传递后的模型可用于实际预测.%This study is focus on the orthogonal signal correction based on different pretreatment methods for near infrared spectroscopy (NIRS) calibration transfer of straw silage. Up to 141 straw silage samples with known crude protein content and two Fourier transform instruments (master instrument: SPECTRUM ONE NTS. Slave instrument 1: ANTARIS), one scanning grating instrument (slave instrument 2: FOSS 6500) were involved. Four pretreatment methods, control, local centering, global centering and Z-score standardization, for orthogonal signal correction were tested and evaluated. The results showed that the orthogonal signal correction with the local centering, the global centering and the Z-score standardization methods worked successful for calibration transfer between two Fourier transform instruments. The local centering and the global centering gave almost equally good results which was better than that of Z-score standardization method. However, for calibration transfer between Fourier transform and scanning grating instruments, only global centering method gave acceptable result while the other three methods were hard for application of practical prediction.
展开▼
