鱼粉、豆粕样本-样本二维相关近红外光谱判别

摘要

In order to investigate the feasibility of sample-sample 2-D correlation near infrared diffuse reflectance spectroscopy method for different protein feeds rapid identification,52 fish meal samples and 60 soybean meal samples were collected, including 74 calibration samples and 38 independent validation samples. Samples were measured by spectrometer. Calibration model was established by sample-sample 2-D correlation spectroscopy method with the pretreatment of 3-point first derivative. This method performed 100% of correctly classified samples. Compared with partial least squares discriminant analysis method, both of them can correctly identify the fish meal and soybean meal. But the sample-sample 2-D correlation spectroscopy method is fast for calculation.%为了探讨利用二维相关近红外光谱法快速判别鱼粉和豆粕饲料原料的可行性,共收集了52个鱼粉、60个豆粕样品,其中34个鱼粉和40个豆粕样品作为定标集,其余18个鱼粉和20个豆粕样品作为独立验证集,用Spectrum400型近红外光谱仪扫描获取光谱,并对其特征峰进行解析.采用样本-样本二维相关分析方法定性判别鱼粉、豆粕,当预处理方法为一阶导数时,判断正确率为100％.与偏最小二乘判别分析方法相比,二者均可正确检测鱼粉、豆粕,其中二维相关方法的建模及验证算法简单、计算量小、运算速度快.