基于SVM算法的近红外光谱技术在鱼糜水分和蛋白质检测中的应用

采用支持向量机(support vector machine,SVM)建立了鱼糜样品中水分和蛋白质含量的近红外光谱校正模型,并采用独立样本集进行了预测。光谱数据经间隔两点一阶导数(DB1G2)、标准正态变换(SNV)、多元散射校正(MSC)相结合的方法预处理后,用偏最小二乘(PLS)降维处理,取前15个投影变量为自变量。获得水分模型的校正相关系数Rc、预测相关系数Rv、定标标准差SEE、预测标准差SEP;蛋白质模型的校正相关系数Rc、预测相关系数Rv、定标标准差SEE、预测标准差SEP,有较好的预测准确性。基于SVM算法的近红外光谱技术可用于鱼糜水分和蛋白质含量的快速检测。

Application of Near Infrared Spectroscopy Technique Based on Support Vector Machine in the Measurement of Moisture and Protein Contents in Surimi

Near infrared spectroscopy calibration models for moisture and protein contents of surimi were established by support vector machine(SVM),and predicted by independent sample set.The spectra processed by first derivative gap two(DB1G2),standard normal variation(SNV) and multiplicative scatter correction(MSC),then compressed by partial least squares(PLS),and the first fifteen were taken as variables.The correlation coefficient of calibration(Rc),correlation coefficient of validation(Rv),standard error of calibration(SEC) and standard error of prediction(SEP) of moisture model were 0.97,0.95,0.53 and 0.58,respectively.The Rc,Rv,SEC and SEP of protein model were 0.99,0.98,0.36 and 0.39,respectively.The two models had good recurrence.Therefore,NIR spectroscopy based on SVR can be applied to rapidly predict moisture and protein contents in surimi.