玉米主要营养成分的多元回归分析

Multiple Variant Regression Analysis of the Main Nutritional Components of Corn

运用多元散射校正(MSC)方法,随机抽样调查100个玉米样品,测出光谱响应数据及蛋白质、纤维素和脂肪的成分含量值.对所有样品的光谱响应数据进行预处理,通过相关系数法(CC)选取光谱响应数据与蛋白质含量相关系数的绝对值最大的波数,利用多元线性回归分析(MLR)、逐步多元回归分析(SMLR)和典型相关分析建立蛋白质的光谱分析模型,从而得到蛋白质的光谱分析初步模型.同时从相关系数、内部交叉验证均方差(RMSECV)和信噪比三个方面进行综合考虑,对模型预测结果的准确性进行了评价分析.按照建立蛋白质光谱分析模型的流程,类似地建立纤维素和脂肪的光谱分析模型,利用上述各自的模型估算玉米样品中蛋白质、纤维素、脂肪三种营养成分的含量.

By way of MSC,giving random sampling 100 corn samples,measuring the spectral response data and protein,fiber and fat content value. And making pre-processing the spectral response data with multiple scatter correction method,and selecting the absolute value of the most large wave number of the correlation coefficients between the spectral response data and the protein content,then using Multiple variant linear regression analysis,gradually multiple regression analysis and canonical correlation analysis to establish the protein spectrum analysis model,so the preliminary model of protein spectrum analysis is obtained. At the same time,considering the correlation coefficient,internal cross validation of variance and signal to noise ratio,the paper makes and analysis of the accuracy of the model prediction results. According to establishment of the protein spectrum analysis model,cellulose and fat spectral analysis model is established similarly,so using the respective model to estimate the nutritional contents of corn protein,fiber and fat.