利用主成分权重重置实现牛奶成分浓度快速检测中近红外光谱的净化去噪

发布时间：2018-03-01 03:18

本文关键词： 光谱学近红外光谱光谱分析直方图规定化模糊层次分析法权重　出处：《光学学报》2017年10期 　论文类型：期刊论文

【摘要】：为了利用近红外光谱技术对牛奶中脂肪、蛋白质和乳糖的浓度进行快速检测,提出了基于直方图规定化和模糊层次分析法的近红外光谱主成分权重重置预处理方法。针对牛奶样品的光谱进行主成分分析,确定光谱数据中的最佳主成分数,得到了各主成分的得分和权重。采用直方图规定化的数学统计思想实现二维光谱矩阵的滤波及净化去噪,利用模糊层次分析方法对有效主成分信息的权重进行重置映射,滤掉无关的主成分干扰信息,建立权重重置后的光谱。在此基础上,对预处理后的光谱数据进行脂肪、蛋白质和乳糖的偏最小二乘回归1(PLS1)建模,得到脂肪的相关系数为0.980,预测均方根误差为0.158×10~(-2) g·mL~(-1);蛋白质的相关系数为0.997,实际预测偏差为0.050×10~(-2) g·mL~(-1);乳糖的相关系数为0.985,实际预测偏差为0.152×10~(-2) g·mL~(-1)。由模型预测结果可知,基于直方图规定化和模糊层次分析法的近红外光谱主成分权重重置预处理方法比常规预处理方法具有更好的滤波和去噪效果,说明将直方图规定化和模糊层次分析法相结合对牛奶近红外光谱进行预处理具有可行性。
[Abstract]:In order to detect the concentration of fat, protein and lactose in milk by near infrared spectroscopy, Based on histogram specification and fuzzy analytic hierarchy process (FAHP), a preprocessing method for resetting the weight of principal components of near infrared spectrum is proposed. The spectrum of milk samples is analyzed by principal component analysis to determine the best principal component fraction in the spectral data. The scores and weights of each principal component are obtained, the filtering and de-noising of two-dimensional spectral matrix are realized by using the mathematical statistics idea of histogram specification, and the weight of effective principal component information is reset and mapped by using fuzzy hierarchy analysis method. Filtering out the irrelevant principal component interference information, establishing the spectrum after the weight reset. On this basis, the pretreated spectral data are modeled by partial least squares regression of fat, protein and lactose. The correlation coefficient of fat is 0.980, the root mean square error of prediction is 0.158 脳 10 ~ (-2) g 路mL ~ (-1), the correlation coefficient of protein is 0.997, the actual prediction deviation is 0.050 脳 10 ~ (-2) g 路mL ~ (-1), the correlation coefficient of lactose is 0.985, and the actual prediction deviation is 0.152 脳 10 ~ (-2) g 路mL ~ (-1). Based on histogram specification and fuzzy analytic hierarchy process, the resetting preprocessing method of principal component weight of near infrared spectrum has better filtering and denoising effect than conventional preprocessing method. It is feasible to pretreat milk near infrared spectrum by histogram specification and fuzzy analytic hierarchy process.
【作者单位】：哈尔滨理工大学测控技术与通信工程学院;哈尔滨理工大学测控技术与仪器黑龙江省高校重点实验室;
【基金】：国家自然科学基金(11574065)
【分类号】：O433

【相似文献】