论文部分内容阅读
发展了一种多光谱融合新技术,该技术充分利用拉曼光谱与红外光谱的互补特性,并借助数据融合手段,高效实现奶粉掺假检测.为进一步提升数据融合算法的准确性,有机结合离散小波变换(DWT)多尺度特性及竞争性自适应重加权偏最小二乘线性判别(CARS-PLSDA)算法,以有效扣除光谱建模中的干扰信息.为验证多光谱融合技术的有效性,对4种典型奶粉掺假体系分别建立分类判别模型.结果表明,基于DWT-CARS-PLSDA多光谱融合算法所建的面粉、淀粉、糊精和大豆分离蛋白奶粉掺假模型灵敏度分别为94.74%、100%、84.21%和100%,正确率分别为99.42%、98.83%、98.25%和98.83%.与单独对拉曼光谱或红外光谱建立模型相比,4种模型能够显著提高奶粉掺假检测灵敏度和准确性,为奶粉掺假快速诊断提供了一种有效工具.
Developed a new multi-spectral fusion technology, which takes full advantage of the complementary characteristics of Raman spectroscopy and infrared spectroscopy, and data fusion to effectively detect the milk powder adulteration.In order to further improve the accuracy of data fusion algorithm, organic combination of discrete Wavelet transform (DWT) multi-scale features and competitive adaptive weighted partial least-squares linear discriminant (CARS-PLSDA) algorithm to effectively eliminate interference information in spectral modeling.In order to verify the effectiveness of multispectral fusion, Four kinds of typical milk powder adulteration systems were established discriminant models.The results showed that the sensitivity of the adulteration model of flour, starch, dextrin and soy protein isolate based on DWT-CARS-PLSDA multi-spectral fusion algorithm were 94.74%, 100 %, 84.21% and 100%, respectively, with correct rates of 99.42%, 98.83%, 98.25% and 98.83%, respectively.Compared with Raman or IR alone, the four models can significantly improve the sensitivity of milk powder adulteration detection and Accuracy, for the rapid diagnosis of adulteration of milk provides an effective tool.