Abstract

Detection of additives in milk is important, especially preservatives. Potassium sorbate was taken as an example to study the feasibility of hyperspectral imaging as a rapid detection tool for preservatives in milk. Samples with different concentrations of potassium sorbate were divided into four groups. Pretreatment and variable selection methods were studied based on the support vector machine model. The model with standard normalized variable pretreatment and competitive adaptive reweighted sampling variable selection had the best performance. Testing set accuracy for different groups from low concentration to high concentration were 92.16%, 95.7%, 99.46%, and 98.33%. Hyperspectral imaging can be a rapid detection tool for potassium sorbate in milk.

摘要

检测牛奶中的添加剂很重要，尤其是防腐剂。以山梨酸钾为例，研究了高光谱成像作为牛奶中防腐剂快速检测工具的可行性。不同浓度山梨酸钾的样品分为四组。研究了基于支持向量机模型的预处理和变量选择方法。具有标准归一化变量预处理和竞争性自适应重加权采样变量选择的模型性能最好。从低浓度到高浓度不同组的测试集准确率分别为92.16%、95.7%、99.46%和98.33%。高光谱成像可以成为牛奶中山梨酸钾的快速检测工具。