基于多元线性回归的食用油掺假检测模式识别算法设计与优化

Design and Optimization of Edible Oil Adulteration Detection Mode Recognition Algorithm Based on Multiple Linear Regression

摘要: 目前食用油市场发展迅速但掺假情况严重，既损害了消费者的利益，同时又干扰了市场秩序。且国内缺乏一套具有自主知识产权的智能便携式食用油掺假检测系统。本文设计了一款基于多元线性回归算法的便携式食用油掺假检测系统，该系统在多元线性回归算法(MLR)的基础上进行拓展，对电子鼻采集到的食用油的气味信号进行处理，实现对食用油的掺假成分的定性分析和定量计算。实验结果表明所设计的基于MLR的算法对食用油的定性分析准确率可达97%，对食用油掺假定量识别误差率小于4%。

Abstract: The edible oil market is currently developing rapidly but is plagued by severe adulteration issues, which harm consumer interests and disrupt market order. Moreover, there is a lack of an intelligent portable edible oil adulteration detection system with independent intellectual property rights in China. This paper designs a portable edible oil adulteration detection system based on the multiple linear regression (MLR) algorithm. The system expands on the MLR algorithm to process the odor signals of edible oils collected by an electronic nose, achieving qualitative analysis and quantitative calculation of the adulteration components in edible oils. Experimental results show that the designed system based on the MLR algorithm has an accuracy rate of up to 97% for qualitative analysis of edible oils and a quantitative recognition error rate of less than 4% for edible oil adulteration.