Application Progress of Spectral Detection Technology of Melamine in Food

Ru-lin LÜ; Hong-yuan HE; Zhen JIA; Shu-yue WANG; Neng-bin CAI; Xiao-bin WANG

doi:10.3964/j.issn.1000-0593(2022)07-1999-08

Journals >Spectroscopy and Spectral Analysis >Volume 42 >Issue 7 >Page 1999 > Article

Spectroscopy and Spectral Analysis
Vol. 42, Issue 7, 1999 (2022)

Application Progress of Spectral Detection Technology of Melamine in Food

Ru-lin LÜ^1、*, Hong-yuan HE^{1、1; *;}, Zhen JIA^1、1;, Shu-yue WANG^1、1;, Neng-bin CAI^2、2;, and Xiao-bin WANG^1、1;

Author Affiliations

¹1. College of Investigation, People’s Public Security University of China, Beijing 100038, China

²2. Shanghai Key Laboratory of on Site Material Evidence, Shanghai Public Security Bureau Material Evidence Identification Center, Shanghai 200000, China

show less

DOI: 10.3964/j.issn.1000-0593(2022)07-1999-08 Cite this Article

Ru-lin LÜ, Hong-yuan HE, Zhen JIA, Shu-yue WANG, Neng-bin CAI, Xiao-bin WANG. Application Progress of Spectral Detection Technology of Melamine in Food[J]. Spectroscopy and Spectral Analysis, 2022, 42(7): 1999 Copy Citation Text

show less

Fig. 1. SEM images of biosensor before (a) and after (b) cross-linking reaction^[29]

Download full size | View in the Article

Fig. 2. Color-coded chemical image of sample mixture (6% concentration) generated by binary detection^[32]

Download full size | View in the Article

Fig. 3. Comparison of Raman spectra before and after airPLS^[45]

Download full size | View in the Article

预处理方法	原理	特点	参考文献
导数校正	$X_{i}^{*} = \frac{X_{i + j} - X_{i}}{g}$	消除荧光背景干扰, 增强特征峰, 但会改变光谱峰形。	[43]
多元散射校正	$X_{i}^{*} = \frac{{\overset{̅}{X}}_{i} - b_{i}}{k_{i}}$	有效地消除不同散射能级引起的光谱差异, 增强光谱与数据的相关性。	[47]
标准正态变量变换	$X_{i}^{*} = \frac{X_{i} - \overset{̅}{X}}{\sqrt[]{\frac{\overset{m}{\sum_{k = 1}} (X_{k} {- \overset{̅}{X})}^{2}}{(m - 1)}}}$	有效地消除散射和粒径干扰, 校正基线漂移和旋转变化, 常用于粉末或填充密实样品的反射光谱。	[48]
高斯平滑滤波	$G_{(x_{i}^{}, y_{i}^{})} = \frac{1}{2 π σ^{2}} e^{- \frac{x_{i}^{2} + y_{i}^{2}}{2 σ^{2}}}$	适用于消除高斯噪声, 广泛应用于高光谱图像去噪。
S-G平滑滤波	$X_{i}^{*} = \frac{\overset{l}{\sum_{j = - r}} X_{i + j} W_{j}}{\overset{i}{\sum_{j = - l}} W_{j}}$	消声效果随窗口大小而变化, 可应用于多种场合。	[42]
自适应迭代加权惩罚最小二乘法	$W_{i}^{t} = \{\begin{array}{l} 0 & x_{i} ⩾ z_{i}^{t - 1} \\ e^{\frac{t (x_{i} - z_{i}^{t - 1})}{\| d^{t} \|}} & x_{i} < z_{i}^{t - 1} \end{array}$	从高维、非零变量的复杂数据中滤除背景基线, 得到与分析对象相对应的纯光谱特征数据。	[49]
中心归一化	$X_{i}^{*} = \frac{X_{i} - \overset{̅}{X}}{σ}$	对数据按比例进行缩放和变换, 使数据落入一个固定的区间内, 消除数据维数的影响。	[29]

Table 1. Main spectral pretreatment methods

View in the Article

模型名称	仪器	线性范围、定量限或检测限	模型评价	参考文献
一元线性回归	表面增强拉曼光谱	LOQ=0.5 μg·mL^-1	R²=0.999 8	[62]
多元线性回归	太赫兹光谱	LOD=4.55% 0.5%~19.99%	R²=0.97 RMSEP=1.38%	[33]
主成分分析回归	近红外光谱	0.1%~2%	R²=0.923 7 RMSEP=0.289%	[22]
偏最小二乘回归	表面增强拉曼光谱	LOD=0.016 5 mmol·L^-1 LOQ=0.055 mmol·L^-1	R²=0.97	[63]
卷积神经网络	近红外光谱	0%~10%	RMSEP=0.064 R²=0.995	[58]
支持向量机	表面增强拉曼光谱	LOQ=0.5 ppm	RMSEP=1.963 6 R²=0.973 6	[57]
直接硬建模回归	表面增强拉曼光谱	0.5~15 mg·kg^-1	R²=0.947 RMSEP=0.893%

Table 2. Melamine quantification regression model

View in the Article

模型名称	仪器	测试结果	参考文献
一类偏最小二乘	近红外光谱	训练集准确率94% 验证集准确率89%	[23]
分类和回归树	傅里叶变换红外光谱	训练集准确率95.5%	[56]
分类和回归树		验证集准确率88.5%
偏最小二乘判别		训练集准确率95.85%
偏最小二乘判别		验证集准确率93.87%
K-最邻近模型	训练集准确率95.21%
K-最邻近模型	验证集准确率91.37%
软独立分类	近红外光谱	训练集准确率100% 验证集准确率100%	[59]

Table 3. Classification model for melamine detection

Download Citation

Save the article for my favorites

Paper Information