Nondestructive Testing and Grading of Milk Quality Based on Fourier Transform Mid-Infrared Spectroscopy

Shi-jie XIAO; Qiao-hua WANG; Chun-fang LI; Chao DU; Zeng-po ZHOU; Sheng-chao LIANG; Shu-jun ZHANG

doi:10.3964/j.issn.1000-0593(2022)04-1243-07

Journals >Spectroscopy and Spectral Analysis >Volume 42 >Issue 4 >Page 1243 > Article

Spectroscopy and Spectral Analysis
Vol. 42, Issue 4, 1243 (2022)

Nondestructive Testing and Grading of Milk Quality Based on Fourier Transform Mid-Infrared Spectroscopy

Shi-jie XIAO^1、*, Qiao-hua WANG^{1、1; 2; *;}, Chun-fang LI^{3、3; 4;}, Chao DU^3、3;, Zeng-po ZHOU^4、4;, Sheng-chao LIANG^4、4;, and Shu-jun ZHANG^{3、3; *;}

Author Affiliations

¹1. College of Engineering, Huazhong Agricultural University, Wuhan 430070, China

³3. Key Laboratory of Animal Breeding and Reproduction of Minstry of Education, Huazhong Agricultural University, Wuhan 430070, China

⁴4. Hebei Animal Husbandry Association, Shijiazhuang 050000, China

show less

DOI: 10.3964/j.issn.1000-0593(2022)04-1243-07 Cite this Article

Shi-jie XIAO, Qiao-hua WANG, Chun-fang LI, Chao DU, Zeng-po ZHOU, Sheng-chao LIANG, Shu-jun ZHANG. Nondestructive Testing and Grading of Milk Quality Based on Fourier Transform Mid-Infrared Spectroscopy[J]. Spectroscopy and Spectral Analysis, 2022, 42(4): 1243 Copy Citation Text

show less

Fig. 1. Mean spectra of special quality milk, high protein characteristic milk, high fat characteristic milk and ordinary milk

Download full size | View in the Article

Fig. 2. Screening characteristic wavelengths by UVE

Download full size | View in the Article

Fig. 3. (a) Number of sampling variables; (b) RMSECV; (c) Regression coefficient path

Download full size | View in the Article

Fig. 4. Hierarchical model based on the second order difference-UVE-SCARS-RF

Download full size | View in the Article

牧场标记	样本数量
牧场标记	特优优质奶	高蛋白特色奶	高乳脂特色奶	普通奶
1	118	183	50	119
2	102	56	151	92
3	144	106	338	249
4	40	22	112	44
5	258	132	167	143
6	187	131	130	141
7	63	66	41	62
8	169	149	81	147
9	139	221	57	176
10	122	89	165	159

Table 1. Sample distribution statistics of each pasture

View in the Article

牛奶标记	级别	脂肪/%	蛋白质/%	体细胞/ (10⁴个·mL^-1)
A	特优优质奶	≥4且≤9	≥3.7且≤7	≤20
B	高蛋白特色奶	≥1.5且≤3.4	≥3.7且≤7	≤50
C	高乳脂特色奶	≥4且≤9	≥1且≤3.1	≤50
D	普通奶	3.14	2.83.4	≤100

Table 2. Standard of classification

View in the Article

模型	预处理	训练集准确率/%					测试集准确率/%
模型	预处理	Total	A	B	C	D	Total	A	B	C	D
NB	全光谱	84.50	87.77	81.09	85.86	82.85	84.22	89.80	82.66	83.46	80.70
	SNV	90.72	89.47	94.56	92.71	86.71	88.46	87.06	92.20	89.66	85.46
	MSC	90.97	89.89	94.93	92.27	87.35	88.98	89.05	92.20	89.92	85.21
	一阶导数	90.69	92.34	88.63	91.27	90.25	88.85	94.03	84.68	89.15	86.97
	二阶导数	93.76	95.74	94.68	93.26	91.43	92.05	96.77	93.06	90.70	87.72
	一阶差分	90.38	92.34	87.64	90.94	90.25	88.85	93.78	84.39	89.41	87.22
	二阶差分	94.31	96.28	95.92	93.59	91.64	92.11	96.52	94.51	90.44	87.22
RF	全光谱	99.86	100	100	99.56	99.89	95.83	97.26	96.53	95.09	94.49
	SNV	99.86	100	100	99.56	99.89	94.52	97.01	95.66	95.09	90.48
	MSC	99.86	100	100	99.56	99.89	94.33	97.01	95.66	95.09	89.72
	一阶导数	99.86	100	100	99.67	99.79	96.15	98.01	97.69	96.12	92.98
	二阶导数	99.86	100	100	99.56	99.89	96.41	98.01	97.69	96.12	93.99
	一阶差分	99.86	100	100	99.56	99.89	96.09	98.01	97.69	95.87	92.98
	二阶差分	99.86	100	100	99.56	99.89	96.87	98.76	97.69	96.90	94.24

Table 3. Full spectrum prediction model using different pre-processing methods

View in the Article

模型	筛选方法	变量数	训练集准确率/%					测试集准确率/%
模型	筛选方法	变量数	Total	A	B	C	D	Total	A	B	C	D
NB	全光谱	514	94.31	96.28	95.92	93.59	91.64	92.11	96.52	94.51	90.44	87.22
	UVE	229	94.17	96.06	95.92	93.26	91.64	92.70	97.76	92.49	90.96	89.47
	CARS	37	93.73	95.21	95.43	94.14	90.35	92.50	96.77	93.06	92.25	87.97
	SCARS	20	94.45	95.96	95.30	94.92	91.75	93.94	97.26	93.93	93.02	91.48
	UVE-CARS	30	93.95	95.11	95.30	93.59	91.96	93.42	97.01	93.35	92.51	90.73
	UVE-SCARS	47	94.68	95.96	95.92	94.59	92.39	93.61	97.26	93.35	93.28	90.48
RF	全光谱	514	99.86	100	100	99.56	99.89	96.87	98.76	97.69	96.90	94.24
	UVE	229	99.86	100	100	99.56	99.89	96.74	98.51	97.69	96.38	94.49
	CARS	37	99.86	100	100	99.56	99.89	95.76	97.76	95.95	95.87	93.48
	SCARS	20	99.86	100	100	99.56	99.89	95.57	98.51	95.95	94.83	92.98
	UVE-CARS	30	99.86	100	100	99.67	99.79	95.83	97.26	97.40	96.12	92.73
	UVE-SCARS	47	99.86	100	100	99.56	99.89	96.48	98.26	97.40	95.87	94.49

Table 4. Prediction results by NB and RF models

Download Citation

Save the article for my favorites

Paper Information