Geographical Origin Discrimination of Taiping Houkui Tea Using Convolutional Neural Network and Near-Infrared Spectroscopy

Qi CHEN; Tian-hong PAN; Yu-qiang LI; Hong LIN

doi:10.3964/j.issn.1000-0593(2021)09-2776-06

Spectroscopy and Spectral Analysis, Volume. 41, Issue 9, 2776(2021)

Geographical Origin Discrimination of Taiping Houkui Tea Using Convolutional Neural Network and Near-Infrared Spectroscopy

Qi CHEN^{1、1; 3;}, Tian-hong PAN^{2、2; 4; *;}, Yu-qiang LI^4、4;, and Hong LIN^4、4;

Author Affiliations

¹1. School of Food and Biological Engineering, Hefei University of Technology, Hefei 230601, China

²2. School of Electrical Engineering and Automation, Anhui University, Hefei 230601, China

⁴4. School of Electrical and Information Engineering, Jiangsu University, Zhenjiang 212013, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(11)

Fig. 1. Feature selection process of 1-D CNN

Download full size

View in Article

Fig. 2. Spectra of Taiping Houkui tea
(a): Original data; (b): Preprocessed data

Download full size

View in Article

Fig. 3. Loss function values of training set for different 1-D CNN structure

Download full size

View in Article

Fig. 4. CIR of different convolution kernel sizes

Download full size

View in Article

Fig. 5. Model structure of 1-D CNN model

Download full size

View in Article

Fig. 6. Prediction results of 1-D CNN model

Download full size

View in Article

Fig. 7. Spectral feature distribution
(a): Original spectrum; (b): First convolutional layer; (c): Second convolutional layer; (d): Third convolutional layer

Download full size

View in Article

Table 1. Sample information

View table

View in Article

Table 1. Sample information

产地	样品规格数量	采摘时间	样品编号
猴坑	50 g×20个	2018.4.18	101-120
猴岗	50 g×20个	2018.4.17	201-220
颜家	50 g×20个	2018.4.17	301-320
三合	50 g×20个	2018.4.16	401-420
石河坑	50 g×20个	2018.4.16	501-520
汪王岭	50 g×20个	2018.4.16	601-620

Table 2. Prediction results with different sampling intervals
View table
View in Article
Table 2. Prediction results with different sampling intervals
采样训练集/% 测试集/% 时间/s
0 100 75.83 217.83
2 100 87.50 112.74
4 100 88.33 60.40
6 100 96.67 43.02
8 100 91.67 35.27
10 100 87.50 29.30

Table 3. CIR of different convolution kernel number
View table
View in Article
Table 3. CIR of different convolution kernel number
训练集/% 测试集/% 时间/s
16 100 82.49 55.77
32 100 96.16 46.80
64 100 98.33 96.84
128 100 96.67 218.82

Table 5. Comparison of Monte Carlo experimental results
View table
View in Article
Table 5. Comparison of Monte Carlo experimental results
方法变量
数累计
贡献
率/% 训练集
精度平均
值/% 预测集精度
平均值
/% 标准差
原始光谱数据 2 074 100 41.29 40.57 7.06
PCA 3 99.99 30.96 31.93 6.96
1-D CNN 64 - 98.48 97.73 3.47

Tools

Get Citation

Copy Citation Text

Qi CHEN, Tian-hong PAN, Yu-qiang LI, Hong LIN. Geographical Origin Discrimination of Taiping Houkui Tea Using Convolutional Neural Network and Near-Infrared Spectroscopy[J]. Spectroscopy and Spectral Analysis, 2021, 41(9): 2776

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites