Identification of Xinjiang Jujube Varieties Based on Hyperspectral Technique and Machine Learning

Liu Lixin; He Di; Li Mengzhu; Liu Xing; Qu Junle

doi:10.3788/CJL202047.1111002

Journals >Chinese Journal of Lasers >Volume 47 >Issue 11 >Page 1111002 > Article

Chinese Journal of Lasers
Vol. 47, Issue 11, 1111002 (2020)

Identification of Xinjiang Jujube Varieties Based on Hyperspectral Technique and Machine Learning

Liu Lixin^1,2,*, He Di¹, Li Mengzhu¹, Liu Xing³, and Qu Junle⁴

Author Affiliations

¹School of Physics and Optoelectronic Engineering, Xidian University, Xi''an, Shaanxi 710071, China

²State Key Laboratory of Transient Optics and Photonics, Xi''an Institute of Optics and Precision Mechanics of Chinese Academy of Sciences, Xi''an, Shaanxi 710119, China

³Sino-German College of Intelligent Manufacturing, Shenzhen Technology University, Shenzhen, Guangdong 518118, China

⁴College of Physics and Optoelectronic Engineering, Key Laboratory of Optoelectronic Devices and Systems of Ministry of Education and Guangdong Province, Shenzhen University, Shenzhen, Guangdong 518060, China

show less

DOI: 10.3788/CJL202047.1111002 Cite this Article Set citation alerts

Liu Lixin, He Di, Li Mengzhu, Liu Xing, Qu Junle. Identification of Xinjiang Jujube Varieties Based on Hyperspectral Technique and Machine Learning[J]. Chinese Journal of Lasers, 2020, 47(11): 1111002 Copy Citation Text

show less

Fig. 1. Schematic of hyperspectral system

Download full size

Spectral profiles of Jujube samples before and after preprocess. (a) Average spectra; (b) original spectra; (c) preprocessed spectra by MSC; (d) preprocessed spectra by SNV; (e) preprocessed spectra by 1-Der; (f) preprocessed spectra by SG smoothing

Fig. 2. Spectral profiles of Jujube samples before and after preprocess. (a) Average spectra; (b) original spectra; (c) preprocessed spectra by MSC; (d) preprocessed spectra by SNV; (e) preprocessed spectra by 1-Der; (f) preprocessed spectra by SG smoothing

Download full size

Fig. 3. Extracting characteristic bands by PCA. (a) Scores of the first three principal components; (b) variance contribution rate of the first ten principal components

Download full size

Fig. 4. Root-mean-square error calculated according to the number of selected characteristic variables

Download full size

Fig. 5. Extracting characteristic bands by CARS. (a) Variation curve of the number of variables with the number of sampling; (b) variation curve of RMSECV with the number of sampling; (c) variation path of variable regression coefficient

Download full size

Pretreatment method	Number of misjudgment			Overall accuracy /%
Pretreatment method	Jinsi jujube	Jun jujube	Tan jujube	Overall accuracy /%
Original	37	22	20	65.35
SG smoothing	37	22	20	65.35
MSC	29	26	35	60.53
SNV	29	25	35	60.96
1-Der	29	14	11	76.32

Table 1. Identification results of LDA model with different pretreatment methods

Pretreatment method	Number of misjudgment			Overall accuracy /%
Pretreatment method	Jinsi jujube	Jun jujube	Tan jujube	Overall accuracy /%
Original	18	16	5	82.89
SG Smoothing	18	16	5	82.89
MSC	14	12	12	83.33
SNV	14	12	12	83.33
1-Der	0	0	0	100

Table 2. Identification results of KNN model with different pretreatment methods

Pretreatment method	Number of misjudgment			Overall accuracy /%
Pretreatment method	Jinsi jujube	Jun jujube	Tan jujube	Overall accuracy /%
Original	5	4	1	95.61
SG Smoothing	3	3	2	96.49
MSC	2	4	2	96.49
SNV	2	4	2	96.49
1-Der	0	0	0	100

Table 3. Identification results of SVM model with different pretreatment methods

Characteristic bandsextraction method	Number of misjudgment			Overall accuracy /%
Characteristic bandsextraction method	Jinsi jujube	Jun jujube	Tan jujube	Overall accuracy /%
FS	29	14	11	76.32
PCA	13	20	19	77.19
SPA	18	16	10	80.70
CARS	15	12	6	85.53

Table 4. Identification results of LDA model with different characteristic bands extraction methods

Characteristic bandsextraction method	Number of misjudgment			Overall accuracy /%
Characteristic bandsextraction method	Jinsi jujube	Jun jujube	Tan jujube	Overall accuracy /%
FS	0	0	0	100
PCA	10	10	9	87.28
SPA	9	11	6	88.60
CARS	0	3	0	98.68

Table 5. Identification results of KNN model with different characteristic bands extraction methods

Characteristic bandsextraction method	Number of misjudgment			Overall accuracy /%
Characteristic bandsextraction method	Jinsi jujube	Jun jujube	Tan jujube	Overall accuracy /%
FS	0	0	0	100
PCA	8	10	7	89.04
SPA	4	5	3	94.74
CARS	2	2	0	98.25

Table 6. Identification results of SVM model with different characteristic bands extraction methods

Characteristic bands extraction method	Number of characteristic bands	Accuracy /%	Runtime /s
FS	1356	100	1.497
PCA	10	89.04	0.026
SPA	13	94.74	0.032
CARS	275	98.25	0.167

Table 7. Accuracy and runtime of SVM model based on different characteristic bands extraction methods

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information

微信扫一扫：分享

微信扫一扫：分享