Fast Classification Method of Black Goji Berry (<i>Lycium Ruthenicum</i> Murr.) Based on Hyperspectral and Ensemble Learning

Wei LU; Miao-miao CAI; Qiang ZHANG; Shan LI

doi:10.3964/j.issn.1000-0593(2021)07-2196-09

Spectroscopy and Spectral Analysis, Volume. 41, Issue 7, 2196(2021)

Fast Classification Method of Black Goji Berry (Lycium Ruthenicum Murr.) Based on Hyperspectral and Ensemble Learning

Wei LU^1、1;, Miao-miao CAI^1、1;, Qiang ZHANG^2、2;, and Shan LI^3、3;

¹1. Jiangsu Provincial Laboratory of Modern Facility Agriculture Technology and Equipment Engineering, College of Artificial Intelligence, Nanjing Agricultural University, Nanjing 210031, China

²2. School of Water Resources and Hydropower, Qinghai University, Xining 810016, China

³3. School of Life Science and Technology, Tongji University, Shanghai 200092, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(9)

Fig. 1. Hyperspectral imaging system

Download full size

View in Article

Fig. 2. Automatic ROI extraction in hyperspectral images

Download full size

View in Article

Fig. 3. Average reflection informations of carpopodium, sarcocarp and background

Download full size

View in Article

Fig. 4. Stacking ensemble learning process
(a): Training and prediction models for metamodels; (b): General flowchart

Download full size

View in Article

Fig. 5. Flow chart of fast and non-destructive grading model of black goji berry

Download full size

View in Article

Fig. 6. Spectral curves of black goji berries before and after pretreatment
(a): Raw spectra of carpopodium; (b): Raw spectra of sarcocarp; (c): Spectra of carpopodium after FD treatment; (d): Spectra of sarcocarp after FD treatment; (e): Average spectra of different grades of black goji berrycarpopodium after FD treatment; (f): Average spectra of different grades of black goji berrysarcocarp after FD treatment

Download full size

View in Article

Table 1. Anthocyanin content of four grades of black goji berry
View table
View in Article
Table 1. Anthocyanin content of four grades of black goji berry
等级名称花青素含量H/(mg·L^-1)
NMH-grade1 30.57
NMH-grade2 24.32
NMH-grade3 23.44
NMH-grade4 20.37

Table 2. PCA-based modeling results

View table

View in Article

Table 2. PCA-based modeling results

特征提取方法				PCA				SPA				CARS
分类器		预处理		果柄		果肉		果柄		果肉		0.783 3		0.933 3
LIBSVM		FD		0.800 0		0.779 2		0.666 7		0.883 3		0.300 0		0.250 0
		FFT		0.250 0		0.266 7		0.250 0		0.304 2		0.483 3		0.395 8
		HT		0.316 7		0.266 7		0.533 3		0.475 0		0.350 0		0.395 8
		SG		0.300 0		0.333 3		0.516 7		0.333 3		0.483 3		0.787 5
		Normalize		0.366 7		0.508 3		0.483 3		0.725 0		0.783 3		0.937 5
		SNV		0.683 3		0.775 0		0.766 7		0.895 8		0.516 7		0.800 0
LDA	FD		0.733 3		0.920 8		0.650 0		0.900 0		0.366 7		0.679 2
	FFT		0.400 0		0.337 5		0.500 0		0.554 2		0.466 7		0.654 2
	HT		0.416 7		0.579 2		0.516 7		0.729 2		0.450 0		0.495 8
	SG		0.633 3		0.687 5		0.833 3		0.820 8		0.912 5		0.933 3
	Normalize		0.383 3		0.333 3		0.912 5		0.941 7		0.383 3		0.795 8
	SNV		0.716 7		0.691 7		0.816 7		0.929 2		0.700 0		0.812 5
KNN	FD		0.716 7		0.758 3		0.716 7		0.787 5		0.616 7		0.858 3
	FFT		0.366 7		0.383 3		0.516 7		0.595 8		0.600 0		0.775 0
	HT		0.416 7		0.520 8		0.516 7		0.816 7		0.683 3		0.845 8
	SG		0.666 7		0.737 5		0.683 3		0.829 2		0.333 3		0.566 7
	Normalize		0.383 3		0.475 0		0.300 0		0.491 7		0.633 3		0.812 5
	SNV		0.516 7		0.737 5		0.600 0		0.812 5		0.783 3		0.941 7
RF	FD		0.666 7		0.829 2		0.766 7		0.912 5		0.550 0		0.812 5
	FFT		0.483 3		0.245 8		0.583 3		0.800 0		0.583 3		0.904 2
	HT		0.450 0		0.550 0		0.500 0		0.904 2		0.683 3		0.854 2
	SG		0.566 7		0.716 7		0.683 3		0.875 0		0.350 0		0.612 5
	Normalize		0.176 7		0.575 0		0.383 3		0.570 8		0.733 3		0.762 5
	SNV		0.566 7		0.854 2		0.616 7		0.783 3		0.750 0		0.900 0
NB	FD		0.733 3		0.883 3		0.816 7		0.883 3		0.483 3		0.641 7
	FFT		0.483 3		0.433 3		0.433 3		0.629 2		0.483 3		0.829 2
	HT		0.516 7		0.608 3		0.466 7		0.891 7		0.633 3		0.750 0
	SG		0.683 3		0.691 7		0.566 7		0.750 0		0.250 0		0.366 7
	Normalize		0.233 3		0.658 3		0.333 3		0.354 2		0.666 7		0.800 0
	SNV		0.666 7		0.716 7		0.483 3		0.766 7		0.783 3		0.933 3

Table 3. Modeling results of Stacking ensemble learning

View table

View in Article

Table 3. Modeling results of Stacking ensemble learning

特征提取方法	预处理	训练集		测试集
特征提取方法	预处理	果柄	果肉	果柄	果肉
PCA	FD	0.944 6	0.955 4	0.766 7	0.908 3
	FFT	0.326 8	0.300 0	0.316 7	0.262 5
	HT	0.598 2	0.337 5	0.383 3	0.312 5
	SG	0.557 1	0.726 8	0.616 7	0.687 5
	Normalize	0.408 9	0.935 7	0.350 0	0.512 5
	SNV	0.930 4	0.935 7	0.683 3	0.762 5
SPA	FD	0.837 5	0.975 0	0.750 0	0.983 3
	FFT	0.632 1	0.678 6	0.533 3	0.537 5
	HT	0.553 6	0.846 4	0.516 7	0.729 2
	SG	0.905 4	0.914 3	0.833 3	0.820 8
	Normalize	0.364 3	0.817 9	0.416 7	0.633 3
	SNV	0.955 4	0.950 0	0.816 7	0.912 5
CARS	FD	0.980 4	0.894 6	0.500 0	0.800 0
	FFT	0.455 4	0.554 2	0.266 7	0.483 3
	HT	0.728 6	0.782 1	0.450 0	0.604 2
	SG	0.450 0	0.612 5	0.300 0	0.404 2
	Normalize	0.941 1	0.950 0	0.916 7	0.804 2
	SNV	0.926 8	0.962 5	0.733 3	0.916 7

Tools

Get Citation

Copy Citation Text

Wei LU, Miao-miao CAI, Qiang ZHANG, Shan LI. Fast Classification Method of Black Goji Berry (Lycium Ruthenicum Murr.) Based on Hyperspectral and Ensemble Learning[J]. Spectroscopy and Spectral Analysis, 2021, 41(7): 2196

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites