Fusion of Visible Near-Infrared (VNIR) Hyperspectral Imaging and Texture Feature for Prediction of Total Phenolics Content in Tan Mutton

You-rui SUN; Mei GUO; Gui-shan LIU; Nai-yun FAN; Hao-nan ZHANG; Yue LI; Fang-ning PU; Shi-hu YANG; Hao WANG

doi:10.3964/j.issn.1000-0593(2022)11-3631-06

Journals >Spectroscopy and Spectral Analysis >Volume 42 >Issue 11 >Page 3631 > Article

Spectroscopy and Spectral Analysis
Vol. 42, Issue 11, 3631 (2022)

Fusion of Visible Near-Infrared (VNIR) Hyperspectral Imaging and Texture Feature for Prediction of Total Phenolics Content in Tan Mutton

You-rui SUN^1、*, Mei GUO^1、*, Gui-shan LIU^{1、1; *;}, Nai-yun FAN^{1、1; *;}, Hao-nan ZHANG^2、2;, Yue LI^1、1;, Fang-ning PU^2、2;, Shi-hu YANG^1、1;, and Hao WANG^2、2;

Author Affiliations

¹1. College of Food and Wine, Ningxia University, Yinchuan 750021, China

²2. School of Physics and Electronic-Electrical Engineering, Ningxia University, Yinchuan 750021, China

show less

DOI: 10.3964/j.issn.1000-0593(2022)11-3631-06 Cite this Article

You-rui SUN, Mei GUO, Gui-shan LIU, Nai-yun FAN, Hao-nan ZHANG, Yue LI, Fang-ning PU, Shi-hu YANG, Hao WANG. Fusion of Visible Near-Infrared (VNIR) Hyperspectral Imaging and Texture Feature for Prediction of Total Phenolics Content in Tan Mutton[J]. Spectroscopy and Spectral Analysis, 2022, 42(11): 3631 Copy Citation Text

show less

Fig. 1. Spectral curves of Tan mutton samples

Download full size | View in the Article

Fig. 2. Performances of full wavelength models based on different pre-processing methods

Download full size | View in the Article

Fig. 3. Selection of the characteristic wavelengths
(a): Change curve of mean weight value and the distribution map based on BOSS algorithm;(b): Change curve of mean weight value by CARS algorithm; (c): Changes map of RMSECV by VCPA-IRIV algorithm;(d): Distribution maps based on the characteristic wavelengths extracted by VCPA-IRIV and iVISSA

Download full size | View in the Article

Fig. 4. The images of first three PCs of mutton samples

Download full size | View in the Article

Fig. 5. Visualizaion maps of TPC content distributions

Download full size | View in the Article

Modeling methods	Selection method	LVs	Calibration set		Prediction set
Modeling methods	Selection method	LVs	$R_{C}^{2}$	RMSEC	$R_{P}^{2}$	RMSEP
PLSR	CARS	16	0.763 4	0.147 5	0.657 8	0.175 5
	BOSS	20	0.825 2	0.126 3	0.738 6	0.154 1
	iVISSA	20	0.852 6	0.116 0	0.735 7	0.156 4
	VCPA-IRIV	19	0.737 0	0.154 9	0.609 6	0.186 8
LSSVM	CARS	/	0.790 6	0.138 4	0.700 2	0.164 9
	BOSS	/	0.851 3	0.116 8	0.745 9	0.155 0
	iVISSA	/	0.896 8	0.097 5	0.714 2	0.162 6
	VCPA-IRIV	/	0.740 0	0.154 4	0.586 1	0.192 4

Table 1. Model performances based on different feature-wavelength methods

View in the Article

Modeling methods	Selection method	LVs	Calibration set		Prediction set
Modeling methods	Selection method	LVs	$R_{C}^{2}$	RMSEC	$R_{P}^{2}$	RMSEP
PLSR	BOSS-COR	20	0.825 6	0.126 2	0.718 4	0.159 8
	BOSS-ASM-CON	20	0.812 7	0.131 0	0.755 8	0.152 7
	BOSS-ASM-ENT-CON	20	0.808 5	0.131 4	0.761 4	0.151 0
	BOSS-COR-ASM-ENT-CON	20	0.807 4	0.132 8	0.742 4	0.157 8
LSSVM	BOSS-COR	/	0.858 1	0.114 0	0.751 2	0.150 5
	BOSS-ASM-CON	/	0.854 4	0.115 4	0.747 7	0.152 2
	BOSS-ASM-ENT-CON	/	0.850 0	0.116 0	0.770 9	0.144 7
	BOSS-COR-ASM-ENT-CON	/	0.857 2	0.114 3	0.757 8	0.148 5

Table 2. Model performance based on image and spectroscopy fusion

Download Citation

Save the article for my favorites

Paper Information