Chlorophyll Content Estimation of Northeast Japonica Rice Based on Improved Feature Band Selection and Hybrid Integrated Modeling

Tan LIU; Tong-yu XU; Feng-hua YU; Qing-yun YUAN; Zhong-hui GUO; Bo XU

doi:10.3964/j.issn.1000-0593(2021)08-2556-09

Journals >Spectroscopy and Spectral Analysis >Volume 41 >Issue 8 >Page 2556 > Article

Spectroscopy and Spectral Analysis
Vol. 41, Issue 8, 2556 (2021)

Chlorophyll Content Estimation of Northeast Japonica Rice Based on Improved Feature Band Selection and Hybrid Integrated Modeling

Tan LIU^*, Tong-yu XU^{1; 2; *;}, Feng-hua YU^{1; 2;}, Qing-yun YUAN^{1; 2;}, Zhong-hui GUO^1;, and Bo XU^1;

Author Affiliations

1. College of Information and Electrical Engineering, Shenyang Agricultural University, Shenyang 110161, China

show less

DOI: 10.3964/j.issn.1000-0593(2021)08-2556-09 Cite this Article

Tan LIU, Tong-yu XU, Feng-hua YU, Qing-yun YUAN, Zhong-hui GUO, Bo XU. Chlorophyll Content Estimation of Northeast Japonica Rice Based on Improved Feature Band Selection and Hybrid Integrated Modeling[J]. Spectroscopy and Spectral Analysis, 2021, 41(8): 2556 Copy Citation Text

show less

Fig. 1. Location map of test site

Download full size | View in the Article

Fig. 2. Spectral curves after SNV pretreatment

Download full size | View in the Article

Fig. 3. Overall structure diagram of GPR-P model

Download full size | View in the Article

Fig. 4. Results of correlation coefficient analysis

Download full size | View in the Article

Fig. 5. Feature band selection results based on SPA algorithm
(a): Number of the best feature bands in the sample model; (b): Distribution of extracted feature bands

Download full size | View in the Article

Fig. 6. Each band selection probability based on fpb-RF algorithm

Download full size | View in the Article

Fig. 7. Feature band selection results based on RF and fpb-RF algorithms
(a): RF algorithm; (b): fpb-RF algorithm

Download full size | View in the Article

Fig. 8. GPR-P model prediction results of different feature band extraction methods
(a): CC-SPA algorithm; (b): RF algorithm; (c): fpb-RF algorithm

Download full size | View in the Article

Fig. 9. PLSR model prediction results of different feature band extraction methods
(a): CC-SPA algorithm; (b): RF algorithm; (c): fpb-RF algorithm

Download full size | View in the Article

Fig. 10. LSSVM model prediction results of different feature band extraction methods
(a): CC-SPA algorithm; (b): RF algorithm; (c): fpb-RF algorithm

Download full size | View in the Article

Fig. 11. BP model prediction results of different feature band extraction methods
(a): CC-SPA algorithm; (b): RF algorithm; (c): fpb-RF algorithm

Download full size | View in the Article

样本集	样本数	最大值/ (mg·L^-1)	最小值/ (mg·L^-1)	平均值/ (mg·L^-1)	标准差/ (mg·L^-1)
总样本	102	12.811 6	1.702 2	6.772 9	2.338 6
训练集	77	12.811 6	1.702 2	6.700 6	2.416 8
测试集	25	11.686 3	1.934 1	6.989 7	2.117 8

Table 1. Statistics of chlorophyll content in rice

View in the Article

方法	迭代次数	收敛时间/min	均方根误差
RF	5 000	15.2	1.314 0
fpb-RF	800	2.1	1.232 8

Table 2. Optimization results of RF and fpb-RF algorithms

View in the Article

特征波段选取方法	模型	R²		RMSE
特征波段选取方法	模型	建模集	测试集	建模集	测试集
CC-SPA	CC-SPA-PLSR	0.670 7	0.592 5	1.377 9	1.398 2
RF	RF-PLSR	0.690 6	0.701 2	1.314 0	1.330 4
fpb-RF	fpb-RF-PLSR	0.717 9	0.704 7	1.232 8	1.275 5
CC-SPA	CC-SPA-LSSVM	0.736 3	0.739 2	1.215 1	1.156 9
RF	RF-LSSVM	0.760 1	0.757 9	1.184 5	1.071 7
fpb-RF	fpb-RF-LSSVM	0.775 2	0.767 5	1.067 6	1.025 6
CC-SPA	CC-SPA-BP	0.740 5	0.746 4	1.207 2	1.066 0
RF	RF-BP	0.763 9	0.765 0	1.184 1	1.048 9
fpb-RF	fpb-RF-PB	0.770 4	0.766 2	1.162 8	1.022 3
CC-SPA	CC-SPA-GPR-P	0.755 3	0.760 1	1.090 2	1.035 7
RF	RF-GPR-P	0.776 8	0.774 5	1.012 2	0.959 7
fpb-RF	fpb-RF-GPR-P	0.781 5	0.779 6	0.904 1	0.928 3

Table 3. Evaluation index of prediction model

Download Citation

Save the article for my favorites

Paper Information