A Hyperspectral Image-Based Method for Estimating Water and Chlorophyll Contents in Maize Leaves under Drought Stress

doi:10.12133/j.smartag.SA202308018

Abstract

Abstract:

[Objectives] Chlorophyll content and water content are key physiological indicators of crop growth, and their non-destructive detection is a key technology to realize the monitoring of crop growth status such as drought stress. This study took maize as an object to develop a hyperspectral-based approach for the rapid and non-destructive acquisition of the leaf chlorophyll content and water content for drought stress assessment. [Methods] Drought treatment experiments were carried out in a greenhouse of the College of Agriculture, Guangxi University. Maize plants were subjected to drought stress treatment at the seedling stage (four leaves). Four drought treatments were set up for normal water treatment [CK], mild drought [W1], moderate drought [W2], and severe drought [W3], respectively. Leaf samples were collected at the 3^rd, 6^th, and 9^th days after drought treatments, and 288 leaf samples were collected in total, with the corresponding chlorophyll content and water content measured in a standard laboratory protocol. A pair of push-broom hyperspectral cameras were used to collect images of the 288 seedling maize leaf samples, and image processing techniques were used to extract the mean spectra of the leaf lamina part. The algorithm flow framework of "pre-processing - feature extraction - machine learning inversion" was adopted for processing the extracted spectral data. The effects of different pre-processing methods, feature wavelength extraction methods and machine learning regression models were analyzed systematically on the prediction performance of chlorophyll content and water content, respectively. Accordingly, the optimal chlorophyll content and water content inversion models were constructed. Firstly, 70% of the spectral data was randomly sampled and used as the training dataset for training the inversion model, whereas the remaining 30% was used as the testing dataset to evaluate the performance of the inversion model. Subsequently, the effects of different spectral pre-processing methods on the prediction performance of chlorophyll content and water content were compared. Different feature wavelengths were extracted from the optimal pre-processed spectra using different algorithms, then their capabilities in preserve the information useful for the inversion of leaf chlorophyll content and water content were compared. Finally, the performances of different machine learning regression model were compared, and the optimal inversion model was constructed and used to visualize the chlorophyll content and water content. Additionally, the construction of vegetation coefficients were explored for the inversion of chlorophyll content and water content and evaluated their inversion ability. The performance evaluation indexes used include determination coefficient and root mean squared error (RMSE). [Results and Discussions] With the aggravation of stress, the reflectivity of leaves in the wavelength range of 400~1700 nm gradually increased with the degree of drought stress. For the inversion of leaf chlorophyll content and water content, combining stepwise regression (SR) feature extraction with Stacking regression could obtain an optimal performance for chlorophyll content prediction, with an R² of 0.878 and an RMSE of 0.317 mg/g. Compared with the full-band stacking model, SR-Stacking not only improved R² by 2.9%, reduced RMSE by 0.0356mg/g, but also reduced the number of model input variables from 1301 to 9. Combining the successive projection algorithm (SPA) feature extraction with Stacking regression could obtain the optimal performance for water content prediction, with an R² of 0.859 and RMSE of 3.75%. Compared with the full-band stacking model, SPA-Stacking not only increased R² by 0.2%, reduced RMSE by 0.03%, but also reduced the number of model input variables from 1301 to 16. As the newly constructed vegetation coefficients, normalized difference vegetation index(NDVI) [(R₄₁₀-R₅₅₉)/(R₄₁₀+R₅₅₉)] and ratio index (RI) (R₄₀₀/R₁₁₇₁) had the highest accuracy and were significantly higher than the traditional vegetation coefficients for chlorophyll content and water content inversion, respectively. Their R² were 0.803 and 0.827, and their RMSE were 0.403 mg/g and 3.28%, respectively. The chlorophyll content and water content of leaves were visualized. The results showed that the physiological parameters of leaves could be visualized and the differences of physiological parameters in different regions of the same leaves can be found more intuitively and in detail. [Conclusions] The inversion models and vegetation indices constructed based on hyperspectral information can achieve accurate and non-destructive measurement of chlorophyll content and water content in maize leaves. This study can provide a theoretical basis and technical support for real-time monitoring of corn growth status. Through the leaf spectral information, according to the optimal model, the water content and chlorophyll content of each pixel of the hyperspectral image can be predicted, and the distribution of water content and chlorophyll content can be intuitively displayed by color. Because the field environment is more complex, transfer learning will be carried out in future work to improve its generalization ability in different environments subsequently and strive to develop an online monitoring system for field drought and nutrient stress.

Key words: drought stress, hyperspectral, inversion of chlorophyll content, inversion of water content, mechine learning

WANG Jingyong, ZHANG Mingzhen, LING Huarong, WANG Ziting, GAI Jingyao. A Hyperspectral Image-Based Method for Estimating Water and Chlorophyll Contents in Maize Leaves under Drought Stress[J]. Smart Agriculture, 2023, 5(3): 142-153.

Figures/Tables 20

Fig. 1

Table 1

Fig. 2

Fig. 3

Table 2

Traditional vegetation indices

植被名称及计算公式		描述	参考文献
GI= $R 554 / R 667$	（6）	叶绿素含量相关的植被系数	［13］
CI_730 = $R 800 / R 730 ‒ 1.0$	（7）		［14］
CI_709 = $R 800 / R 709 ‒ 1.0$	（8）		［14］
Chl_green = $R 800 / R 550 ‒ 1.0$	（9）		［15］
NDRE = $(R 790 ‒ R 720) / (R 790 + R 720)$	（10）		［16］
RGVI = $R 550 / R 670$	（11）		［17］
MTCI = $(R 754 - R 709) / (R 709 + R 681)$	（12）		［18］
NDWI = $(R 860 ‒ R 1240) / (R 860 + R 1240)$	（13）	含水量相关的植被系数	［19］
MSI = $R 1600 / R 820$	（14）		［20］
hNDVI = $(R 900 ‒ R 680) / (R 900 + R 680)$	（15）		［21］
WI = $R 900 / R 970$	（16）		［22］
SRWI = $R 820 / R 1200$	（17）		［23］
NDII = $(R 820 ‒ R 1649) / (R 820 + R 1649)$	（18）		［21］

Table 2

Table 3

Vegetation index names and formulas

植被系数名称及计算公式		参考文献
$D I =$ $R i / R j$	（19）	［24］
$R I =$ $R i / R j$	（20）	［24］
$N D V I =$ $(R i + R j) / (R i + R j)$	（21）	［24］

Table 3

Fig. 4

Fig. 5

Table 4

Inversion performance of different pre-treated and original spectra for chlorophyll content and water content

反演目标	预处理方法	主成分数	训练集		验证集
反演目标	预处理方法	主成分数	$R t 2$	$R M S E t$	$R c v 2$	$R M S E c v$
叶绿素含量/（mg·g^-1）	无	10	0.857	0.307	0.839	0.312
	SG	10	0.866	0.298	0.832	0.318
	MSC	11	0.867	0.297	0.831	0.312
	SNV	12	0.873	0.289	0.823	0.326
	FD	6	0.908	0.247	0.838	0.312
	SD	4	0.908	0.246	0.809	0.339
	SG+MSC	11	0.857	0.307	0.839	0.311
	SG+SNV	12	0.872	0.291	0.825	0.324
	SG+FD	6	0.884	0.277	0.830	0.320
	SG+SD	4	0.898	0.260	0.807	0.341
含水量/%	无	11	0.877	3.05	0.843	3.57
	SG	11	0.877	3.09	0.843	3.58
	MSC	13	0.856	3.34	0.796	4.07
	SNV	14	0.863	3.26	0.702	4.92
	FD	5	0.849	3.42	0.624	5.53
	SD	4	0.868	3.20	0.710	4.85
	SG+MSC	13	0.855	3.35	0.796	4.07
	SG+SNV	14	0.861	3.28	0.706	4.89
	SG+FD	5	0.839	3.53	0.632	5.47
	SG+SD	4	0.862	3.27	0.734	4.65

Table 4

Fig. 6

Fig.7

Table 5

Inversion results of different models for chlorophyll content of maize leaves

特征波长提取方法	回归方法	训练集		测试集
特征波长提取方法	回归方法	$R t 2$	$R M S E t$ /（mg·g^-1）	$R p 2$	$R M S E p$ /（mg·g^-1）
无	ANN	0.849	0.292	0.812	0.394
	SVR	0.854	0.288	0.865	0.334
	KNN	0.831	0.309	0.766	0.439
	Stacking	0.870	0.272	0.849	0.353
SPA	ANN	0.898	0.241	0.816	0.390
	SVR	0.845	0.297	0.806	0.400
	KNN	0.844	0.298	0.750	0.455
	Stacking	0.869	0.273	0.840	0.364
相关系数	ANN	0.833	0.308	0.773	0.433
	SVR	0.752	0.375	0.741	0.462
	KNN	0.596	0.479	0.587	0.584
	Stacking	0.761	0.368	0.775	0.431
RF	ANN	0.807	0.331	0.774	0.432
	SVR	0.764	0.366	0.735	0.468
	KNN	0.615	0.467	0.594	0.579
	Stacking	0.791	0.343	0.778	0.428
SR	ANN	0.847	0.295	0.867	0.331
	SVR	0.828	0.312	0.849	0.353
	KNN	0.810	0.328	0.796	0.410
	Stacking	0.855	0.287	0.878	0.317

Table 5

Fig.8

Table 6

Inversion performance of different models for the water content of maize leaves

特征波长提取方法	回归方法	训练集		测试集
特征波长提取方法	回归方法	$R t 2$	$R M S E t$ /%	$R p 2$	$R M S E p$ /%
无	ANN	0.767	3.99	0.835	4.06
	SVR	0.709	4.47	0.799	4.47
	KNN	0.777	3.92	0.804	4.41
	Stacking	0.792	3.78	0.857	3.78
SPA	ANN	0.746	4.19	0.809	4.37
	SVR	0.710	4.47	0.784	4.64
	KNN	0.812	3.60	0.827	4.16
	Stacking	0.815	3.58	0.859	3.75
相关系数	ANN	0.597	5.28	0.711	5.37
	SVR	0.508	5.83	0.595	6.36
	KNN	0.643	4.96	0.699	5.48
	Stacking	0.639	4.99	0.724	5.24
RF	ANN	0.604	5.23	0.699	5.47
	SVR	0.536	5.66	0.539	6.77
	KNN	0.639	4.99	0.695	5.51
	Stacking	0.633	5.04	0.762	4.87
SR	ANN	0.753	4.13	0.821	4.22
	SVR	0.719	4.40	0.805	4.41
	KNN	0.764	4.03	0.781	4.67
	Stacking	0.785	3.85	0.848	3.90

Table 6

Fig. 9

Table 7

Performance of different vegetation indices in the inversion of chlorophyll content and water content

反演目标	植被系数	训练集		测试集
反演目标	植被系数	$R c 2$	$R M S E c$	$R p 2$	$R M S E p$
叶绿素含量/（mg·g^-1）	GI	0.130	0.704	0.015	0.902
	CI_730	0.281	0.640	0.150	0.838
	CI_709	0.367	0.601	0.292	0.765
	Chl_green	0.523	0.522	0.438	0.687
	NDRE	0.300	0.632	0.187	0.819
	RGVI	0.126	0.706	0.012	0.903
	MTCI	0.271	0.645	0.208	0.809
含水量/%	NDWI	0.034	8.25	0.015	10.74
	MSI	0.324	6.14	0.150	9.51
	hNDVI	0.204	7.01	0.293	8.20
	WI	0.400	5.68	0.338	7.80
	SRWI	0.300	6.31	0.187	9.13
	NDII	0.126	7.55	0.012	10.77

Table 7

Fig. 10

Table 8

The maximum correlation coefficient and wavelength position corresponding to the newly constructed vegetation coefficient

植被系数	叶绿素含量		含水量
植被系数	$R m a x$	波长位置（i，j）/nm	$R m a x$	波长位置（i，j）/nm
DI	0.882	（420，558）	0.783	（742，1681）
RI	0.892	（420，559）	0.898	（400，1171）
NDVI	0.889	（410，559）	0.890	（410，1348）

Table 8

Table 9

The inversion performance of the newly constructed vegetation coefficients for chlorophyll content and water content

反演目标	植被系数	回归方程	训练集		测试集
反演目标	植被系数	回归方程	$R c 2$	$R M S E c$	$R p 2$	$R M S E p$
叶绿素含量/（mg·g^-1）	DI（420，558）	y=28.96x+1.959 （22）	0.769	0.363	0.785	0.421
	RI（420，559）	y= ‒8.691x+10.914 （23）	0.774	0.358	0.791	0.415
	NDVI（410，559）	y=15.13x+2.268 （24）	0.784	0.351	0.803	0.403
含水量/%	DI（742，1681）	y=2.014x+1.340 （25）	0.731	3.42	0.799	3.53
	RI（400，1171）	y= ‒0.9413x+1.5327 （26）	0.788	3.04	0.827	3.28
	NDVI（410，1348）	y= ‒1.526x+0.5752 （27）	0.775	3.12	0.807	3.46

Table 9

Fig. 11

References 25

1	吕梦琪, 宋宇杰, 翁海勇, 等. 近红外高光谱成像扫描速度对拟南芥冠层含水率预测的影响[J]. 光谱学与光谱分析, 2020, 40(11): 3508-3514.
	LYU M Q, SONG Y J, WENG H Y, et al. Effect of near infrared hyperspectral imaging scanning speed on prediction of water content in Arabidopsis [J]. Spectroscopy and spectral analysis, 2020, 40(11): 3508-3514.
2	张金富, 汤斌, 王建旭, 等. 基于近红外光谱的苹果可溶性固形物特征波长筛选建模对比分析[J/OL].激光与光电子学进展. (2022-11-01)[2023-08-12].
	ZHANG J F, TANG B, WANG J X, et al. Comparative analysis of characteristic wavelength screening of apple soluble solids based on near-infrared spectroscopy[J/OL]. Laser & Optoelectronics Progress. (2022-11-01) [2023-08-12].
3	杨晶晶, 卢俊玮. 光谱技术在花生品质检测中的应用研究进展[J]. 作物研究, 2018, 32(4): 345-348.
	YANG J J, LU J W. Research progress of spectral technology in peanut quality detection[J]. Crop research, 2018, 32(4): 345-348.
4	孙俊, 金夏明, 毛罕平, 等. 基于高光谱图像的生菜叶片氮素含量预测模型研究[J]. 分析化学, 2014, 42(5): 672-677.
	SUN J, JIN X M, MAO H P, et al. A model for predicting nitrogen of lettuce leaves based on hyperspectral imaging[J]. Chinese journal of analytical chemistry, 2014, 42(5): 672-677.
5	GAO D H, LI M Z, ZHANG J Y, et al. Improvement of chlorophyll content estimation on maize leaf by vein removal in hyperspectral image[J]. Computers and electronics in agriculture, 2021, 184: ID 106077.
6	TAN K Z, WANG S W, SONG Y Z, et al. Estimating nitrogen status of rice canopy using hyperspectral reflectance combined with BPSO-SVR in cold region[J]. Chemometrics and intelligent laboratory systems, 2018, 172: 68-79.
7	SHU M Y, DONG Q Z, FEI S P, et al. Improved estimation of canopy water status in maize using UAV-based digital and hyperspectral images[J]. Computers and electronics in agriculture, 2022, 197: ID 106982.
8	LICHTENTHALER H K, WELLBURN A R. Determinations of total carotenoids and chlorophylls a and b of leaf extracts in different solvents[J]. Biochemical society transactions, 1983, 11(5): 591-592.
9	MASSAOUDI M, REFAAT S S, ABU-RUB H, et al. PLS-CNN-BiLSTM: An end-to-end algorithm-based savitzky-golay smoothing and evolution strategy for load forecasting[J]. Energies, 2020, 13(20): ID 5464.
10	BAO Y D, KONG W W, LIU F, et al. Detection of glutamic acid in oilseed rape leaves using near infrared spectroscopy and the least squares-support vector machine[J]. International journal of molecular sciences, 2012, 13(11): 14106-14114.
11	CHEN Q S, JIANG P, ZHAO J W. Measurement of total flavone content in snow lotus (Saussurea involucrate) using near infrared spectroscopy combined with interval PLS and genetic algorithm[J]. Spectrochimica acta part A: Molecular and biomolecular spectroscopy, 2010, 76(1): 50-55.
12	ARAÚJO M C U, SALDANHA T C B, GALVÃO R K H, et al. The successive projections algorithm for variable selection in spectroscopic multicomponent analysis[J]. Chemometrics and intelligent laboratory systems, 2001, 57(2): 65-73.
13	ZARCO-TEJADA P J, BERJÓN A, LÓPEZ-LOZANO R, et al. Assessing vineyard condition with hyperspectral indices: Leaf and canopy reflectance simulation in a row-structured discontinuous canopy[J]. Remote sensing of environment, 2005, 99(3): 271-287.
14	MEHDAOUI R, ANANE M. Exploitation of the red-edge bands of sentinel 2 to improve the estimation of durum wheat yield in Grombalia region (Northeastern Tunisia)[J]. International journal of remote sensing, 2020, 41(23): 8986-9008.
15	GITELSON A A, KEYDAN G P, MERZLYAK M N. Three-band model for noninvasive estimation of chlorophyll, carotenoids, and anthocyanin contents in higher plant leaves[J]. Geophysical research letters, 2006, 33(11): 431-433.
16	FITZGERALD G, RODRIGUEZ D, O'LEARY G. Measuring and predicting canopy nitrogen nutrition in wheat using a spectral index: The canopy chlorophyll content index (CCCI)[J]. Field crops research, 2010, 116(3): 318-324.
17	JORDAN C F. Derivation of leaf-area index from quality of light on the forest floor[J]. Ecology, 1969, 50(4): 663-666.
18	DASH J, CURRAN P J. The MERIS terrestrial chlorophyll index[J]. International journal of remote sensing, 2004, 25(23): 5403-5413.
19	GAO B C. NDWI: A normalized difference water index for remote sensing of vegetation liquid water from space[J]. Remote sensing of environment, 1996, 58(3): 257-266.
20	HUNT E R, ROCK B N. Detection of changes in leaf water content using near- and middle-Infrared reflectances[J]. Remote sensing of environment, 1989, 30(1): 43-54.
21	ZHANG J J, ZHANG W, XIONG S P, et al. Comparison of new hyperspectral index and machine learning models for prediction of winter wheat leaf water content[J]. Plant methods, 2021, 17(1): 1-14.
22	PENUELAS J, PINOL J, OGAYA R, et al. Estimation of plant water concentration by the reflectance Water Index WI (R900/R970)[J]. International journal of remote sensing, 1997, 18(13): 2869-2875.
23	ZARCO-TEJADA P J, USTIN S L. Modeling canopy water content for carbon estimates from MODIS data at land EOS validation sites[C]// IGARSS 2001. Scanning the Present and Resolving the Future. Proceedings. IEEE 2001 International Geoscience and Remote Sensing Symposium (Cat. No.01CH37217). Piscataway, New Jersey, USA: IEEE, 2002: 342-344.
24	SUN H, FENG M C, XIAO L J, et al. Assessment of plant water status in winter wheat (Triticum aestivum L.) based on canopy spectral indices[J]. PLoS one, 2019, 14(6): ID e0216890.
25	MENZE B.H., KELM B.M., MASUCH R . et al. A comparison of random forest and its Gini importance with standard chemometric methods for the feature selection and classification of spectral data[J]. BMC Bioinformatics, 2009, 10(1): 1-16.

技术参数	相机型号
技术参数	FX10	FX17
光谱范围/nm	400~1000	900~1700
光谱分辨率/nm	5.5	8
狭缝宽度/µm	30	30
空间像素数	1024	640
像素大小/µm	8×8	15×15

[1]	FAN Chengzhi, WANG Ziwen, YANG Xingchao, LUO Yongkai, XU Xuexin, GUO Bin, LI Zhenhai. Machine Learning Inversion Model of Soil Salinity in the Yellow River Delta Based on Field Hyperspectral and UAV Multispectral Data [J]. Smart Agriculture, 2022, 4(4): 61-73.
[2]	GUO Zhiming, WANG Junyi, SONG Ye, ZOU Xiaobo, CAI Jianrong. Research Progress of Sensing Detection and Monitoring Technology for Fruit and Vegetable Quality Control [J]. Smart Agriculture, 2021, 3(4): 14-28.
[3]	YANG Feifei, LIU Shengping, ZHU Yeping, LI Shijuan. Identification and Level Discrimination of Waterlogging Stress in Winter Wheat Using Hyperspectral Remote Sensing [J]. Smart Agriculture, 2021, 3(2): 35-44.
[4]	LIU Tianlin, ZHU Xicun, BAI Xueyuan, PENG Yufeng, LI Meixuan, TIAN Zhongyu, JIANG Yuanmao, YANG Guijun. Hyperspectral Estimation Model Construction and Accuracy Comparison of Soil Organic Matter Content [J]. Smart Agriculture, 2020, 2(3): 129-138.
[5]	Yang Chenghai. Airborne remote sensing systems for precision agriculture applications [J]. Smart Agriculture, 2020, 2(1): 1-22.
[6]	Yu Fenghua, Xu Tongyu, Guo Zhonghui, Du Wen, Wang Dingkang, Cao Yingli. Remote sensing inversion of chlorophyll content in rice leaves in cold region based on Optimizing Red-edge Vegetation Index (ORVI) [J]. Smart Agriculture, 2020, 2(1): 77-86.
[7]	Fu Yuanyuan, Yang Guijun, Duan Dandan, Zhang Yongtao, Gu Xiaohe, Yang Xiaodong, Xu Xingang, Li Zhenhai. Comparison analysis of spatial and spectral feature in vegetation classification based on AVIRIS hyperspectral image [J]. Smart Agriculture, 2020, 2(1): 68-76.