Estimating cotton leaf nitrogen by combining the bands sensitive to nitrogen concentration and oxidase activities using hyperspectral imaging

doi:10.1016/j.compag.2021.106390

Computers and Electronics in Agriculture

Volume 189, October 2021, 106390

https://doi.org/10.1016/j.compag.2021.106390 Get rights and content

Highlights

•
Leaf nitrogen concentration was significantly correlated with MDA and POD activities.
•
Band extraction by SPA reduced the collinearity and redundancy of N-sensitive bands.
•
Combining of LNC and OA-sensitive bands improved the model accuracy.

Abstract

Oxidase activities (OA) are highly correlated with the nitrogen concentration in crop leaves. To improve the universality and inter-annual repeatability of the model for estimating cotton leaf nitrogen concentration (LNC), a method for model construction was proposed based on the combination of the bands sensitive to LNC with the bands sensitive to OA. In this plot experiment, 320 and 250 sets of hyperspectral data of cotton leaves in seedling stage, bud stage, initial flowering stage, full flowering stage, and boll setting stage were collected in 2019 and 2020, respectively by using hyperspectral technology, and the LNC and OA were also measured in indoor biochemical experiments. Then, successive projection algorithm (SPA) was used to analyze the LNC and OA-sensitive bands in the original spectrum and five kinds of spectral conversions in 2019, to construct the partial least squares regression (PLSR) and principal component regression (PCR) models. Finally, the accuracy of the models were verified using the spectral data in 2020. The results showed that the selection of LNC and OA-sensitive bands could greatly reduce the collinearity and redundant information among bands. The accuracy of the models based on the LNC and OA-sensitive bands in all stages were higher than those of the models based on the LNC-sensitive bands. The optimal was the model based on the malondialdehyde (MDA), peroxidase (POD), and LNC sensitive bands in full flowering stage, with determination coefficients (R²) of 0.846, root mean squared error (RMSE) of 3.081, and residual prediction deviation (RPD) of 2.975. The universality and inter-annual repeatability of the optimal model were significantly improved, with R² increasing by 12.24%-79.89% and RMSE reducing by 19.80%-72.52%, compared with those of the model based on the LNC-sensitive bands. Besides, the accuracy and stability of PLSR models were significantly higher than those of PCR models. In conclusion, the combination of LNC and OA-sensitive bands could obviously improve the accuracy and universality of the LNC estimation model. This study provides a new method for improving the accuracy and universality of crop nitrogen estimation model.

Introduction

Nitrogen (N) is one of the macronutrient elements necessary for cotton growth, but also a key factor in yield formation. Accurate monitoring of cotton nitrogen status is the basis for yield increase and fertilizer reduction (Wei et al., 2002). Traditional nitrogen-monitoring methods are time-consuming and laborious (Guo et al., 2017), while hyperspectral imaging technology could combine the spectral data and texture information of the samples to extract the information, and accurately and non-destructively identify the external characteristics (Chen and Liang, 2019). By establishing linear or nonlinear relationship between leaf nitrogen concentration (LNC) and spectral reflectance, hyperspectral imaging has been widely used in agricultural production (Liu et al., 2015, Li et al., 2020).

Scholars have proposed that forty-two absorption characteristics in the visible and near infrared region induced by bending and stretching of chemical bonds in crop leaves are closely related to nitrogen (Curran, 1989). Many studies have constructed regression models based on partial least squares, successive projection algorithm (SPA), independent component analysis (ICA), and artificial neural network (ANN) (Wang et al., 2018, Sabzi et al., 2021) for the applications in rice, wheat, cotton, and other crops (Sun et al., 2019, Blackmer et al., 1994, Li et al., 2021). Moreover, at present, many studies construct LNC estimation model using the sensitive bands selected based on the correlation between leaf spectra and LNC. However, the correlations between leaf spectra and N-related components such as proteins, amino acids, oxidases, etc. in the crop were neglected, resulting in that the sensitive bands selected were not completely the response bands of the chemical bonds of N (Tian et al., 2007). The changes of time and space bring to the vibrational shift in the radiation or light energy on the leaf, leading to the differentials in the reflection and absorption spectra. Therefore, the “drift” phenomenon always appeared for sensitive bands (Kong et al., 2017), which may decrease the accuracy of the nitrogen estimation model, especially in the inter-annual tests (Kokaly and Clark, 1999, Sripada et al., 2005). In-depth analysis of the characterization of biochemical components closely related to crop nitrogen, and the synergistic responses between nitrogen and its related physical and chemical characteristics, are of great significance to improve the accuracy and universality of nitrogen estimation model.

Oxidase is a protease closely related to nitrogen in crops. It has been used to indicate nitrogen status. Under nitrogen stress, the balance of reactive oxygen species defense system is destroyed, and malondialdehyde (MDA) with cytotoxicity is produced, which could inhibit the crop growth (Ma et al., 2015). However, peroxidase (POD), superoxide dismutase (SOD), and catalase (CAT) play an important role in the elimination of reactive oxygen species. Scholars have selected the bands sensitive to MDA content (Wang et al., 2020) and POD and SOD activities (Sgherri et al., 2001) to construct estimation models, achieving real-time and non-destructive detection of crop nitrogen status through using the chemical components closely related to nitrogen to characterize the nitrogen status.

In this study, the bands sensitive to MDA content and SOD, POD, and CAT activities in the leaves of nitrogen-stressed cotton were selected using SPA and combined with the bands sensitive to LNC to construct nitrogen estimation models, aiming to improve the stability and inter-annual repeatability of nitrogen estimation model. The specific objectives were: (1) to select the bands sensitive to LNC and OA and make clear their distribution based on hyperspectral imaging technology, (2) to compare the accuracy of the model based on the LNC-sensitive bands and the model based on the LNC and OA-sensitive bands, and (3) to verify the accuracy of the model based on the LNC-sensitive bands and the model based on the LNC and OA-sensitive bands using inter-annual spectral data. This study provides guidance for improving the stability and inter-annual repeatability of nitrogen estimation model, which is helpful for the cotton nitrogen monitoring and fertilization management.

Section snippets

Experimental design

The experiment was carried out at the Experimant Station of Shihezi University in Shihezi City, Xinjiang Province, China (45°19′N, 86°3′E) in 2019 and 2020, where there has a continental arid climate. The soil is grey desert soil, with organic matter concentration of 11.5 g·kg⁻¹, total nitrogen concentration of 750 mg·kg⁻¹, available nitrogen concentration of 93.6 mg·kg⁻¹, available phosphorus concentration of 18.7 mg·kg⁻¹, and available potassium concentration of 242 mg·kg⁻¹ in the 0–20 cm

Effect of different nitrogen application rates on the nitrogen concentration of cotton leaves

The LNC in all nitrogen treatments increased and then decreased over time. The LNC in the N_0.8 treatment in the full flowering stage was 6.68 mg·kg⁻¹, which was 48.78% higher than that in the seedling stage (P < 0.05). With the increase of nitrogen application rate, the LNC showed an upward trend. There was no difference in the N_0.8 and N_c treatment (P > 0.05) (Fig. 3).

Effect of different nitrogen application rates on the oxidase activities of cotton leaves

The MDA content of cotton leaves in all treatments showed a trend of first decreasing and then increasing over time (Fig. 4a).

Discussion

Spectroscopic detection of crop nitrogen is mainly through the vibration of chemical bonds in the molecular structure of different forms of nitrogen such as protein, amino acid, rubisco carboxylase, etc., which occurs under different radiation levels or wavelength light, forming different reflection and absorption spectra (Grossman et al., 1996). In this study, it was found that the LNC-sensitive bands were mainly distributed in the visible region and less in the near-infrared region. Other

Conclusion

To improve the accuracy and inter-annual repeatability of the model for estimating cotton leaf nitrogen, the bands sensitive to oxidase activities (MDA, CAT, and POD) which are highly correlated with leaf nitrogen, were selected to construct the estimation model. The data conversion, especially the (1/SG)'' conversion, could obviously improve the correlations of R, LNC, and OA, and decrease the collinearity and redundant information among bands. The model based on the combination of

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work is supported by the International Science and Technology Cooperation Project (2015DFA11660), Corps Science and Technology Project (2018AA004, 2018AA005, and 2020AB018), and Shihezi University Project (RCZX201425 and RCZK20208).

References (41)

M.C.U. Araújo et al.
The successive projections algorithm for variable selection in spectroscopic multicomponent analysis
Chemometrics Intelligent Lab. Syst.
(2001)
P.J. Curran
Remote sensing of foliar chemistry
Remote Sens. Environ.
(1989)
Y.L. Grossman et al.
Critique of stepwise multiple linear regression for the extraction of leaf biochemistry information from leaf reflectance data
Remote Sens. Environ.
(1996)
B.B. Guo et al.
Remotely assessing leaf N uptake in winter wheat based on canopy hyperspectral red-edge absorption
Eur. J. Agron.
(2017)
W. Kong et al.
Estimation of canopy carotenoid content of winter wheat using multi-angle hyperspectral data
Adv. Space Res.
(2017)
R.F. Kokaly et al.
Spectroscopic determination of leaf biochemistry using band-depth analysis of absorption features and stepwise multiple linear regression
Remote Sens. Environ.
(1999)
D.L. Peterson et al.
Remote sensing of forest canopy and leaf biochemical contents
Remote Sens. Environ.
(1988)
T. Pholpho et al.
Classification of longan fruit bruising using visible spectroscopy
J. Food Eng.
(2011)
C. Sgherri et al.
Antioxidative enzymes in two wheat cultivars, differently sensitive to drought and subjected to subsymptomatic copper doses
J. Plant Physiol.
(2001)
J. Wang et al.
Wavelet-based coupling of leaf and canopy reflectance spectra to improve the estimation accuracy of foliar nitrogen concentration
Agric. For. Meteorol.
(2018)

B.J. Yoder et al.

Predicting nitrogen and chlorophyll content and concentrations from reflectance spectra (400–2500 nm) at leaf and canopy scales

Remote Sens. Environ.

(1995)

S.D. Bao

Soil and Agricultural Chemistry Analysis

(2000)

T.M. Blackmer et al.

Light reflectance compared with other nitrogen stress measurements in corn leaves

Agron. J.

(1994)

X. Cai et al.

The development of biosensors and biochips in IECAS. Network & Parallel Computing, Ifip International Conference, Npc, Wuhan, China

October. DBLP.

(2004)

J. Chen et al.

N status monitoring model in winter wheat based on image processing

Trans. Chinese Soc. Agri. Eng.

(2016)

P.F. Chen et al.

Diagnosis of cotton nitrogen nutrition based on low-altitude drone image spectrum and texture characteristics

Chinese Agri. Sci.

(2019)

L. Chen et al.

Effects of nitrogen forms and ratios on plant growth, seed antioxidant enzyme activities and reactive oxygen metabolism of vegetable soybean

Plant Nutrition and Fertilizer Science.

(2010)

Inoue, Y., Guérif, M., Baret, F., Skidmore, A., Gitelson, A., Schlerf, M., Darvishzadeh., Olioso., 2016. Simple and...

T.S. Li et al.

Monitoring of leaf nitrogen content of winter wheat using multi-angle hyperspectral data

Int. J. Remote Sens.

(2021)

Li, D. L., Li, C., Y, Yao., Li, M. D., Liu, L. C., 2020. Modern imaging techniques in plant nutrition analysis: a...

Cited by (10)

Pixel-class prediction for nitrogen content of tea plants based on unmanned aerial vehicle images using machine learning and deep learning
2023, Expert Systems with Applications
Nitrogen management in tea (Camellia sinensis L.) fields has deeply depended on manual work for a long term due to low-cost labor. Because of the decreasing population, costs of labor have been rising. Computer technology recently has made great success in many fields, and intelligent tea fields expert system will be updated again. Nitrogen management with computer technology instead of manual work is an inevitable attempt. In this study, UAVs were used to take photos of tea plants, meanwhile, the elder tea leaves and tea shoots were collected to determine the nitrogen content. According to the HSV color system, pixels of background, elder tea leaves, and tea shoots were screened respectively. The experimental strategy I and II then were established. In experimental strategy I, pixel-class information of elder tea leaves and tea shoots were used to calculate weighted average values on RGB and HSV color systems, respectively. The values then were respectively combined with the nitrogen content of elder tea leaves and tea shoots to build datasets. In experimental strategy II, after suppressing the background, not only the weighted average values on RGB and HSV color systems in the rest each photo but also the proportion of elder tea leaves and tea shoots were computed. The proportion values were respectively multiplied by the nitrogen content to obtain a new weighted nitrogen content, and then with weighted RGB and HSV values together formed datasets. The ordinary least squares (OLS), extreme gradient boosting (XGBoost), long short-term memory (RNN-LSTM), convolution neural network (CNN), and residual network (ResNet) models were designed and carried out on predicting nitrogen content. Generally, the performance of HSV datasets was better than that of RGB datasets. ResNet and CNN models performed best in experimental strategy I; the ResNet model had the best results in experimental strategy II. Both experimental strategies performed well when predicting nitrogen content in spring tea; as for summer tea, experimental strategy II had better performance. This method had the advantages of convenience and accuracy and could elevate intelligent tea field management. As a part of precision agriculture, it could improve intelligent tea fields expert system, and further deal with challenges caused by the labor population decrease.
Combination of hyperspectral imaging and entropy weight method for the comprehensive assessment of antioxidant enzyme activity in Tan mutton
2023, Spectrochimica Acta - Part A: Molecular and Biomolecular Spectroscopy
The antioxidant enzymes play the crucial role in inhibiting mutton spoilage. In this study, visible near-infrared (Vis-NIR) hyperspectral imaging (HSI) combined with entropy weight method (EWM) was developed for the first time to evaluate the antioxidant properties of Tan mutton. The comprehensive index of antioxidant enzymes (AECI) consisting of peroxidase (49.34%), catalase (37.97%) and superoxidase (12.69%) was constructed by the EWM. Partial least squares regression, least squares support vector machine and artificial neural networks (ANN) were developed based on characteristic wavelengths extracted by successful projections algorithm, uninformative variable selection, iteratively retains informative variables (IRIV), regression coefficient and competitive adaptive reweighted sampling (CARS). The textural features (TF) were extracted by the gray level co-occurrence matrix and fused with the spectral data to establish models. Visualization of the changes in antioxidant enzyme activity was constructed from the optimal model. In addition, two-dimensional correlation spectra (2D-COS) with AECI as a perturbation variable was used to identify spectral features, revealing chemical bond changes order under the characteristic peaks at 612–799–473–708–559 nm. The results showed that the IRIV-CARS-TF-ANN model performed the best, with prediction set coefficient of determination (R_P²) of 0.8813, which improved 2.12%, 1.11% and 2.77% over the R_P² of full band, IRIV and IRIV-CARS, respectively. It was suggested that fusion data of HSI may effectively predict the activity of antioxidant enzymes in Tan mutton.
Combining multispectral and hyperspectral data to estimate nitrogen status of tea plants (Camellia sinensis (L.) O. Kuntze) under field conditions
2022, Computers and Electronics in Agriculture
Citation Excerpt :
Sabzi et al. (2021) estimated the N content in cucumber with hyperspectral data, and obtained a positive result by PLSR model. Wang et al. (2021) proposed a model for estimating leaf N content based on the bands sensitive to N concentration and oxidase activities by hyperspectral imaging. Yu et al. (2018) combined the hyperspectral data and deep learning method to detect N concentration of oilseed rape leaf.
Nitrogen (N) plays a pivotal role in management of tea plantation, with significant impacts on the growth, productivity, and nutrition status of tea plants. The existing methods for N content monitoring of tea leaves are complicated and can not realize in suite and in real time way. This study proposed a method for estimating the N content of tea plants in field conditions based on a combination of a multispectral imaging system and hyperspectral data. A total of 32 parameters were extracted from five tea gardens using calibrated multispectral images of the tea plant canopy, and 27 indices were selected by Pearson correlation analysis. A total of 28 wavelengths selected by competitive adaptive reweighted sampling from hyperspectral data were combined with 27 multispectral indices as the original data. Subsequently, five variables of fused data (H, VOG, BGI, 1664 nm and 1665 nm) were selected by variable combination population analysis based on the 55 combination parameters. Partial least squares regression, random forest regression, and support vector machine regression (SVR) models all showed excellent performance for both the calibration and prediction sets. The overall results indicated that the infused data of multispectral and hyperspectral data combined with SVR are effective in monitoring the N level under field conditions, and the R² (coefficient of determination) and root mean square error values of the prediction were 0.9186 and 0.0560, respectively. The findings of this study are important in retaining the nutritional and quality attributes of agricultural commodities.
A local model based on environmental variables clustering for estimating foliar phosphorus of rubber trees with vis-NIR spectroscopic data
2022, Heliyon
Citation Excerpt :
The larger value of RPD, the better is the model performance. If RPD <1.4, the prediction accuracy of the model is unacceptable; if 1.4 < RPD <2.0, the prediction accuracy is acceptable but needs improvement; if RPD >2.0, the prediction accuracy is high (Dong et al., 2022; Li et al., 2018; Wang et al., 2013, 2021). Therefore, a good model should have higher R2 and RPD, but lower RMSE.
Existing local models based on multiple environmental variables clustering (LM-MEVC) treat the influences of environmental factors on leaf phosphorus concentration (LPC) of rubber trees (Hevea brasiliensis) equally when grouping samples. In fact, the effects that environmental factors assert on LPC are different. So, environmental factors need to be treated differently so that the different effects can be taken into consideration when dividing samples into clusters or groups. According to this basic idea, a local model based on weighted environmental variables clustering (LM-WEVC) was developed. This approach consists of four steps. Firstly, the most important environmental variables that influence LPC were selected. Then, the weights of the selected environmental variables were determined. In the following, the selected environmental variables were weighted and used as clustering variables to group samples. Finally, within each cluster or group of samples, an estimation model was established. In order to verify its effectiveness in predicting LPC of rubber trees, the proposed method was applied to a case study in Hainan Island, China. Rubber tree (cultivar CATAS-7-33-97) leaf samples were collected from three different sampling periods. Spectral reflectance of the collected leaf samples was measured using an ASD spectroradiometer, FieldSpec 3. Leaf samples collected from the three different sampling periods were used separately to test LM-WEVC. Coefficient of determination (R²), root mean squared error (RMSE), and ratio of prediction deviation (RPD) were employed as evaluation criterion. Performance of LM-WEVC was compared with that of the existing LM-MEVC. Results indicated that for the three sampling periods, the prediction accuracies of LM-WEVC were always higher than those of LM-MEVC. The values of R² and RPD for LM-WEVC were increased by 8.15%–36.68%, and by 11.33%–59.40% respectively, while values of RMSE were reduced by 9.09%–37.5%, compared with those for LM-MEVC. These results demonstrate that LM-WEVC was effective in estimating LPC of rubber trees, and also confirmed our hypothesis that environmental factors unequally influenced LPC of rubber trees.
Detection of Key Indicators of Ripening Quality in Table Grapes Based on Visible near-infrared Spectroscopy
2024, Nongye Jixie Xuebao/Transactions of the Chinese Society for Agricultural Machinery
Estimation of Biochemical Pigment Content in Poplar Leaves Using Proximal Multispectral Imaging and Regression Modeling Combined with Feature Selection
2024, Sensors

View all citing articles on Scopus

View full text

Original papersEstimating cotton leaf nitrogen by combining the bands sensitive to nitrogen concentration and oxidase activities using hyperspectral imaging

Highlights

Abstract

Introduction

Section snippets

Experimental design

Effect of different nitrogen application rates on the nitrogen concentration of cotton leaves

Effect of different nitrogen application rates on the oxidase activities of cotton leaves

Discussion

Conclusion

Declaration of Competing Interest

Acknowledgments

Chemometrics Intelligent Lab. Syst.

Remote Sens. Environ.

Remote Sens. Environ.

Eur. J. Agron.

Adv. Space Res.

Remote Sens. Environ.

Remote Sens. Environ.

J. Food Eng.

J. Plant Physiol.

Agric. For. Meteorol.

Remote Sens. Environ.

Soil and Agricultural Chemistry Analysis

Light reflectance compared with other nitrogen stress measurements in corn leaves

Agron. J.

The development of biosensors and biochips in IECAS. Network & Parallel Computing, Ifip International Conference, Npc, Wuhan, China

October. DBLP.

N status monitoring model in winter wheat based on image processing

Trans. Chinese Soc. Agri. Eng.

Diagnosis of cotton nitrogen nutrition based on low-altitude drone image spectrum and texture characteristics

Chinese Agri. Sci.

Effects of nitrogen forms and ratios on plant growth, seed antioxidant enzyme activities and reactive oxygen metabolism of vegetable soybean

Plant Nutrition and Fertilizer Science.

Monitoring of leaf nitrogen content of winter wheat using multi-angle hyperspectral data

Int. J. Remote Sens.

Original papers
Estimating cotton leaf nitrogen by combining the bands sensitive to nitrogen concentration and oxidase activities using hyperspectral imaging