Estimating Leaf Nitrogen Concentration Research Articles

A novel framework to assess apple leaf nitrogen content: Fusion of hyperspectral reflectance and phenology information through deep learning

Rapid and accurate assessment of apple-tree leaf nitrogen content (LNC) plays an important role in smart orchards for accurate fertilization. Hyperspectral remote sensing technology provides a reliable, economical, and fast method to estimate LNC. However, high-dimensional spectral data contain redundant information that is irrelevant to the target parameters, necessitating the development of robust hyperspectral analysis techniques to estimate apple-tree LNC. Additionally, the role of phenological information in improving LNC estimates across multiple growth stages remains uncertain. The present study thus proposes a novel hyperspectral data-processing framework (Boruta-Iteration-DNN) to optimize estimates of apple-tree LNC and investigates the potential of phenological information to improve LNC estimates of multi-fertility apple trees. The proposed framework incorporates the Boruta feature-selection method for identifying important spectral bands and assessing how they affect model performance through iterative feature analysis. A deep neural network (DNN) regression model is then developed for LNC estimation. In addition, a widely used Pearson-DNN method is used for comparison to demonstrate the superiority of the proposed approach. Furthermore, the potential of incorporating phenological information (Day After Anthesis, DAA) into the estimation models is examined to evaluate its ability to better estimate LNC. The results demonstrate that the proposed framework not only enhances the accuracy of apple-tree LNC estimates with respect to the full-spectrum model (without DAA: validation R2 improved from 0.69 to 0.75, NRMSE reduced from 11.92 % to 9.97 %; with DAA: validation R2 improved from 0.72 to 0.79, NRMSE reduced from 11.09 % to 9.06 %), but also reduces the complexity and dimensionality of the models by eliminating 96 % of the redundant spectral bands (2015 bands). Moreover, the inferior results of the Pearson-Iteration-DNN framework further underscore the effectiveness of the Boruta method in selecting relevant spectral bands. Additionally, including the DAA phenological information enhances the accuracy of all models (the total R2 of training and validation increased by 0.05 ∼ 0.12, and the total NRMSE reduced by 1.49 % ∼ 1.93 %). These findings highlight the ability of DNN models integrated with hyperspectral data and phenological information to accurately assess LNC in apple trees, thereby offering valuable guidance to orchard management for developing precise fertilization strategies.

Construction of cotton leaf nitrogen content estimation model based on the PROSPECT model

Leaf nitrogen content (LNC) is an important index to measure the nitrogen deficiency in cotton. The rapid and accurate monitoring of LNC is of great significance for understanding the growth status of cotton and guiding precise fertilization in the field. At present, the hyperspectral technology monitoring of LNC is very mature, but it is interfered with by external factors such as shadow and soil, and data acquisition is still dependent on manpower. Therefore, on the basis of clarifying the correlation and quantitative relationship between physiological parameters and cotton LNC, the 400-2500 nm spectral curve was simulated based on PROSPECT-5 model. Combined with the measured spectra, the sensitive bands of leaf nitrogen content were screened, and four machine learning algorithms based on the reflectance of the sensitive bands were compared to construct a model for the estimation of LNC in cotton and determine the optimal model. The results show the following: (1) The parameter with the best correlation with nitrogen content was Cab, and the linear relationship was y=0.3942x+12.521, R2=0.81, RMSE=12.87 g/kg. (2) The shuffled frog leaping algorithm (SFLA) and the successive projections algorithm (SPA) were used to screen the relevant bands sensitive to LNC. SFLA selected nine characteristic bands, mainly distributed between 700 and 750 nm. SPA screened seven characteristic bands, mainly distributed between 670 and 760 nm. The characteristic bands of both screening methods were distributed near the red edge. (3) Based on the sensitive bands, the four machine learning algorithms were compared. Among them, the band modeling of SFLA screening under the random forest (RF) algorithm was the best (modeling set R2=0.973, RMSE=1.001 g/kg, rRMSE=3.41%, validation set R2=0.803, RMSE=3.191 g/kg, rRMSE=10.85%). In summary, this study proposes an optimal estimation model of cotton leaf nitrogen content based on the radiative transfer model, which provides a theoretical basis for the dynamic, accurate, and non-destructive monitoring of cotton leaf nitrogen content.

Monitoring of Nitrogen Concentration in Soybean Leaves at Multiple Spatial Vertical Scales Based on Spectral Parameters.

Nitrogen is a fundamental component for building amino acids and proteins, playing a crucial role in the growth and development of plants. Leaf nitrogen concentration (LNC) serves as a key indicator for assessing plant growth and development. Monitoring LNC provides insights into the absorption and utilization of nitrogen from the soil, offering valuable information for rational nutrient management. This, in turn, contributes to optimizing nutrient supply, enhancing crop yields, and minimizing adverse environmental impacts. Efficient and non-destructive estimation of crop LNC is of paramount importance for on-field crop management. Spectral technology, with its advantages of repeatability and high-throughput observations, provides a feasible method for obtaining LNC data. This study explores the responsiveness of spectral parameters to soybean LNC at different vertical scales, aiming to refine nitrogen management in soybeans. This research collected hyperspectral reflectance data and LNC data from different leaf layers of soybeans. Three types of spectral parameters, nitrogen-sensitive empirical spectral indices, randomly combined dual-band spectral indices, and "three-edge" parameters, were calculated. Four optimal spectral index selection strategies were constructed based on the correlation coefficients between the spectral parameters and LNC for each leaf layer. These strategies included empirical spectral index combinations (Combination 1), randomly combined dual-band spectral index combinations (Combination 2), "three-edge" parameter combinations (Combination 3), and a mixed combination (Combination 4). Subsequently, these four combinations were used as input variables to build LNC estimation models for soybeans at different vertical scales using partial least squares regression (PLSR), random forest (RF), and a backpropagation neural network (BPNN). The results demonstrated that the correlation coefficients between the LNC and spectral parameters reached the highest values in the upper soybean leaves, with most parameters showing significant correlations with the LNC (p < 0.05). Notably, the reciprocal difference index (VI6) exhibited the highest correlation with the upper-layer LNC at 0.732, with a wavelength combination of 841 nm and 842 nm. In constructing the LNC estimation models for soybeans at different leaf layers, the accuracy of the models gradually improved with the increasing height of the soybean plants. The upper layer exhibited the best estimation performance, with a validation set coefficient of determination (R2) that was higher by 9.9% to 16.0% compared to other layers. RF demonstrated the highest accuracy in estimating the upper-layer LNC, with a validation set R2 higher by 6.2% to 8.8% compared to other models. The RMSE was lower by 2.1% to 7.0%, and the MRE was lower by 4.7% to 5.6% compared to other models. Among different input combinations, Combination 4 achieved the highest accuracy, with a validation set R2 higher by 2.3% to 13.7%. In conclusion, by employing Combination 4 as the input, the RF model achieved the optimal estimation results for the upper-layer LNC, with a validation set R2 of 0.856, RMSE of 0.551, and MRE of 10.405%. The findings of this study provide technical support for remote sensing monitoring of soybean LNCs at different spatial scales.

Integrating Unmanned Aerial Vehicle-Derived Vegetation and Texture Indices for the Estimation of Leaf Nitrogen Concentration in Drip-Irrigated Cotton under Reduced Nitrogen Treatment and Different Plant Densities

The accurate assessment of nitrogen (N) status is important for N management and yield improvement. The N status in plants is affected by plant densities and N application rates, while the methods for assessing the N status in drip-irrigated cotton under reduced nitrogen treatment and different plant densities are lacking. Therefore, this study was conducted with four different N treatments (195.5, 299, 402.5, and 506 kg N ha−1) and three sowing densities (6.9 × 104, 13.8 × 104, and 24 × 104 plants ha−1) by using a low-cost Unmanned Aerial Vehicle (UAV) system to acquire RGB imagery at a 10 m flight altitude at cotton main growth stages. We evaluated the performance of different ground resolutions (1.3, 2.6, 5.2, 10.4, 20.8, 41.6, 83.2, and 166.4 cm) for image textures, vegetation indices (VIs), and their combination for leaf N concentration (LNC) estimation using four regression methods (stepwise multiple linear regression, SMLR; support vector regression, SVR; extreme learning machine, ELM; random forest, RF). The results showed that combining VIs (ExGR, GRVI, GBRI, GRRI, MGRVI, RGBVI) and textures (VAR, HOM, CON, DIS) yielded higher estimation accuracy than using either alone. Specifically, the RF regression models had a higher accuracy and stability than SMLR and the other two machine learning algorithms. The best accuracy (R2 = 0.87, RMSE = 3.14 g kg−1, rRMSE = 7.00%) was obtained when RF was applied in combination with VIs and texture. Thus, the combination of VIs and textures from UAV images using RF could improve the estimation accuracy of drip-irrigated cotton LNC and may have a potential contribution in the rapid and non-destructive nutrition monitoring and diagnosis of other crops or other growth parameters.

Estimation of leaf nitrogen content in winter wheat based on continuum removal and discrete wavelet transform

ABSTRACT Accurate and rapid estimation of leaf nitrogen content (LNC) in winter wheat using hyperspectral techniques is important for growth monitoring and accurate fertilization. Based on field experiments conducted over four years at three ecological sites with four different nitrogen (N) application treatments and four different N-efficient wheat varieties in Henan Province, the canopy spectra and LNC of winter wheat were obtained simultaneously in the main growth stages. The original spectrum was subjected to continuum removal (CR), the correlation between LNC and various vegetation indices was systematically analysed, and the optimal vegetation index estimation model for LNC was constructed. Simultaneously, discrete wavelet transform (DWT) was used to further compress and extract the CR spectrum. The model was combined with partial least squares regression (PLSR) and K-nearest neighbour (KNN) algorithms to estimate LNC in winter wheat. The results showed that the spectrum treated with CR was significantly improved in its correlation with LNC and that the model established based on normalized vegetation index NDVI (CR728, CR977) combined with the CR spectrum was better compared to existing vegetation indices. The calibration and validation coefficients of determination (R2) and root mean square error (RMSE) were 0.816 and 0.799, and 0.352% and 0.342%, respectively. DWT was used to transform the CR spectrum, and the results showed that a combination of the CR spectrum and a sym8 wavelet function with PLSR based on the approximation coefficients at decomposition level 2 predicted LNC most accurately. The coefficient of determination RC 2, root mean square error RMSEC, and relative percent deviation (RPDC) of the calibration set were 0.884, 0.279%, and 2.932, respectively, and RV 2, RMSEV, and RPDV of the validation set were 0.855, 0.291%, and 2.619, respectively, indicating that the model had good stability and predictive ability. Combination of CR and DWT improved the modelling accuracy of wheat LNC and showed better prediction results for multi-year and multi-point samples. The results of the study can provide a basis and reference for rapid monitoring of LNC in winter wheat.

Semi-supervised cooperative regression model for small sample estimation of citrus leaf nitrogen content with UAV images

ABSTRACT Nitrogen is an essential nutrient element for the growth of citrus tree. The accurate estimation of leaf nitrogen content (LNC) is important to guarantee fruit quality and yield. The unmanned aerial vehicle (UAV) remote sensing has shown great potential for monitoring the LNC of crops. However, the number of training samples is always limited due to the high cost and time consumption of acquiring LNC samples. Obtaining satisfactory results is difficult for most machine-learning models with insufficient samples. Thus, a semi-supervised regression model is designed in this paper to improve the performance of estimating citrus LNC from UAV images under the condition of small samples. First, the local binary pattern (LBP) operator is employed to fully extract textural features in UAV images. Second, semi-supervised cooperative regression models based on ridge regression (Ridge), support vector regression (SVR), and random forest (RF) are constructed by combining spectral and textural features. Finally, the best learning model is used to realize the accurate limited sample LNC estimation from the available training samples. The experiments confirm that the semi-supervised cooperative regression model has better and promising results than other machine-learning methods under the condition of small samples. The RF-based cooperative regression model (CoRF) performs best among other models, with a determination coefficient (R2) of 0.716 and a root mean square error (RMSE) of 0.620 g·kg−1. The CoRF model also achieves superior performance when combining LBP textural features. The results estimated by the semi-supervised cooperative regression model is capable of providing instructions for the production and management of citrus orchard.

Monitoring leaf nitrogen content in rice based on information fusion of multi-sensor imagery from UAV

Timely and accurately monitoring leaf nitrogen content (LNC) is essential for evaluating crop nutrition status. Currently, Unmanned Aerial Vehicles (UAV) imagery is becoming a potentially powerful tool of assessing crop nitrogen status in fields, but most of crop nitrogen estimates based on UAV remote sensing usually use single type imagery, the fusion information from different types of imagery has rarely been considered. In this study, the fusion images were firstly made from the simultaneously acquired digital RGB and multi-spectral images from UAV at three growth stages of rice, and then couple the selecting methods of optimal features with machine learning algorithms for the fusion images to estimate LNC in rice. Results showed that the combination with different types of features could improve the models’ accuracy effectively, the combined inputs with bands, vegetation indices (VIs) and Grey Level Co-occurrence Matrices (GLCMs) have the better performance. The LNC estimation of using fusion images was improved more obviously than multispectral those, and there was the best estimation at jointing stage based on Lasso Regression (LR), with R2 of 0.66 and RMSE of 11.96%. Gaussian Process Regression (GPR) algorithm used in combination with one feature-screening method of Minimum Redundancy Maximum Correlation (mRMR) for the fusion images, showed the better improvement to LNC estimation, with R2 of 0.68 and RMSE of 11.45%. It indicates that the information fusion from UAV multi-sensor imagery can significantly improve crop LNC estimates and the combination with multiple types of features also has a great potential for evaluating LNC in crops.

Accuracy Comparison of Estimation on Cotton Leaf and Plant Nitrogen Content Based on UAV Digital Image under Different Nutrition Treatments

The rapid, accurate estimation of leaf nitrogen content (LNC) and plant nitrogen content (PNC) in cotton in a non-destructive way is of great significance to the nutrient management of cotton fields. The RGB images of cotton fields in Shihezi (China) were obtained by using a low-cost unmanned aerial vehicle (UAV) with a visible-light digital camera. Combined with the data of LNC and PNC in different growth stages, the correlation between N content and visible light vegetation indices (VIs) was analyzed, and then the Random Forest (RF), Support Vector Machine (SVM), Back Propagation Neural Network (BP), and stepwise multiple linear regression (SMLR) were used to develop N content estimation models at different growth stages. The accuracy of the estimation model was assessed by coefficient of determination (R2), root mean squared error (RMSE), and relative root mean square error (rRMSE), so as to determine the optimal estimated growth stage and the best model. The results showed that the correlation between VIs and LNC was stronger than that between PNC, and the estimation accuracy of different models decreased continuously with the development of growth stages, with higher estimation accuracy in the peak squaring stage. Among the four algorithms, the best accuracy (R2 = 0.9001, RMSE = 1.2309, rRMSE = 2.46% for model establishment, and R2 = 0.8782, RMSE = 1.3877, rRMSE = 2.82% for model validation) was obtained when applying RF at the peak squaring stage. The LNC model for whole growth stages could be used in the later growth stage due to its higher accuracy. The results of this study showed that there is a potential for using an affordable and non-destructive UAV-based digital system to produce predicted LNC content maps that are representative of the current field nitrogen status.

Study on Estimating Total Nitrogen Content in Sugar Beet Leaves Under Drip Irrigation Based on Vis-NIR Hyperspectral Data and Chlorophyll Content

The relationship between the leaf nitrogen content (LNC) and hyperspectral remote sensing imagery (HYP) was determined to construct an estimation model of the LNC of drip-irrigated sugar beets, aiming to provide supports for the in-time monitoring of sugar beet growth and nitrogen management in arid areas. In this study, a field hyperspectrometer was used to collect the leaf reflectance at the 350–2500 nm for each treatment on the 65th, 85th, 104th, 124th, and 140th day after emergence, and the LNC and leaf chlorophyll content (CHL) of sugar beets were also determined. The spectral characteristic parameters were selected to construct the vegetation indices. The LNC estimation model using HYP as the independent variable (HYP-LNC), and that using CHL and HYP as the independent variables (HYP-CHL-LNC), were compared. The results shows that the HYP-CHL-LNC models had a better linear relationship and a higher fitting accuracy than the HYP-LNC models.

Estimation of Leaf Nitrogen Content in Rice Using Vegetation Indices and Feature Variable Optimization with Information Fusion of Multiple-Sensor Images from UAV

LNC (leaf nitrogen content) in crops is significant for diagnosing the crop growth status and guiding fertilization decisions. Currently, UAV (unmanned aerial vehicles) remote sensing has played an important role in estimating the nitrogen nutrition of crops at the field scale. However, many existing methods of evaluating crop nitrogen based on UAV imaging techniques usually have used a single type of imagery such as RGB or multispectral images, seldom considering the usage of information fusion from different types of UAV imagery for assessing the crop nitrogen status. In this study, GS (Gram–Schmidt Pan Sharpening) was utilized to fuse images from two sensors of digital RGB and multispectral cameras mounted on UAV, and the specific bands of the multispectral cameras are blue, green, red, rededge and NIR. The color space transformation method, HSV (Hue-Saturation-Value), was used to separate soil background noise from crops due to the high spatial resolution of UAV images. Two methods of optimizing feature variables, the Successive Projection Algorithm (SPA) and the Competitive Adaptive Reweighted Sampling method (CARS), combined with two regularization regression algorithms, LASSO and RIDGE, were adopted to estimate the LNC, compared to the commonly used Random Forest algorithm. The results showed that: (1) the accuracy of LNC estimation using the fusion image is improved distinctly by a comparison to the original multispectral image; (2) the denoised images performed better than the original multispectral images in evaluating LNC in rice; (3) the RIDGE-SPA combined method, using SPA to select the MCARI, SAVI and OSAVI, had the best performance for LNC in rice, with an R2 of 0.76 and an RMSE of 10.33%. It can be demonstrated that the information fusion of multiple-sensor imagery from UAV coupling with the methods of optimizing feature variables can estimate the rice LNC more effectively, which can also provide a reference for guiding the decision making of fertilization in rice fields.

Integration of Canopy Water Removal and Spectral Triangle Index for Improved Estimations of Leaf Nitrogen and Grain Protein Concentrations in Winter Wheat

Previous studies on estimating leaf nitrogen concentration (LNC) and grain protein concentration (GPC) from canopy reflectance did not pay particular attention to the nitrogen (N) and protein absorption features, which are mostly located in the shortwave infrared (SWIR) region and challenging to use due to the adverse effect of water absorption. This study aimed to develop a new approach to mitigate the effect of water signals from canopy reflectance spectra of winter wheat and enhance the absorption features of N and protein in the SWIR region. The water effect at the canopy level was removed by using the simulated dry-canopy reflectance (SDR) spectra based on the radiative transfer models and a three-band spectral triangle index (STI) was constructed to capture the chemical-specific absorption variations of N and protein. The results demonstrated that these absorption features became sufficiently apparent by canopy water removal. With the SDR spectra, the STIs with three bands related to N and protein absorption features in SWIR sub-regions exhibited remarkably stronger correlations with LNC and GPC than those with the measured reflectance (MR) spectra. Specifically, STI <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2060-2180-2240</sub> from the SDR spectra achieved the top sensitivity to LNC and GPC. The combination of plant pigment ratio (PPR) and STI <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2060-2180-2240</sub> yielded significantly lower root mean square error (RMSE) values (LNC: RMSE = 0.19%; GPC: RMSE = 0.67%) than either one alone. The integration of water removal and STI could help us better understand the underlying mechanism on the relationships of LNC and GPC with N and protein absorption features.

Remote estimation of leaf nitrogen concentration in winter oilseed rape across growth stages and seasons by correcting for the canopy structural effect

Estimating crop leaf nitrogen concentration (LNC, %) with the canopy bidirectional reflectance factor (BRF) is an effective method for detecting the nitrogen (N) deficiency in crops. It is challenging to remotely estimate LNC across growth stages and seasons with a general empirical model, since the complex change in canopy structure under N deficiencies and across growth stages affects the accuracy of the estimations. The canopy scattering coefficient (CSC), the ratio of the BRF to the directional area scattering factor (DASF), has been suggested to reduce the canopy structural effect on BRF. However, the DASF can only be calculated for closed canopies and is not applicable to the early growth stages of crops when the fields are sparsely vegetated. This study proposed a new method for decoupling the canopy structural effect and canopy BRF using the near-infrared reflectance of vegetation (NIRV). NIRV is driven by the change in canopy structure while mitigates the soil contribution. The method was tested through six field experiments on ten farmers' fields of winter oilseed rape (Brassica napus L.) using both in-situ hyperspectral data and unmanned aerial vehicle (UAV) multispectral images. The results demonstrated that NIRV was closely related to the leaf area index (LAI) (R2 = 0.79) across growth stages and seasons. The CSC was derived with NIRV based on the linear relationship between NIRV and DASF for the closed canopies. The LNC predicted by the NIRV-derived CSC (R2 = 0.69, RMSE = 0.51 for in-situ hyperspectral data and R2 = 0.65, RMSE = 0.49 for UAV multispectral images) was more accurate than the results derived from the BRF (R2 = 0.55, RMSE = 0.62 for in-situ hyperspectral data and R2 = 0.59, RMSE = 0.60 for UAV multispectral images) with the independent dataset, suggesting that correcting for the canopy structural effect with NIRV provided a new alternative for suppressing the impact of canopy structure on the canopy BRF. As NIRV is easily calculated with diverse remote sensing data sources, this study proved the potential of applying NIRV to improving the accuracy and transferability of the LNC prediction model at different scales.

Monitoring Wheat Leaf Nitrogen Content Using HJ-CCD Images and Ridge Regression

Remote sensing has long been used in agricultural applications, especially crop growth monitoring. Leaf nitrogen content (LNC) of field crop It is an important indicator of crop quality final grain yield. Many studies have used remote sensing technology to estimate the LNC of various crops. However, the performances of these estimations vary. To further improve the estimation accuracy, this research investigated the quantifiable relationships between satellite remote sensing variable images acquired from the Chinese four-band HJ-CCD sensor and wheat LNC. The ridge regression algorithms were used to build and verify multivariate remote sensing modelling of wheat LNC estimation. Results revealed that collinearities existed between wheat LNC and most of the chosen remote sensing variables. The ridge regression model for monitoring of wheat LNC adopted NDVI, GNDVI, NRI, SIPI, PSRI, DVI, RVI and EVI as independent variables and obtained optimal regularization coefficient (lambda, λ) 0.024 and RMSE 0.128 using cross validation method. Through validation from data sets of different years and regions, the coefficients of determination (R2) of wheat LNC monitoring model were 0.701 and 0.641, respectively, while its RMSE were 0.114 and 0.121, respectively. The results demonstrated that this model could be used for monitoring wheat LNC with high accuracy and confirmed that model was not limited by years and regions of wheat planting.

Construction of spectral index based on multi-angle spectral data for estimating cotton leaf nitrogen concentration

Rapid, non-destructive, and accurate monitoring is of great significance for determining crop nitrogen status and improving nitrogen fertilizer management. Multiple angle spectral data contains comprehensive and abundant information of crop canopy. In this study, the spectral data of cotton leaves at different leaf inclination angles (0°, 15°, 30°, and 45°) were collected using hyperspectral imaging technology at the seedling stage, bud stage, flowering stage, and boll-forming stage of cotton from 2018 to 2020, and indoor determination of cotton leaf nitrogen concentration (LNC) was conducted simultaneously. Based on the spectral features in blue edge and green edge regions, a new multi-angle blue-edge absorption vegetation index (MBEAVI) was constructed. Then, the MBEAVI and 16 vegetation indices (VI) reported in previous studies were constructed using single-angle and multi-angle spectral data of 2018 (320 sets) and 2019 (320 sets), and the inter-annual test was conducted with the data of 2020 (320 sets). The results showed that the VI models based on multi-angle spectral data (multi-angle models) had higher accuracy than the models based on single-angle spectral data (single-angle models). Among single-angle models, the model based on the optimized red edge absorption index (OREA) had the highest accuracy in cotton LNC estimation, with R2 of 0.735. However, the accuracy of multi-angle model, MBEAVI, was significantly higher than that of the OREA model, with R2 of 0.812. Besides, the R2 of the MBEAVI model reached 0.789 in the inter-annual test. Therefore, the MBEAVI model based on multi-angle spectral data had a higher accuracy and stability in predicting cotton LNC compared with the single-angle models. This study provides theoretical support for improving the accuracy of monitoring cotton nitrogen status by using multi-angle hyperspectral data.

Estimating Leaf Nitrogen Content in Wheat Using Multimodal Features Extracted from Canopy Spectra

The leaf nitrogen content (LNC) of wheat is one of key bases for wheat nitrogen fertilizer management and nutritional diagnosis, which is of great significance to the sustainable development of precision agriculture. The canopy spectrum provides an effective way to monitor the nitrogen content of wheat. Previous studies have shown that features extracted from the canopy spectrum, such as vegetation indices (VIs) and band positions (BPs), have successfully achieved the monitoring of crop nitrogen nutrition. However, the features mentioned above are spectral features extracted on the basis of linear or nonlinear combination models with a simple structure, which limits the general applicability of the model. In addition, models based on spectral features are prone to overfitting, which also reduces the accuracy of the model. Therefore, we propose an estimation model based on multimodal features (convolutional features and VIs, BPs) of the canopy spectrum, which aim to improve accuracy in estimating wheat LNC. Among these, the convolutional features (CFs) extracted by the designed convolutional neural network represent the deep semantic information of the canopy reflection spectrum, which can make up for the lack of robustness of the spectral features. The results showed that the accuracy of the model based on the fusion features (VIs + BPs + CFs) was higher than that of the feature of single modality. Moreover, the particle swarm optimization–support vector regression (PSO-SVR) model based on multimodal features had the best prediction effect (R2 = 0.896, RMSE = 0.188 for calibration, R2 = 0.793, RMSE = 0.408 for validation). Therefore, the method proposed in this study could improve performance in the estimation of wheat LNC, which provides technical support for wheat nitrogen nutrition monitoring.

An assessment of background removal approaches for improved estimation of rice leaf nitrogen concentration with unmanned aerial vehicle multispectral imagery at various observation times

Background effect is a crucial limitation for the monitoring of leaf nitrogen concentration (LNC) in crops with unmanned aerial vehicle (UAV) multispectral imagery. Some background removal approaches have been developed for improve the estimation of LNC, but their performances are not compared in one study and it is unclear whether they are sensitive to the observation time of UAV imagery. This study evaluated three background removal approaches, i.e., the soil-adjusted vegetation index (SAVI) approach, the green pixel vegetation index approach (GPVI) and abundance adjusted vegetation index (AAVI), for estimating rice LNC from UAV-based multispectral imagery at individual and across growth stages as well as different observation times of the day. The red edge chlorophyll index (CIre) was chosen as the common basis for the last two approaches. In particular, the AAVI approach was refined with a higher number of endmembers and automated endmember extraction, and further evaluated for assessing the effect of separating sunlit components from shaded components of the canopy. Our results demonstrated that the vegetation indices (VIs) for off-noon observation times showed better relationships with LNC than those for noon at individual and across growth stages. Compared to both SAVI and CIre-green, the AACIre for all pixels (AACIre-all) exhibited the weakest sensitivity to observation time and yielded the best relationships for single-stage (jointing: r2=0.70, booting: r2=0.76, heading: r2=0.70) and across-stage (r2=0.66) models. Among the AAVIs derived from three categories of pixels, the AACIre-sunlit (R2 =0.90, RMSE=0.17%, Bias=0.03%) outperformed AACIre-all (R2 =0.85, RMSE=0.23%, Bias=0.08%) and then AACIre-shaded (R2 =0.38, RMSE=0.49%, Bias=0.40%) remarkably for the estimation accuracy of LNC. This study suggests that the refined AAVI approach has great value in reducing the background effect for more accurate monitoring of growth parameters and could be extended to other crops and regions for improved precision crop management and field-based high-throughput phenotyping.

Combining spectral and texture features of UAV hyperspectral images for leaf nitrogen content monitoring in winter wheat

ABSTRACT The unmanned aerial vehicle (UAV) image spectral information and texture feature (TF) information were fused to develop an improved model for winter wheat leaf nitrogen content (LNC) monitoring model to provide a reference for wheat nitrogen status monitoring and accurate management. The data of wheat LNC and UAV-hyperspectral imaging were simultaneously obtained at the main growth stages (jointing, booting, and filling stages) of various winter wheat varieties under various nitrogen fertilizer treatments. The correlation between the vegetation indexes (VIs) in combination of any two bands, the TFs, and LNC were systematically analyzed. Then, the optimal VIs and TFs without multicollinearity problems were screened using a variance inflation factor (VIF) to form a ‘graph–spectrum’ fusion index. Four machine learning methods, namely ridge regression (RR), partial least squares regression (PLSR), support vector machine regression (SVR), and random forest (RF), were used to construct respective quantitative winter wheat LNC estimation models. The results revealed that the model for estimating LNC constructed using the ‘graph–spectrum’ information formed by eight parameters, including the normalized vegetation index NDVI (R 578, R 490), the difference vegetation index DVI (R 830, R 778), MEA490, MEA778, VAR490, VAR578, VAR778, and HOM578 as input and the RR algorithm, performed the best. It outperformed the models developed by the implementation of VIs and TFs as input. The coefficient of determination (R 2 ), root mean square error (RMSE), and relative percent deviation (RPD) of the calibration set were 0.84, 0.25, and 2.50, correspondingly, and those of the validation set were 0.87, 0.27, and 2.33, respectively. The model of winter wheat LNC constructed by fusing spectral and TF information considerably improved the prediction accuracy. The present research results provide a basis and reference for the application of UAV hyperspectral technology in wheat nitrogen nutrient monitoring.

Combining Multiangular, Polarimetric, and Hyperspectral Measurements to Estimate Leaf Nitrogen Concentration From Different Plant Species

Optical remote sensing is one of the most popular methods for estimating leaf nitrogen concentration (LNC). This nondestructive approach based on reflected intensity measurements has been applied to estimate the variation and distribution of nitrogen concentration in leaf and canopy levels in numerous studies. However, both intensity and polarization are necessary to describe the optical properties of light reflected from leaves and to estimate LNC estimation. In this study, based on the Stokes parameters, the total reflectance, the polarized reflectance, and the nonpolarized reflectance factors (NpRFs) were simultaneously obtained through polarimetric hyperspectral measurements under varied source-viewing geometries in both laboratory and field conditions. Several published hyperspectral indices based on the NpRF showed much better LNC estimation accuracy than those using the total reflectance factor. A clear improvement was found in the viewing directions dominated by specular reflection. Thus, using multi-angular polarimetric hyperspectral measurements not only improves the accuracy of hyperspectral indices on LNC estimation using the NpRF, but also enables the hyperspectral indices to be effective for a wide range of viewing angles. Moreover, polarimetric measurements deepen the understanding of the optical properties of light reflected from leaves. These results indicate that the combination of multiangular, polarimetric, and hyperspectral measurements may play a key role in the estimation of LNC.

Estimation model of leaf nitrogen content based on GEP and leaf spectral reflectance

High-precision estimation of leaf nitrogen content is essential for monitoring and evaluating the quality of crop products. So, it is necessary to find out the bands that affect the estimation of leaf nitrogen content from the more original hyperspectral bands, which is very important to improve the accuracy of leaf nitrogen estimation. However, the calculation complexity based on multiple regression is relatively high and the multiple regression model needs to rely on prior knowledge. This paper proposes an estimation model of leaf nitrogen content based on Gene Expression Programming (GEP) and leaf spectral reflectance. The experimental results show that compared with the other regression algorithms, the first derivative local band model obtained based on the GEP has better estimation accuracy, and the selected local band is less affected by moisture.

Combining transfer learning and hyperspectral reflectance analysis to assess leaf nitrogen concentration across different plant species datasets

Accurate estimation of leaf nitrogen concentration (LNC) is critical to characterize ecosystem and plant physiological processes for example in carbon fixation. Remote sensing can capture LNC, while interrelated traits and spectral diversity across plant species prevent development of transferable LNC assessment models based on leaf reflectance. Here, we developed a new transfer learning method by coupling transfer component analysis with the support vector regression, namely TCA-SVR, to transfer LNC assessment models across different plant species. We benchmarked the performance of TCA-SVR against a well-established partial least squares regression (PLSR) model with five remote sensing datasets on 60 plant species measured from three spectroradiometers with varied spectral resolutions and illumination and viewing angles. The result showed that leaf reflectance presented the high spectral diversity in different spectral regions, plant species, and growth stages. The combination of visible (VIS), near infrared (NIR), and shortwave infrared (SWIR) reflectance (e.g. 550–2300 nm) achieved the optimal LNC assessment across all datasets. Results on the testing datasets showed that the transferability of the PLSR models highly depended on the LNC distribution and spectral features, which were associated with the differences in plant species, spectral measurements, and growth conditions between datasets. These differences led to the large variations in LNC and leaf reflectance, which thus produced the overestimations and underestimations of LNC. Compared to the PLSR model, TCA-SVR greatly improved the transferability of the LNC assessment model by reducing the average root mean square error by 36.76%. Further, the implementation of modeling updating can help TCA-SVR learn the features related to the difference in plant species and LNC ranges by transferring samples from the target dataset to the source dataset. Our model updating approach improved the performance of TCA-SVR and only needed 5% of the off-site samples to supplement the source dataset to achieve an effective assessment of LNC. Refining the proposed method with new remote sensing datasets will aid rapid monitoring of plant nitrogen status and may improve carbon‑nitrogen interactions in existing ecosystem models.