Leaf nitrogen content is one of the most important indicators of rice health. Non-destructive detection of the leaf nitrogen content is crucial to ensure the growth, development, and yield of rice. However, the existing detectors are expensive and large in size. It is a high requirement for the operating environment. This study aims to fully meet the demand for the low-cost acquisition of the field crop nutrition, in order to reduce the application cost for the portability. Multispectral detectors were also designed using spectral features. Therefore, the experimental area was taken as the Precision Agriculture Aviation Research Base of Shenyang Agricultural University in Gengzhuang Town, Haicheng City, Anshan City, Liaoning Province, China. Ocean Optics HR2000+ was used to collect the spectral reflectance of "Shen Nong 9816" rice leaves. The nitrogen concentration of the leaves was measured using indoor experiments. Then, the competitive adaptive weighted sampling (CARS) was used to extract the characteristic bands. The nitrogen characteristic transfer index (NCTI) was also selected to extract the nitrogen spectral features of the rice leaves. According to the optimization and combination of the nitrogen spectral features, a suitable multispectral sensor was selected to design a portable non-destructive detector of the multispectral rice nitrogen. The hardware system of the device consisted of a host and an external leaf clamp. The host included a spectral acquisition, control, and display, as well as an external power module. The external leaf clamp was used to fix the rice leaves. The host interface of the computer operation was designed using the PyQt5 framework in the Python3.11.4 environment. The acquisition and storage of the spectral data were realized for the display of the rice nitrogen. The detection equipment was developed to collect the spectral reflectance of the rice leaves. The characteristic bands with the vegetation index were used as the inputs, while the nitrogen concentration was used as the output. The rice nitrogen concentration models were constructed using partial least squares regression (PLSR), extreme learning machine (ELM), and extreme learning machine using bat optimization algorithm (BA-ELM). Among them, the accuracy of the BA-ELM inversion was the highest with the spectral features, where the characteristic bands and NCTI were the inputs. The determination coefficient of the model on the training set was 0.792, and the root mean square error was 0.423%. While on the test set, the determination coefficient was 0.783, and the root mean square error was 0.423%. The prediction model was imported into the detection equipment. The accuracy and stability of the equipment were verified after prediction. According to the residual device accuracy of the true and predicted values, the maximum absolute residual value was 0.737%. The coefficient of variation was used to verify the stability of the equipment, and the maximum coefficient of variation was 1.901%. The portable detection equipment of the rice nitrogen was independently developed for high accuracy and stability. The finding can fully meet the requirement for the real-time detection of the rice nitrogen concentration.

Nitrogen is a vital nutrient that affects rice growth and yield, playing a key role in photosynthesis, protein synthesis, and carbon-nitrogen metabolism. Effective fertilization decisions depend on nitrogen nutritional status. Traditional methods rely on field sampling and biomass measurement, which are inefficient and lack real-time data. This study proposes a nitrogen nutrition diagnosis method using UAV remote sensing technology combined with a critical nitrogen concentration dilution curve based on Leaf Area Index to guide fertilization during the rice tillering stage. UAV-acquired multi-source remote sensing data, including visible light and hyperspectral images, are used to construct LAI inversion models and nitrogen concentration inversion models optimized by ZOA-KELM and DBO-KELM, respectively. A critical nitrogen concentration dilution curve for rice based on LAI (R²=0.87) was established. Using this method, nitrogen deficiency was calculated, guiding fertilization decisions. Compared to traditional methods, this approach reduces fertilization by 7.8% while ensuring stable yield. In conclusion, fertilization based on the LAI-based nitrogen dilution curve provides an efficient solution for precision fertilization in modern agriculture.

Rice is one of the major grain crops in the world. It is of great significance to accurately inverse the chlorophyll content during growth and development in the field. However, there are some deficiencies using conventional interpretability that are driven by data inversion of the physical and chemical parameters. The leaf radiative transfer model can be expected to effectively simulate the spectral information of rice leaves, and then describe the effects of leaf parameters on spectral reflectance. A strong mechanism model can be used for the inversion of chlorophyll content of rice leaves, according to the physically driven method. The PIOSL (PROSPECT considering the internal optical structure of the leaves) model can be utilized to assume that the internal structure of the leaf blade is composed of the superposition of two layers of optical properties, which is much more in line with the actual growth of the plant. This study aims to verify the feasibility of retrieving rice leaf chlorophyll content by the PIOSL model. Firstly, the sensitivity analysis was carried out on each input parameter into the PIOSL model. The parameters C_ab and C_ab12 were determined as the high sensitivity in the 400-750 nm band. The inversion of chlorophyll content (C_ab) was then implemented. The projection (SPA) was used to extract the features of spectral data. Five characteristic bands were sensitive to chlorophyll: 490, 575, 645, 675, and 725 nm. A multiple regression model was constructed to realize the inversion of rice leaf chlorophyll. The inversion performance of the mechanism model was then evaluated using the PIOSL model. The chlorophyll content of rice leaves was also inverted to construct an initial lookup table. LSE (least squares estimate) was used to screen the simulated sample data in the initial lookup table closer to the measured spectra. A classification prediction model was constructed using a support vector machine (SVM), in order to determine whether the combination of the screened sample parameters was in line with the actual growth of the leaves. The improved look-up table was randomly split into the training and test sets in the ratio of 7:3. The hierarchical inversion of chlorophyll content of rice leaves was performed to construct the WOA-ELM (whale optimization, WOA; extreme learning machine, ELM) inversion model. A comparison was made on the commonly-used PROSPECT model with the same data. An improved lookup table was then obtained using the PROSPECT model. The WOA-ELM model was used for the chlorophyll inversion. The results showed that the R² and RMSE of the inversion model using PIOSL-WOA-ELM were 0.977 and 2.356 μg/cm², respectively, and the inversion accuracies of both models were above 0.9, compared with the PROSPECT-WOA-ELM model. According to the numerically driven inversion model and the physically-driven radiative transfer model, the inversion accuracy of the PIOSL-WOA-ELM model was higher than that of the multivariate regression with continuous projection. Therefore, it is feasible for a more effective inversion with the radiative transfer mechanism. The finding can provide new ideas for the accurate inversion of the distribution of rice chlorophyll in the leaf, further effectively retrieving the crop physicochemical parameters.

Phosphorus plays a vital role in the growth and development of rice in the cold northern regions, affecting the yield and quality of rice. The phosphorus content of leaves can indicate the nutritional status of rice. Rapid and accurate acquisition of the phosphorus content in leaves is the basis for ensuring healthy rice growth and maintaining stable and high rice yield. Hyperspectral technology can reflect the shape of rice leaves and then evaluate the phosphorus content in the leaves, so hyperspectral technology has the potential to estimate the phosphorus content in plant leaves quickly and accurately. The hyperspectral data of the rice leaves were pretreated using the SG smoothing method. The spectral characteristics of pretreated spectral data were extracted using principal component analysis (PCA) and linear discriminant analysis (LDA). Extreme learning machine (ELM) and Bat algorithm optimized extreme learning machine (BA-ELM) were constructed to retrieve the phosphorus content in rice leaves. The results show that there are seven feature vectors produced by the two methods, and the feature vectors selected by the two methods are used as inputs, respectively. The verification sets R² and RMSE of the two models constructed using the feature reflectivity chosen by the LDA algorithm as input were between 0.603 and 0.604, and 0.025 and 0.032, respectively. Under the condition of the same inversion model, the model constructed by using the reflectivity of the features selected by the PCA algorithm as input has a better prediction effect, and the verification set R² of the two models was between 0.685-0.765, and RMSE was between 0.022-0.038. In addition, when using the features selected by these two algorithms to model, comparing the prediction results of the two models, it was found that the accuracy of the BA-ELM was higher than that of ELM. Its determination coefficient R² and RMSE of the verification set were 0.765 and 0.022, respectively. Because of this, the ELM optimized by principal component analysis and BA has certain advantages in the hyperspectral inversion of phosphorus content in rice leaves in cold regions, and can provide some reference for rapid and accurate detection of phosphorus content in rice leaves.

Nitrogen is one of the important elements for the growth and development of rice，and accurate estimation of nitrogen concentration is crucial for guiding precise fertilization and assisting in the selection of nitrogen efficient varieties in rice. Traditional field sampling methods make it difficult to obtain real-time nitrogen concentration in rice. With the rapid development of information technology，establishing the relationship between unmanned aerial vehicle hyperspectral data and nitrogen concentration through machine learning methods is currently one of the main technical routes for crop nitrogen nutrition diagnosis. This study constructs an inversion model by using the feature bands of unmanned aerial vehicle canopy hyperspectral data selected by the continuous projection algorithm as input and the measured nitrogen concentration data as output. Extreme learning machine（ELM）has the advantages of fast speed and strong generalization ability compared to similar machine learning methods. However，due to its randomly generated connection weights and neuron thresholds，its training stability is insufficient and it is prone to falling into local optima. The beluga whale optimization（BWO）is a competitive algorithm inspired the behavior of beluga whales to solve single modal and multimodal optimization problems. In this study，the BWO-ELM rice nitrogen concentration unmanned aerial vehicle hyperspectral inversion model was constructed to achieve rapid estimation of rice nitrogen concentration by optimizing the connection weight between the input layer and the hidden layer of the ELM，as well as the initial weight of the hidden layer through the BWO. The research results show that the continuous projection algorithm filters out 10 feature bands，which are 673，703，727，823，850，877，895，952，961，and 985 nm，respectively. The training set R² and RMSE of the nitrogen concentration inversion model constructed based on BWO-ELM are 0.7425 and 0.3826%，respectively，while the testing set R² and RMSE are 0.7028 and 0.4877%，respectively. The predictive ability is superior to that of the nitrogen concentration inversion model constructed based on ELM. In summary，the rice nitrogen concentration unmanned aerial vehicle hyperspectral inversion model based on BWO-ELM can quickly and accurately obtain rice nitrogen concentration，providing a new method for rice nutrition monitoring.