A method for monitoring nitrogen content of corn leaves at multiple growth stages based on PROSAIL-PRO and multispectral unmanned aerial vehicle remote sensing

Abstract

The application discloses a kind of corn multi-growth period leaf nitrogen content monitoring method based on PROSAIL-PRO and multispectral unmanned aerial vehicle remote sensing, applied to crop growth nutrition quantitative remote sensing monitoring technical field.Based on unmanned aerial vehicle platform, the remote sensing image of target corn multi-growth period is obtained, and the ground measured data is collected at the same time;The preprocessed ground measured data is input into the radiation transfer model based on PROSAIL-PRO, and the simulated spectrum is obtained;According to the simulated spectrum, the input parameters with high sensitivity are selected, and the target monitoring model is constructed by combining the random forest model;The nitrogen content data of corn leaf at different growth stages in the ground measured data is used to compare and evaluate the optimal target monitoring model of each corn growth stage;The preprocessed remote sensing image is input into the corresponding optimal target monitoring model, and the monitoring result of the nitrogen content of corn leaf at each growth stage is obtained.The application constructs the target monitoring model of leaf nitrogen content based on area of air-ground integration, and has higher monitoring precision at farm scale.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of remote sensing monitoring of crop growth nutrition, in particular to a method for monitoring nitrogen content of corn leaves in multiple growth periods based on PROSAIL-PRO and multispectral unmanned aerial vehicle remote sensing. BACKGROUND

[0002] Nitrogen is a key nutrient element for crop growth, and nitrogen content of crops in different growth periods has an important influence on final yield and quality. Traditional crop nitrogen detection methods usually adopt outdoor sampling and indoor detection, which are not only time-consuming and laborious, but also destructive and lagging. With the development of remote sensing spectral monitoring technology, non-destructive remote sensing has become an attractive tool for crop nitrogen assessment. Leaf nitrogen content (LNC_A, g / m 2 ) based on area is a key indicator for quantitatively evaluating nitrogen nutrition status of corn. There are mainly two methods for estimating crop nitrogen content from canopy scale by using remote sensing means: empirical model method and physical model method.

[0003] It is not accurate enough to monitor nitrogen content of corn leaves in multiple growth periods by relying on existing empirical models or physical models, and it is difficult to put into practical monitoring at the farm scale. SUMMARY

[0004] The application aims to provide a method for monitoring nitrogen content of corn leaves in multiple growth periods based on PROSAIL-PRO and multispectral unmanned aerial vehicle remote sensing, so as to solve the problems of inaccurate estimation by relying on existing empirical models or physical models and difficulty in putting into practical monitoring at the farm scale, and make up for the low resolution of satellite remote sensing and the deficiency that images are easily covered by clouds in field monitoring.

[0005] To achieve the above-mentioned purpose, a method for monitoring nitrogen content of corn leaves in multiple growth periods based on PROSAIL-PRO and multispectral unmanned aerial vehicle remote sensing provides the following technical solutions:

[0006] Remote sensing images of target corn in multiple growth periods are obtained based on an unmanned aerial vehicle platform, and ground measured data are collected at the same period, which are used to represent ground farm conditions corresponding to the remote sensing images; the remote sensing images and the ground measured data are preprocessed;

[0007] The preprocessed ground measured data are input into a radiation transfer model based on PROSAIL-PRO to obtain simulated spectra; the radiation transfer model based on PROSAIL-PRO is used to analyze the sensitivity of the ground measured data, calibrate key model parameters, and invert crop canopy spectra to construct simulated spectra and calculate important physiological and biochemical parameters;

[0008] According to the simulated spectrum, high-sensitive input parameters are selected, and a target monitoring model of leaf nitrogen content based on area is constructed by combining a random forest model;

[0009] The leaf nitrogen content data of corn at different growth stages in the ground measured data are used to compare and evaluate the optimal target monitoring model of leaf nitrogen content based on area at each corn growth stage from the target inversion model.

[0010] The pre-processed remote sensing images obtained at each corn growth stage are input into the corresponding optimal inversion model to obtain the monitoring results of leaf nitrogen content at each corn growth stage; the monitoring results are cm-level resolution.

[0011] Optionally, the remote sensing images of the target corn at multiple growth stages are obtained based on a UAV platform, and the ground measured data are collected at the same time, including:

[0012] The UAV remote sensing images of the key corn growth stages in the multiple corn growth stages are determined and photographed, the UAV remote sensing images are pre-processed such as radiation calibration and geometric correction, and the cm-level resolution remote sensing images of the farm scale are obtained by splicing.

[0013] The field corn plants and the ground measured data are collected, the ground measured data of each sampling point are sorted and analyzed, the corn leaves are fully ground to pass through a 100-mesh sieve, and the total nitrogen parameters of the leaves are obtained by using an elemental analyzer.

[0014] Optionally, the PROSAIL-PRO radiation transfer model is obtained by coupling PROSAIL-PRO and 4SAIL, which is used for the physical process based on the interaction of solar radiation and vegetation canopy, and establishes the physical relationship between the crop state and its bidirectional reflection characteristics.

[0015] In the PROSAIL-PRO radiation transfer model, the absorption process of vegetation to light is determined by the absorption coefficient k(λ) of each basic layer, the scattering process is determined by the leaf structure parameter N and the refractive index n(λ), and the reflectivity R(λ) and the transmittance T(λ) of each wave band are:

[0016] [R(λ), T(λ) = PROSPECT_PRO(N, k(λ), n(λ))

[0017] Wherein, k(λ) is the product of each physiological and biochemical parameter value Ci of the leaf and the corresponding absorption coefficient ki(λ):

[0018] k(λ) = ∑ki(λ)·Ci·N·i

[0019] Preferably, in order to improve the accuracy of remote sensing images, the unmanned aerial images to be spliced need to be optimized in camera parameters, images not meeting the requirements are screened out, three-dimensional coordinate information of GCP is introduced into Pix4Dmapper to ensure the horizontal accuracy and vertical accuracy of the images, corresponding calibration is performed for each control point, and the corresponding relationship between DN value and actual reflectivity is established to adjust the image data, so as to meet the requirements of image accuracy, integrity and compatibility.

[0020] Optionally, the preprocessed ground measured data is input into the PROSAIL-PRO based radiation transfer model to obtain simulated spectra, including:

[0021] The ground measured data and the prior knowledge of the multi-growth period of corn are input into the PROSAIL-PRO based radiation transfer model, the spectral sensitivity of the physiological and biochemical parameters in the ground measured data is analyzed, the key model parameters in the PROSAIL-PRO based radiation transfer model are calibrated to accurately simulate the reflectance spectrum characteristics of the actual vegetation; the Markov chain Monte Carlo method is used for sampling by using the differential evolution adaptive Metropolis algorithm, the Markov parallel chain is introduced, the likelihood function is constructed, and finally the simulated spectrum of the corn canopy is generated.

[0022] Preferably, in order to improve the accuracy of the corn canopy spectrum, the spectral sensitivity of PROSAIL-PRO is evaluated, and based on the parameter samples of the Markov chain Monte Carlo method MCMC sampling stationary distribution, it is judged whether the maximum likelihood value of each parameter is in the high density area of the parameter distribution, so as to correct the parameter value of the model by combining the measured data and the prior knowledge, and construct the accurate canopy simulation spectrum.

[0023] Optionally, the area-based leaf nitrogen content target monitoring model includes a corn leaf nitrogen content monitoring model based on spectral features and vegetation index, a corn leaf nitrogen content monitoring model based on chlorophyll content, and a corn leaf nitrogen content monitoring model based on protein; according to the simulated spectrum, high-sensitive input parameters are selected, a random forest model is combined, and an area-based leaf nitrogen content target monitoring model is constructed, including:

[0024] Step one, using the Sobol model, the contribution of the total variance and partial variance of the input parameters in the simulated spectrum to the output of the PROSAIL-PRO based radiation transfer model is calculated to evaluate the importance of each parameter;

[0025] Step two: a random forest model is directly established to monitor the nitrogen content of corn leaves based on spectral characteristics and vegetation index, the nitrogen content of corn leaves based on chlorophyll content, and the nitrogen content of corn leaves based on protein; ten-fold cross-validation is used to adjust the parameters of the monitoring model of the nitrogen content of corn leaves based on spectral characteristics and vegetation index, the monitoring model of the nitrogen content of corn leaves based on chlorophyll content, and the monitoring model of the nitrogen content of corn leaves based on protein.

[0026] Preferably, to improve the representativeness of the inversion of physiological and biochemical parameters, the Sobol model is used to consider the differences between the characteristics of corn at different growth stages, analyze the sensitivity of the main index parameters to different wavebands, and use three methods to construct the monitoring model of the nitrogen content of corn leaves. Random forest machine learning is used to establish the monitoring model of the nitrogen content of corn leaves based on spectral characteristics and vegetation index, the monitoring model of the nitrogen content of corn leaves based on chlorophyll content, and the monitoring model of the nitrogen content of corn leaves based on protein.

[0027] Optionally, the nitrogen content data of corn leaves at different growth stages in the ground measured data is measured based on an elemental analyzer, and the nitrogen content data of corn leaves at different growth stages is used to verify the accuracy of the target monitoring model of the area-based leaf nitrogen content; the statistical indicators of the accuracy verification process of the target monitoring model of the area-based leaf nitrogen content include the determination coefficient, the root mean square error, and the relative root mean square error.

[0028] Preferably, to screen the highest accuracy, that is, the optimal target monitoring model of the area-based leaf nitrogen content at each growth stage of corn, the determination coefficient R 2 , the root mean square error RMSE, and the relative root mean square error RRMSE are combined with the ground measured data to evaluate the accuracy of the three target monitoring models of the area-based leaf nitrogen content, and finally the optimal monitoring method of the nitrogen content of corn leaves at each growth stage is obtained for the output of the monitoring result. The present application provides a corn multi-growth-stage leaf nitrogen content monitoring method based on PROSAIL-PRO and multi-spectral unmanned aerial remote sensing, which has the following advantages:

[0029] It has a centimeter-level resolution, and the monitoring method is easy to apply in actual production, and has good social benefits. In the known technology, a hyperspectral sensor is used to monitor the nitrogen content of corn leaves at different growth stages, but the hyperspectral sensor has high cost, complex operation, and is prone to data redundancy and poor migration, so it is difficult to popularize in actual agricultural production. Compared with the known technology, the present application uses a multi-spectral sensor with lower cost and good reliability, which is suitable for large-scale application.

[0030] The application combines the PROSAIL-PRO radiation transfer model to analyze the sensitivity of each input parameter in the model. This helps to reduce the redundancy of model parameters, establish a more accurate connection between the model and the actual remote sensing data, and further improve the accuracy of important parameter measurement and estimation.

[0031] The actual measured leaf area index (LAI) value is used as input model instead of leaf area scanner, which is more accurate. LAI is a highly sensitive parameter in the PROSAIL-PRO radiation transfer model, and the accuracy of field measurement will have a significant impact on the construction of corn canopy spectrum. The application uses actual measured values instead of leaf area scanner, which can greatly reduce the measurement error of LAI.

[0032] The screening of various area-based leaf nitrogen content target monitoring models improves the model accuracy and is suitable for different growth periods. Unmanned aerial vehicle images and measured data are collected at key growth stages of corn, and a random forest model is used to establish multiple area-based leaf nitrogen content target monitoring models. Then, according to the general statistical indicators, the model of each growth period is compared, and the integrated leaf nitrogen content monitoring model of different growth periods is selected. BRIEF DESCRIPTION OF DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0034] Figure 1 A flowchart of the corn multi-growth period leaf nitrogen content monitoring method based on PROSAIL-PRO and multispectral unmanned aerial vehicle remote sensing provided by the present application is shown in the figure.

[0035] Figure 2 A technical framework diagram of the corn multi-growth period leaf nitrogen content monitoring method based on PROSAIL-PRO and multispectral unmanned aerial vehicle remote sensing provided by the present application is shown in the figure.

[0036] Figure 3 A posterior parameter distribution result graph of different physiological and biochemical indicators provided by the present application is shown in the figure.

[0037] Figure 4 The optimal area-based leaf nitrogen content target monitoring result of the three corn growth periods of large bell mouth, detasseling and milk ripening is shown in the figure. DETAILED DESCRIPTION

[0038] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the protection scope of the present application.

[0039] Nitrogen is a key nutrient element for crop growth, and the nitrogen content of crops at different growth stages has an important influence on the final yield and quality. There are mainly two methods for estimating crop nitrogen content at the canopy scale using remote sensing methods: empirical model method and physical model method.

[0040] The empirical model method is to use the full-band original spectrum, transformed spectrum, or various vegetation indices calculated from the spectrum as input, and the field-measured nitrogen content as output, to describe the linear or nonlinear relationship between the input and output through a series of modeling methods. Common modeling methods mainly include linear regression, stepwise regression, partial least squares regression, etc.

[0041] The physical model method often uses a light radiation transfer model, which assumes that the crop leaf is a uniform medium layer, simulates the reflection, refraction and scattering process of light inside the leaf and canopy. The input parameters required by the light radiation transfer model can generate simulated hyperspectral data. Among them, the PROSPECT model is one of the most commonly used leaf-scale light radiation transfer models, which includes several important parameters such as leaf structure parameters, chlorophyll content, water content, and dry matter content. The 4SAIL model is also the most widely used canopy-scale radiation transfer model, which can effectively utilize the principle of canopy bidirectional reflection. The PROSAIL model obtained by coupling the PROSPECT and canopy reflection model SAI can effectively improve the estimation accuracy based on chlorophyll content and leaf area index.

[0042] The empirical model method is convenient to apply, but lacks physical basis support, the characteristic spectrum is not universal, and the migration between different growth stages of crops and different crops is unstable, which is difficult to promote in actual agricultural production. While the physical model has the advantages of strong stability and easy expansion, but the existing empirical model or physical model is not accurate enough for monitoring the nitrogen content of corn leaves at multiple growth stages, which is difficult to put into practical monitoring at the farm scale.

[0043] Therefore, in order to solve the above problems, the application provides a corn multi-growth period leaf nitrogen content monitoring method based on PROSAIL-PRO and multi-spectral unmanned aerial vehicle remote sensing, which combines an empirical model and a physical model to provide a new method for monitoring leaf nitrogen content. The spatial resolution of the unmanned aerial vehicle remote sensing image can reach centimeter scale, the time resolution is high, and the ground remote sensing information can be flexibly obtained according to actual needs. Therefore, the corn multi-growth period leaf nitrogen content monitoring method based on PROSAIL-PRO and multi-spectral unmanned aerial vehicle remote sensing provided by the application fully utilizes the advantages of unmanned aerial vehicles, such as low cost, high efficiency and easy investment, realizes the application of unmanned aerial vehicle corn nitrogen content inversion in a farm scale, and can be applied to actual production, thereby solving the problem of insufficient monitoring of corn nitrogen utilization in a field.

[0044] The application provides a corn multi-growth period leaf nitrogen content monitoring method based on PROSAIL-PRO and multi-spectral unmanned aerial vehicle remote sensing. Figures 1-4 The application provides a corn multi-growth period leaf nitrogen content monitoring method based on PROSAIL-PRO and multi-spectral unmanned aerial vehicle remote sensing. Figure 1 The application provides a corn multi-growth period leaf nitrogen content monitoring method based on PROSAIL-PRO and multi-spectral unmanned aerial vehicle remote sensing. Figure 2 The application provides a corn multi-growth period leaf nitrogen content monitoring method based on PROSAIL-PRO and multi-spectral unmanned aerial vehicle remote sensing.

[0045] Step 1, based on an unmanned aerial vehicle platform, remote sensing images of target corn multi-growth periods are obtained, and ground measured data is collected at the same period, the ground measured data is used for characterizing ground farm conditions corresponding to the remote sensing images; the remote sensing images and the ground measured data are preprocessed.

[0046] In the key growth period of corn, sunny weather is selected, and unmanned aerial vehicle multi-spectral sensors are used to collect sample field remote sensing images. The shooting time period is 10:00-15:00, the unmanned aerial vehicle lens is perpendicular to the ground during the collection process, the gimbal angle is-90°, the unmanned aerial vehicle flight speed is 2 m / s, the flight height of the unmanned aerial vehicle is 120 m, the flight heading overlap degree and the lateral overlap degree are 85% and 76% respectively, and a calibration plate is shot before each take-off, which is used for the radiation calibration of the unmanned aerial vehicle image.

[0047] Pix4Dmapper software is used to splice the unmanned aerial vehicle shooting images. The GCP three-dimensional coordinate information is imported into Pix4Dmapper to ensure the horizontal accuracy and vertical accuracy of the image, the corresponding calibration is performed for each control point, and the digital value (DN) of the image is converted into the corresponding physical value to obtain the final ground reflectivity, and the calibration principle is as follows:

[0048]

[0049] In the formula, R is the surface reflectance. DN is the digital value of the original image. O is the dark current of the sensor, i.e. the output of the sensor when no light is incident. G is the gain coefficient of the sensor, which is used to convert the digital value to radiance brightness. E is the solar irradiance at a certain height above the sensor. θ is the solar zenith angle, i.e. the angle between the sun's rays and the vertical direction.

[0050] The ground measured data collection includes leaf area index, leaf nitrogen content, chlorophyll content and control point position information. Among them, all sample leaves of corn in each test plot are taken to measure leaf length and width, and the true leaf area value of corn is calculated by combining the empirical model. The ratio of the statistical leaf area to the area of the experimental plot is obtained to obtain the true LAI:

[0051]

[0052] In the formula, a is the long axis of the leaf, b is the short axis of the leaf, k is the proportionality coefficient, and the corn leaf is taken as 0.75. ρ is the planting density of corn in the plot.

[0053] The leaf nitrogen content based on area (LNC_A) is to place the corn leaves in an oven at 105℃ for 2h to kill green, then dry at 80℃ constant temperature until the mass is constant, grind and take 70-80mg powder, use Elementar vario MACRO cube element analyzer to determine the mass-based total nitrogen content (LNC_M) in the corn leaves, and then convert to LNC_A

[0054]

[0055]

[0056]

[0057] In the formula, LMAdry is the specific leaf weight based on dry weight (g / cm 2 ), Mdry is the total dry weight of the dried leaves (g), and S is the leaf area (cm 2 ).

[0058] The SPAD-502Plus instrument is used to measure the chlorophyll content, and 9 samples are taken uniformly at each sample point for measurement and calculation of the average value.

[0059] II. The pretreated ground measured data is input into the PROSAIL-PRO based radiation transfer model to obtain the simulated spectrum; the PROSAIL-PRO based radiation transfer model is used to analyze the sensitivity of the ground measured data, calibrate the key model parameters, invert the crop canopy spectrum to construct the simulated spectrum, and calculate the important physiological and biochemical parameters.

[0060] (1) Use PROSAIL-PRO radiation transfer model obtained by coupling PROSPECT-PRO and 4SAIL model. Based on the model, calculate leaf-level reflectance R(λ) and transmittance T(λ) of each waveband:

[0061] [R(λ), T(λ)] = PROSPECT-PRO (N, Cab, Car, Anth, Cbrown, Cw, Cm, Prot, CBC)

[0062] In the formula, N is a leaf structure parameter, Car is carotenoid content (μg / cm 2 ), Cab is chlorophyll content (μg / cm 2 ), Anth is anthocyanin content (μg / cm 2 ), Cw is water content (μg / cm 2 ), Cbrown is the content of gray matter (μg / cm 2 ), Cm is dry matter content (μg / cm 2 ), CBC is the content of other carbon-based components (μg / cm 2 ), and Prot is protein content (μg / cm 2 ).

[0063] Simulate canopy-level bidirectional reflectance ρi in the wavelength range of 400-2500 nm by 4SAIL model:

[0064] ρi = 4SAIL (LAI, LAD, R(λ), T(λ), hspot, tts, tto, psi, psoil)

[0065] In the formula, LAI is leaf area index, LAD is leaf angle distribution function, hspot is hotspot parameter, tts is solar elevation angle, tto is observation zenith angle, psi is azimuth angle, psoil is soil factor value, R(λ) is leaf reflectance, and T(λ) is leaf transmittance.

[0066] Use the leaf reflectance and transmittance data provided by PROSPECT-PRO model as input of 4SAIL model, and use SAIL model to simulate output canopy-level spectral reflectance by combining vegetation structure parameters and environmental conditions, so as to realize coupling of leaf-level and canopy-level simulation.

[0067] (2) Analyze the influence of input parameters on simulated spectrum and parameter sensitivity. Figure 3The posterior parameter distribution result figures of different physiological and biochemical indicators provided in the application are provided, and the parameters include chlorophyll content (Cab), nitrogen structure parameter (N), other carbon-based component content (CBC), protein content (Prot), carotenoid content (Car), anthocyanin content (Anth), dry matter content (Cm), water content (Cw), brown matter content (Cbrown), leaf inclination distribution factor (LIDF), leaf area index (LAI), solar zenith angle (tts), dry and wet soil index (psoil), hot spot parameter (hspot), and observed zenith angle (tto).

[0068] The MCMC (Markov Chain Monte Carlo) parameter calibration method is selected to calibrate the key parameters, the model parameters are adjusted to fit the spectral reflectance characteristics of a specific type of vegetation or vegetation under specific environmental conditions, and the posterior distribution range of the parameters is determined according to the Bayesian statistical analysis theory:

[0069]

[0070] θ represents the model parameters, d represents the data we need to fit, and M represents the selected model. P(θ|d, M) represents the probability of obtaining the parameter θ under the condition of selecting the data and the model M; P(θ|M) represents the probability of obtaining the parameter θ under the condition of selecting the model M, that is, the prior probability.

[0071] The DREAM (Differential Evolution Adaptive Metropolis algorithm) algorithm is used to optimize the parameter estimation range, the Markov chain parameter space convergence diagnostic value is set according to the prior knowledge, the model converges when the diagnostic value is less than 1.1, and then 100 times of sampling are continued to ensure that the maximum likelihood value of each parameter is in the high-density area of the parameter distribution, and the posterior parameter distribution result figure is generated to judge the parameter calibration requirement.

[0072] III. According to the simulated spectrum, the input parameters with high sensitivity are selected, a random forest model is combined, and an area-based leaf nitrogen content target monitoring model is constructed.

[0073] (1) The random forest method is used to estimate LNC_A in combination with vegetation index. Based on the differences between the characteristics of corn at different growth stages, Sobol sensitivity analysis method is used to determine the spectral sensitivity of LAI, chlorophyll and other physiological and biochemical parameters, and the vegetation index and reflectance band with high sensitivity in different growth stages are selected and adjusted by ten-fold cross-validation, which are used as the input features of the random forest model to inverse LNC_A.

[0074] (2) Estimation of LNC_A according to chlorophyll content. According to the results of sensitivity analysis, the vegetation index (NDVI, RVI, SAVI) and spectral band (Blue, Red) for retrieving Cab were selected as input parameters of the model, and after obtaining the output chlorophyll value, a retrieval model between the two was established using random forest, and ten-fold cross-validation was used for parameter tuning.

[0075] (3) Estimation of LNC_A according to protein content (Prot). Based on the simulated spectral data set, the vegetation index (NDVI, EVI) and spectral band (NIR) for retrieving protein content were selected as input variables, and the corresponding Prot value was used as the output variable to build a training data set using the random forest method, and the model parameters were optimized through ten-fold cross-validation. According to the retrieved protein content (Prot), the following conversion relationship was used to calculate LNC_A:

[0076] Prot = LNC_A * 4.41

[0077] After that, the accuracy of the retrieved LNC_A was verified by comparing it with the measured leaf nitrogen content.

[0078] Four. Using the measured leaf nitrogen content data of corn at different growth stages in the ground data, the optimal area-based leaf nitrogen content target monitoring model of each corn growth stage was compared and evaluated from the target retrieval model. The optimal area-based leaf nitrogen content target monitoring model can also be called the optimal retrieval model. Using the measured leaf nitrogen content data of corn at different growth stages, the optimal retrieval model at each growth stage was compared and evaluated. In different growth stages, combined with the measured leaf nitrogen content data, the area-based leaf nitrogen content target monitoring model in step 3 was compared, and the coefficient of determination (R 2 ), root mean square error (RMSE) and relative root mean square error (RRMSE) were calculated to evaluate the accuracy of the model, and the R 2 The closer to 1, the closer to 0, the higher the model accuracy.

[0079]

[0080]

[0081]

[0082] In the formula, i represents the ith sample, y i is the measured value of leaf nitrogen content, is the predicted value of nitrogen content, is the average value of the measured nitrogen content, and n represents the number of samples.

[0083] Preprocessed remote sensing images obtained at each maize growth stage are input into the corresponding optimal area-based leaf nitrogen content target monitoring model to obtain the monitoring results of leaf nitrogen content at each maize growth stage; the monitoring results are at centimeter resolution. Thus, the centimeter-resolution maize LNC_A monitoring results for that growth stage can be obtained. The optimal area-based leaf nitrogen content target monitoring results for the three maize growth stages—large trumpet stage, tasseling stage, and milk stage—provided in this application are as follows: Figure 4 As shown.

[0084] The embodiments of this application are described below with reference to the accompanying drawings. The terminology used in the implementation section of this application is for explaining specific embodiments only and is not intended to limit the scope of this application.

[0085] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.

[0086] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not explicitly listed or inherent to those processes, methods, products, or apparatuses.

Claims

1. A method for monitoring nitrogen content in maize leaves at multiple growth stages based on PROSAIL-PRO and multispectral UAV remote sensing, characterized in that, include: Remote sensing images of target maize at multiple growth stages are acquired using an unmanned aerial vehicle (UAV) platform, and ground-measured data are collected simultaneously. The ground-measured data is used to characterize the ground agricultural conditions corresponding to the remote sensing images. The remote sensing images and the ground-measured data are preprocessed. The preprocessed ground-based measured data is input into the PROSAIL-PRO-based radiative transfer model to obtain a simulated spectrum. The PROSAIL-PRO-based radiative transfer model is used to analyze the sensitivity of the ground-based measured data, calibrate key model parameters, retrieve crop canopy spectra to construct simulated spectra, and calculate important physiological and biochemical parameters. Based on the simulated spectrum, highly sensitive input parameters are selected, and combined with a random forest model, a target monitoring model for leaf nitrogen content based on area is constructed. Using the nitrogen content data of maize leaves at different growth stages from the ground-measured data, the optimal target monitoring model for leaf nitrogen content at each maize growth stage is compared and evaluated from the target monitoring models for leaf nitrogen content based on area. The preprocessed remote sensing images are input into the corresponding optimal target monitoring model for leaf nitrogen content based on area to obtain monitoring results of leaf nitrogen content at each maize growth stage; the monitoring results are at centimeter-level resolution.

2. The method for monitoring nitrogen content in maize leaves at multiple growth stages based on PROSAIL-PRO and multispectral UAV remote sensing according to claim 1, characterized in that, The method involves acquiring remote sensing images of the target maize at multiple growth stages using an unmanned aerial vehicle (UAV) platform, while simultaneously collecting ground-based measured data, including: Unmanned aerial vehicle (UAV) remote sensing images of key maize growth stages in the multi-growth stage of maize were identified and captured. The UAV remote sensing images were preprocessed by radiometric calibration and geometric correction, and then stitched together to obtain a farm-scale centimeter-resolution remote sensing image. Field corn plant data and ground measurement data were collected. The ground measurement data from each sampling point were sorted and analyzed. The corn leaves were thoroughly ground and passed through a 100-mesh sieve. The total nitrogen parameters of the leaves were obtained using an elemental analyzer.

3. The method for monitoring nitrogen content in maize leaves at multiple growth stages based on PROSAIL-PRO and multispectral UAV remote sensing according to claim 1, characterized in that, The PROSAIL-PRO radiative transfer model is obtained by coupling PROSAIL-PRO and 4SAIL, and is used to establish the physical relationship between crop status and its bidirectional reflectance characteristics based on the physical process of interaction between solar radiation and vegetation canopy. In the PROSAIL-PRO radiative transfer model, the light absorption process of vegetation is determined by the absorption coefficient k(λ) of each basic layer, and the scattering process is determined by the leaf structure parameter N and the refractive index n(λ). The reflectivity R(λ) and transmittance T(λ) of each wavelength band are: [R(λ), T(λ)=PROSPECT_PRO(N, k(λ), n(λ)) Where k(λ) is the product of the leaf's physiological and biochemical parameter values ​​Ci and the corresponding absorption coefficient ki(λ): k(λ)=∑ki(λ)·Ci·N·i.

4. The method for monitoring nitrogen content in maize leaves at multiple growth stages based on PROSAIL-PRO and multispectral UAV remote sensing according to claim 3, characterized in that, The step of inputting the preprocessed ground-measured data into a PROSAIL-PRO-based radiative transfer model to obtain a simulated spectrum includes: The measured ground data and prior knowledge of the multiple growth stages of maize are input into the PROSAIL-PRO-based radiative transfer model. The spectral sensitivity of physiological and biochemical parameters in the measured ground data is analyzed. By calibrating the key model parameters in the PROSAIL-PRO-based radiative transfer model, the reflectance spectral characteristics of actual vegetation are accurately simulated. The differential evolution adaptive Metropolis algorithm is used to sample using the Markov chain Monte Carlo method, and a Markov parallel chain is introduced to construct a likelihood function, finally generating the simulated spectrum of the maize canopy.

5. The method for monitoring nitrogen content in maize leaves at multiple growth stages based on PROSAIL-PRO and multispectral UAV remote sensing according to claim 1, characterized in that, The target monitoring models for leaf nitrogen content based on area include monitoring models for nitrogen content in maize leaves based on spectral characteristics and vegetation indices, monitoring models for nitrogen content in maize leaves based on chlorophyll content, and monitoring models for nitrogen content in maize leaves based on protein. The step involves selecting highly sensitive input parameters based on the simulated spectrum, combining them with a random forest model, and constructing a target monitoring model for leaf nitrogen content based on area, including: Step 1: Using the Sopol model, the importance of each parameter is assessed by calculating the contribution of the total variance and partial variance of the input parameters in the simulated spectrum to the output of the PROSAIL-PRO-based radiative transfer model. Step 2: Directly establish monitoring models for nitrogen content in maize leaves based on spectral features and vegetation indices, chlorophyll content, and protein content using the random forest model; use 10-fold cross-validation to adjust the parameters of these monitoring models.

6. The method for monitoring nitrogen content in maize leaves at multiple growth stages based on PROSAIL-PRO and multispectral UAV remote sensing according to claim 1, characterized in that, The nitrogen content data of maize leaves at different growth stages in the ground-based measured data were obtained based on elemental analysis. The nitrogen content data of maize leaves at different growth stages were used to verify the accuracy of the target monitoring model of leaf nitrogen content based on area. The statistical evaluation indicators for the verification process of the target monitoring model of leaf nitrogen content based on area included the coefficient of determination, root mean square error, and relative root mean square error.

Images