Precise corn topdressing prescription map generation method based on unmanned aerial vehicle multispectral remote sensing

Abstract

The invention relates to the technical field of agricultural information, discloses a corn precise topdressing prescription map generation method based on unmanned aerial vehicle multispectral remote sensing, and effectively solves the key problems of insufficient monitoring precision, poor environmental adaptability, low decision intelligent level and the like in the traditional fertilization technology. Through the unmanned aerial vehicle multispectral imaging technology, efficient and accurate monitoring of the corn nitrogen nutrition condition is achieved; in combination with a machine learning algorithm, an intelligent fertilization decision model adapted to different environmental conditions is established, and the fertilization precision in a complex farmland scene is remarkably improved; and a multi-source data fusion technology is adopted, and crop growth, soil characteristics and meteorological factors are comprehensively considered, so that the fertilization scheme is more scientific and reasonable. Compared with a traditional method, the method has the advantages that accurate variable fertilization can be realized according to actual needs of crops, the utilization rate of the fertilizer is increased, environmental pollution is reduced, and reliable technical support is provided for sustainable development of modern agriculture.

Description

Technical Field

[0001] The present invention relates to the field of agricultural information technology, and specifically to a method for generating a prescription map for precise topdressing of corn based on unmanned aerial vehicle (UAV) multispectral remote sensing. Background Art

[0002] Since the development of precision agriculture technology in the 1990s, a technology system centered on GPS positioning, geographic information system (GIS), and remote sensing technology has been formed. Early research combined GPS with soil sensors to achieve location-based dynamic fertilization management, significantly improving fertilization accuracy. However, limited by data acquisition costs and efficiency, it was difficult to promote in large-scale farmland. In recent years, the popularization of hyperspectral and UAV remote sensing technologies has promoted the innovation of crop monitoring. For example, using hyperspectral reflectance and partial least squares regression (PLSR) to invert the nitrogen concentration in corn leaves, and evaluating the relationship between crop canopy coverage and NDVI through UAV multispectral data, demonstrating the potential of remote sensing technology in nitrogen monitoring and fertilization decision-making. In addition, combining remote sensing with variable fertilization systems has increased the nitrogen use efficiency of wheat by 20%-30%, showing the application prospects of multi-technology integration.

[0003] Despite the significant progress made in existing research, the existing methods still have the following limitations:

[0004] Poor environmental adaptability: Traditional models are mostly based on a single vegetation index (such as NDVI) or linear regression, and do not fully consider the interference of cloudy weather, complex terrain, and soil heterogeneity on spectral data, resulting in a decrease in accuracy in variable climates or non-uniform farmland;

[0005] Insufficient data integration: Most studies rely on a single data source (such as satellites or ground sensors), lacking the in-depth integration of multi-source data (UAV remote sensing, soil fertility, meteorological information), and it is difficult to comprehensively reflect the dynamic needs of the farmland system;

[0006] Limited large-scale application: Satellite remote sensing is restricted by the revisit cycle and cloud cover, and the deployment cost of ground sensors is high, unable to meet the real-time monitoring needs of large-scale farmland;

[0007] Low level of decision-making intelligence: The generation of existing fertilization prescription maps relies on manual experience or simple interpolation algorithms, lacking the dynamic optimization ability based on machine learning, and it is difficult to achieve precise regulation of "one plant, one strategy". The above problems restrict the further promotion of precision fertilization technology, and there is an urgent need to develop an intelligent solution with high adaptability and multi-source integration.

[0008] Therefore, a method for generating a prescription map for precise topdressing of corn based on UAV multispectral remote sensing is proposed. Summary of the Invention

[0009] In view of the deficiencies of the prior art, the present invention provides a method for generating a precise topdressing prescription map for corn based on unmanned aerial vehicle (UAV) multispectral remote sensing to solve the problems in the background art.

[0010] To achieve the above object, the present invention provides the following technical solutions: A method for generating a precise topdressing prescription map for corn based on UAV multispectral remote sensing, comprising the following steps:

[0011] Step 1: Obtain the canopy multispectral image data of the corn field by carrying a multispectral sensor on a UAV. The multispectral sensor covers the green, red, red edge, and near-infrared bands, and the reflectance data of each band is corrected in real time by a light sensor.

[0012] Step 2: Preprocess the multispectral image, including:

[0013] Geometric correction: Taking the UAV orthophoto as a reference, uniformly select 20 - 30 control points for registration, and the geometric error ≤ 0.5 pixel.

[0014] Radiometric correction: Adopt the pseudo-standard object radiometric correction method, and combine with the reflectance of the ground white reference board to convert the image DN value into a standardized reflectance.

[0015] Step 3: Extract the reflectances of the green, red, red edge, and near-infrared bands after preprocessing, and calculate at least 11 vegetation indices, including the normalized difference vegetation index (NDVI), difference vegetation index (DVI), green normalized difference vegetation index (GNDVI), modified non-linear vegetation index (MNLI), modified soil adjusted vegetation index 2 (MSAVI2), and optimized soil adjusted vegetation index (OSAVI).

[0016] Step 4: Construct an inversion model for the corn nitrogen nutrition index (LNC) based on a machine learning algorithm. The algorithm is one of multiple linear regression (MLR), support vector machine (SVM), and random forest (RF). The input of the model is the spectral variables in Step 3, and the output is the predicted value of LNC.

[0017] Step 5: Calculate the topdressing amount per plant according to the difference between the predicted value of LNC and the preset threshold, combined with the corn planting density, nitrogen fertilizer utilization rate, and topdressing loss correction coefficient.

[0018] Step 6: Perform spatial interpolation on the topdressing amount data through ArcGIS software, and use the ordinary Kriging algorithm to generate a rasterized fertilization prescription map and mark the geographical coordinates.

[0019] Preferably, the radiometric correction formula in Step 2 is:

[0020]

[0021] R目标 : Reflectance of the target object (dimensionless);

[0022] DN 目标 : Original DN value of the multi - spectral image of the target object (digital quantization value);

[0023] R 参考板 : Reflectance of the white reference board (set to 0.1, dimensionless);

[0024] DN 参考板 : Mean DN value of the white reference board (digital quantization value).

[0025] Preferably, the model verification criteria in step four are the coefficient of determination R 2 , root mean square error RMSE, and normalized root mean square error nRMSE. The larger the R 2 in model building and verification, and the smaller the corresponding RMSE and nRMSE, the better the model estimation ability, which are respectively expressed as:

[0026]

[0027] In the formula, X i , X, Y i and Y represent the measured value, mean of measured values, estimated value, and mean of estimated values respectively; n represents the number of samples in the estimation model or verification model, and p represents the number of independent variables in the model.

[0028] Preferably, the machine learning model in step four preferably selects multiple linear regression, and its regression equation is:

[0029] y = β0 + β1x1 + β2x2 + … + β p x p + θ

[0030] Among them, y is the dependent variable, x1, x2, … x p are independent variables, β0 is the intercept, β1, β2, …, β p are regression coefficients, and θ is the error term.

[0031] Preferably, the calculation formula for the top - dressing amount per plant in step five is:

[0032]

[0033] Q: Top - dressing amount of urea per plant (unit: g / plant);

[0034] LNC 标准 : Lower limit of the standard value of the nitrogen nutrition index of maize during the filling period (2.8%);

[0035] LNC 实测:Maize nitrogen nutrition index predicted by the model (unit: %).

[0036] x kg / ha: Urea requirement per 0.1% LNC gap per hectare.

[0037] y m 2 / plant: Floor area occupied by a single maize plant.

[0038] 1.15: Topdressing loss correction coefficient (15% additional compensation).

[0039] Preferably, in step six, the rasterization process uses a 1m×1m cell resolution to ensure that the fertilization prescription map is aligned with the geographic coordinates of the original image.

[0040] Preferably, the method further includes an environmental adaptability correction module, specifically:

[0041] In cloudy weather, historical light data is used to dynamically compensate for radiometric correction;

[0042] For different soil types (black soil, sandy soil, clay), the LNC threshold range is adjusted by the soil organic matter content, and the adjustment formula is:

[0043]

[0044] Where the unit of organic matter content is %.

[0045] Preferably, based on the fertilization prescription map generated in step six, the output of two fertilization modes is provided:

[0046] Solid urea granule broadcasting: According to the topdressing amount per hectare (kg / ha) marked on the prescription map, guide the mechanical fertilization equipment to operate as needed;

[0047] Foliar spraying of urea solution (0.5% concentration): According to the spraying amount per plant (mL / plant) marked on the prescription map, guide the drone or spraying equipment to spray precisely.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] The maize precise topdressing method based on unmanned aerial vehicle (UAV) multispectral remote sensing provided by the present invention effectively solves the key problems existing in traditional fertilization technologies, such as insufficient monitoring accuracy, poor environmental adaptability, and low level of decision-making intelligence. Through the UAV multispectral imaging technology, the efficient and precise monitoring of the nitrogen nutritional status of maize is realized; combined with machine learning algorithms, an intelligent fertilization decision-making model adapted to different environmental conditions is established, significantly improving the fertilization accuracy in complex farmland scenarios; the multi-source data fusion technology is adopted to comprehensively consider crop growth, soil characteristics, and meteorological factors, making the fertilization plan more scientific and reasonable. Compared with the traditional method, the present invention can achieve precise variable fertilization according to the actual needs of crops, reduce environmental pollution while improving the fertilizer utilization rate, and provide a reliable technical support for the sustainable development of modern agriculture.

[0050] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be realized and obtained by the structure pointed out in the specification, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flow chart of the method for generating a maize precise topdressing prescription map based on UAV multispectral remote sensing of the present invention;

[0052] Figure 2 It is a diagram of the pseudo-standard ground object radiation correction method of the present invention;

[0053] Figure 3 It is a UAV remote sensing image of the present invention;

[0054] Figure 4 It is a schematic diagram of the correlation coefficient between multispectral image variables and LNC of the present invention;

[0055] Figure 5 It is a verification diagram of the relationship between the predicted value and the measured value of the LNC model of the present invention;

[0056] Figure 6 It is a urea solution spraying topdressing prescription map of the present invention;

[0057] Figure 7 It is a urea topdressing prescription map of the present invention;

[0058] Figure 8 It is a reflectance image of different bands of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art in the technical field of the present invention without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0060] Please refer to Figure 1 , the method for generating a precise topdressing prescription map for corn based on multi-spectral remote sensing of unmanned aerial vehicles in the present invention specifically includes the following contents:

[0061] 1. Test area and experimental design

[0062] 1.1 General situation of the test area:

[0063] Geographical location: xx Town, with an average altitude of 180 meters, belonging to the temperate continental monsoon climate, with an average annual precipitation of 425 - 500 mm, and the precipitation from June to August accounting for 60% - 70%, and 2000 hours of annual sunshine.

[0064] Soil type: Black soil, with an organic matter content of 2.8%, a pH value of 6.5, uniform fertility, and good drainage.

[0065] Test field design: The total area is 100 hectares, divided into 50 rectangular plots of 2 hectares each, and a completely randomized block design is adopted. The initial nitrogen fertilizer application rate for all plots is uniformly 150 kg / ha (urea, nitrogen content 46%), and topdressing is only carried out during the filling period.

[0066] 1.2 Experimental objectives:

[0067] Verify the accuracy of the inversion of the nitrogen nutrition index (LNC) of corn by multi-spectral remote sensing of unmanned aerial vehicles;

[0068] Evaluate the impact of precise topdressing based on the prescription map on yield and nitrogen fertilizer utilization rate.

[0069] 2. Data collection and preprocessing

[0070] 2.1 Unmanned aerial vehicle multi-spectral data collection

[0071] 2.11 Equipment configuration:

[0072] Unmanned aerial vehicle: A multi-spectral version is adopted, equipped with a Parrot Sequoia 4-channel multi-spectral camera (green light 560 ± 16 nm, red light 650 ± 16 nm, red edge 730 ± 16 nm, near-infrared 860 ± 26 nm), and equipped with a light sensor (Table 1);

[0073] Table 1 shows the band parameters of the multi-spectral sensor

[0074] Band Band center / nm Bandwidth Green (G) 560 16 Red (R) 650 16 Red (RE) 730 16 Near Infrared (NIR) 860 26

[0075] Ground reference board: A 0.1m×0.1m white reference board with a reflectivity calibrated to 0.1, as Figure 2 shown;

[0076] 2.12 Flight parameters:

[0077] Flight altitude is 50 meters, and the image resolution is 5 cm / pixel;

[0078] North-south flight path, with a forward overlap of 70% and a side overlap of 60%;

[0079] Collection time: At the initial stage of the corn filling period, from 9:00 to 15:00 every day (in sunny and cloudless weather, with a light intensity ≥ 800 W / m 2 )(The obtained pictures are as Figure 3 and Figure 8 shown).

[0080] 2.2 Ground data collection

[0081] LNC determination:

[0082] Sampling time: At the initial stage of the filling period;

[0083] Method: Randomly select 2 corn plants from each plot, collect the top fully expanded leaves, for a total of 100 groups of samples;

[0084] Laboratory treatment: Fix at 105°C for 30 minutes, dry at 80°C to constant weight, and determine LNC by the Kjeldahl method after grinding (as shown in Table 2).

[0085] Soil data: Collect soil samples from the 0 - 20 cm soil layer and determine the organic matter content (2.8%) and pH value (6.5).

[0086] Table 2 shows the data of 100 groups of samples

[0087]

[0088] 3. Multispectral image preprocessing

[0089] 3.1 Geometric correction

[0090] Reference image: UAV remote sensing image ( Figure 3 ) and reflectivity images of different bands (8);

[0091] Selection of control points: Uniformly select 25 ground control points (GCP) for each scene of the image, and use ENVI 5.6 software for polynomial correction, with a geometric error ≤ 0.4 pixels;

[0092] Output result: Generate geometrically corrected images of 4 bands: green, red, red edge, and near-infrared (asFigure 2 as shown

[0093] 3.2 Radiometric Calibration

[0094] Calibration formula:

[0095]

[0096] R 目标 : Reflectance of the target ground object (dimensionless);

[0097] DN 目标 : Original DN value (digital quantization value) of the multispectral image of the target ground object;

[0098] R 参考板 : Reflectance of the white reference panel (set to 0.1, dimensionless);

[0099] DN 参考板 : Mean DN value of the white reference panel (digital quantization value).

[0100] Band processing: Radiometric calibration is performed on 4 bands respectively, and a multispectral image is synthesized (as Figure 2 shown).

[0101] 3.3 Vegetation Index Calculation

[0102] Extract the reflectances of 4 bands (GRE, RED, REG, NIR), and calculate 11 vegetation indices, as shown in Table 3:

[0103] Table 3

[0104]

[0105]

[0106] 4. Construction and Verification of the LNC Inversion Model

[0107] 4.1 Evaluation Criteria

[0108] Select the coefficient of determination R 2 , root mean square error RMSE, and normalized root mean square error nRMSE as the indicators for evaluating the modeling and verification accuracy of the estimation model. The larger the R 2 of the model modeling and verification, the smaller the corresponding RMSE and nRMSE, and the better the model estimation ability, which are respectively expressed as:

[0109]

[0110]

[0111] where X i , X, Y iY and Y represent the measured value, the mean of the measured values, the estimated value, and the mean of the estimated values respectively; n represents the number of samples of the estimation model or the verification model, and p represents the number of independent variables in the model.

[0112] 4.2 Data Matching and Division

[0113] Associate 100 groups of measured LNC values with spectral variables (4-band reflectance + 11 vegetation indices) at the corresponding coordinates;

[0114] Randomly divide 80 groups as the training set and 20 groups as the verification set.

[0115] Based on the drone multispectral image data, use the multispectral image after radiometric calibration to extract the average reflectance of the 4 bands of GRE, RED, REG, and NIR in the measured area and the various vegetation indices constructed, and conduct a correlation analysis with the LNC corresponding to the measured area. The results are as Figure 4 shown. It can be seen that the correlation coefficients of GRE, GNDVI, and LNC are the highest; the correlations of REG, DVI, MSR, NLI, etc. are relatively good, and the correlation coefficient of REG and NIR is the lowest.

[0116] 4.3 Model Training

[0117] Multiple Linear Regression (MLR)

[0118] Use the MATLAB fitlm function to construct the model;

[0119] Input variables: GRE, GNDVI, NLI, MSR;

[0120] Regression equation:

[0121] LNC = -0.3332 + 4.9361×GRE - 7.9367×GNDVI + 5.2177×NLI

[0122] + 1.2197×MSR

[0123] GRE: Reflectance of the green light band;

[0124] GNDVI: Green Normalized Difference Vegetation Index;

[0125] NLI: Nonlinear Vegetation Index, and the calculation formula is

[0126] MSR: Modified Simple Ratio Vegetation Index, and the calculation formula is

[0127] Comprehensive analysis: R 2 = 0.869, RMSE = 0.184%, nRMSE = 0.125%.

[0128] The regression method completed the establishment of the maize LNC inversion model and compared it with the measured LNC to verify the accuracy of the inversion model. The results are as follows Figure 5 shown. The linear relationship between the LNC estimated by the inversion model and the actual LNC is good, and the distribution of residuals approximately follows a normal distribution, proving that the inversion model has a significant effect on estimating the LNC during the maize filling period.

[0129] 5. Precise estimation of topdressing amount

[0130] Based on the inversion model of maize leaf nitrogen content, the difference between the current maize and the minimum standard of the LNC level (2.8% - 3.2%) of maize during the normal filling period was derived, and the nitrogen requirement for topdressing was calculated.

[0131] Based on applying urea with a nitrogen content of 46%, according to 2.17 kg of urea required per hectare for every 0.1% LNC gap, which is 0.217 g / m2 per square meter. Considering the losses during topdressing, a 15% correction is given; and according to the nitrogen utilization rate of 40% for nitrogen fertilizer, the nitrogen demand ratio of 20% during the maize filling period, and the average floor area of 0.25 m per plant 2 , the calculation formula is:

[0132]

[0133] Q: Urea topdressing amount per plant (unit: g / plant);

[0134] LNC 标准 : Lower limit of the standard value of the nitrogen nutrition index of maize during the filling period (2.8%);

[0135] LNC 实测 : Nitrogen nutrition index of maize predicted by the model (unit: %);

[0136] 2.17 kg / ha: Urea requirement per hectare for every 0.1% LNC gap;

[0137] 0.25 m 2 / plant: Floor area per maize plant.

[0138] Finally, through calculation, it can be obtained that for maize at the initial stage of the filling period, for every 0.1 percentage point less than the normal level of LNC of 2.8%, 0.03125 g / plant of urea needs to be applied (single plant occupies 0.25 m2). If topdressing is carried out by spraying urea solution (0.5% concentration) on the leaves, then for every 0.1 percentage point lower than the normal level, an additional 6.25 ml of solution needs to be sprayed per plant.

[0139] 6. Spatial interpolation and prescription map output

[0140] Based on the results of the fertilizer demand analysis described in the previous article on accurate estimation of the amount of topdressing, the amount of topdressing required for the plant corresponding to each coordinate was calculated, and an accurate fertilization prescription map was generated using ArcGIS software and spatial analysis methods.

[0141] First, the data is rasterized through ArcGIS, each discrete point data is converted into spatial data composed of raster cells, and the data of each coordinate point is aligned with the geographic information of the original image.

[0142] The ArcGIS geostatistical analysis module was used to perform ordinary Kriging interpolation analysis to spatially interpolate the fertilizer demand levels in different regions and obtain a continuous fertilizer demand distribution. Figure 6 and Figure 7 shown.

[0143] Ultimately, the generated prescription map can clearly show the amount of topdressing required for each area, providing decision support for precise fertilization.

[0144] 7. Environmental adaptability correction

[0145] 7.1 Cloudy weather compensation

[0146] Dynamic adjustment: call historical illumination data for the same period and correct the radiation value:

[0147]

[0148] 7.2 Soil type adaptation

[0149] Sandy soil area (organic matter content 1.5%): Adjust LNC threshold:

[0150]

[0151] 8. Summary

[0152] The invention discloses a method for generating a precise topdressing prescription map for corn based on multispectral remote sensing of unmanned aerial vehicles. The method can reflect the growth status and fertilizer demand of corn fields in real time through remote sensing data, and generate a precise topdressing prescription map. Experimental results show that the proposed method can improve the utilization efficiency of corn fertilizer, reduce fertilizer waste, and effectively reduce environmental pollution.

Claims

1. A method for generating a prescription map for precise topdressing of corn based on multi-spectral remote sensing of unmanned aerial vehicles, characterized in that, It includes the following steps: Step 1: Obtain the canopy multi-spectral image data of the corn field by using a drone equipped with a multi-spectral sensor. The multi-spectral sensor covers the green, red, red-edge, and near-infrared bands, and the reflectance data of each band is corrected in real time by a light sensor; Step 2: Preprocess the multi-spectral image, including: Geometric correction: Taking the orthoimage of the drone as a reference, uniformly select 20 - 30 control points for registration, and the geometric error ≤ 0.5 pixel; Radiometric correction: Adopt the pseudo-standard ground object radiometric correction method, and combine with the reflectance of the ground white reference board to convert the image DN value into a normalized reflectance; Step 3: Extract the reflectances of the green, red, red-edge, and near-infrared bands after preprocessing, and calculate at least 11 vegetation indices, including the Normalized Difference Vegetation Index (NDVI), Difference Vegetation Index (DVI), Green Normalized Difference Vegetation Index (GNDVI), Modified Nonlinear Vegetation Index (MNLI), Modified Second Soil Adjusted Vegetation Index (MSAVI2), and Optimized Soil Adjusted Vegetation Index (OSAVI); Step 4: Construct an inversion model for the corn nitrogen nutrition index (LNC) based on a machine learning algorithm. The algorithm is one of Multiple Linear Regression (MLR), Support Vector Machine (SVM), and Random Forest (RF). The input of the model is the spectral variables in Step 3, and the output is the predicted value of LNC; Step 5: According to the difference between the predicted value of LNC and the preset threshold, combined with the corn planting density, nitrogen fertilizer utilization rate, and topdressing loss correction coefficient, calculate the topdressing amount per plant; Step 6: Perform spatial interpolation on the topdressing amount data through ArcGIS software, use the ordinary Kriging algorithm to generate a rasterized fertilization prescription map, and mark the geographic coordinates.

2. The method for generating a precise topdressing prescription map of corn based on UAV multispectral remote sensing according to claim 1, wherein The radiometric correction formula in Step 2 is: R 目标 : Reflectivity of the target object (dimensionless); DN 目标 : The original DN value (digital quantization value) of the multispectral image of the target feature; R 参考板 : Reflectivity of the white reference plate (set to 0.1, dimensionless); DN 参考板 : DN mean value (digital quantization value) of the white reference board.

3. The method for generating a precise topdressing prescription map for corn based on multi-spectral remote sensing by an unmanned aerial vehicle according to claim 1, wherein, The model verification criteria in Step 4 are the coefficient of determination R 2 , root mean square error RMSE, and normalized root mean square error nRMSE. For model building and verification, the larger the R 2 , and the smaller the corresponding RMSE and nRMSE, the better the model estimation ability, which are respectively expressed as: where X i , X, Y i and Y represent the measured value, the average value of the measured values, the estimated value, and the average value of the estimated values, respectively; n represents the number of samples of the estimation model or the verification model, and p represents the number of independent variables in the model.

4. The method for generating a precise topdressing prescription map for corn based on multi-spectral remote sensing of unmanned aerial vehicles according to claim 1, characterized in that, In Step 4, the machine learning model preferably selects Multiple Linear Regression, and its regression equation is: y = β0 + β1x1 + β2x2 + … + β p x p + θ Among them, y is the dependent variable, x1, x2, …, x p are independent variables, β0 is the intercept, β1, β2, …, β p are regression coefficients, and θ is the error term.

5. The method for generating a prescription map for precise topdressing of corn based on multi-spectral remote sensing by an unmanned aerial vehicle according to claim 1, wherein The formula for calculating the topdressing amount per plant in Step 5 is: Q: Topdressing amount of urea per plant (unit: g / plant); LNC 标准 : Lower limit of the standard value of the nitrogen nutrition index of maize during the filling period (2.8%); LNC 实测 : Maize nitrogen nutrition index predicted by the model (unit: %); x kg / ha: Urea demand per hectare per 0.1% LNC gap; y m 2 / plant: floor area occupied by a single maize plant; 1.15: Topdressing loss correction coefficient (15% additional compensation).

6. The method for generating a corn precise topdressing prescription map based on drone multispectral remote sensing according to claim 1, wherein In Step 6, the rasterization process uses a 1m × 1m cell resolution to ensure that the fertilization prescription map is aligned with the geographic coordinates of the original image.

7. The method for generating a corn precise topdressing prescription map based on drone multispectral remote sensing according to claim 1, wherein The method further includes an environmental adaptability correction module, specifically: Under cloudy weather conditions, use historical light data to perform dynamic compensation on radiometric correction; For different soil types (black soil, sandy soil, clay soil), adjust the LNC threshold range through the soil organic matter content, and the adjustment formula is: Where the unit of organic matter content is %.

8. The method for generating a precise topdressing prescription map of corn based on multi-spectral remote sensing of unmanned aerial vehicles according to claim 7, wherein, Based on the fertilization prescription map generated in Step 6, provide the output of two fertilization modes: Solid urea granule spreading: According to the topdressing amount per hectare (kg / ha) marked on the prescription map, guide the mechanical fertilization equipment to operate as needed; Foliar spraying of urea solution (0.5% concentration): According to the spraying amount per plant (mL / plant) marked on the prescription map, guide the drone or spraying equipment to spray precisely.

Images