A peanut nitrogen nutrition index diagnosis method based on critical nitrogen concentration

Abstract

The application discloses a peanut nitrogen nutrition index diagnosis method based on critical nitrogen concentration and belongs to the technical field of plant nutrition diagnosis. The application constructs a peanut critical nitrogen dilution model by using field measured agronomic parameters, establishes a general critical nitrogen dilution model suitable for peanuts, and the model is Nc=4.193xW ‑0.189 The nitrogen nutrition of peanut plants is diagnosed through the peanut plant nitrogen nutrition index PNNI, the relationship between PNNI and yield indexes is further analyzed, the nitrogen status of peanut plants can be accurately reflected, a peanut ANNI regression prediction model is constructed and screened by using a machine learning algorithm, and the nitrogen nutrition of the aboveground part of peanuts is further diagnosed. The nitrogen nutrition of the plants and the aboveground part of peanuts is comprehensively analyzed by comprehensively analyzing the two constructed models, the peanut fertilization strategy is adjusted, the growth and yield of peanuts are predicted, and a more convenient and efficient method is provided for early nitrogen nutrition diagnosis of peanuts.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of plant nutrition diagnosis, and particularly relates to a peanut nitrogen nutrition index diagnosis method based on critical nitrogen concentration. BACKGROUND

[0002] Peanut is an important economic crop and oil crop, and its growth and yield formation are directly restricted by nitrogen level. The improper application of nitrogen fertilizer, which is widespread in current production practice, not only restricts the increase of peanut yield, but also may cause secondary environmental problems. Precise management of nitrogen fertilizer has become an important part of modern agriculture. Diagnosis methods based on plant nitrogen nutrition dynamic monitoring can assess the nitrogen demand status of crops in real time, provide scientific support for optimizing nitrogen fertilizer use, and effectively avoid resource waste and environmental pollution risks. Therefore, the research and application of peanut nitrogen nutrition diagnosis technology are of great significance for improving nitrogen fertilizer utilization rate and environmental protection.

[0003] However, there are few studies on peanut nitrogen nutrition diagnosis at present. As a crop that flowers above ground and fruits below ground, peanut has different biological characteristics and harvesting methods from other food crops. Therefore, how to construct a nitrogen nutrition diagnosis technology suitable for peanut based on critical nitrogen concentration and how to estimate the nitrogen nutrition index of peanut based on unmanned aerial vehicle multispectral data are current problems to be solved. SUMMARY

[0004] To solve the above problems, the purpose of the present application is to provide a peanut nitrogen nutrition index diagnosis method based on critical nitrogen concentration, which uses field measured agronomic parameters to construct a peanut critical nitrogen dilution model, and diagnoses nitrogen nutrition through peanut nitrogen nutrition index, and further combines unmanned aerial vehicle multispectral data to retrieve nitrogen nutrition index, so as to realize rapid, non-destructive and accurate estimation of peanut nitrogen nutrition diagnosis.

[0005] In order to achieve the above purpose, the present application is realized by the following scheme:

[0006] The present application provides a peanut nitrogen nutrition index diagnosis method based on critical nitrogen concentration, which comprises the following steps:

[0007] (1) constructing a peanut above-ground critical nitrogen dilution model and a plant critical nitrogen dilution model;

[0008] (2) constructing a peanut general critical nitrogen dilution model according to the plant critical nitrogen dilution model obtained above, and verifying the peanut general critical nitrogen dilution model by using model R² and RMSE as evaluation indexes;

[0009] (3) constructing a plant nitrogen nutrition index model PNNI based on the peanut general critical nitrogen dilution model, and diagnosing peanut plant nitrogen nutrition by combining PNNI index;

[0010] (4) Based on the above-ground critical nitrogen dilution model, the above-ground nitrogen nutrition index ANNI of peanut is calculated, the multi-spectral data collected by the unmanned aerial vehicle is processed, the machine learning algorithm is used to construct the ANNI regression prediction model of peanut, and the above-ground nitrogen nutrition index of peanut is predicted;

[0011] (5) Based on the constructed ANNI regression prediction model of peanut and the plant nitrogen nutrition index model PNNI, the nitrogen nutrition index of peanut in the experimental block is predicted, and the peanut fertilization strategy is adjusted through comprehensive analysis.

[0012] Further, the above-ground critical nitrogen dilution model and the plant critical nitrogen dilution model in step (1) are Nc=a×W -b , wherein Nc is the critical nitrogen concentration value, W is the maximum value of peanut leaf dry matter accumulation, a and b are parameters of the equation, a represents the critical nitrogen concentration when the above-ground dry matter is 1 t / ha, and b is a statistical parameter that determines the slope of the curve.

[0013] Further, the formula of the general critical nitrogen dilution model of peanut in step (2) is Nc=4.193×W -0.189 , wherein Nc is the critical nitrogen concentration value, and W is the maximum value of peanut plant dry matter accumulation.

[0014] Further, the , wherein: is the measured value, is the predicted value, is the average of the measured values, and n is the sample size.

[0015] Further, the verification result of the general critical nitrogen dilution model of peanut is R²>0.8 and RMSE<0.25.

[0016] Further, in step (3), the plant nitrogen nutrition index model PNNI=Na / Nc, wherein Na represents the actual nitrogen concentration of the plant, and Nc represents the critical nitrogen concentration of the plant.

[0017] Further, the relationship between the plant nitrogen nutrition index model PNNI and the nitrogen nutrition of peanut plant is as follows: 0.95≤PNNI≤1.05 indicates that the nitrogen nutrition of peanut plant is in the best state; PNNI<0.95 indicates that the nitrogen nutrition of peanut plant is insufficient; and PNNI>1.05 indicates that the nitrogen nutrition of peanut plant is excessive.

[0018] Further, the peanut ANNI regression prediction model in the step (4) comprises a random forest regression prediction model, a BP neural network regression prediction model and a support vector regression prediction model, and the prediction ability of the three models is: random forest regression prediction model > BP neural network regression prediction model > support vector regression prediction model.

[0019] Further, the random forest regression prediction model R² > 0.85 and RMSE < 0.016.

[0020] Further, the random forest regression model for predicting the nitrogen nutrition index of the above-ground part specifically operates as follows:

[0021] S1: preparing a sample set;

[0022] S2: selecting 75% of the samples from the sample set as a training set by using a bagging method with sampling and returning;

[0023] S3: generating a decision tree by using the sample set obtained by sampling;

[0024] S4: repeating steps S2 to S3 to train and form an ANNI random forest inversion model;

[0025] S5: predicting and inverting the test samples by using the random forest prediction model obtained by training, and determining the prediction result by using a voting method;

[0026] S6: calling the trained model to predict and invert the whole remote sensing image pixel by pixel, and splicing and adding geographic reference coordinates to the inversion result to obtain an inversion result map of the nitrogen nutrition index of the research area.

[0027] Further, the ANNI value of the random forest regression prediction model ranges from 0.9 to 1.2.

[0028] Compared with the prior art, the application has the advantages that:

[0029] 1. Through two years of field nitrogen fertilizer experiments, eight main varieties of peanuts in Shandong are selected, and five different nitrogen application levels are set. The present experiment covers multiple varieties and multiple nitrogen fertilizer gradients, and provides strong data support for in-depth exploration of the influence of nitrogen fertilizer on peanut growth.

[0030] 2. Based on the leaf dry matter, the above-ground dry matter and the plant dry matter (above-ground part + pod), critical nitrogen dilution models (Nc) of eight different varieties are constructed. The most suitable critical nitrogen dilution model for peanuts is explored from three angles. The results show that the Nc model based on the plant dry matter is the most suitable model for peanuts, and the model is: Nc=4.193×W -0.189 , R²=0.885. It has important application value and wide popularization significance.

[0031] 3. The application provides the relationship between PNNI and yield indicators, and the saturation range of PNNI is 0.907~1.068, which indicates that PNNI can accurately reflect the nitrogen status of plants, further verifying the effectiveness of the nitrogen nutrition index in evaluating the nitrogen status of peanuts and proving the reliability of the nitrogen nutrition index as a peanut nitrogen nutrition diagnosis tool.

[0032] 4. The application provides a critical nitrogen dilution model for calculating the nitrogen nutrition index of peanuts, and a random forest regression model is used to construct a regression prediction model of peanut ANNI, thereby providing a more convenient and efficient method for early nitrogen nutrition diagnosis of peanuts. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 A design diagram for the test area;

[0034] Figure 2 A critical nitrogen dilution curve of leaves of different varieties, wherein A is Rihua No.1, B is Huayu 22, C is Huayu 25, D is Huayu 9116, E is Huayu 23, F is Huayu 39, G is Huayu 917, and H is Huayu 9518;

[0035] Figure 3 A critical nitrogen dilution curve of aboveground parts of different varieties, wherein A is Rihua No.1, B is Huayu 22, C is Huayu 25, D is Huayu 9116, E is Huayu 23, F is Huayu 39, G is Huayu 917, and H is Huayu 9518;

[0036] Figure 4 A critical nitrogen dilution curve of plants of different varieties, wherein A is Rihua No.1, B is Huayu 22, C is Huayu 25, D is Huayu 9116, E is Huayu 23, F is Huayu 39, G is Huayu 917, and H is Huayu 9518;

[0037] Figure 5 A general critical nitrogen dilution model for peanuts;

[0038] Figure 6 Dynamic changes of the nitrogen nutrition index of plants of different varieties, wherein A is Rihua No.1, B is Huayu 22, C is Huayu 25, D is Huayu 9116, E is Huayu 23, F is Huayu 39, G is Huayu 917, and H is Huayu 9518;

[0039] Figure 7 Relationship between the nitrogen nutrition index of plants and relative yield, wherein A, B and C respectively represent the relationship between the nitrogen nutrition index of plants and relative yield at 84 days, 108 days and 124 days after sowing;

[0040] Figure 8 Pearson correlation analysis between the vegetation index and the nitrogen nutrition index;

[0041] Figure 9 The measured value and the predicted value of the aboveground nitrogen nutrition index of different models are related, wherein A is a random forest regression model, B is a BP neural network regression model, and C is a support vector regression model.

[0042] Figure 10 The aboveground nitrogen nutrition index inversion map is obtained. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present application will be clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0044] Embodiment 1: According to the determination method of the critical nitrogen concentration dilution model of Justes, the critical nitrogen dilution models of different varieties of peanut leaves / aboveground parts / plants are constructed

[0045] 1. Materials and methods

[0046] (1) Test materials. Eight main peanut varieties (cultivated by Shandong Peanut Research Institute, Rizhao Donggang Peanut Research Institute, and commercially available) were selected as test materials. Among them, four large peanut varieties: Rihua No. 1, Huayu 22, Huayu 25 and Huayu 9116; two small peanut varieties: Huayu 23 and Huayu 39; two high-oleic large peanut varieties: Huayu 917 and Huayu 9518. The fertilizers used include urea (N≥46.7%), potassium dihydrogen phosphate (P2O5≥52%, K2O≥34%) and potassium sulfate (K≥52%). All fertilizers were mixed thoroughly before use and evenly spread on the soil surface and rotovated.

[0047] (2) Test design. This test carried out two years of field nitrogen fertilizer test in 2023 and 2024. The test set 5 nitrogen application levels: 0 kg / hm² (N0), 75 kg / hm² (N1), 150 kg / hm² (N2), 225 kg / hm² (N3) and 300 kg / hm² (N4), 8 peanut varieties, 40 treatments, a total of 3 times, in addition to 7 blank plots without any fertilizer, a total of 127 plots. The size of each plot is 8m x 3.4m, with an area of 27.2 m², with 4 ridges, a ridge distance of 0.8 m, a ridge surface width of 0.5 m, and a ridge height of 0.14 m. Two rows of peanuts are planted on each ridge, with 2 seeds per hole, and a hole distance of 0.16 m. Apply 230 kg / hm² of potassium dihydrogen phosphate and 150 kg / hm² of potassium fertilizer, all fertilizers are one-time base application, no additional fertilizers are added during the test period. The test area design is shown in Figure 1 : Different colors are used in the test area design to represent different nitrogen levels, where N0: yellow; N1: blue; N2: green; N3: pink; N4: purple; no fertilizer treatment: gray.

[0048] (3) Obtain field agronomic indicators: Dry matter and total nitrogen determination: During the entire growth period of peanuts, take samples every 20 days, and take 12 representative peanuts with uniform growth vigor for each treatment in 5 replicates. The leaves, stems, and pods are removed and packaged in kraft paper bags, then placed in an oven at 105°C for 30 minutes to kill the green, then reduced to 85°C for drying to constant weight. The dry weight of each organ is then weighed. Then put the sample into a pulverizer, pass through a 1 mm sieve and put it into a sealed bag. Weigh 0.1 g of the sample into a conical flask, add 5 mL of concentrated sulfuric acid and let it stand overnight, then use the H2SO4-H2O2 method for digestion. After digestion, the digested liquid is diluted to 100 ml in a volumetric flask, shaken well and divided into 50 mL centrifuge tubes for filtration. The filtered clear solution is divided into 10 mL centrifuge tubes. Use an AA3 flow analyzer to determine the total nitrogen content of each organ of the peanut.

[0049] Yield determination: At the peanut maturation stage (124 days after sowing) in 2024, 10 m 2 of peanut plants with uniform growth vigor were selected in each plot, the fruits were removed and placed in a mesh bag, then naturally air-dried and weighed to record the yield of each plot, and the relative yield of each plot was calculated. The relative yield calculation formula is:

[0050] RY = Ys / Ym;

[0051] Where Ys is the yield of a specific nitrogen treatment, and Ym is the highest yield among all treatments.

[0052] (4) UAV multispectral data acquisition: while collecting ground data, UAV multispectral data acquisition is carried out. According to the task requirements, choose mapping aerial photography as the operation type, and configure the load of the aircraft to visible light and multispectral simultaneous shooting. According to the actual mapping scale and the requirement of ground spatial resolution, combined with the relationship between the number of effective pixels and the flight height, the field of view angle, the appropriate flight height is determined, and the route is automatically planned by using the software. After setting the shooting parameters, the task is uploaded and started to execute, the aircraft automatically operates, and the multispectral image data of the whole study area is successfully acquired.

[0053] The aerial operation time is selected at 10:00-14:00 around noon when the weather condition is good, no cloud and high visibility, the aerial photography area is about 5.2 mu, the flight height is 100 m, the flight speed is 6 m / s, the heading overlap rate is 75%, and the pan-tilt is set to vertical downward.

[0054] Table 1: Multispectral and UAV parameters

[0055]

[0056] 2. Constructing peanut critical nitrogen dilution model based on different parts

[0057] (1) Based on the dry matter and nitrogen concentration of different parts of peanut in 2024, three kinds of critical nitrogen dilution models are constructed: leaf critical nitrogen dilution model (Nlcnc), aboveground part (stem and leaf) critical nitrogen dilution model (Nacnc) and plant (stem, leaf and pod) critical nitrogen dilution model (Npcnc). Through the analysis and comparison of the three models, the most suitable critical nitrogen dilution model for peanut is determined.

[0058] Constructing peanut critical nitrogen dilution model of different parts includes several indicators: leaf dry matter and leaf nitrogen concentration, aboveground part dry matter and aboveground part nitrogen concentration, plant dry matter and aboveground part nitrogen concentration.

[0059] Aboveground dry matter (t / ha) = stem dry matter (t / ha) + leaf dry matter (t / ha);

[0060] Plant dry matter (t / ha) = aboveground dry matter (t / ha) + pod dry matter (t / ha);

[0061] Aboveground nitrogen concentration (%) = stem dry matter (t / ha) x stem nitrogen concentration (%) + leaf dry matter (t / ha) x leaf nitrogen concentration (%) / aboveground dry matter (t / ha);

[0062] Plant nitrogen concentration (%) = aboveground nitrogen concentration (%) + pod dry matter (t / ha) x pod nitrogen concentration (%) / plant dry matter (t / ha).

[0063] (2) Data analysis: Using DPS data processing system, the dry matter of each sampling day under different nitrogen levels was subjected to variance analysis. In the test statistics, find the completely randomized design, click single factor test statistics analysis, and select Duncan new multiple comparison method. According to the letter mark, the results are divided into nitrogen limiting group and non-nitrogen limiting group. The nitrogen limiting group refers to the nitrogen treatment group with significant increase in dry matter with the increase of nitrogen application amount; the non-nitrogen limiting group refers to the nitrogen treatment group with no significant increase in dry matter with the continuous increase of nitrogen application amount.

[0064] (3) Model construction: Taking one variety and one part (leaves) as an example, the critical nitrogen dilution model of peanut was constructed by using OriginPro 2025. First, the dry matter and corresponding nitrogen concentration values of different nitrogen treatments and different sampling days were imported into Book1, the dry matter was taken as the horizontal coordinate, and the nitrogen concentration was taken as the vertical coordinate, and a scatter plot was drawn. Click Analysis-Linear-Linear Fitting in the toolbar in turn, open the dialog box, select the data of nitrogen limiting group in the input data for linear fitting, and get the linear fitting equation y=ax+b. Calculate the average value of the dry matter of the non-nitrogen limiting group as the maximum dry matter, and bring it into the x value in the linear fitting equation to get the critical nitrogen concentration (y) of that sampling day.

[0065] According to the above method, the critical nitrogen concentration value of each sampling day was obtained, and the fitting curve y=ax -b was obtained. Click Analysis-Fitting-Nonlinear Curve Fitting in the toolbar, open the dialog box, select Origin Basic Functions in the category, select Allometricl in the function, select Levenberg-Marquardt optimization algorithm in the iteration algorithm, and then click Fitting until convergence for fitting. The obtained Nc=a×W -b is the critical nitrogen dilution model of one variety and one part. Among them, Nc is the critical nitrogen concentration value, W is the maximum value of peanut leaf dry matter accumulation, a and b are parameters of the equation, a represents the critical nitrogen concentration when the aboveground dry matter is 1 t / ha, and b is a statistical parameter that determines the slope of the curve. Other varieties and parts are constructed according to the above method.

[0066] Example 2: According to the obtained critical nitrogen dilution model of peanut, the more suitable critical nitrogen dilution model for peanut is screened; and the general critical nitrogen dilution model of peanut is constructed

[0067] 1. Screening the most suitable critical nitrogen dilution model for peanut

[0068] On the one hand, the R2 of the model was taken as an evaluation index to comprehensively analyze and compare the R2 of different parts; on the other hand, the scatter plots of the biomass and nitrogen concentration of different parts and 8 varieties in the whole growth period were compared, and the dispersion degree of the scatter plots of different parts was taken as an evaluation index, as shown in Figs. Figure 2 、 3 、4, and the model with the highest R2 and the highest fitting degree of the scatter plot was selected as the most suitable critical nitrogen dilution model for peanuts. Table 2 shows the determination coefficients of the critical nitrogen dilution models of different parts.

[0069] R2 = 1 -

[0070] In the formula: is the measured value, is the predicted value, is the average of the measured values, and n is the sample size. The R2 value ranges between 0 and 1, and the closer it is to 1, the better the model fitting.

[0071] Table 2: Determination coefficients of critical nitrogen dilution models of different parts

[0072]

[0073] According to the analysis in Table 1, the determination coefficients of the aboveground critical nitrogen dilution models (Nacnc) of Rihua No. 1 to Huayu 9518 were increased by 9.8%, 8.4%, 13.9%, 3.5%, 1.7%, -5.3%, 19.1%, and 3.1% compared to the leaf critical nitrogen dilution models (Nlcnc). Among them, the determination coefficients of Huayu 917 and Huayu 25 were significantly improved, while the determination coefficient of Huayu 39 was reduced. Compared with the aboveground critical nitrogen dilution model (Nacnc), the determination coefficients of the plant critical nitrogen dilution model (Npcnc) of Rihua No. 1 to Huayu 9518 were increased by 6.7%, -1.5%, 3.2%, 7.0%, 2.2%, -12.5%, 11.8%, and 8.3%. This change indicates that the determination coefficient of Huayu 917 is further improved, but the determination coefficient of Huayu 39 is further reduced. Compared with the leaf critical nitrogen dilution model (Nlcnc), the determination coefficients of the plant critical nitrogen dilution model (Npcnc) of Rihua No. 1 to Huayu 9518 were increased by 17.2%, 6.7%, 17.6%, 10.7%, 4.0%, -17.1%, 33.1%, and 11.6%. Similarly, the determination coefficient of Huayu 917 is significantly improved, while the determination coefficient of Huayu 39 is significantly reduced.

[0074] Based on the above analysis, it is found through the comparison of the determination coefficients of the three models that the plant critical nitrogen dilution model performs best in most varieties.

[0075] Meanwhile, variety difference is one of the key factors affecting the parameters of critical nitrogen dilution model. It is of great significance to overcome the variety difference and construct a general critical nitrogen dilution model for peanut. According to the data of dry matter and corresponding nitrogen concentration of different parts of peanut under five nitrogen treatments of eight varieties, the scatter plots of dry matter and corresponding nitrogen concentration of leaves, aboveground parts and whole plants of peanut during the whole growth period under different nitrogen treatments of eight varieties were drawn by Origin. By comparing the dispersion degree of scatter plots, it was found that the scatter plot of dry matter and corresponding nitrogen concentration of whole plant had the highest fitting degree, and the scatter plot of dry matter and corresponding nitrogen concentration of leaves had the lowest fitting degree. Therefore, the critical nitrogen dilution model based on dry matter and corresponding nitrogen concentration of whole plant was the most beneficial to overcome the difference among varieties.

[0076] The above results showed that the critical nitrogen dilution model of whole plant was the most suitable model for peanut. The higher R² was, the more effective the critical nitrogen dilution model of whole plant based on aboveground parts and pods was in integrating the physiological needs of each part compared with the single organ model, and the stronger the practical applicability was. The higher the fitting degree of scatter plot was, the more beneficial the general model based on whole plant was in overcoming the difference among varieties.

[0077] 2. Analysis of the influence of different varieties and different parts on the critical nitrogen dilution model of peanut

[0078] Through comprehensive analysis of model parameters, it was found that for different varieties, the variability of parameter b was higher than that of parameter a. In addition, the influence of different parts on the critical nitrogen dilution model of peanut was analyzed, and it was found that there were differences in the values of parameters a and b of each part. The order of parameter a from high to low was: leaves > aboveground parts > whole plant, and the order of parameter b from high to low was: aboveground parts > whole plant > leaves. The Nc model of leaves showed higher a value and lower b value.

[0079] 3. Construction and verification of general critical nitrogen dilution model of peanut

[0080] The critical nitrogen concentration values of whole plants of eight varieties were fitted to the general critical nitrogen dilution model of whole plant Nc = 4.193 x W -0.189 , R² = 0.885, and the fitting accuracy reached a significant level, as shown in Figure 5 .

[0081] Example 3: Verification of general critical nitrogen dilution model of peanut using model R² and RMSE as evaluation indexes

[0082] First, the general model was cross-validated between varieties. The cross-validation process was as follows: seven different varieties were used to build a general plant critical nitrogen dilution model, and the remaining one variety was used for validation, a total of eight validations. The R² and RMSE values were calculated for each validation, and the average values were taken as the evaluation indicators of the model. The results showed that the average R² of the model was 0.874, and the RMSE was 0.237, indicating that the model had high reliability.

[0083] In addition, the general model was validated between years. Based on the data of the independent test in 2023, a variety was randomly selected to build its plant critical nitrogen dilution model as Nc=5.02×W -0.219 The actual nitrogen concentration calculated by the model was linearly fitted with the predicted nitrogen concentration calculated by the general model, with R² of 0.75 and RMSE of 0.51, further indicating that the model had certain reliability and effectiveness.

[0084] RMSE =

[0085] In the formula: is the measured value, is the predicted value, is the mean value of the measured value, and n is the sample size.

[0086] Example 4: Based on the general critical nitrogen dilution model of peanuts, a plant nitrogen nutrition index model PNNI was constructed

[0087] Based on the plant critical nitrogen dilution model of each variety, the plant nitrogen nutrition index PNNI was calculated:

[0088] (1) The plant nitrogen nutrition index PNNI=Na / Nc; where Na represents the actual nitrogen concentration of the plant, and Nc represents the critical nitrogen concentration of the plant.

[0089] Taking the sampling days as the horizontal coordinate and the plant nitrogen nutrition index as the vertical coordinate, the PNNI line graph under different nitrogen treatments was drawn, which more intuitively showed the nitrogen nutrition status of peanut plants under different nitrogen treatments. 0.95≤PNNI≤1.05 indicates that the plant nitrogen nutrition is in the best state; PNNI<0.95 indicates that the plant nitrogen nutrition is insufficient; PNNI>1.05 indicates that the plant nitrogen nutrition is excessive.

[0090] (2) By calculating the proportion of plants with insufficient nitrogen nutrition, optimal nitrogen nutrition, and excessive nitrogen nutrition under different nitrogen treatments (N0, N1, N2, N3, and N4), the optimal nitrogen application amount of peanuts was determined.

[0091] In combination with Figure 6It can be seen that under the treatment of N0, the PNNI values of most of the eight varieties are lower than 0.95, concentrated between 0.72 and 0.94, and 91.7% of the plants are nitrogen deficient, affecting growth and development. In contrast, the PNNI values of the N1 treatment have increased, concentrated between 0.78 and 0.98, but still 52.1% of the plants are nitrogen deficient and have not reached the optimal growth state, indicating that the amount of nitrogen is still insufficient. Under the treatment of N2, 64.6% of the plants have PNNI between 0.95 and 1.05, and 22.9% of the plants are in a state of nitrogen luxury absorption, showing better nitrogen supply; under the treatments of N3 and N4, 0.95 to 1.05 plants account for 58.3% and 56.3% respectively, but the proportion of nitrogen luxury absorption state is higher. Overall, the N2 treatment is more effective, providing the best nitrogen supply conditions.

[0092] (3) Further analysis of the relationship between PNNI and RY at different growth stages, RY is the ratio of actual yield under crop growth conditions to maximum yield, which can well eliminate the influence of some external factors, making the research more objective, universal and easy to compare. For example Figure 7 As shown in the results, PNNI and RY at 84 days, 108 days and 124 days after sowing showed a linear plus platform relationship. From 84 days to 124 days after sowing, the determination coefficient of PNNI and relative yield reached a significant level, and when the yield reached the maximum, the saturation range of PNNI was 0.907-1.068, near 0.95-1.05, indicating that the plant was in the best state of nitrogen, and its yield also reached the highest. It is shown that PNNI can accurately reflect the nitrogen status of the plant. This further verifies the effectiveness of the nitrogen nutrition index in evaluating the nitrogen status of peanuts and proves the reliability of its use as a diagnostic tool for peanut nitrogen nutrition.

[0093] Example 5: Based on the aboveground critical nitrogen dilution model described in Example 1, a regression prediction model of peanut ANNI was constructed

[0094] The application is based on unmanned aerial vehicle multi-spectral data combined with vegetation index to estimate peanut nitrogen nutrition index. The application uses unmanned aerial vehicle to carry visible light and multi-spectral camera to collect multi-spectral image of peanut in early growth period (51 days after sowing), performs image preprocessing, extracts canopy reflectance of 127 plots, calculates 14 vegetation indexes, analyzes the correlation of the indexes with plant nitrogen nutrition index (PNNI), aboveground nitrogen nutrition index (ANNI) and leaf nitrogen nutrition index (LNNI), selects the optimal vegetation index as the independent variable, the optimal nitrogen nutrition index as the dependent variable, adopts random forest regression (Random Forest Regression, RFR), BP neural network regression (Back Propagation Neural Network Regression, BPNN) and support vector regression (Support Vector Regression, SVR) algorithm to establish the inversion model of nitrogen nutrition index, and provides a more convenient and efficient method for early diagnosis of peanut nitrogen nutrition.

[0095] (1) Processing of unmanned aerial vehicle multi-spectral data. After data collection, the image data collected by the unmanned aerial vehicle is copied to the computer first, and PhotoScan (now called Agisoft Metashape) software is used for data splicing processing. Through the operation steps of aligning the photos, generating dense point cloud, constructing grid texture and generating orthographic image, the orthographic image after data processing is the dimensionless ground remote sensing reflectivity value in the study area.

[0096] (2) Calculation of 14 vegetation indexes. The ENVI software is used to process the reflectivity of the image as a whole to obtain standard reflectivity image data, and the "band operation" tool is used to calculate each vegetation index.

[0097] (3) Calculation of leaf nitrogen nutrition index LNNI, aboveground nitrogen nutrition index ANNI and plant nitrogen nutrition index PNNI based on the critical nitrogen dilution model of leaf, aboveground and plant in embodiment 1.

[0098] (4) Correlation analysis of vegetation index and three kinds of nitrogen nutrition index. The Pearson correlation analysis is performed on 14 vegetation indexes and LNNI, ANNI and PNNI. The results show that the correlation of vegetation index and ANNI is the highest, which is inconsistent with the conclusion of taking plant as the optimal model. The reason is that the unmanned aerial vehicle multi-spectral data essentially reflects the canopy spectral characteristics, so it has stronger homology with the aboveground nitrogen nutrition index; and PNNI also contains the information of underground pods, which has no direct contribution to the remote sensing signal, and is affected by the allocation dynamics and measurement error, so the correlation with the vegetation index is relatively reduced.

[0099] (5) Screening the optimal vegetation index. NDRE and LCI are the optimal vegetation index. Combined with the optimal vegetation index, the ANNIs of the whole test plot are predicted. Figure 8 The specific analysis is as follows: 9 vegetation indexes (NDRE, GNDVI, SAVI, CVI, DVI, MSAVI, LCI, MTVI2, OSAVI) are extremely significantly positively correlated with PNNI, among which NDRE has the strongest correlation (0.59). NGI is extremely significantly negatively correlated with PNNI (-0.30), NDVI and RVI are weakly correlated with PNNI (0.24). MCARI and GCI have no significant correlation with PNNI. ANNIs are extremely significantly correlated with 13 vegetation indexes, NDRE has the highest correlation with ANNI (0.67), NGI is extremely significantly negatively correlated with ANNI (-0.40), MCARI has a lower correlation with ANNI (0.29), and GCI has no significant correlation with ANNI. LNNIs are extremely significantly positively correlated with 12 vegetation indexes, NDRE has the strongest correlation with LNNI (0.64), NGI is extremely significantly negatively correlated with LNNI (-0.38), MCARI has a weak correlation with LNNI, and GCI has no significant correlation with LNNI.

[0100] (6) The regression prediction model of peanut ANNI is constructed by using machine learning algorithm. In modeling, 75% of the data is used as the training set, and 25% of the data is used as the validation set. The optimal vegetation index is used as the independent variable, and ANNI is used as the dependent variable. The random forest regression, BP neural network regression and support vector machine regression algorithms are used to construct the ANNI inversion model as shown in Figure 9 . The prediction accuracy of the model is evaluated by using R² and RMSE, and the results show that the random forest regression model has the strongest prediction ability.

[0101] Table 3: Peanut ANNI prediction model

[0102]

[0103] (7) According to the screened random forest regression model, the ANNI of the whole test plot is predicted.

[0104] Random forest regression algorithm can effectively handle high-dimensional data regression problems, has strong generalization ability, anti-interference ability and high computing efficiency, and is not easy to fall into overfitting. In addition, the algorithm can also sort according to the importance of features. The importance of each independent variable is measured by the increase in node purity, which reflects the influence of each independent variable on the heterogeneity of each node observation of the regression tree. The greater the value, the more important the independent variable. In fact, the random forest algorithm is a random forest composed of multiple CART (Classification and Regression Tree) decision trees, with double randomness of data sampling and feature selection. Decision tree is a supervised learning algorithm based on tree structure, widely used in classification and regression problems. Its basic idea is to classify or predict data through a series of judgment rules. Decision tree is composed of nodes and edges, each node represents a feature or attribute, and each edge represents the relationship between nodes, representing decision rules or results. For regression problems, decision trees are based on Bagging (Bootstrap Aggregation) algorithm, which extracts m samples from the training set by resampling, and constructs m regression trees. In each node of each regression tree, a feature is randomly selected from n independent variables (n is less than the total number of independent variables) to divide the data space, and finally the average value of these trees is used to predict the value of the dependent variable.

[0105] The construction of the random forest model of the application is based on the sklearn library in the Python environment, and is realized by using the random forest regressor module. The specific steps are as follows:

[0106] 1. Making sample set: dividing the cell, collecting unmanned aerial vehicle and ground data on site, outlining the cell, and making sample set.

[0107] 2. Use the bagging method to select 75% of the samples from the sample set as a training set.

[0108] 3. Generate a decision tree using the sampled sample set. At each node of the generated tree:

[0109] ① Randomly select 14 features without repetition;

[0110] ② Use the 14 features to divide the sample set respectively, and find the best division feature.

[0111] 4. Repeat steps 2 to 3 for 100 times (100 is the number of decision trees in the random forest), and train to form the ANNI random forest inversion model.

[0112] 5. Use the trained random forest model to predict and invert the test sample, and use the voting method to determine the prediction result, and evaluate the accuracy of the model.

[0113] 6. The trained model is called to predict and retrieve the nitrogen nutrition index of the whole remote sensing image pixel by pixel, and the retrieval results are spliced and added with geographic reference coordinates to obtain the retrieval result map of the nitrogen nutrition index of the study area.

[0114] Figure 10 As shown in the figure, the height of the ANNI value is represented by different colors, ranging from 0.93 to 1.13. The redder the color, the lower the ANNI value, indicating that the peanut aboveground nitrogen is insufficient; the greener the color, the higher the ANNI value, indicating that the peanut aboveground nitrogen is excessive. The color of the N0 treatment plot is red, and the redness of the 7 plots without nitrogen treatment is deepened, indicating that the peanut aboveground is in a typical nitrogen deficiency state. The color of the N4 treatment plot is green, indicating that the peanut aboveground is in a state of excessive nitrogen. This further verifies the effectiveness and prediction ability of the random forest regression model in reflecting the nitrogen nutrition status of peanuts.

[0115] Example 6: Application of the peanut nitrogen nutrition index diagnosis method based on critical nitrogen concentration

[0116] In combination with Examples 1-5, this embodiment combines the optimized plant nitrogen nutrition index model PNNI and the peanut ANNI regression prediction model to comprehensively predict the nitrogen nutrition and yield of peanut plants and aboveground parts, and to analyze and adjust the fertilization strategy as a whole. The specific application steps are as follows:

[0117] Step 1: Constructing the aboveground critical nitrogen dilution model and the plant critical nitrogen dilution model of peanuts;

[0118] The aboveground critical nitrogen dilution model and the plant critical nitrogen dilution model are both Nc = a x W -b , wherein Nc is the critical nitrogen concentration value, W is the maximum value of peanut leaf dry matter accumulation, a and b are both parameters of the equation, a represents the critical nitrogen concentration when the aboveground dry matter is 1 t / ha, and b is a statistical parameter that determines the slope of the curve.

[0119] Step 2: According to the plant critical nitrogen dilution model obtained above, a general critical nitrogen dilution model of peanuts is constructed, and the model R² and RMSE are used as evaluation indexes to verify the general critical nitrogen dilution model of peanuts.

[0120] The formula of the general critical nitrogen dilution model of peanuts is Nc = 4.193W -0.189 , wherein Nc is the critical nitrogen concentration value, and W is the maximum value of peanut plant dry matter accumulation. The R² = 1 - , wherein: is the measured value, is the predicted value, is the mean value of the measured value, and n is the sample size.

[0121] Step 3: Based on the general critical nitrogen dilution model of peanuts, a plant nitrogen nutrition index model PNNI is constructed, and the nitrogen nutrition of peanut plants is diagnosed in combination with the PNNI index; the plant nitrogen nutrition index model PNNI = Na / Nc, wherein Na represents the actual nitrogen concentration of the plant, and Nc represents the critical nitrogen concentration of the plant.

[0122] The plant nitrogen nutrition index model PNNI and the nitrogen nutrition of peanut plants are related as follows: 0.95 ≤ PNNI ≤ 1.05 indicates that the nitrogen nutrition of peanut plants is in the best state; PNNI < 0.95 indicates that the nitrogen nutrition of peanut plants is insufficient; and PNNI > 1.05 indicates that the nitrogen nutrition of peanut plants is excessive.

[0123] Step 4: Based on the aboveground critical nitrogen dilution model of step 1, the aboveground nitrogen nutrition index ANNI is calculated, the multi-spectral data collected by the unmanned aerial vehicle is processed, a peanut ANNI regression prediction model is constructed by using a machine learning algorithm, and the aboveground nitrogen nutrition index of peanuts is predicted.

[0124] The peanut ANNI regression prediction model in step (4) is a random forest regression prediction model, which is used to predict the aboveground nitrogen nutrition index by using the random forest regression model described in embodiment 5, and in combination with the results of the plant nitrogen nutrition index model (PNNI) analysis, the nitrogen nutrition of peanuts is diagnosed as a whole, and the fertilization strategy for peanuts is adjusted.

[0125] According to the PNNI value, the present application sets 0.95-1.05 as the best nitrogen nutrition state of crops. In addition, the present application also analyzes the relationship between PNNI and relative yield at different growth stages, and the results show that PNNI can accurately reflect the nitrogen state of the plant.

[0126] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, the technical solutions recorded in the foregoing embodiments can still be modified by those of ordinary skill in the art, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions claimed by the present application.

Claims

1. A method for diagnosing the nitrogen nutrient index of peanuts based on critical nitrogen concentration, characterized in that, The method for diagnosing the nitrogen nutritional index of peanuts includes the following steps: (1) Construct the aboveground critical nitrogen dilution model and the plant critical nitrogen dilution model of peanut; the aboveground critical nitrogen dilution model and the plant critical nitrogen dilution model are both Nc=a×W -b Where Nc is the critical nitrogen concentration value, a represents the critical nitrogen concentration when the dry matter of the aboveground part or plant is 1 t / ha, and b is a statistical parameter that determines the slope of this curve. (2) Based on the plant critical nitrogen dilution model obtained above, a universal critical nitrogen dilution model for peanuts was constructed, and the model's R² and RMSE were used as evaluation indicators to verify the universal critical nitrogen dilution model for peanuts; the formula of the universal critical nitrogen dilution model for peanuts is Nc = 4.193 × W -0.189 , where Nc is the critical nitrogen concentration value and W is the maximum value of the accumulated dry matter of peanut plants; (3) Based on the peanut general critical nitrogen dilution model, construct the plant nitrogen nutrition index model PNNI, and combine the PNNI index to diagnose the nitrogen nutrition of peanut plants; the plant nitrogen nutrition index model PNNI=Na / Nc, where Na represents the actual nitrogen concentration of the plant and Nc represents the critical nitrogen concentration of the plant. (4) Based on the aboveground critical nitrogen dilution model described in step (1), calculate the aboveground nitrogen nutrition. The ANNI index is used to process multispectral data collected by drones and a machine learning algorithm is used to build a regression prediction model for peanut ANNI to predict the nitrogen nutrition index of the aboveground parts of peanut. (5) Based on the constructed peanut ANNI regression prediction model and plant nitrogen nutrition index model PNNI, the nitrogen nutrition index of peanuts in the experimental block was predicted, and the comprehensive analysis was conducted to adjust the peanut fertilization strategy.

2. The method for diagnosing peanut nitrogen nutrient index based on critical nitrogen concentration according to claim 1, characterized in that, The In the formula: These are measured values. For predicted values, is the mean of the measured values, and n is the number of samples.

3. The peanut nitrogen nutrient index diagnostic method based on critical nitrogen concentration according to claim 1, Its features are, The relationship between the plant nitrogen nutrition index model PNNI and peanut plant nitrogen nutrition is as follows: 0.95≤PNNI≤1.05 indicates that the peanut plant nitrogen nutrition is in the optimal state; PNNI<0.95 indicates that the peanut plant nitrogen nutrition is insufficient; PNNI>1.05 indicates that the peanut plant nitrogen nutrition is excessive.

4. The peanut nitrogen nutrient index diagnostic method based on critical nitrogen concentration according to claim 1, Its features are, In step (4), the peanut ANNI regression prediction model includes random forest regression prediction model, BP neural network regression prediction model and support vector regression prediction model. The prediction capabilities of the three models are: random forest regression prediction model > BP neural network regression prediction model > support vector regression prediction model.

5. The peanut nitrogen nutrient index diagnostic method based on critical nitrogen concentration according to claim 4, Its features are, The specific operation of the random forest regression model for predicting the aboveground nitrogen nutrient index is as follows: S1: Create a sample set; S2: Select 75% of the samples in the sample set as a training set using the sampling replacement method; S3: Generate a decision tree using the sampled data set; S4: Repeat steps S2 to S3 to train and form the ANNI random forest inversion model; S5: The trained random forest prediction model is used to predict and invert the test samples, and the prediction results are determined by voting. S6: Call the trained model to perform pixel-by-pixel prediction and inversion of the global remote sensing image, and stitch the inversion results together and add geographic reference coordinates to obtain the nitrogen nutrient index of the aboveground part.

6. The peanut nitrogen nutrient index diagnostic method based on critical nitrogen concentration according to claim 4, characterized in that, The multispectral data processing method for the UAV collected in step (4) is as follows: copy the image data collected by the UAV to the computer and use Agisoft Metashape software to perform data stitching processing; by aligning the photos, generating dense point clouds, constructing grid textures and generating orthophotos, the orthophoto after data processing is the dimensionless ground remote sensing reflectance value within the study area.

7. The peanut nitrogen nutrient index diagnostic method based on critical nitrogen concentration according to claim 4, Its features are, The ANNI value of the random forest regression prediction model ranges from 0.9 to 1.2.

Images