Soil available nitrogen remote sensing mapping method and system with synergic nitrogen supply and demand

Through Sentinel-2 satellite multispectral data and partial least squares method, combined with Kriging interpolation, the problem of large-scale application and insufficient fineness in soil fast-acting nitrogen remote sensing monitoring is solved, and the rapid, accurate monitoring and fineness of soil fast-acting nitrogen is achieved. and used for soil fertility evaluation and farmland management at the scale of field scale.

CN118916654APending Publication Date: 2024-11-08YANGZHOU UNIV
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410784462.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-18
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The existing technology lacks large-scale application in soil fast-acting nitrogen remote sensing monitoring, satellite multispectral data is insufficiently utilized, and the spatial autocorrelation of soil fast-acting nitrogen is not considered, resulting in insufficient monitoring precision and insufficient description of local spatial characteristics.

Method used

The Sentinel-2 satellite remote sensing multispectral data was used to establish a soil fast-acting nitrogen inversion model through pre-processing and feature parameter extraction, combined with partial least squares method and kriging interpolation, and verified the prediction results through the 50% cross-validation method to generate Final soil fast-acting nitrogen prediction map.

Benefits of technology

It realizes rapid, accurate and non-destructive monitoring of large-scale soil fast-acting nitrogen, improves the spatial information precision and prediction accuracy of soil fast-acting nitrogen, and can be used for field-scale soil fertility evaluation and farmland management zoning to guide crop fertilization.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118916654A_ABST
    Figure CN118916654A_ABST
Patent Text Reader

Abstract

The invention discloses a nitrogen supply and demand coordinated soil available nitrogen remote sensing mapping method and system, and relates to the technical field of agricultural environment and smart agriculture, and the method comprises the steps: obtaining remote sensing multispectral data, and preprocessing the remote sensing data to obtain characteristic parameters; extracting remote sensing variable information of corresponding sampling points, and determining model input variables; establishing a relationship between the soil available nitrogen and the remote sensing information by utilizing a partial least square method, and completing a preliminary prediction drawing of the soil available nitrogen; kriging interpolation is carried out, and a soil available nitrogen residual spatial distribution diagram is predicted; and outputting a final soil available nitrogen prediction map. The method is low in cost of acquiring satellite remote sensing data, and has a relatively great advantage of quickly and nondestructively acquiring the spatial information of the soil available nitrogen in a large range. Soil available nitrogen information obtained by using remote sensing data has rich field space detail information, and can be used for field scale soil fertility evaluation and spatial mapping, farmland management zoning, and guidance of crop variable rate fertilization and implementation of precision agriculture.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of agricultural environment and smart agricultural technology, and specifically to a soil available nitrogen remote sensing mapping method and system for coordinated nitrogen supply and demand. Background Art

[0002] Nitrogen is one of the essential nutrients for plants and is the main component of protein. It is extremely important for crop yield and quality and is also the most active factor in soil fertility. More than half of the total nitrogen accumulated by crops is obtained from the soil. Soil available nitrogen refers to nitrogen that can be directly absorbed by plant roots. As the main nutrient content in the soil, the soil available nitrogen content and its existence state are often positively correlated with crop yield. Therefore, monitoring the soil available nitrogen content is of great guiding significance for soil fertility evaluation, farmland management zoning, and guiding the rational application of nitrogen fertilizers for crops.

[0003] Traditionally, soil nutrient information is mainly obtained by chemical measurement, which is relatively accurate, but time-consuming, costly, and lacks representativeness of "point" sampling, and is not effective, making it difficult to meet the needs of actual production. Spatial interpolation methods based on geostatistics can extrapolate "point" data to the entire plot, meeting the needs of farmland spatial management, but model selection and data quality still lead to insufficient spatial information on soil nutrients and even the lack of spatial detail information.

[0004] Remote sensing technology provides powerful data and technical support for soil nutrient prediction and mapping. Remote sensing data has the advantages of wide coverage, low acquisition cost, fast revisit cycle, and rich spatial information. It effectively makes up for the shortcomings of traditional measurement and provides the possibility of macroscopic, rapid, accurate and non-destructive acquisition of soil physical and chemical properties. However, there are two shortcomings in the current remote sensing monitoring of soil nutrients: (1) Soil nutrient monitoring using remote sensing technology is mostly focused on soil organic matter, total nitrogen, total phosphorus and total potassium, and there is a lack of remote sensing monitoring methods for soil available components (such as available nitrogen); (2) The monitoring of soil available components that has been carried out is mainly based on the use of ground hyperspectral data, and satellite remote sensing multispectral data is rarely used, which still cannot achieve large-scale monitoring and application; (3) When the existing technology uses regression algorithms and remote sensing data to map soil available nitrogen, the spatial autocorrelation of soil available nitrogen is not considered, resulting in insufficiently detailed description of the spatial local characteristics of soil available nitrogen, which leads to the spatial mutation phenomenon of soil available nitrogen that is inconsistent with reality.

[0005] Crop nutrients come from the direct supply of nutrients in the soil, and the nutrient status of crops is a comprehensive reflection of the soil nutrient supply capacity and the crop's ability to absorb nutrients. Therefore, there is a certain supply-demand synergistic relationship between crops and soil nitrogen. Satellite remote sensing data obtains crop canopy spectral data, which can accurately evaluate crop growth and its nitrogen nutritional status, and thus reflect the soil nitrogen supply capacity. Compared with satellite images such as Landsat and Gaofen, Sentinel-2 multispectral images have richer spectral bands such as red edge and short-wave infrared, shorter revisit cycles and higher spatial resolution. Therefore, using Sentinel-2 satellite remote sensing multispectral data to monitor crop growth characteristics and then conduct field-scale soil available nitrogen monitoring has good theoretical support and feasibility. Summary of the invention

[0006] In view of the above-mentioned problems, the present invention is proposed.

[0007] Therefore, the technical problem solved by the present invention is: how to use multi-spectral satellite remote sensing data and crop-soil supply and demand to coordinately map and predict soil available nitrogen.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions: a soil available nitrogen remote sensing mapping method with coordinated nitrogen supply and demand, comprising the following steps:

[0009] Acquire Sentinel-2 remote sensing multispectral data, preprocess the remote sensing data, and obtain characteristic parameters.

[0010] The soil available nitrogen sample point data were obtained in the study area, the remote sensing variable information of the corresponding sampling points was extracted, and exploratory data analysis was carried out on the data set to determine the model input variables.

[0011] The partial least squares method was used to establish the relationship between soil available nitrogen and remote sensing information, determine the number of principal components, obtain the corresponding soil available nitrogen inversion model, and carry out preliminary prediction and mapping of soil available nitrogen.

[0012] The predicted values ​​of soil available nitrogen at the training points were extracted from the preliminary prediction map. The measured values ​​of soil available nitrogen at the training points were subtracted from the preliminary predicted values ​​of soil available nitrogen. Kriging interpolation was performed on the residual values ​​at the points to predict the spatial distribution map of soil available nitrogen residuals.

[0013] The preliminary predicted soil available nitrogen map was added to the residual map obtained by Kriging interpolation to obtain the final soil available nitrogen prediction map, and the prediction results were verified using the five-fold cross-validation method.

[0014] As a preferred solution of the soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand described in the present invention, the Sentinel-2 remote sensing multispectral data obtained is an L2A-level ground reflectance image, and the image formation time and ground sampling time do not exceed the specified time.

[0015] The preprocessing includes format conversion, band synthesis, study area clipping and resampling to obtain band reflectance with spatial resolution.

[0016] The preprocessed Sentinel-2 images were mathematically transformed and processed to extract the crop nitrogen characteristic parameters.

[0017] The characteristic parameters include greenness index, chlorophyll index and nitrogen index.

[0018] As a preferred scheme of the soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand described in the present invention, wherein: the acquisition of soil available nitrogen sample point data in the study area includes selecting terrain with flat terrain, uniform crop growth, and far away from roads and houses, as well as locations with good, medium and poor soil fertility or crop growth, removing dead grass and debris on the soil surface during sampling, and keeping the soil sampling depth and sampling volume of each sampling point consistent.

[0019] At each sampling point, multiple soil samples were collected in a plum blossom shape, mixed and placed in plastic bags. After the soil samples were collected, they were air-dried in time and the soil available nitrogen was determined through laboratory chemical analysis.

[0020] As a preferred solution of the soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand described in the present invention, the remote sensing variable information of the corresponding sampling points is extracted, and exploratory data analysis of the data set is carried out, including using the coordinates of the ground sampling points to extract the corresponding values ​​of the characteristic variable data on the remote sensing image, forming a training sample set, calculating the correlation coefficient matrix, using the t-hypothesis test method, completing the significance test of the correlation coefficient between the remote sensing variable and the soil available nitrogen, removing the variables whose significance level with the soil available nitrogen is greater than the first threshold, and analyzing the correlation coefficient matrix to find the variable pairs with correlation coefficients greater than the second threshold in the remote sensing variable set, removing the variables with small correlation coefficient values ​​with the soil available nitrogen, and screening out the remote sensing variable set for model input.

[0021] The training sample set includes band reflectance, nitrogen characteristic parameters and soil available nitrogen.

[0022] As a preferred scheme of the soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand described in the present invention, wherein: the determination of the model input variables includes screening the remote sensing band reflectance and nitrogen characteristic parameters, taking the ground-observed soil available nitrogen as the output, and establishing the soil available nitrogen model using the partial least squares method.

[0023] The soil available nitrogen dependent variable Y and m remote sensing characteristic independent variables (X1, X2, ..., X m ), collect n sample points and form a data table X of independent variables and dependent variables n×m and Y n×1 , the partial least squares method extracts components t1 and u1 in X and Y respectively. After the first components t1 and u1 are extracted, the partial least squares method performs regression of X and Y on t1 respectively.

[0024] If the regression equation reaches the preset accuracy, it is terminated. Otherwise, the residual information of X and Y explained by t1 is used for the second round of component extraction until the preset accuracy is reached.

[0025] If p components t1,…,t are finally extracted from X p , partial least squares regression will implement Y on t1,…,t p The regression of Y with respect to the original variables X1, X2, …, X m The regression equation of .

[0026] When the cumulative contribution rate of the principal component to the input remote sensing parameters and the output soil available nitrogen variable is greater than the third threshold, and the sum of the variance explained by the remote sensing parameters remains unchanged, the partial least squares model is built as the optimal number of principal components, and the expression is:

[0027] Y=a1×X1+a2×X2+…+a m ×X m

[0028] Where Y is soil available nitrogen, a1,…,a m are regression coefficients, X1, X2, …, X m is the remote sensing parameter variable.

[0029] The model is applied to the selected remote sensing image matrix to predict the preliminary results of soil available nitrogen in the entire study area, expressed as:

[0030] SAN PLSR =a1×S1+a2×S2+…+a n ×S n

[0031] Among them, SAN PLSR is the soil available nitrogen map initially predicted by the partial least squares model, a1, a2, …, a n are regression coefficients, S1, S2, …, S n It is a remote sensing variable parameter image.

[0032] The components t1, u1 are linear combinations of (X1, X2, ..., Xm) and Y, respectively.

[0033] As a preferred embodiment of the soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand of the present invention, the predicted soil available nitrogen residual spatial distribution map includes the soil available nitrogen measured value trained by partial least squares modeling minus the soil available nitrogen value predicted by partial least squares, and outputting the soil available nitrogen residual of the training sample point, which is expressed as:

[0034] SAN resd (s i )=SAN obs (s i )-SAN PLSR (s i )

[0035] Among them, SAN obs (s i ) is the sampling point s i Measured soil available nitrogen, SAN PLSR (s i ) is the sampling point s predicted by the partial least squares method i Soil available nitrogen, SAN resd (s i ) is the sampling point s i Soil available nitrogen residual.

[0036] Ordinary Kriging interpolation is performed on the point residuals, and all output values ​​are integrated to complete the prediction of the spatial distribution map of soil available nitrogen residuals in the entire study area. The expression is:

[0037]

[0038] Among them, SAN resd-OK is the soil available nitrogen residual map after ordinary Kriging interpolation, λ i Weights determined for spatial autocorrelation analysis of soil available nitrogen residuals.

[0039] As a preferred embodiment of the soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand of the present invention, the predicted soil available nitrogen residual spatial distribution map includes superimposing the preliminary predicted soil available nitrogen map with the residual map obtained by ordinary Kriging interpolation, and outputting the final soil available nitrogen prediction map, which is expressed as follows:

[0040] SAN final =SAN PLSR +SAN resd-OK

[0041] Among them, SAN resd-OK It is the soil available nitrogen residual map after ordinary Kriging interpolation, SAN PLSR The soil available nitrogen map, SAN finalThe final predicted soil available nitrogen map.

[0042] The prediction results were verified using the five-fold cross-validation method. The data set containing remote sensing variables and soil available nitrogen was randomly divided into five subsets of equal size. During the training and verification process of the partial least squares model, one of the subsets was repeatedly used as an independent test set, and the remaining four subsets were used as training sets. The partial least squares model constructed with the training set was used to predict the soil available nitrogen at the sample points of the test set. The measured and predicted values ​​of soil available nitrogen on the test set were compared to complete the model evaluation.

[0043] The measured and predicted values ​​of soil available nitrogen on the test set are compared to complete the model evaluation including the coefficient of determination, root mean square error and mean absolute error, and the expression is:

[0044]

[0045] Among them, X i is the measured value of sample i, Y i is the estimated value of sample i, is the mean of the sample measured values, and n is the total sample size.

[0046] Another object of the present invention is to provide a soil available nitrogen remote sensing mapping system that coordinates nitrogen supply and demand, which can work together through four modules: remote sensing data preprocessing, data acquisition and analysis, modeling and prediction, and residual correction and verification, thereby solving the problems of complex data processing and low model prediction accuracy in existing methods.

[0047] To solve the above technical problems, the present invention provides the following technical solutions: a soil available nitrogen remote sensing mapping system with coordinated nitrogen supply and demand, comprising: a remote sensing data preprocessing module, a data acquisition and exploratory analysis module, a modeling and preliminary prediction module, and a residual prediction and correction module.

[0048] The remote sensing data preprocessing module obtains Sentinel-2 remote sensing multispectral data and preprocesses it, including format conversion, band synthesis, study area clipping and resampling, to obtain the band reflectance of spatial resolution.

[0049] The data collection and exploratory analysis module is to obtain soil available nitrogen sample point data in the study area and extract remote sensing variable information of the corresponding sampling points.

[0050] The modeling and preliminary prediction module uses the partial least squares method to establish the relationship between soil available nitrogen and remote sensing information, determine the number of principal components, establish a soil available nitrogen inversion model, and carry out preliminary prediction mapping. By applying the model to the remote sensing image matrix, a preliminary result map of soil available nitrogen in the entire study area is predicted.

[0051] The residual prediction and correction module extracts the soil available nitrogen prediction value of the training sample point from the preliminary prediction map, subtracts the preliminary prediction value from the measured value of the soil available nitrogen of the training sample point to obtain the residual and performs Kriging interpolation on the residual to generate a residual spatial distribution map, adds the preliminary predicted soil available nitrogen map and the residual map to obtain the final soil available nitrogen prediction map, verifies the prediction result using the five-fold cross validation method, and completes the model evaluation.

[0052] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand are implemented.

[0053] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand.

[0054] Beneficial effects of the present invention: Compared with traditional chemical analysis methods, the present invention has low cost in acquiring satellite remote sensing data, simple data processing, strong timeliness, and has great advantages in rapid and non-destructive acquisition of spatial information of available nitrogen in soil over a wide range. The present invention combines statistical models and geostatistical methods to carry out soil available nitrogen mapping, which can not only ensure the prediction accuracy, but also take into account the spatial autocorrelation of available nitrogen in soil, and has high prediction accuracy; the soil available nitrogen information obtained using remote sensing data has rich field spatial detail information, which can be used for field-scale soil fertility evaluation and spatial mapping, farmland management zoning, and guide the implementation of variable fertilization of crops and precision agriculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0056] Figure 1 An overall flow chart of a soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand provided in the first embodiment of the present invention.

[0057] Figure 2 A flowchart of constructing a soil available nitrogen map in a soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand provided in the first embodiment of the present invention.

[0058] Figure 3 A structural diagram of a soil available nitrogen remote sensing mapping system for coordinated nitrogen supply and demand provided in a second embodiment of the present invention.

[0059] Figure 4 A scatter plot for verifying the accuracy of Hulunbuir farmland soil available nitrogen mapping results predicted in a soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand provided in the third embodiment of the present invention.

[0060] Figure 5 A scatter plot for verifying the accuracy of Hulunbuir farmland soil available nitrogen mapping results predicted by traditional partial least squares method in a soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand provided in the third embodiment of the present invention.

[0061] Figure 6 A scatter plot for verifying the accuracy of Hulunbuir farmland soil available nitrogen mapping results predicted by traditional remote sensing greenness index in a soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand provided in the third embodiment of the present invention.

[0062] Figure 7 A spatial distribution map of soil available nitrogen in Hulunbuir farmland predicted in a soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand provided in the third embodiment of the present invention. DETAILED DESCRIPTION

[0063] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0064] Example 1

[0065] Reference Figure 1-2 As one embodiment of the present invention, a soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand is provided, comprising:

[0066] S101: Acquire Sentinel-2 remote sensing multispectral data, pre-process the remote sensing data, and perform mathematical transformation and processing to extract characteristic parameters reflecting crop nitrogen such as spectral band reflectance, greenness index, chlorophyll index and leaf pigment index.

[0067] S102: Obtain soil available nitrogen sample point data in the study area; extract remote sensing variable information of corresponding sampling points, conduct exploratory data analysis on the data set, and determine the model input variables.

[0068] S103: Use partial least squares method to establish the relationship between soil available nitrogen and remote sensing information, determine the number of principal components, obtain the corresponding soil available nitrogen inversion model, and carry out preliminary mapping of soil available nitrogen.

[0069] S104: Subtract the predicted soil available nitrogen from the trained measured value of soil available nitrogen, perform Kriging interpolation on the point residual values, and predict the spatial distribution map of soil available nitrogen residuals.

[0070] S105: Add the preliminary predicted soil available nitrogen map and the residual map obtained by Kriging to obtain the final soil available nitrogen prediction map; and verify the prediction results using a five-fold cross validation method.

[0071] Sentinel-2 remote sensing data were obtained from the European Space Agency's Copernicus Data Space Ecosystem website. The obtained Sentinel-2 remote sensing data were L2A-level ground reflectance images with a time close to the study area and ground sampling. The images were preprocessed, including format conversion, band synthesis, study area cropping, and resampling, to obtain 10 band reflectances with a spatial resolution of 10 meters, including blue, green, red, red edge 1, red edge 2, red edge 3, near infrared, red edge 4, shortwave infrared 1, and shortwave infrared 2.

[0072] The preprocessed Sentinel-2 remote sensing data were used to extract crop nitrogen characteristic parameters, including greenness index (NDVI, RVI, EVI, ARVI), chlorophyll index (NDVIre, mSRre, mNDVIre, LCI) and nitrogen index (SIPI).

[0073] Table 1 Calculation formula for crop nitrogen characteristic parameters

[0074]

[0075] The corresponding values ​​of the characteristic variable data on the remote sensing image were extracted using the coordinates of the ground sampling points to form a training sample set including band reflectance, nitrogen characteristic parameters and soil available nitrogen. The correlation coefficient matrix was calculated, and the significance of the remote sensing variables and soil available nitrogen was tested using the t hypothesis test method. The variables with a significance level greater than 0.05 with soil available nitrogen were removed, and the correlation coefficient matrix was analyzed to find the variable pairs with a correlation coefficient greater than 0.85 in the remote sensing variable set. The variables with a small correlation coefficient value with soil available nitrogen were removed, and the remote sensing variable set for model input was screened out.

[0076] The model input variables include the reflectance of the selected remote sensing bands and nitrogen characteristic parameters. The soil available nitrogen observed on the ground is used as the output, and the soil available nitrogen model is established using the partial least squares method. The partial least squares method is a multivariate statistical analysis method that integrates the functions and advantages of multivariate linear regression analysis, canonical correlation analysis, and principal component analysis. It organically combines the data analysis of modeling with the non-model data cognition method to improve the accuracy, robustness, and practicality of the model. For the soil available nitrogen dependent variable (Y) and m remote sensing characteristic independent variables (X1, X2, ..., Xm), n sample points are observed, thus forming the data tables Xn×m and Yn×1 of independent variables and dependent variables. The partial least squares method extracts components t1 and u1 in X and Y respectively (t1 and u1 are linear combinations of X1, X2, ..., Xm and Y respectively). Two conditions must be met during extraction: t1 and u1 should carry as much variation information in their respective data tables as possible; the correlation between t1 and u1 reaches the maximum. After the first components t1 and u1 are extracted, partial least squares is used to regress X and Y on t1. If the regression equation reaches a satisfactory accuracy, the algorithm terminates; otherwise, the residual information of X and Y after being explained by t1 is used to extract the second round of components, and this process is repeated until a satisfactory accuracy is achieved. If p components t1, ..., tp are finally extracted from X, partial least squares regression will perform regression of Y on t1, ..., tp, and express it as the regression equation of Y on the original variables X1, X2, ..., Xm; for the determination of the number of principal components, when the cumulative contribution rate of the principal components to the input remote sensing parameters and the output soil available nitrogen variables is greater than 85%, and the cumulative explained variance of the independent variables remains basically unchanged, the partial least squares model is built as the optimal number of principal components Y = a1×X1+a2×X2+am×Xm, where Y is soil available nitrogen, a1, ..., am are regression coefficients, and X1, X2, ..., Xm are remote sensing parameter variables; and the model is applied to the selected remote sensing image matrix, refer to Figure 2 The preliminary results map of soil available nitrogen in the entire study area was predicted.

[0077] SANP LSR =a1×S1+a2×S2+…+a n ×S n

[0078] Among them, SAN PLSR is the soil available nitrogen map initially predicted by the partial least squares model, a1, a2, …, a n are regression coefficients, S1, S2, …, S n It is a remote sensing variable parameter image.

[0079] It should be further explained that explained variance is a statistical concept, and the core idea is to measure the model's ability to explain data changes or the accuracy of prediction. In the partial least squares method, explained variance refers to the variability in the entire data set explained by each principal component. This is used to quantify the degree to which each principal component explains the variability (variance) of the original data set, that is, the importance of each principal component. Specifically, the explained variance ratio of the i-th principal component is calculated by dividing the eigenvalue of the principal component by the sum of all eigenvalues, thereby obtaining the proportion of the total variance explained by the principal component. Since the sum of all eigenvalues ​​is equal to the total variance of the original data set, the sum of the explained variance ratios of all principal components must be 1, indicating that they jointly explain all the variability of the data set.

[0080] The soil available nitrogen value predicted by partial least squares is subtracted from the measured value of soil available nitrogen trained by partial least squares modeling, and the soil available nitrogen residual of the training sample point is output. The expression is: SAN resd (s i )=SAN obs (s i )-SAN PLSR (s i ), where SAN obs (s i ) is the soil available nitrogen measured at the sampling point, SAN PLSR (s i ) is the available nitrogen in the soil at the sampling point predicted by the partial least squares method, SAN resd (s i ) is the soil available nitrogen residual at the sampling point; ordinary Kriging interpolation is performed on the point residuals, and all output values ​​are integrated to complete the prediction of the spatial distribution map of soil available nitrogen residuals in the entire study area. The expression is:

[0081]

[0082] Among them, SAN resd-OK is the soil available nitrogen residual map after ordinary Kriging interpolation, λ i Weights determined for spatial autocorrelation analysis of soil available nitrogen residuals.

[0083] The preliminary predicted soil available nitrogen map is superimposed with the residual map obtained by ordinary Kriging interpolation to output the final soil available nitrogen prediction map, which is expressed as:

[0084] SAN final =SAN PLSR +SAN resd-OK

[0085] Among them, SAN resd-OK It is the soil available nitrogen residual map after ordinary Kriging interpolation, SAN PLSRThe soil available nitrogen map, SAN final The final predicted soil available nitrogen map.

[0086] The prediction results were verified using the five-fold cross-validation method. The data set containing remote sensing variables and soil available nitrogen was randomly divided into five subsets of equal size. During the training and verification process of the partial least squares model, one of the subsets was repeatedly used as an independent test set, and the remaining four subsets were used as training sets. The partial least squares model constructed with the training set was used to predict the soil available nitrogen at the sample points of the test set, and the measured and predicted values ​​of soil available nitrogen on the test set were compared to achieve model evaluation. The determination coefficient (R2), mean absolute error (MAE) and root mean square error (RMSE) were used to evaluate the model accuracy.

[0087] Example 2

[0088] Reference Figure 3 , which is an embodiment of the present invention, provides a system for a soil available nitrogen remote sensing mapping method with coordinated nitrogen supply and demand, including: a remote sensing data preprocessing module, a data acquisition and exploratory analysis module, a modeling and preliminary prediction module, and a residual prediction and correction module.

[0089] The remote sensing data preprocessing module is to obtain Sentinel-2 remote sensing multispectral data and preprocess it, including format conversion, band synthesis, study area cropping and resampling, to obtain the band reflectance of spatial resolution.

[0090] The data collection and exploratory analysis module is to obtain soil available nitrogen sample point data in the study area and extract remote sensing variable information of the corresponding sampling points.

[0091] The modeling and preliminary prediction module uses the partial least squares method to establish the relationship between soil available nitrogen and remote sensing information, determine the number of principal components, establish a soil available nitrogen inversion model, and carry out preliminary prediction mapping. By applying the model to the remote sensing image matrix, the preliminary result map of soil available nitrogen in the entire study area is predicted.

[0092] The residual prediction and correction module extracts the predicted values ​​of soil available nitrogen of the training points from the preliminary prediction map, subtracts the preliminary prediction values ​​from the measured values ​​of soil available nitrogen of the training points, obtains the residuals and performs Kriging interpolation on the residuals to generate a residual spatial distribution map. The preliminary predicted soil available nitrogen map is added to the residual map to obtain the final soil available nitrogen prediction map. The prediction results are verified using the five-fold cross-validation method to complete the model evaluation.

[0093] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

[0094] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0095] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0096] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0097] Example 3

[0098] Reference Figures 4 to 7 This is another embodiment of the present invention. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments. This embodiment experiments the existing traditional method and the method of this embodiment respectively.

[0099] Step 1: Obtain remote sensing data on the platform

[0100] Sentinel-2 remote sensing data were obtained from the European Space Agency's Copernicus Data Space Ecosystem website. The obtained Sentinel-2 remote sensing data were L2A-level ground reflectance images with a time close to the study area and ground sampling. The images were preprocessed, including format conversion, band synthesis, study area cropping, and resampling, to obtain 10 band reflectances with a spatial resolution of 10 meters, including blue, green, red, red edge 1, red edge 2, red edge 3, near infrared, red edge 4, shortwave infrared 1, and shortwave infrared 2.

[0101] Step 2: Mathematical transformation of preprocessed Sentinel-2 remote sensing data

[0102] The preprocessed Sentinel-2 remote sensing data were used to extract crop nitrogen characteristic parameters, including greenness index (NDVI, RVI, EVI, ARVI), chlorophyll index (NDVIre, mSRre, mNDVIre, LCI) and nitrogen index (SIPI).

[0103] Table 1 Calculation formula for crop nitrogen characteristic parameters

[0104]

[0105]

[0106] Step 3: Conduct exploratory data analysis on the dataset

[0107] The corresponding values ​​of the characteristic variable data on the remote sensing image were extracted using the coordinates of the ground sampling points to form a training sample set including band reflectance, nitrogen characteristic parameters and soil available nitrogen. The correlation coefficient matrix was calculated, and the significance of the remote sensing variables and soil available nitrogen was tested using the t hypothesis test method. The variables with a significance level greater than 0.05 with soil available nitrogen were removed, and the correlation coefficient matrix was analyzed to find the variable pairs with a correlation coefficient greater than 0.85 in the remote sensing variable set. The variables with a small correlation coefficient value with soil available nitrogen were removed, and the remote sensing variable set for model input was screened out.

[0108] Step 4: Establish soil available nitrogen inversion model

[0109] The model input variables include the reflectance of the selected remote sensing bands and nitrogen characteristic parameters. The soil available nitrogen observed on the ground is used as the output, and the soil available nitrogen model is established using the partial least squares method. The partial least squares method is a multivariate statistical analysis method that integrates the functions and advantages of multivariate linear regression analysis, canonical correlation analysis, and principal component analysis. It organically combines the data analysis of modeling with the non-model data cognition method to improve the accuracy, robustness, and practicality of the model. For the soil available nitrogen dependent variable (Y) and m remote sensing characteristic independent variables (X1, X2, ..., Xm), n sample points are observed, thus forming the data tables Xn×m and Yn×1 of independent variables and dependent variables. The partial least squares method extracts components t1 and u1 in X and Y respectively (t1 and u1 are linear combinations of X1, X2, ..., Xm and Y respectively). Two conditions must be met during extraction: t1 and u1 should carry as much variation information in their respective data tables as possible; the correlation between t1 and u1 reaches the maximum. After the first components t1 and u1 are extracted, partial least squares is used to regress X and Y on t1. If the regression equation reaches a satisfactory accuracy, the algorithm terminates; otherwise, the residual information of X and Y after being explained by t1 is used to extract the second round of components, and this process is repeated until a satisfactory accuracy is achieved. If p components t1, ..., tp are finally extracted from X, partial least squares regression will perform regression of Y on t1, ..., tp, and express it as the regression equation of Y on the original variables X1, X2, ..., Xm; for the determination of the number of principal components, when the cumulative contribution rate of the principal components to the input remote sensing parameters and the output soil available nitrogen variables is greater than 85%, and the cumulative explained variance of the independent variables remains basically unchanged, the partial least squares model Y = a1×X1+a2×X2+am×Xm is used as the optimal number of principal components for modeling, where Y is soil available nitrogen, a1, ..., am are regression coefficients, and X1, X2, ..., Xm are remote sensing parameter variables; and the model is applied to the selected remote sensing image matrix to predict the preliminary result map of soil available nitrogen in the entire study area.

[0110] SAN PLSR =a1×S1+a2×S2+…+a n×S n

[0111] Among them, SAN PLSR is the soil available nitrogen map initially predicted by the partial least squares model, a1, a2, …, a n are regression coefficients, S1, S2, …, S n It is a remote sensing variable parameter image.

[0112] Step 5: Predict the spatial distribution of soil available nitrogen residuals

[0113] The soil available nitrogen value predicted by partial least squares is subtracted from the measured value of soil available nitrogen trained by partial least squares modeling, and the soil available nitrogen residual of the training sample point is output. The expression is: SAN resd (s i )=SAN obs (s i )-SAN PLSR (s i ), where SAN obs (s i ) is the soil available nitrogen measured at the sampling point, SAN PLSR (s i ) is the available nitrogen in the soil at the sampling point predicted by the partial least squares method, SAN resd (s i ) is the soil available nitrogen residual at the sampling point; ordinary Kriging interpolation is performed on the point residuals, and all output values ​​are integrated to complete the prediction of the spatial distribution map of soil available nitrogen residuals in the entire study area. The expression is:

[0114]

[0115] Among them, SAN resd-OK is the soil available nitrogen residual map after ordinary Kriging interpolation, λ i Weights determined for spatial autocorrelation analysis of soil available nitrogen residuals.

[0116] Step 6: Map the soil available nitrogen

[0117] The preliminary predicted soil available nitrogen map is superimposed with the residual map obtained by ordinary Kriging interpolation to output the final soil available nitrogen prediction map, which is expressed as:

[0118] SAN final =SAN PLSR +SAN resd-OK

[0119] Among them, SAN resd-OK It is the soil available nitrogen residual map after ordinary Kriging interpolation, SAN PLSR The soil available nitrogen map, SANfinal The final predicted soil available nitrogen map.

[0120] Step 7: Verify the results of soil available nitrogen mapping

[0121] The prediction results were verified using the five-fold cross-validation method. The data set containing remote sensing variables and soil available nitrogen was randomly divided into five subsets of equal size. During the training and verification process of the partial least squares model, one of the subsets was repeatedly used as an independent test set, and the remaining four subsets were used as training sets. The partial least squares model constructed with the training set was used to predict the soil available nitrogen at the sample points of the test set, and the measured and predicted values ​​of soil available nitrogen on the test set were compared to achieve model evaluation. The determination coefficient (R2), mean absolute error (MAE) and root mean square error (RMSE) were used to evaluate the model accuracy. The results are shown in Figure 2. Figure 4 , Figure 5 and Figure 6 shown.

[0122] The results show that compared with the traditional remote sensing greenness index and the traditional partial least squares method, the soil available nitrogen predicted by the present invention has the highest accuracy, R2 reaches 0.90, RMSE is 0.83mg / kg, and MAE is 0.67mg / kg; the soil available nitrogen predicted by the traditional remote sensing greenness index has the worst accuracy, R2 is 0.67, RMSE is 1.55mg / kg, and MAE is 1.33mg / kg; the soil available nitrogen predicted by the traditional partial least squares method has an R2 of 0.89, RMSE of 0.94mg / kg, and MAE of 0.76mg / kg. The present invention has better performance in the three accuracy evaluation indicators, indicating that it can reduce the error in the prediction of soil available nitrogen and improve the prediction accuracy.

[0123] Prediction of spatial distribution of soil available nitrogen Figure 7 It can be seen that the soil available nitrogen content map predicted by the method proposed in the present invention can very well reflect the difference in soil available nitrogen under different crop management gradients, which has important indicative significance for precise fertilization and crop management within the field.

[0124] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A soil available nitrogen remote sensing mapping method with coordinated nitrogen supply and demand, characterized in that: include: Acquire Sentinel-2 remote sensing multispectral data, preprocess the remote sensing data, and obtain characteristic parameters; Obtain soil available nitrogen sample point data in the study area, extract remote sensing variable information of the corresponding sampling points, conduct exploratory data analysis on the data set, and determine the model input variables; The partial least squares method was used to establish the relationship between soil available nitrogen and remote sensing information, determine the number of principal components, obtain the corresponding soil available nitrogen inversion model, and carry out preliminary prediction and mapping of soil available nitrogen; Extract the predicted value of soil available nitrogen of the training sample point from the preliminary prediction map, subtract the preliminary predicted value of soil available nitrogen from the measured value of soil available nitrogen of the training sample point, perform Kriging interpolation on the residual value of the point, and predict the spatial distribution map of soil available nitrogen residual; The preliminary predicted soil available nitrogen map was added to the residual map obtained by Kriging interpolation to obtain the final soil available nitrogen prediction map; the prediction results were verified using the five-fold cross-validation method.

2. The method for remote sensing mapping of soil available nitrogen with coordinated nitrogen supply and demand as claimed in claim 1, characterized in that: The Sentinel-2 remote sensing multispectral data obtained is L2A-level ground reflectance image, and the image formation time and ground sampling time do not exceed the specified time; The preprocessing includes format conversion, band synthesis, research area clipping and resampling to obtain the band reflectance of spatial resolution; Perform mathematical transformation and processing on the pre-processed Sentinel-2 images to extract crop nitrogen characteristic parameters; The characteristic parameters include greenness index, chlorophyll index and nitrogen index.

3. The method for remote sensing mapping of soil available nitrogen with coordinated nitrogen supply and demand as claimed in claim 2, characterized in that: The acquisition of soil available nitrogen sample point data in the study area includes selecting a flat terrain, uniform crop growth, and terrain far away from roads and houses, as well as a location where soil fertility or crop growth is good, medium, and poor. When sampling, dead grass and debris on the soil surface are removed, and the soil sampling depth and sampling volume of each sampling point are kept consistent; At each sampling point, multiple soil samples were collected in a plum blossom shape, mixed and placed in plastic bags. After the soil samples were collected, they were air-dried in time and the soil available nitrogen was determined through laboratory chemical analysis.

4. The method for remote sensing mapping of soil available nitrogen with coordinated nitrogen supply and demand as claimed in claim 3, characterized in that: The extracting of remote sensing variable information of corresponding sampling points and conducting exploratory data analysis on the data set include extracting corresponding values ​​of characteristic variable data on remote sensing images using the coordinates of ground sampling points, forming a training sample set, calculating a correlation coefficient matrix, and using a t-hypothesis test method to complete a significance test of the correlation coefficient between remote sensing variables and soil available nitrogen, removing variables with a significance level greater than a first threshold with respect to soil available nitrogen, and analyzing the correlation coefficient matrix to find variable pairs with correlation coefficients greater than a second threshold in the remote sensing variable set, removing variables with small correlation coefficients with soil available nitrogen, and screening out a remote sensing variable set for model input; The training sample set includes band reflectance, nitrogen characteristic parameters and soil available nitrogen.

5. The method for remote sensing mapping of soil available nitrogen with coordinated nitrogen supply and demand as claimed in claim 4, characterized in that: Determining the model input variables includes screening the remote sensing band reflectance and nitrogen characteristic parameters, taking the ground-observed soil available nitrogen as output, and establishing a soil available nitrogen model using the partial least squares method; The soil available nitrogen dependent variable Y and m remote sensing characteristic independent variables (X1, X2, ..., X m ), collect n sample points and form a data table X of independent variables and dependent variables n×m and Y n×1 , the partial least squares method extracts components t1 and u1 in X and Y respectively. After the first components t1 and u1 are extracted, the partial least squares method regresses X and Y on t1 respectively; If the regression equation reaches the preset accuracy, it is terminated, otherwise the residual information of X and Y explained by t1 is used for the second round of component extraction until the preset accuracy is reached; If p components t1,…,t are finally extracted from X p , partial least squares regression will implement Y on t1,…,t p The regression of Y with respect to the original variables X1, X2, …, X m The regression equation of When the cumulative contribution rate of the principal component to the input remote sensing parameters and the output soil available nitrogen variable is greater than the third threshold, and the sum of the variance explained by the remote sensing parameters remains unchanged, the partial least squares model is built as the optimal number of principal components, and the expression is: And=a1×X1+a2×X2+…+a m ×X m Where Y is soil available nitrogen, a1,…,a m are regression coefficients, X1, X2, …, X m is the remote sensing parameter variable; The model is applied to the selected remote sensing image matrix to predict the preliminary results of soil available nitrogen in the entire study area, expressed as: SAN PLSR =a1×S1+a2×S2+…+a n ×S n Among them, SAN PLSR is the soil available nitrogen map initially predicted by the partial least squares model, a1, a2, …, a n are regression coefficients, S1, S2, …, S n It is a remote sensing variable parameter image; The components t1, u1 are linear combinations of (X1, X2, ..., Xm) and Y, respectively.

6. The method for remote sensing mapping of soil available nitrogen with coordinated nitrogen supply and demand as claimed in claim 5, characterized in that: The predicted soil available nitrogen residual spatial distribution map includes the soil available nitrogen measured value trained by partial least squares modeling minus the soil available nitrogen value predicted by partial least squares, and outputting the soil available nitrogen residual of the training sample point, which is expressed as: SAN resd (s i )=SAN obs (s i )-SAN PLSR (s i ) Among them, SAN obs (s i ) is the sampling point s i Measured soil available nitrogen, SAN PLSR (s i ) is the sampling point s predicted by the partial least squares method i Soil available nitrogen, SAN resd (s i ) is the sampling point s i soil available nitrogen residual; Ordinary Kriging interpolation is performed on the point residuals, and all output values ​​are integrated to complete the prediction of the spatial distribution map of soil available nitrogen residuals in the entire study area. The expression is: Among them, SAN resd-OK is the soil available nitrogen residual map after ordinary Kriging interpolation, λ i Weights determined for spatial autocorrelation analysis of soil available nitrogen residuals.

7. A soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand as claimed in claim 6, characterized in that: The predicted soil available nitrogen residual spatial distribution map includes superimposing the initially predicted soil available nitrogen map with the residual map obtained by ordinary Kriging interpolation, and outputting the final soil available nitrogen prediction map, which is expressed as: SAINT final =SAN PLSR +SAINT resd-OK Among them, SAN resd-OK It is the soil available nitrogen residual map after ordinary Kriging interpolation, SAN PLSR The soil available nitrogen map, SAN final The final predicted soil available nitrogen map; The prediction results were verified using the five-fold cross-validation method. The data set containing remote sensing variables and soil available nitrogen was randomly divided into five subsets of equal size. During the training and verification process of the partial least squares model, one of the subsets was repeatedly used as an independent test set, and the remaining four subsets were used as training sets. The partial least squares model constructed using the training set was used to predict the soil available nitrogen at the sample points of the test set, and the measured and predicted values ​​of soil available nitrogen on the test set were compared to complete the model evaluation. The measured and predicted values ​​of soil available nitrogen on the test set are compared to complete the model evaluation including the coefficient of determination, root mean square error and mean absolute error, and the expression is: Among them, X i is the measured value of sample i, Y i is the estimated value of sample i, is the mean of the sample measured values, and n is the total sample size.

8. A system using a soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand as claimed in any one of claims 1 to 7, characterized in that: It includes remote sensing data preprocessing module, data acquisition and exploratory analysis module, modeling and preliminary prediction module, and residual prediction and correction module; The remote sensing data preprocessing module is to obtain Sentinel-2 remote sensing multispectral data and preprocess it, including format conversion, band synthesis, study area clipping and resampling, to obtain the band reflectance of spatial resolution; The data collection and exploratory analysis module is to obtain soil available nitrogen sample point data in the study area and extract remote sensing variable information of the corresponding sampling points; The modeling and preliminary prediction module uses the partial least squares method to establish the relationship between soil available nitrogen and remote sensing information, determine the number of principal components, establish a soil available nitrogen inversion model, and carry out preliminary prediction mapping. Through the application of the model to the remote sensing image matrix, a preliminary result map of soil available nitrogen in the entire study area is predicted; The residual prediction and correction module extracts the soil available nitrogen prediction value of the training sample point from the preliminary prediction map, subtracts the preliminary prediction value from the measured value of the soil available nitrogen of the training sample point to obtain the residual and performs Kriging interpolation on the residual to generate a residual spatial distribution map, adds the preliminary predicted soil available nitrogen map and the residual map to obtain the final soil available nitrogen prediction map, verifies the prediction result using the five-fold cross validation method, and completes the model evaluation.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of a soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a soil available nitrogen remote sensing mapping method for coordinated nitrogen supply and demand as described in any one of claims 1 to 7 are implemented.

Citation Information

Cited By

  • Straw returning effect analysis method combining soil nutrients and crop yield

    CN120910583A