Soil comprehensive nutrient evaluation method and device based on satellite remote sensing image and medium

Abstract

The invention provides a soil comprehensive nutrient evaluation method and device based on a satellite remote sensing image, equipment and a medium. The method comprises the following steps: acquiring a Sentinel-2 multispectral image, cultivated land vector boundary data and actually measured soil nutrient data of a target area; establishing a multiple linear regression model by utilizing the wave band reflectivity of the Sentinel-2 multispectral image and actually measured soil nutrient data, generating a region-level nutrient content grid map, and extracting a nutrient content grid map of soil in a cultivated land range of a research region; classifying the grid map by adopting a spatial constraint natural breakpoint method clustering target function; a three-digit decimal code is adopted to represent a grading result, a soil comprehensive nutrient spatial distribution diagram is generated, coordinated color matching is set, various areas and proportions of soil comprehensive nutrients are counted, and fertilization suggestions are given, so that the defects of single sensing dimension and disjunction of decision support during remote sensing monitoring of the soil nutrients are overcome; and a closed loop serving precision agricultural practice is difficult to form.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural remote sensing applications, and in particular to a method, device, equipment and medium for evaluating soil comprehensive nutrients based on satellite remote sensing images. Background Art

[0002] At present, one of the core contradictions facing sustainable agricultural development is the prominent contradiction between the demand for precise soil nutrient management and the limitations of traditional monitoring methods. As the degree of agricultural intensification in my country continues to increase, the phenomenon of excessive input of nutrients such as nitrogen, phosphorus, and potassium is widespread. Taking the black soil region of Northeast China as an example, the application of nitrogen fertilizer has increased by nearly 40% in the past 20 years, but the nitrogen fertilizer utilization rate of corn is only 30% to 35%. The nutrients that are not absorbed continue to accumulate in the soil, which not only causes waste of resources, but also leads to ecological problems such as soil acidification and salinization. This situation exposes the deep-seated defects of the traditional nutrient management system:

[0003] On the one hand, while laboratory-based soil testing methods offer high accuracy, their inherent lag and high cost make them difficult to meet the real-time decision-making needs of modern agriculture. In Northeast China, due to the vast cultivated land area and the sampling period being affected by seasonal freezing, traditional methods have lower spatial coverage. Sparse sampling prevents accurate capture of the spatial variability of nutrients in black soil regions. For example, the alkaline nitrogen content of corn fields in the Songnen Plain can fluctuate by 35% to 50% within a 100-meter range. Furthermore, existing remote sensing monitoring technologies have yet to overcome the limitations of single-indicator, static assessments. Most studies can only provide rough estimates of a single nutrient and are significantly affected by vegetation cover. More importantly, these methods generally employ fixed threshold grading with equal intervals, ignoring the nonlinear spatial distribution of soil nutrients. For example, in the black soil belt of central Jilin Province, traditional equal interval grading incorrectly classified 30% of high organic matter areas (>4%) as medium or low, leading to significant bias in organic fertilizer application recommendations.

[0004] The essence of this technical dilemma lies in the lack of a comprehensive soil nutrient evaluation system that integrates multi-source data and dynamically adapts to spatiotemporal changes. While existing patented technologies attempt to incorporate machine learning algorithms, they still suffer from two fundamental flaws: First, feature engineering relies too heavily on spectral information and lacks the ability to mine texture features and spatial contextual relationships; second, grading strategies are disconnected from agronomic management requirements, failing to generate decision-making codes that can directly guide production. This technological gap makes current soil nutrient monitoring difficult to support the practical needs of precision agriculture.

[0005] Therefore, it is urgent to propose a soil comprehensive nutrient evaluation method based on satellite remote sensing images to solve the structural defects of soil nutrient remote sensing monitoring, such as the single perception dimension and the disconnection between decision support, which makes it difficult to form a closed loop to serve precision agriculture practice. Summary of the Invention

[0006] In order to overcome the problems existing in the related technologies, the present disclosure provides a soil comprehensive nutrient evaluation method, device, equipment and medium based on satellite remote sensing images to solve the structural defects of soil nutrient remote sensing monitoring in the related technologies, such as the single perception dimension and the disconnection between decision support, which makes it difficult to form a closed loop serving precision agriculture practice.

[0007] One or more embodiments of this specification provide a method for comprehensive soil nutrient evaluation based on satellite remote sensing images, comprising the following steps:

[0008] Download Sentinel-2 multispectral imagery of the target area and simultaneously obtain cultivated land vector boundary data and measured soil nutrient data, including nitrogen, phosphorus, and potassium.

[0009] A multiple linear regression model is established using the band reflectance of the Sentinel-2 multispectral image and the measured soil nutrient data. A regional nitrogen, phosphorus, and potassium nutrient content raster map is generated based on the multiple linear regression model. The nitrogen, phosphorus, and potassium nutrient content raster map of the soil in the cultivated land area of ​​the study area is extracted using the cultivated land vector data mask.

[0010] The spatially constrained natural breakpoint clustering objective function was used to classify the nitrogen, phosphorus, and potassium nutrient content grid map of the cultivated land in the study area after normalization.

[0011] The three-digit decimal code is used to represent the classification results of the nitrogen, phosphorus and potassium nutrient content grid map of the soil in the cultivated land area of ​​the study area, and the soil comprehensive nutrient spatial distribution map is calculated;

[0012] The spatial distribution map of soil comprehensive nutrients is analyzed, coordinated color matching is set, the nutrient conditions of nitrogen, phosphorus and potassium in each soil are analyzed, the areas and proportions of various types of soil comprehensive nutrients are counted, and fertilization recommendations are given.

[0013] Preferably, the method of using the spatially constrained natural breakpoint clustering objective function to classify the nitrogen, phosphorus, and potassium nutrient content grid map of the cultivated land in the study area after normalization processing also includes determining the inflection point with the largest inter-class difference through the spatial distribution of nutrient content, specifically including the following steps:

[0014] The semivariogram of soil nutrients in the study area was calculated based on geostatistical methods to determine the spatial autocorrelation range of nutrient content, which was used as the spatial constraint radius of the natural breakpoint method.

[0015] The improved Jenks natural breakpoint algorithm is used to solve the optimal classification threshold, and the objective function is:

[0016]

[0017] Among them, C k represents the kth category, μ k represents the mean of the kth class, w ij represents the spatial weight of pixels i and j, I represents the indicator function, and λ represents the balance factor;

[0018] Bootstrap resampling was performed 1000 times to calculate the 95% confidence interval of each classification threshold, and the threshold was considered stable when the CV was < 5%.

[0019] Preferably, the method of using three-digit decimal codes to respectively characterize the grading results of the nitrogen, phosphorus and potassium nutrient content grid map of the soil in the cultivated land area of ​​the study area specifically includes the following steps:

[0020] Each digit in the three-digit decimal code independently expresses an indicator, wherein the value range of each digit is 1-3, and the whole constitutes a three-dimensional discrete vector;

[0021] Use three digits to represent the grading levels of the three nutrients, and the value of each digit corresponds to three levels of grading:

[0022] A: Available phosphorus level, 1 = low phosphorus, 2 = medium phosphorus, 3 = high phosphorus;

[0023] B position: alkaline nitrogen level, 1 = low nitrogen, 2 = medium nitrogen, 3 = high nitrogen;

[0024] C position: fast-acting potassium level, 1 = low potassium, 2 = medium potassium, 3 = high potassium;

[0025] The number of bits of the encoding can be expanded to N bits to be compatible with future newly added indicators.

[0026] Preferably, providing fertilization suggestions specifically includes the following steps:

[0027] When the last digit of the code is 1, it triggers an increase in potassium fertilizer application; when the middle digit of the code is 3, it triggers a decrease in nitrogen fertilizer application; and when the first digit of the code is 2, it triggers the application of phosphorus fertilizer.

[0028] Preferably, the setting of coordinated color matching specifically includes the following steps:

[0029] Red indicates areas with severe nutrient deficiency, which require priority management; yellow indicates areas with moderate nutrient deficiency; green indicates areas with suitable nutrients; and blue indicates areas with excess nutrients.

[0030] The CIE LAB color space was used to ensure that the color difference between adjacent codes was ΔE>12. The lightness gradient increased according to the nutrient level, with low-level lightness at 40% to 50%, medium-level lightness at 60% to 70%, and high-level lightness at 80% to 90%. The saturation gradient was positively correlated with the nutrient management priority.

[0031] Display consistency is ensured through HSV to RGB color space conversion, the color scheme is verified by WCAG 2.1 accessibility standards, and supports direct call from GIS platforms.

[0032] One or more embodiments of this specification provide a soil comprehensive nutrient evaluation device based on satellite remote sensing images, including a data acquisition module, a nutrient inversion module, a classification module, a distribution calculation module, and an evaluation module;

[0033] The data acquisition module is used to download Sentinel-2 multispectral images of the target area and simultaneously obtain cultivated land vector boundary data and measured soil nutrient data, wherein the soil nutrients include nitrogen, phosphorus, and potassium;

[0034] The nutrient inversion module is used to establish a multiple linear regression model using the band reflectance of the Sentinel-2 multispectral image and the measured soil nutrient data, generate a regional nitrogen, phosphorus, and potassium nutrient content grid map based on the multiple linear regression model, and use the cultivated land vector data mask to extract the nitrogen, phosphorus, and potassium nutrient content grid map of the soil within the cultivated land area of ​​the study area;

[0035] The classification module is used to classify the nitrogen, phosphorus and potassium nutrient content grid map of the cultivated land in the study area after normalization using the spatial constraint natural breakpoint method clustering objective function;

[0036] The distribution calculation module is used to use three-digit decimal codes to respectively represent the classification results of the nitrogen, phosphorus and potassium nutrient content grid map of the soil in the cultivated land area of ​​the study area, and calculate and generate a soil comprehensive nutrient spatial distribution map;

[0037] The evaluation module is used to analyze the spatial distribution map of the soil comprehensive nutrients, set coordinated colors, analyze the nutrient conditions of nitrogen, phosphorus, and potassium in each soil, count the areas and proportions of various types of soil comprehensive nutrients, and provide fertilization recommendations.

[0038] Preferably, the classification module is further configured to determine the inflection point with the largest inter-class difference through the spatial distribution of nutrient content, and includes a spatial heterogeneity analysis unit, an optimal classification threshold solution unit, and a threshold verification unit;

[0039] The spatial heterogeneity analysis unit is used to calculate the semivariogram of soil nutrients in the study area based on geostatistical methods, determine the spatial autocorrelation range of nutrient content, and serve as the spatial constraint radius of the natural breakpoint method;

[0040] The optimal classification threshold solving unit is used to solve the optimal classification threshold using the improved Jenks natural breakpoint algorithm, and the objective function is:

[0041]

[0042] Among them, C k represents the kth category, μ k represents the mean of the kth class, w ij represents the spatial weight of pixels i and j, I represents the indicator function, and λ represents the balance factor;

[0043] The threshold verification unit is used to calculate the 95% confidence interval of each classification threshold through Bootstrap resampling 1000 times, and determine that the threshold is stable when CV<5%.

[0044] Preferably, the distributed computing module is configured as follows:

[0045] Use three digits to represent the grading levels of the three nutrients, and the value of each digit corresponds to three levels of grading:

[0046] A: Available phosphorus level, 1 = low phosphorus, 2 = medium phosphorus, 3 = high phosphorus;

[0047] B position: alkaline nitrogen level, 1 = low nitrogen, 2 = medium nitrogen, 3 = high nitrogen;

[0048] C position: fast-acting potassium level, 1 = low potassium, 2 = medium potassium, 3 = high potassium;

[0049] The number of bits of the encoding can be expanded to N bits to be compatible with future newly added indicators.

[0050] One or more embodiments of the present specification provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for comprehensive soil nutrient evaluation based on satellite remote sensing images as described above is implemented.

[0051] One or more embodiments of this specification provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-mentioned soil comprehensive nutrient evaluation method based on satellite remote sensing images.

[0052] The present disclosure provides a soil comprehensive nutrient evaluation method, device, equipment and medium based on satellite remote sensing images. The advantages are that by downloading Sentinel-2 multispectral images of the target area, cultivated land vector boundary data and measured soil nutrient data are synchronously obtained, wherein the soil nutrients include nitrogen, phosphorus and potassium, and the core basic data required for soil nutrient remote sensing inversion and evaluation are obtained, ensuring the spatiotemporal consistency and target targeting of the data source; a multivariate linear regression model is established using the band reflectance of the Sentinel-2 multispectral image and the measured soil nutrient data, and a regional-level nitrogen, phosphorus and potassium nutrient content grid map is generated based on the multivariate linear regression model, and the cultivated land vector data mask is used to extract the cultivated land in the study area. The nitrogen, phosphorus and potassium nutrient content grid map of the soil in the study area was obtained, and a quantitative mapping relationship between spectral reflectance and soil nutrients was established, realizing the preliminary conversion from remote sensing images to nutrient content, and expanding the discrete point measured data into a continuous spatially distributed nutrient content grid map. Through cultivated land mask extraction, the nitrogen, phosphorus and potassium nutrient spatial distribution map of the cultivated land in the study area was accurately obtained, eliminating the interference of non-cultivated land; the spatially constrained natural breakpoint method clustering objective function was used to classify the nitrogen, phosphorus and potassium nutrient content grid maps of the cultivated land in the study area after normalization. The normalization treatment eliminated the dimensional differences of nitrogen, phosphorus and potassium, making multiple nutrients comparable. The "spatially constrained natural breakpoint method" not only takes into account the numerical distribution, but also incorporates the spatial relationship between adjacent pixels. The system makes the grading results more continuous and smooth in space, reduces the "salt and pepper" phenomenon, and is more in line with the actual spatial variation pattern of farmland nutrients. The optimal segmentation point is found based on the statistical characteristics of the data itself, so that the grading results objectively reflect the inherent distribution law of the data; a three-digit decimal code is used to represent the grading results of the nitrogen, phosphorus and potassium nutrient content grid map of the soil in the cultivated land area of ​​the study area, and the soil comprehensive nutrient spatial distribution map is calculated and generated. The three independent nutrient levels of nitrogen, phosphorus and potassium are efficiently and intuitively integrated into a single comprehensive evaluation index. The coding method is directly related to the agronomic management needs. Different coding combinations directly correspond to different nutrient abundance and deficiency conditions, which lays the foundation for the subsequent generation of operational fertilization decision-making codes. The spatial distribution map clearly displays The spatial pattern of comprehensive nutrients was analyzed; the spatial distribution map of soil comprehensive nutrients was analyzed, coordinated color matching was set, the nutrient conditions of nitrogen, phosphorus and potassium in each soil were analyzed, the area and proportion of each type of soil comprehensive nutrients were counted, and fertilization recommendations were given. The coordinated color matching scheme made the spatial distribution map of comprehensive nutrients easier to interpret and identify areas with different nutrient combinations. The statistical analysis provided a global quantitative description of the nutrient status of cultivated land in the study area, and was able to generate differentiated and specific fertilization recommendations for different nitrogen-phosphorus-potassium combinations represented by different codes, thereby improving the accuracy and reliability of the soil comprehensive nutrient evaluation results. It provided new ideas for regional soil comprehensive nutrient evaluation and provided basic data and reference opinions for agricultural managers to formulate precision agriculture variable fertilization and sustainable management plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0054] Figure 1 A schematic flow chart of a method for comprehensive soil nutrient evaluation based on satellite remote sensing images provided in one or more embodiments of this specification;

[0055] Figure 2 A flow chart of a soil comprehensive nutrient evaluation method based on satellite remote sensing images provided in one or more embodiments of this specification;

[0056] Figure 3 The Jiansanjiang Sentinel-2 multispectral image provided in one or more embodiments of this specification;

[0057] Figure 4 A five-point sampling diagram provided for one or more embodiments of this specification;

[0058] Figure 5 The spatial distribution map of soil available potassium, soil alkaline nitrogen, and soil available phosphorus content based on Sentinel-2 image inversion provided in one or more embodiments of this specification;

[0059] Figure 6 A fertility level chart of soil available potassium, soil alkaline nitrogen, and soil available phosphorus based on the natural breakpoint method on April 23, 2025, in the experiments provided in one or more embodiments of this specification;

[0060] Figure 7 A spatial distribution map of soil comprehensive nutrients in Jiansanjiang area of ​​Heilongjiang Province based on three-digit decimal coding on April 23, 2025 in the experiments provided in one or more embodiments of this specification;

[0061] Figure 8 The area and proportion of various types of comprehensive soil nutrients in the cultivated land area of ​​Jiansanjiang region provided in one or more embodiments of this specification;

[0062] Figure 9 A schematic diagram of the structure of a soil comprehensive nutrient evaluation device based on satellite remote sensing images provided in one or more embodiments of this specification;

[0063] Figure 10 A schematic diagram of the structure of a computer device provided in one or more embodiments of this specification. DETAILED DESCRIPTION

[0064] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below in conjunction with the drawings in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this invention document.

[0065] The present invention will be described in detail below with reference to specific implementation methods and the accompanying drawings.

[0066] Method Example

[0067] According to an embodiment of the present invention, a soil comprehensive nutrient evaluation method based on satellite remote sensing images is provided. Figure 1 FIG. 1 is a flow chart of a method for evaluating soil comprehensive nutrients based on satellite remote sensing images according to an embodiment of the present invention. The method for evaluating soil comprehensive nutrients based on satellite remote sensing images according to an embodiment of the present invention includes the following steps:

[0068] S110. Field investigations revealed that the bare soil window period in Jiansanjiang, Heilongjiang Province, is from April to May, and the image spectral characteristics reflect those of the soil, excluding the influence of vegetation. The satellite remote sensing imagery described in this embodiment is a satellite remote sensing imagery collected during the bare soil window period in Jiansanjiang. Sentinel-2 multispectral imagery of the target area is downloaded, and cultivated land vector boundary data and measured soil nutrient data are simultaneously acquired. This data is obtained through field collection, wherein the soil nutrients include nitrogen, phosphorus, and potassium.

[0069] S120. Establish a multiple linear regression model using the band reflectance of the Sentinel-2 multispectral image and the measured soil nutrient data. Generate a regional-level nitrogen, phosphorus, and potassium nutrient content raster map based on the multiple linear regression model. Use the cultivated land vector data mask to extract the nitrogen, phosphorus, and potassium nutrient content raster map of the soil in the cultivated land area of ​​the study area. Implement collaborative inversion and spatial visualization of nitrogen, phosphorus, and potassium nutrients through Sentinel-2 multispectral remote sensing.

[0070] S130. Using Python, a spatially constrained natural breakpoint method clustering objective function is used to classify the nitrogen, phosphorus, and potassium nutrient content grid map of the cultivated land area of ​​the study area after normalization.

[0071] S140, the grading results of the nitrogen, phosphorus and potassium nutrient content grid maps of the soil of the cultivated land range in the study area are represented respectively by using ternary decimal coding, soil comprehensive nutrient data are obtained, and a soil comprehensive nutrient spatial distribution map is generated according to the soil comprehensive nutrient data.

[0072] S150, the soil comprehensive nutrient spatial distribution map is analyzed, coordinated color setting is uniformly set, the nutrient conditions of each soil nitrogen, phosphorus and potassium are analyzed, the areas and proportions of various soil comprehensive nutrients are simultaneously counted, and fertilization suggestions are given.

[0073] The method provided by the embodiment, by downloading the Sentinel-2 multispectral image of the target area, synchronously acquiring the cultivated land vector boundary data and the measured soil nutrient data, wherein the soil nutrients include nitrogen, phosphorus and potassium, the core basic data required for soil nutrient remote sensing inversion and evaluation are acquired, and the spatio-temporal consistency and target pertinence of the data source are ensured; a multiple linear regression model is established by using the band reflectivity of the Sentinel-2 multispectral image and the measured soil nutrient data, a regional nitrogen, phosphorus and potassium nutrient content grid map is generated based on the multiple linear regression model, the nitrogen, phosphorus and potassium nutrient content grid map of the soil in the cultivated land range of the study area is extracted by using the cultivated land vector data mask, the quantitative mapping relationship between the spectral reflectivity and the soil nutrient is established, the preliminary conversion from the remote sensing image to the nutrient content is realized, the discrete point-shaped measured data is expanded into the continuous spatial distribution nutrient content grid map, the nitrogen, phosphorus and potassium nutrient spatial distribution map of the cultivated land range in the study area is accurately obtained through the cultivated land mask extraction, and the interference of non-cultivated land is excluded; the normalized soil nitrogen, phosphorus and potassium nutrient content grid map of the cultivated land range in the study area is classified by using a spatial constraint natural breakpoint method clustering target function, the normalization processing eliminates the dimension difference of nitrogen, phosphorus and potassium, so that the multiple nutrients are comparable, the spatial constraint natural breakpoint method considers the numerical distribution and integrates the spatial relationship of adjacent pixels, so that the classification result is more continuous and smooth in space, the "pepper and salt" phenomenon is reduced, and the actual spatial variation mode of the farmland nutrient is more in line with the actual situation, the optimal segmentation point is found based on the statistical characteristics of the data itself, so that the classification result objectively reflects the internal distribution law of the data; the classification results of the nitrogen, phosphorus and potassium nutrient content grid map of the soil in the cultivated land range of the study area are represented by using three decimal encoding, and a soil comprehensive nutrient spatial distribution map is calculated and generated, three independent nutrient grades of nitrogen, phosphorus and potassium are efficiently and intuitively fused into a single comprehensive evaluation index, the encoding mode is directly related to the agronomic management demand, different encoding combinations are directly corresponding to different nutrient abundance and deficiency combinations, which lays a foundation for generating an operable fertilization decision code in the subsequent process, and the spatial distribution map clearly shows the spatial pattern of the comprehensive nutrient; the soil comprehensive nutrient spatial distribution map is analyzed, a coordinated color scheme is set, the nutrient conditions of the soil nitrogen, phosphorus and potassium are analyzed, the area and proportion of each soil comprehensive nutrient are counted, and fertilization suggestions are given, the coordinated color scheme makes the comprehensive nutrient spatial distribution map easier to interpret and identify different nutrient combination regions, the statistical analysis provides a global quantitative description of the cultivated land nutrient conditions in the study area, different fertilization suggestions can be generated for different nitrogen-phosphorus-potassium combination conditions represented by different codes, the accuracy and reliability of the soil comprehensive nutrient evaluation result are improved, the regional soil comprehensive nutrient evaluation provides a new idea, and provides basic data and reference opinions for the precision agriculture variable fertilization and the formulation of a sustainable management scheme for the agricultural manager.

[0074] In one embodiment, S130, using a spatially constrained natural breakpoint clustering objective function to classify the normalized nitrogen, phosphorus, and potassium nutrient content grid map of the cultivated land in the study area, further includes determining the inflection point with the largest inter-class difference based on the spatial distribution of nutrient content, specifically comprising the following steps:

[0075] The semivariogram of soil nutrients in the study area was calculated based on geostatistical methods to determine the spatial autocorrelation range of nutrient content, which was used as the spatial constraint radius of the natural breakpoint method.

[0076] The improved Jenks natural breakpoint algorithm is used to solve the optimal classification threshold, and the objective function is:

[0077]

[0078] Among them, C k represents the kth category, μ k represents the mean of the kth class, w ij represents the spatial weight of pixels i and j, I represents the indicator function (1 when the categories are different), and λ represents the balance factor.

[0079] Bootstrap resampling was performed 1000 times to calculate the 95% confidence interval of each classification threshold. The threshold was considered stable when the CV was <5%. This optimization method took into account the statistical distribution characteristics and spatial autocorrelation, and the threshold reliability was evaluated through the confidence interval.

[0080] The method provided in this example uses the geostatistically quantified spatial autocorrelation range of nutrients as a constraint radius. A spatial weight matrix and a balance factor are integrated into the improved Jenks algorithm to jointly optimize the objective function, simultaneously minimizing the within-class statistical variance and maximizing the between-class spatial continuity. Bootstrap resampling verifies the threshold stability, fundamentally addressing the problem of traditional methods neglecting spatial context. The output combines statistical rationality, spatial continuity, and engineering robustness, generating hierarchical units that conform to the distribution patterns of farmland nutrients. This provides a quantifiable and actionable spatial decision-making basis for variable-rate fertilization, achieving an upgrade from numerical statistics to spatial decision-making.

[0081] In one embodiment, S140, using three-digit decimal codes to represent the classification results of the nitrogen (alkaline-hydrolyzed nitrogen), phosphorus (available phosphorus), and potassium (fast-acting potassium) nutrient content grid map of the soil in the cultivated land area of ​​the study area, specifically includes the following steps:

[0082] Each digit in the three-digit decimal code (ABC) independently expresses an indicator, wherein the value range of each digit is 1-3, and the whole constitutes a three-dimensional discrete vector. Its advantage is that it intuitively reflects the combination of nutrient states and is convenient for computer bit-by-bit analysis and calculation.

[0083] Each of the three nutrients is represented by a three-digit code, and the value of each digit corresponds to the three levels of classification:

[0084] A bit: effective phosphorus level, 1 = low phosphorus, 2 = medium phosphorus, 3 = high phosphorus;

[0085] B bit: alkali nitrogen level, 1 = low nitrogen, 2 = medium nitrogen, 3 = high nitrogen;

[0086] C bit: available potassium level, 1 = low potassium, 2 = medium potassium, 3 = high potassium;

[0087] For example: the code "213" represents the level of available phosphorus content + the level of alkali nitrogen content + the level of available potassium content. Cluster analysis is performed on adjacent grids with the same code to generate a nutrient management zoning map.

[0088] The given fertilization recommendation specifically includes the following steps:

[0089] When the last digit of the code is 1, potassium fertilizer is increased (application amount = k x 1.5), when the middle digit of the code is 3, nitrogen fertilizer is reduced (application amount = k x 0.7), and when the first digit of the code is 2, phosphorus fertilizer is applied (application amount = k), where k is the base amount.

[0090] Table 1 Soil comprehensive nutrient code, color matching and fertilization recommendation

[0091]

[0092]

[0093]

[0094] The design of the coding system has certain scalability, and the number of bits in the code can be expanded to N bits (N ≥ 4) to accommodate future additions of indicators (such as organic matter, etc.). For example, the fourth bit in the four-digit code "3124" represents the organic matter level.

[0095] Compared with traditional coding methods, the three-digit decimal coding has the following advantages:

[0096] Table 2 Comparison of coding methods

[0097]

[0098] The method provided in the embodiment fuses the soil nitrogen, phosphorus and potassium nutrient grading results into a single comprehensive decision unit through three-digit decimal coding (A-B-C), each digit independently represents the agronomic grading (1-3 levels) of a specific nutrient, forms an intuitive discrete decision vector, and directly maps the fertilization formula, which can be extended to provide a modular digital expression base for farmland precision management, and fundamentally bridges the gap between grading results and agronomic practices, and realizes the precision decision-making closed loop.

[0099] In one embodiment, the coordinated color matching is used by using a coordinated unified setting, a standardized color matching scheme is used to visually express the soil comprehensive nutrient evaluation results of the three-digit decimal coding, and fertilization suggestion analysis is provided by analyzing the soil nitrogen, phosphorus and potassium nutrient conditions. The setting coordinated color matching specifically includes the following steps:

[0100] (1) Color semantic principle: the red color system represents the nutrient severely deficient area which needs to be managed first, the yellow color system represents the nutrient moderately deficient area, the green color system represents the nutrient suitable area, and the blue color system represents the nutrient excessive area.

[0101] (2) Visual perception principle: the CIE LAB color space is used to ensure that the color difference ΔE of adjacent codes is greater than 12, the lightness gradient increases according to the nutrient level, the lightness of the low level (1 level) is 40% to 50%, the lightness of the medium level (2 level) is 60% to 70%, the lightness of the high level (3 level) is 80% to 90%, and the saturation gradient is positively correlated with the nutrient management priority.

[0102] (3) Technical implementation principle: the color space conversion from HSV to RGB is used to ensure display consistency, the color matching scheme is verified by the WCAG 2.1 accessibility standard, and the GIS platform (ArcGIS, QGIS) can directly call.

[0103] The method provided in the embodiment realizes management priority visualization through the red-yellow-green-blue four-color system, ensures the recognition degree by forcibly setting ΔE>12 color difference based on the CIELAB color space, sets the step lightness according to the nutrient level and associates the saturation with the agricultural urgency, ensures the cross-platform universality through the lossless conversion from HSV to RGB and the WCAG 2.1 accessibility authentication, and finally forms a standardized visual coding system which can be directly called by the GIS system, so that the complex nutrient spatial distribution is converted into a "color vision agricultural guide" which can be responded by the field workers in seconds.

[0104] The following experiment is used to verify the effect of the application:

[0105] Experiment: Taking the Jianjiang region in Heilongjiang Province as an example, the soil comprehensive nutrients of the farmland in Jianjiang were evaluated and analyzed, such as Figure 2FIG. 1 is a flow chart of a soil comprehensive nutrient evaluation method based on satellite remote sensing images provided in this embodiment. The steps of a multi-index hierarchical fusion soil comprehensive nutrient evaluation method based on satellite remote sensing images are as follows:

[0106] S1. Data acquisition: Download Sentinel-2 multispectral images of the target area, such as Figure 2 As shown, the Jiansanjiang Sentinel-2 multispectral image provided in this embodiment simultaneously obtains cultivated land vector boundary data and measured soil nitrogen, phosphorus, and potassium nutrient data;

[0107] S2. Nutrient inversion: A multivariate linear regression model was used to map the band reflectance of the Sentinel-2 imagery with the measured soil nutrient data to generate regional nitrogen, phosphorus, and potassium nutrient content raster maps. These maps were then clipped using cultivated land vector data to obtain soil nitrogen, phosphorus, and potassium nutrient content raster maps for the cultivated land area of ​​the study area.

[0108] S3. Normalize the nitrogen, phosphorus, and potassium nutrient contents of the soil in the cultivated land area of ​​the study area and classify them using the spatially constrained natural breakpoint clustering objective function using Python;

[0109] S4. Using three-digit decimal codes to represent the classification results of alkaline-hydrolyzable nitrogen, available potassium, and available phosphorus, respectively, perform calculations to generate a spatial distribution map of soil comprehensive nutrients;

[0110] S5. Finally, the soil comprehensive nutrient data is analyzed, the coordinated color scheme is unified, and the nitrogen, phosphorus and potassium nutrient conditions of each soil are analyzed. At the same time, the area and proportion of each type of soil comprehensive nutrients are counted, and fertilization recommendation analysis is performed.

[0111] In step S1, the Sentinel-2 multispectral image of the target area needs to be downloaded during a specific period - the bare soil window period. Due to the high latitude of Jiansanjiang area in Heilongjiang Province, it is still winter in March with snow cover, while May has entered the sowing and transplanting period. At this time, soil inversion will be affected by soil and vegetation factors. Therefore, the best bare soil window period in Jiansanjiang area of ​​Heilongjiang Province is April every year. This experiment downloaded Sentinel-2 multispectral images from the official website of ESA (https: / / dataspace.copernicus.eu / ) on April 23, 2025. There is no cloud cover in the cultivated land area, such as Figure 3 As shown, this is the Jiansanjiang Sentinel-2 multispectral image provided by this embodiment.

[0112] During the same period, uniform soil nitrogen, phosphorus and potassium nutrient data collection was carried out in Jiansanjiang area, with a total of 300 samples collected. When collecting each soil sample, a five-point sampling (plum blossom sampling) method was adopted, that is, when collecting a soil sample point, a 10*10m sampling area was made with the location of the soil sample point as the center, and then 4 soil samples were collected from the 4 corners of the sampling area and mixed with the center sample to serve as the soil sample of the sample point. Figure 4 As shown in FIG, a schematic diagram of five-point sampling provided in this embodiment is provided.

[0113] In step S2, the soil nitrogen, phosphorus, and potassium nutrient inversion is based on a multivariate linear regression model to establish a mapping relationship between band reflectance and measured soil nutrient data, thereby generating a regional nitrogen, phosphorus, and potassium nutrient content grid map. The multivariate linear regression model considers the impact of multiple input variables on the output variable and has the advantages of strong interpretability and high computational efficiency. Its basic formula is as follows:

[0114] y=k1x1+k2x2+…+k d x d +b;

[0115] In this experiment, the model was constructed using a 2:1 ratio of sample selection for the modeling set to the validation set. Using satellite multispectral reflectance data as the independent variable and soil nutrient content as the dependent variable, a multivariate stepwise regression model was used to construct an inversion model for soil alkaline nitrogen, soil available phosphorus, and soil available potassium. The model formula is as follows:

[0116] AN=0.053*B1-0.205*B2+0.095*B3+265.976;

[0117] AP=0.026*B1-0.023*B3+0.007*B4-0.010*B5+0.012*B6+11.802;

[0118] AK=-0.103*B1+0.044*B2+0.085*B3-0.034*B4+-0.05*B5+0.055*B6+261.416;

[0119] Among them, B1 is the reflectivity of the blue band, B2 is the reflectivity of the green band, B3 is the reflectivity of the red band, B4 is the reflectivity of the near-red band, B5 is the reflectivity of the shortwave infrared I band, and B6 is the reflectivity of the shortwave infrared II band.

[0120] Using the adjusted coefficient of determination (R 2 adj ), root mean square error (RMSE) is the accuracy evaluation index, and the experimental results are as follows:

[0121] Table 3 Accuracy of different soil nutrient inversion models

[0122]

[0123] Note: R 2 R 2 adj R Figure 5 The spatial distribution maps of soil available potassium, soil alkali hydrolysis nitrogen and soil available phosphorus content based on Sentinel-2 image inversion provided in this embodiment are shown in the following table.

[0124] In step S3, the spatial constraint natural break method is used to cluster the target function of the soil available potassium, soil alkali hydrolysis nitrogen and soil available phosphorus content raster map inverted by S2, and the classification results are shown in the following table. Figure 6 The soil available potassium, soil alkali hydrolysis nitrogen and soil available phosphorus fertility grade map based on the natural break method on April 23, 2025 in the experiment provided in this embodiment is shown in the following table. This step is based on the traditional Jenks natural break method, and a spatial smoothing constraint term is introduced to make the classification results meet the minimum difference within the group and the adjacent pixels belong to the same grade as possible. The optimization objective function is as follows:

[0125]

[0126] Wherein, C k : the kth class, μ k : the mean of the kth class, w ij : the spatial weight of pixels i and j, I represents the indicator function (1 when the class is different), and λ represents the balance factor.

[0127] In this experiment, the spatial radius R is set to 500 meters (about one grid width), the Delaunay triangular net is used to establish the topological relationship of the pixels to ensure that the constraint acts on the real adjacent fields, and the cross-validation is used to determine λ = 0.3. The threshold stability is analyzed by Bootstrap verification, and after resampling 1000 times, taking the alkali hydrolysis nitrogen classification as an example, the threshold distribution is as follows:

[0128] Table 4 Alkali hydrolysis nitrogen classification threshold

[0129]

[0130] The coefficient of variation CV of the classification threshold is less than 5%, and the classification result is stable.

[0131] In this experiment, the application effect boundary fitting degree is significantly improved, and after the spatial constraint classification, the matching degree of the fertilization prescription map and the agricultural operation width is improved from 70% to 92%.

[0132] Table 5 Evaluation of application effect of spatial constraint Jenks algorithm

[0133]

[0134] In step S4, the three-digit decimal coding is used to represent the classification results of alkali-hydrolyzable nitrogen, available potassium and available phosphorus respectively, and the soil comprehensive nutrient spatial distribution map is generated by operation, as shown in the following figure. Figure 7

[0135] A: available phosphorus level (1 = low phosphorus, 2 = medium phosphorus, 3 = high phosphorus);

[0136] B: alkali-hydrolyzable nitrogen level (1 = low nitrogen, 2 = medium nitrogen, 3 = high nitrogen);

[0137] C: available potassium level (1 = low potassium, 2 = medium potassium, 3 = high potassium);

[0138] For example, the code "213" represents the level of available phosphorus content, the level of alkali-hydrolyzable nitrogen content, and the level of available potassium content. Cluster analysis is performed on adjacent grids with the same code to generate a nutrient management zoning map. As shown in the following figure, the soil comprehensive nutrient spatial distribution map of Jian Sanjiang area in Heilongjiang Province on April 23, 2025 based on three-digit decimal coding is provided in the experiment of the present embodiment. Figure 8

[0139] In step S5, the soil comprehensive nutrient data is analyzed. In the experiment, the soil comprehensive nutrients are uniformly set and coordinated according to the color semantics principle, visual perception principle and technical realization principle, and the soil nitrogen, phosphorus and potassium nutrient conditions of each plot are analyzed, and the area, proportion and fertilization recommendation analysis of soil comprehensive nutrients are also analyzed.

[0140] Fertilization recommendation: when the last digit of the code is 1, potassium fertilizer is increased (application amount = k x 1.5), when the middle digit of the code is 3, nitrogen fertilizer is reduced (application amount = k x 0.7), and when the first digit of the code is 2, phosphorus fertilizer is applied (application amount = k), where k is the basic amount.

[0141] Table 6 Area and proportion of soil comprehensive nutrients

[0142]

[0143]

[0144] ​​From the data analysis, it is known that in Jiansanjiang area as a whole, the soil nutrients with low phosphorus, high nitrogen and medium potassium have the largest area, followed by medium phosphorus, medium nitrogen and medium potassium and low phosphorus, medium nitrogen and medium potassium. Therefore, when applying fertilizer, the main approach is to increase the application of phosphorus fertilizer, reduce the application of nitrogen fertilizer and combine with conventional potassium fertilizer.

[0145] This experiment took Jiansanjiang, Heilongjiang Province, as an example to evaluate and analyze the comprehensive nutrients of the soil in Jiansanjiang's cultivated land. Through field investigation and variable rate fertilization, the accuracy was verified. The experiment showed that providing fertilization recommendations based on comprehensive soil nutrient analysis can reduce fertilizer application by 5% to 10%, saving 6-10 yuan per mu in fertilizer costs. Figures 5 to 8 shown.

[0146] Device embodiment

[0147] According to an embodiment of the present invention, a soil comprehensive nutrient evaluation device based on satellite remote sensing images is provided. Figure 10 As shown, it is a structural diagram of the soil comprehensive nutrient evaluation device based on satellite remote sensing images provided in this embodiment. According to the embodiment of the present invention, the soil comprehensive nutrient evaluation device based on satellite remote sensing images includes a data acquisition module 101, a nutrient inversion module 102, a classification module 103, a distribution calculation module 104 and an evaluation module 105.

[0148] The data acquisition module 101 is used to download Sentinel-2 multispectral images of the target area and simultaneously obtain cultivated land vector boundary data and measured soil nutrient data, wherein the soil nutrients include nitrogen, phosphorus, and potassium.

[0149] The nutrient inversion module 102 is used to establish a multiple linear regression model using the band reflectance of the Sentinel-2 multispectral image and the measured soil nutrient data, generate a regional nitrogen, phosphorus, and potassium nutrient content grid map based on the multiple linear regression model, and use the cultivated land vector data mask to extract the nitrogen, phosphorus, and potassium nutrient content grid map of the soil in the cultivated land area of ​​the study area.

[0150] The classification module 103 is used to classify the nitrogen, phosphorus and potassium nutrient content grid map of the cultivated land in the study area after normalization using the spatial constraint natural breakpoint method clustering objective function.

[0151] The distribution calculation module 104 is used to use three-digit decimal codes to represent the classification results of the nitrogen, phosphorus and potassium nutrient content grid map of the cultivated land in the study area, and calculate and generate a soil comprehensive nutrient spatial distribution map.

[0152] The evaluation module 105 is used to analyze the spatial distribution map of the soil comprehensive nutrients, set coordinated colors, analyze the nutrient conditions of nitrogen, phosphorus, and potassium in each soil, calculate the area and proportion of each type of soil comprehensive nutrients, and provide fertilization recommendations.

[0153] The device provided in this embodiment downloads the Sentinel-2 multispectral image of the target area through the data acquisition module 101, and simultaneously obtains the cultivated land vector boundary data and the measured soil nutrient data, wherein the soil nutrients include nitrogen, phosphorus, and potassium, and obtains the core basic data required for remote sensing inversion and evaluation of soil nutrients, thereby ensuring the spatiotemporal consistency and target targeting of the data source; the nutrient inversion module 102 uses the band reflectance of the Sentinel-2 multispectral image and the measured soil nutrient data to establish a multiple linear regression model, generates a regional nitrogen, phosphorus, and potassium nutrient content grid map based on the multiple linear regression model, and uses the cultivated land vector data mask to extract the nitrogen, phosphorus, and potassium nutrient content grid map of the soil in the cultivated land area of ​​the study area. The quantitative mapping relationship between spectral reflectance and soil nutrients was established, and the preliminary conversion from remote sensing images to nutrient content was achieved. The discrete point measured data was expanded into a continuous spatial distribution nutrient content grid map. Through the cultivated land mask extraction, the nitrogen, phosphorus and potassium nutrient spatial distribution map of the cultivated land in the study area was accurately obtained, eliminating the interference of non-cultivated land; the classification module 103 used the spatial constraint natural breakpoint method clustering objective function to classify the nitrogen, phosphorus and potassium nutrient content grid map of the cultivated land in the study area after normalization. The normalization eliminated the dimensional differences of nitrogen, phosphorus and potassium, making multiple nutrients comparable. The "spatial constraint natural breakpoint method" not only took into account the numerical distribution, but also incorporated the spatial relationship of adjacent pixels, making the classification results spatially consistent. It is more continuous and smooth, reduces the "salt and pepper" phenomenon, and is more in line with the actual spatial variation pattern of farmland nutrients. It finds the optimal segmentation point based on the statistical characteristics of the data itself, so that the grading results objectively reflect the inherent distribution law of the data; the distribution calculation module 104 uses three-digit decimal coding to respectively characterize the grading results of the nitrogen, phosphorus and potassium nutrient content grid map of the soil in the cultivated land area of ​​the study area, calculates and generates the soil comprehensive nutrient spatial distribution map, and efficiently and intuitively integrates the three independent nutrient levels of nitrogen, phosphorus and potassium into a single comprehensive evaluation index. The coding method is directly related to the agronomic management needs. Different coding combinations directly correspond to different nutrient abundance and deficiency conditions, which lays the foundation for the subsequent generation of operational fertilization decision coding. The spatial distribution map clearly shows the comprehensive nutrients. Spatial pattern; the evaluation module 105 analyzes the spatial distribution map of the soil comprehensive nutrients, sets coordinated colors, analyzes the nutrient conditions of nitrogen, phosphorus, and potassium in each soil, counts the areas and proportions of various types of soil comprehensive nutrients, and gives fertilization recommendations. The coordinated color scheme makes the spatial distribution map of the comprehensive nutrients easier to interpret and identify areas with different nutrient combinations. The statistical analysis provides a global quantitative description of the nutrient status of cultivated land in the study area, and can generate differentiated and specific fertilization recommendations for different nitrogen-phosphorus-potassium combinations represented by different codes, thereby improving the accuracy and reliability of the soil comprehensive nutrient evaluation results, providing new ideas for regional soil comprehensive nutrient evaluation, and providing basic data and reference opinions for agricultural managers to formulate precision agricultural variable fertilization and sustainable management plans.

[0154] In one embodiment, the classification module 103 is further configured to determine the inflection point with the largest inter-class difference through the spatial distribution of nutrient content, and includes a spatial heterogeneity analysis unit, an optimal classification threshold solution unit, and a threshold verification unit.

[0155] The spatial heterogeneity analysis unit is used to calculate the semivariogram of soil nutrients in the study area based on geostatistical methods, determine the spatial autocorrelation range of nutrient content, and serve as the spatial constraint radius of the natural breakpoint method.

[0156] The optimal classification threshold solving unit is used to solve the optimal classification threshold using the improved Jenks natural breakpoint algorithm, and the objective function is:

[0157]

[0158] Among them, C k represents the kth category, μ k represents the mean of the kth class, w ij represents the spatial weight of pixels i and j, I represents the indicator function (1 when the categories are different), and λ represents the balance factor.

[0159] The threshold verification unit is used to calculate the 95% confidence interval of each classification threshold through Bootstrap resampling 1000 times, and determine that the threshold is stable when CV<5%.

[0160] The device provided in this embodiment uses the geostatistically quantified spatial autocorrelation range of nutrients as a constraint radius. It then integrates a spatial weight matrix and a balance factor into an improved Jenks algorithm to jointly optimize the objective function, simultaneously minimizing the intra-class statistical variance and maximizing the inter-class spatial continuity. Bootstrap resampling verifies threshold stability, fundamentally addressing the problem of traditional methods neglecting spatial context. Its output combines statistical rationality, spatial continuity, and engineering robustness, generating hierarchical units that conform to the laws of farmland nutrient distribution. This provides a quantifiable and actionable spatial decision-making basis for variable-rate fertilization, achieving an upgrade from numerical statistics to spatial decision-making.

[0161] In one embodiment, the distributed computing module 104 is configured to:

[0162] Use three digits to represent the grading levels of the three nutrients, and the value of each digit corresponds to three levels of grading:

[0163] A: Available phosphorus level, 1 = low phosphorus, 2 = medium phosphorus, 3 = high phosphorus;

[0164] B position: alkaline nitrogen level, 1 = low nitrogen, 2 = medium nitrogen, 3 = high nitrogen;

[0165] C position: fast-acting potassium level, 1 = low potassium, 2 = medium potassium, 3 = high potassium;

[0166] The number of bits of the encoding can be expanded to N bits to be compatible with future newly added indicators.

[0167] The device provided in this embodiment integrates the soil nitrogen, phosphorus, and potassium nutrient grading results into a single comprehensive decision-making unit through a three-digit decimal code (ABC). Each digit independently represents the agronomic grade (grades 1-3) of a specific nutrient, forming an intuitive discrete decision vector that directly maps the fertilization formula. Its scalable architecture provides a modular digital expression foundation for precise farmland management, fundamentally bridging the gap between grading results and agronomic practice, and realizing a closed-loop precise decision-making process.

[0168] The embodiment of the present invention is an apparatus embodiment corresponding to the above-mentioned method embodiment. The specific operations of the processing steps of each module can be understood by referring to the description of the method embodiment, and will not be repeated here.

[0169] like Figure 10 As shown, the present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the soil comprehensive nutrient evaluation method based on satellite remote sensing images in the above-mentioned embodiment is implemented, or when the computer program is executed by a processor, the soil comprehensive nutrient evaluation method based on satellite remote sensing images in the above-mentioned embodiment is implemented.

[0170] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0171] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the device or system embodiments, since they are basically similar to the method embodiments, the description is relatively simple. For the relevant parts, refer to the partial description of the method embodiments. The device and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. A person of ordinary skill in the art can understand and implement it without making any creative efforts.

[0172] Finally, 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 above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and the contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A soil comprehensive nutrient evaluation method based on satellite remote sensing images, characterized in that: The following steps are involved: Download Sentinel-2 multispectral imagery of the target area and simultaneously obtain cultivated land vector boundary data and measured soil nutrient data, including nitrogen, phosphorus, and potassium. A multiple linear regression model is established using the band reflectance of the Sentinel-2 multispectral image and the measured soil nutrient data. A regional nitrogen, phosphorus, and potassium nutrient content raster map is generated based on the multiple linear regression model. The nitrogen, phosphorus, and potassium nutrient content raster map of the soil in the cultivated land area of ​​the study area is extracted using the cultivated land vector data mask. The spatially constrained natural breakpoint clustering objective function was used to classify the nitrogen, phosphorus, and potassium nutrient content grid map of the cultivated land in the study area after normalization. The three-digit decimal code is used to represent the classification results of the nitrogen, phosphorus and potassium nutrient content grid map of the soil in the cultivated land area of ​​the study area, and the soil comprehensive nutrient spatial distribution map is calculated; The spatial distribution map of soil comprehensive nutrients is analyzed, coordinated color matching is set, the nutrient conditions of nitrogen, phosphorus and potassium in each soil are analyzed, the areas and proportions of various types of soil comprehensive nutrients are counted, and fertilization recommendations are given.

2. The method for comprehensive soil nutrient evaluation based on satellite remote sensing images according to claim 1, wherein: The method of using the spatially constrained natural breakpoint clustering objective function to classify the nitrogen, phosphorus, and potassium nutrient content grid map of the cultivated land in the study area after normalization processing specifically includes the following steps: The semivariogram of soil nutrients in the study area was calculated based on geostatistical methods to determine the spatial autocorrelation range of nutrient content, which was used as the spatial constraint radius of the natural breakpoint method. The improved Jenks natural breakpoint algorithm is used to solve the optimal classification threshold, and the objective function is: Among them, C k represents the kth category, μ k represents the mean of the kth class, w ij represents the spatial weight of pixels i and j, I represents the indicator function, and λ represents the balance factor; Bootstrap resampling was performed 1000 times to calculate the 95% confidence interval of each classification threshold. When the CV was less than 5%, the threshold was considered stable and the threshold was taken as the inflection point with the maximum difference between the classes. The classification level is determined by the inflection point where the difference between the classes is the largest.

3. The method for comprehensive soil nutrient evaluation based on satellite remote sensing images according to claim 1, wherein: The method of using three-digit decimal codes to respectively represent the classification results of the nitrogen, phosphorus and potassium nutrient content grid map of the cultivated land in the study area specifically includes the following steps: Each digit in the three-digit decimal code independently expresses an indicator, wherein the value range of each digit is 1-3, and the whole constitutes a three-dimensional discrete vector; Use three digits to represent the grading levels of the three nutrients, and the value of each digit corresponds to three levels of grading: A: Available phosphorus level, 1 = low phosphorus, 2 = medium phosphorus, 3 = high phosphorus; B position: alkaline nitrogen level, 1 = low nitrogen, 2 = medium nitrogen, 3 = high nitrogen; C position: fast-acting potassium level, 1 = low potassium, 2 = medium potassium, 3 = high potassium; The number of bits of the encoding can be expanded to N bits to be compatible with future newly added indicators.

4. The method for comprehensive soil nutrient evaluation based on satellite remote sensing images according to claim 3, wherein: The provision of fertilization suggestions specifically includes the following steps: When the last digit of the code is 1, it triggers an increase in potassium fertilizer application; when the middle digit of the code is 3, it triggers a decrease in nitrogen fertilizer application; and when the first digit of the code is 2, it triggers the application of phosphorus fertilizer.

5. The method for comprehensive soil nutrient evaluation based on satellite remote sensing images according to claim 1, wherein: The method of setting the coordinated color matching specifically includes the following steps: Red indicates areas with severe nutrient deficiency, which require priority management; yellow indicates areas with moderate nutrient deficiency; green indicates areas with suitable nutrients; and blue indicates areas with excess nutrients. The CIE LAB color space was used to ensure that the color difference between adjacent codes was ΔE>12. The lightness gradient increased according to the nutrient level, with low-level lightness at 40% to 50%, medium-level lightness at 60% to 70%, and high-level lightness at 80% to 90%. The saturation gradient was positively correlated with the nutrient management priority. Display consistency is ensured through HSV to RGB color space conversion, the color scheme is verified by WCAG 2.1 accessibility standards, and supports direct call from GIS platforms.

6. A soil comprehensive nutrient evaluation device based on satellite remote sensing images, characterized in that: It includes data acquisition module, nutrient inversion module, classification module, distribution calculation module and evaluation module; The data acquisition module is used to download Sentinel-2 multispectral images of the target area and simultaneously obtain cultivated land vector boundary data and measured soil nutrient data, wherein the soil nutrients include nitrogen, phosphorus, and potassium; The nutrient inversion module is used to establish a multiple linear regression model using the band reflectance of the Sentinel-2 multispectral image and the measured soil nutrient data, generate a regional nitrogen, phosphorus, and potassium nutrient content grid map based on the multiple linear regression model, and use the cultivated land vector data mask to extract the nitrogen, phosphorus, and potassium nutrient content grid map of the soil within the cultivated land area of ​​the study area; The classification module is used to classify the nitrogen, phosphorus and potassium nutrient content grid map of the cultivated land in the study area after normalization using the spatial constraint natural breakpoint method clustering objective function; The distribution calculation module is used to use three-digit decimal codes to respectively represent the classification results of the nitrogen, phosphorus and potassium nutrient content grid map of the soil in the cultivated land area of ​​the study area, and calculate and generate a soil comprehensive nutrient spatial distribution map; The evaluation module is used to analyze the spatial distribution map of the soil comprehensive nutrients, set coordinated colors, analyze the nutrient conditions of nitrogen, phosphorus, and potassium in each soil, count the areas and proportions of various types of soil comprehensive nutrients, and provide fertilization recommendations.

7. The device for comprehensive soil nutrient evaluation based on satellite remote sensing images according to claim 6, wherein: The classification module is further configured to determine the inflection point with the largest inter-class difference based on the spatial distribution of nutrient content, and includes a spatial heterogeneity analysis unit, an optimal classification threshold solution unit, a threshold verification unit, and a classification unit; The spatial heterogeneity analysis unit is used to calculate the semivariogram of soil nutrients in the study area based on geostatistical methods, determine the spatial autocorrelation range of nutrient content, and serve as the spatial constraint radius of the natural breakpoint method; The optimal classification threshold solving unit is used to solve the optimal classification threshold using the improved Jenks natural breakpoint algorithm, and the objective function is: Among them, C k represents the kth category, μ k represents the mean of the kth class, w ij represents the spatial weight of pixels i and j, I represents the indicator function, and λ represents the balance factor; The threshold verification unit is used to calculate the 95% confidence interval of each classification threshold by Bootstrap resampling 1000 times, and when the CV is less than 5%, the threshold is determined to be stable, and the threshold is used as the inflection point with the maximum difference between classes; The grading unit is used to determine the classification grade according to the inflection point where the difference between the classes is the largest.

8. The device for comprehensive soil nutrient evaluation based on satellite remote sensing images according to claim 6, wherein: The distributed computing module is configured as follows: Use three digits to represent the grading levels of the three nutrients, and the value of each digit corresponds to three levels of grading: A: Available phosphorus level, 1 = low phosphorus, 2 = medium phosphorus, 3 = high phosphorus; B position: alkaline nitrogen level, 1 = low nitrogen, 2 = medium nitrogen, 3 = high nitrogen; C position: fast-acting potassium level, 1 = low potassium, 2 = medium potassium, 3 = high potassium; The number of bits of the encoding can be expanded to N bits to be compatible with future newly added indicators.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the soil comprehensive nutrient evaluation method based on satellite remote sensing images as described in any one of claims 1 to 5 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the soil comprehensive nutrient evaluation method based on satellite remote sensing images as claimed in any one of claims 1 to 5 are implemented.

Images