Spectral identification method for realizing early warning of nitrogen phosphorus and potassium nutrition imbalance of summer corn

Abstract

The invention discloses a spectrum identification method for early warning of summer corn nitrogen phosphorus and potassium nutrition imbalance, and relates to the technical field of agricultural information monitoring and spectral analysis, and the method comprises the following steps: S1, obtaining hyperspectral reflectivity data of leaves at different leaf positions of a summer corn target plant in a key growth period; and S2, based on the hyperspectral reflectivity data, calculating a nitrogen-phosphorus-potassium collaborative diagnosis spectral index. According to the spectrum identification method for realizing early warning of the summer corn nitrogen-phosphorus-potassium nutrition imbalance, comprehensive judgment and early risk quantification of the summer corn nitrogen-phosphorus-potassium nutrition imbalance condition are realized by constructing a collaborative diagnosis spectrum index fused with nitrogen, phosphorus and potassium multi-element information and introducing a dynamic decision matrix based on a leaf position response time sequence rule. According to the method, the limitation of traditional single element diagnosis is overcome, and spectrum weak abnormal signals of different leaf positions of crops before visible symptoms appear can be captured, so that the early warning opportunity is advanced.

Description

Technical Field

[0001] This invention relates to the field of agricultural information monitoring and spectral analysis technology, specifically a spectral identification method for early warning of nitrogen, phosphorus, and potassium nutrient imbalance in summer maize. Background Technology

[0002] Summer maize, an important food crop in my country, has its yield and quality directly affected by its nutritional status during its growth process. Nitrogen, phosphorus, and potassium are the three essential nutrients for crop growth and development, and their balanced supply is crucial for the healthy growth of summer maize. Nutritional imbalances, especially deficiencies or excesses of nitrogen, phosphorus, and potassium, will lead to physiological metabolic disorders and decreased stress resistance in the plant, thus affecting yield and quality. Therefore, early identification and warning of nutrient imbalances in summer maize are of great significance for guiding precision fertilization, improving fertilizer utilization, and reducing environmental pollution.

[0003] Currently, the authorized publication number CN115343249A describes "a hyperspectral diagnostic method for nitrogen nutrition at the leaf scale throughout the entire life cycle of summer maize." This method achieves accurate assessment of the nitrogen nutrition status of summer maize by constructing an improved absorption area index and establishing nitrogen diagnostic models and critical nitrogen concentration dilution models for different leaf positions. Starting from the leaf scale, this technology clarifies the optimal diagnostic leaf position for different growth stages, improving the accuracy and practicality of nitrogen diagnosis.

[0004] However, the existing technologies mentioned above still have the following shortcomings: First, the method mainly targets the diagnosis of a single nutrient element (nitrogen) and has not yet achieved comprehensive identification and early warning of imbalances in multiple nutrients such as nitrogen, phosphorus, and potassium; second, the technology focuses on nutrient diagnosis throughout the entire life cycle, and its ability to provide early warning of nutrient imbalances has not yet been clearly defined and optimized; in addition, in actual field applications, when crops are under nutrient stress, the effects are often not obvious in the canopy, while the middle and lower leaves show characteristic changes earlier, and the existing methods still have room for improvement in multi-level, multi-angle three-dimensional monitoring and early identification.

[0005] Therefore, existing technologies lack a spectral identification method capable of simultaneously providing early warning of imbalances in the three key nutrients—nitrogen, phosphorus, and potassium—in summer maize, hindering true comprehensive nutritional diagnosis and precise regulation. Thus, there is an urgent need to conduct more comprehensive research on spectral identification technologies and construct a spectral model and methodology system suitable for early warning of nitrogen, phosphorus, and potassium nutrient imbalances in summer maize, providing technical support for achieving precision nutritional management of summer maize. Summary of the Invention

[0006] The purpose of this invention is to provide a spectral identification method for early warning of nitrogen, phosphorus and potassium nutrient imbalance in summer maize, so as to solve the problems mentioned in the background art.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a spectral identification method for early warning of nitrogen, phosphorus, and potassium nutrient imbalance in summer maize, comprising the following steps: S1: Obtain hyperspectral reflectance data of leaves at different leaf positions of summer maize target plants during key growth stages; S2: Calculate the nitrogen, phosphorus, and potassium synergistic diagnostic spectral index based on the hyperspectral reflectance data; S3: Based on the preset dynamic diagnostic decision matrix that associates leaf position, growth period and nutrient elements, select the specified leaf position and corresponding diagnostic spectral index that need to be monitored for different potential nutrient imbalances under the current growth period, and calculate the early risk index based on the deviation between the actual spectral index value of the selected leaf position and the corresponding normal growth benchmark value. S4: If the early risk index exceeds the primary warning threshold, initiate the first-level verification process and recalculate and judge the consistency of the same sample using a different spectral index or wavelength combination than in step S3. S5: After passing the first level of verification, a preliminary warning signal is generated, and the second level of verification process is initiated to conduct rapid field verification of the area that triggered the warning. The verification results are then fed back to the model optimization module to adjust the parameters in the dynamic diagnostic decision matrix.

[0008] Furthermore, in step S2, the construction process of the nitrogen, phosphorus, and potassium synergistic diagnostic spectral index includes: From the hyperspectral reflectance data, a first characteristic wavelength combination, a second characteristic wavelength combination, and a third characteristic wavelength combination that have specific responses to changes in the concentrations of nitrogen, phosphorus, and potassium elements were selected respectively. A first primary spectral index is constructed based on the reflectance data of the first characteristic wavelength combination, a second primary spectral index is constructed based on the reflectance data of the second characteristic wavelength combination, and a third primary spectral index is constructed based on the reflectance data of the third characteristic wavelength combination. The first, second, and third primary spectral indices are coupled using a nonlinear function that reflects the physiological coupling relationship between nitrogen, phosphorus, and potassium elements to form the nitrogen-phosphorus-potassium synergistic diagnostic spectral index.

[0009] Furthermore, in step S3, the dynamic diagnostic decision matrix is ​​constructed based on historical test data, and this matrix defines: During each predetermined summer maize growth stage, the leaf level that should be prioritized for monitoring should be for each potential type of nutrient imbalance. For the identified leaf level, one or more target spectral indices are recommended for assessing the nutrient imbalance. The change in the warning threshold and the weighting coefficient corresponding to the target spectral index; The calculation method of the early risk index is as follows: for the current reproductive period and suspected imbalance type, for one or more leaf level and their corresponding target spectral index specified by the dynamic diagnostic decision matrix, calculate the difference between the actual value of each index and the normal growth benchmark value, divide the difference by the corresponding early warning threshold change, multiply by the corresponding weight coefficient, and sum all items.

[0010] Furthermore, in step S4, the first-level verification process is specifically as follows: When the early risk index exceeds the primary warning threshold for the first time, at least one alternative spectral evaluation path that differs from the spectral index or wavelength combination used in calculating the early risk index in step S3 in terms of spectral formation mechanism is automatically selected according to the dynamic diagnostic decision matrix or preset rules. The backup spectral evaluation path is used to calculate the spectral index of the same leaf position of the target plant, and a verification index or verification conclusion is generated based on the calculation results. Determine whether the verification index or verification conclusion is consistent with the risk direction indicated by the early risk index; The first level of verification is confirmed to be passed only if the judgment result is consistent.

[0011] Furthermore, in step S5, the second-level verification process and model optimization process include: For the areas where the preliminary warning signals are generated, perform non-destructive or minimally destructive rapid field diagnostics to obtain verification data on nitrogen, phosphorus, and potassium nutrition status; The verification data is associated with the spectral data corresponding to the triggering of the initial warning signal, the early risk index, and the parameters used in the dynamic diagnostic decision matrix at that time, forming a feedback data pair; Using an incremental learning algorithm, the changes in the warning threshold and weight coefficients in the dynamic diagnostic decision matrix are dynamically adjusted and optimized based on the feedback data pairs.

[0012] Furthermore, the selection of the first characteristic wavelength combination, the second characteristic wavelength combination, and the third characteristic wavelength combination is completed using a continuous projection algorithm or a competitive adaptive reweighted sampling method to ensure that the wavelengths within each combination are highly sensitive to the target element and that the spectral interference between combinations is low.

[0013] Furthermore, in the dynamic diagnostic decision matrix, for the early warning of potential potassium deficiency, the monitoring priority of the lower leaves is set higher than that of the middle and upper leaves in multiple growth stages; for the early warning of potential nitrogen deficiency, the monitoring priority of the upper leaves is set relatively high in the vegetative growth stage.

[0014] Furthermore, the alternative spectral evaluation path includes: using a vegetation index calculation formula different from the main evaluation path, or using a combination of wavelengths obtained based on different characteristic wavelength screening algorithms to construct a spectral index.

[0015] Furthermore, the non-destructive or minimally destructive rapid field diagnostics include: measuring the relative chlorophyll content of leaves using a portable chlorophyll meter, and / or acquiring canopy multispectral data using a canopy multispectral sensor, and / or collecting a small number of leaf samples for rapid determination of nutrient elements.

[0016] Furthermore, the key growth period includes at least the seedling stage, the jointing stage, and the trumpet stage; the different leaf positions include at least the upper leaves, the middle leaves, and the lower leaves.

[0017] This invention provides a spectral identification method for early warning of nitrogen, phosphorus, and potassium nutrient imbalance in summer maize. It has the following beneficial effects: This spectral identification method for early warning of nitrogen, phosphorus, and potassium nutrient imbalance in summer maize constructs a collaborative diagnostic spectral index integrating information from multiple elements (nitrogen, phosphorus, and potassium) and introduces a dynamic decision matrix based on the temporal patterns of leaf position response. This enables a comprehensive assessment and early risk quantification of nitrogen, phosphorus, and potassium nutrient imbalance in summer maize. This method overcomes the limitations of traditional single-element diagnosis, capturing weak spectral abnormal signals at different leaf positions before visible symptoms appear, thus advancing the warning time.

[0018] This spectral identification method enables early warning of nitrogen, phosphorus, and potassium nutrient imbalances in summer maize. Through a built-in two-level independent verification process and a model adaptive optimization mechanism based on field verification, the method effectively improves the reliability and stability of the warning results. The entire solution forms a complete closed loop from spectral monitoring, risk warning, field verification to model self-correction, enhancing the applicability and practicality of the technology in different field environments and providing a directly operable tool for the refined nutrient management of summer maize. Attached Figure Description

[0019] Figure 1 This is a system architecture diagram of a spectral identification method for early warning of nitrogen, phosphorus, and potassium nutrient imbalance in summer maize according to the present invention. Figure 2 This is a two-level verification flowchart of a spectral identification method for early warning of nitrogen, phosphorus and potassium nutrient imbalance in summer maize according to the present invention. Figure 3 This is a schematic flowchart of a spectral identification method for early warning of nitrogen, phosphorus, and potassium nutrient imbalance in summer maize according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figures 1 to 3 This invention provides a technical solution: a spectral identification method for early warning of nitrogen, phosphorus, and potassium nutrient imbalance in summer maize, comprising the following steps: S1: Obtain hyperspectral reflectance data of leaves at different leaf positions of summer maize target plants during key growth stages; S2: Calculate the nitrogen, phosphorus, and potassium synergistic diagnostic spectral index based on hyperspectral reflectance data; S3: Based on the preset dynamic diagnostic decision matrix that associates leaf position, growth period and nutrient elements, select the specified leaf position and corresponding diagnostic spectral index that need to be monitored for different potential nutrient imbalances under the current growth period, and calculate the early risk index based on the deviation between the actual spectral index value of the selected leaf position and the corresponding normal growth benchmark value. S4: If the early risk index exceeds the primary warning threshold, initiate the first-level verification process and recalculate and judge the consistency of the same sample using a different spectral index or wavelength combination than in step S3. S5: After passing the first level of verification, a preliminary warning signal is generated, and the second level of verification process is initiated to conduct rapid field verification of the area that triggered the warning. The verification results are then fed back to the model optimization module to adjust the parameters in the dynamic diagnostic decision matrix.

[0022] It should be further explained that, firstly, during the key growth stages of summer maize, such as the seedling stage, jointing stage, and tasseling stage, hyperspectral measurement equipment was used to collect reflectance data representing the upper, middle, and lower layers of leaves of the plant, respectively, to obtain continuous spectral information covering the visible to near-infrared bands.

[0023] Next, based on the acquired hyperspectral data, a spectral index for the synergistic diagnosis of nitrogen, phosphorus, and potassium nutritional status is constructed. The construction of this index involves analyzing the entire spectral band, selecting three sets of characteristic wavelengths that are most sensitive to changes in nitrogen, phosphorus, and potassium content and have the least mutual interference, and forming three basic spectral parameters based on the reflectance values ​​of these sets. These three parameters are then fused through a preset nonlinear function to obtain a composite index that can comprehensively reflect the balance relationship between the three elements.

[0024] Then, based on a pre-established dynamic diagnostic decision matrix, the leaf positions and corresponding spectral indices that need to be monitored in the current growth stage are determined. This matrix is ​​based on historical field trial data and specifies the leaf position levels to be examined first and the recommended diagnostic indices for different nutrient imbalances that may occur at different growth stages. It also stores the corresponding normal growth benchmark values ​​and early warning threshold parameters. Using the guidance of this matrix, the deviation between the actual spectral index value of the selected leaf position and the benchmark value is calculated. Combined with the pre-set weighting coefficients in the matrix, a quantitative early risk index is calculated by weighted summation.

[0025] When the calculated early risk index exceeds the set initial warning line, the system automatically triggers the first-level verification process. This process requires the same leaf sample to be re-evaluated using a different spectral analysis path than the one used to calculate the risk index. This path may use another set of vegetation index calculation formulas or wavelength combinations selected based on different principles, and determine whether the risk trends indicated by the two evaluation results are consistent. Only when the trends are consistent is it considered to have passed the verification.

[0026] For cases that pass the first layer of verification, the system generates an initial warning and then initiates the second layer of verification. This process involves conducting rapid on-site verification of the field area where the warning was issued, such as using portable equipment to measure chlorophyll or collecting trace leaf samples for rapid testing, to obtain actual nutritional status verification information. Finally, these verification results, along with the spectral data, risk index, and matrix parameters used when the warning was triggered, are fed back to the model optimization module. This module uses an incremental learning algorithm to adaptively adjust the warning threshold and weight parameters in the decision matrix, thereby completing a complete closed-loop process from diagnosis and verification to model optimization.

[0027] In step S2, the construction process of the nitrogen, phosphorus, and potassium synergistic diagnostic spectral index includes: From hyperspectral reflectance data, the first characteristic wavelength combination, the second characteristic wavelength combination, and the third characteristic wavelength combination that have specific responses to changes in the concentrations of nitrogen, phosphorus, and potassium elements were selected respectively. A first primary spectral index is constructed based on reflectance data of a first characteristic wavelength combination, a second primary spectral index is constructed based on reflectance data of a second characteristic wavelength combination, and a third primary spectral index is constructed based on reflectance data of a third characteristic wavelength combination. The first, second, and third primary spectral indices are coupled using a nonlinear function that reflects the physiological coupling relationship between nitrogen, phosphorus, and potassium elements to form a nitrogen-phosphorus-potassium synergistic diagnostic spectral index.

[0028] It should be further explained that in the process of constructing the nitrogen, phosphorus, and potassium synergistic diagnostic spectral index, the acquired hyperspectral reflectance data was first preprocessed, including noise removal and smoothing. Subsequently, a continuous projection algorithm was used to extract characteristic wavelengths from the full-spectrum data: this algorithm, through iterative projection operations, finds a set of wavelengths that are most correlated with the nitrogen concentration in the leaves and have the lowest collinearity among them, forming the first characteristic wavelength combination; using the same principle, it finds two sets of independent wavelengths that are most correlated with the phosphorus concentration and potassium concentration respectively and have low internal redundancy, forming the second and third characteristic wavelength combinations.

[0029] Based on the selected wavelength groups, specific mathematical combinations of their reflectance are calculated. For example, for the first combination, the reflectance ratio or normalized difference of two wavelengths can be used to form the first primary spectral index, which mainly responds to changes in nitrogen content. Similarly, using the reflectance data of the second and third combinations, through similar but different mathematical operations, the second and third primary spectral indices, which are more sensitive to changes in phosphorus and potassium content, are constructed, respectively.

[0030] Finally, the three primary indices are fed into a predefined nonlinear function for fusion. The structure of this function reflects the interaction between nitrogen, phosphorus, and potassium in plant physiological metabolism. For example, a weighted product form can be used, where the weight coefficient of each primary index is determined through training with historical data, or a quadratic polynomial form with cross terms can be used. This ensures that the final synergistic diagnostic index not only contains independent information about each element, but also amplifies the spectral response anomalies caused by imbalances in the proportions between elements, thereby achieving a comprehensive assessment of nitrogen, phosphorus, and potassium nutritional status and a sensitive indication of imbalances.

[0031] In step S3, the dynamic diagnostic decision matrix is ​​constructed based on historical experimental data. This matrix defines: During each predetermined summer maize growth stage, the leaf level that should be prioritized for monitoring should be for each potential type of nutrient imbalance. For the identified leaf level, one or more target spectral indices are recommended for assessing the nutrient imbalance. The change in the warning threshold and its weighting coefficient corresponding to the target spectral index; The calculation method for the early risk index is as follows: for the current reproductive period and suspected imbalance type, one or more leaf level and their corresponding target spectral index specified by the dynamic diagnostic decision matrix are used to calculate the difference between the actual value of each index and the normal growth benchmark value. The difference is divided by the corresponding early warning threshold change and then multiplied by the corresponding weight coefficient. All items are then summed.

[0032] It should be further explained that the construction of the dynamic diagnostic decision matrix is ​​based on the historical field trial data of the system. The trial set up multiple treatment groups covering different application levels of nitrogen, phosphorus and potassium, and simultaneously collected detailed hyperspectral data of the upper, middle and lower leaves of each treatment group and the corresponding chemical values ​​of nitrogen, phosphorus and potassium concentrations in the leaves during multiple preset growth stages of summer maize.

[0033] By analyzing these historical data, the temporal order and intensity differences of spectral abnormal signals appearing at different leaf positions when different nutrients are subjected to mild stress (at which time the plant has not yet shown typical deficiency symptoms) are quantitatively summarized. Specifically, the matrix is ​​stored in the form of a table or database. Its row index corresponds to different growth stages and potential combinations of nutrient imbalance types. Its column information defines the leaf position level that should be prioritized for monitoring in this specific situation, one or more target spectral indices recommended (such as the constructed synergistic diagnostic index or its specific sub-indices), the baseline value of the index under normal growth conditions, the threshold change used to determine whether the deviation constitutes a warning, and the weight coefficient reflecting the importance of the leaf position-index combination for diagnosing the current imbalance type.

[0034] The calculation process of the early risk index is as follows: First, based on the current reproductive stage and the type of imbalance to be screened, the corresponding monitoring scheme, i.e., one or more (leaf position, target spectral index) pairs, is retrieved from the matrix; then, the spectral data of these specified leaf positions are measured and the actual values ​​of the corresponding target spectral indices are calculated; next, for each (leaf position, index) pair, the difference between its actual value and the corresponding benchmark value stored in the matrix is ​​calculated, and the difference is divided by the corresponding threshold change amount pre-stored in the matrix to obtain a scalarized ratio, and then the ratio is multiplied by the corresponding weight coefficient assigned in the matrix; finally, the weighted results of all (leaf position, index) pairs calculated in the above manner are summed, and the resulting comprehensive evaluation value is the early risk index. This index realizes the weighted fusion and quantitative expression of the weak abnormal information reflected by different leaf positions and different indices.

[0035] In step S4, the first-level verification process is as follows: When the early risk index exceeds the primary warning threshold for the first time, at least one backup spectral assessment path that differs from the spectral index or wavelength combination used in calculating the early risk index in step S3 in terms of spectral formation mechanism is automatically selected according to the dynamic diagnostic decision matrix or preset rules. The spectral index of the same leaf position of the target plant is calculated using the backup spectral evaluation path, and a verification index or verification conclusion is generated based on the calculation results. Determine whether the verification index or verification conclusion is consistent with the risk direction indicated by the earlier risk index; The first level of verification is confirmed to be passed only if the judgment result is consistent.

[0036] It should be further explained that the specific implementation of the first-level verification process is as follows: When the early risk index value calculated by the system exceeds the preset primary warning threshold, the process is automatically started; the system first accesses the preset verification rule base according to the potential nutrient imbalance type and the current growth stage corresponding to the triggering of the warning. The rule base is configured with at least one backup spectral assessment path for different scenarios. The core calculation features used by the path, whether it is the specific vegetation index formula selected (e.g., using the red edge position index to replace the original ratio index) or the characteristic wavelength combination on which the index is constructed (e.g., another set of wavelengths selected by the competitive adaptive reweighted sampling method), are all independent of the main assessment path used in step S3 when calculating the early risk index in terms of physical principles or data. The system then uses this backup path to process the hyperspectral reflectance data of the same leaf position of the same target plant, calculates the corresponding verification spectral index value, and generates an independent verification conclusion or quantified verification index based on the comparison of the verification index value with the corresponding benchmark value. Subsequently, the system compares the risk direction indicated by the early risk index obtained from the main assessment path (e.g., indicating "increased risk of nitrogen deficiency") with the trend of the verification conclusion or verification index obtained from the backup path. The specific criterion for judging whether the two are consistent is: if the early risk index indicates the risk of a specific nutrient deficiency, the verification index value calculated by the backup path should also be lower than its normal range or show the same downward trend. Only when the risk judgment directions obtained from the two independent analysis paths are consistent, the system confirms that the warning signal has passed the first level of verification; otherwise, the warning is regarded as a false alarm caused by accidental interference and is excluded.

[0037] In step S5, the second-level verification process and model optimization process include: For areas where preliminary warning signals are generated, perform non-destructive or minimally destructive rapid field diagnostics to obtain verification data on nitrogen, phosphorus, and potassium nutrition status; The verification data is linked and recorded with the spectral data, early risk index, and parameters used in the dynamic diagnostic decision matrix at the time when the initial warning signal was triggered, forming a feedback data pair; Using an incremental learning algorithm, the changes in warning thresholds and weight coefficients in the dynamic diagnostic decision matrix are dynamically adjusted and optimized based on feedback data pairs.

[0038] It should be further explained that the specific implementation of the second-level verification process and model optimization process includes the following steps: For specific field areas where preliminary warning signals have been generated, operators use portable chlorophyll meters to measure the SPAD value of leaves at marked points, and / or use handheld multispectral sensors to collect multispectral reflectance data of the canopy layer, and / or use sampling tools to collect a small number of leaf samples, and obtain approximate measured values ​​of nitrogen, phosphorus and potassium content through portable leaf nutrient rapid testers, which are used as rapid field verification data; The system binds the verification data with the complete spectral data recorded at the time of triggering the warning, the calculated early risk index value, all parameters called in the dynamic diagnostic decision matrix at that time (including the specified leaf position, target spectral index, benchmark value, threshold change and weight coefficient), as well as the specific timestamp and geographical location information, to form a structured feedback data pair and store it in the historical feedback database. The model optimization module is activated periodically or after accumulating a certain number of feedback data pairs. The incremental learning algorithm used in this module is an online sequence learning algorithm. Its processing is as follows: the field verification results in the feedback data pair are used as the true label, and the prediction judgment made by the system based on spectral data at the time of warning (determined by the early risk index and threshold) is used as the prediction label of the current output of the model. The error between the two is calculated. Based on this error, the algorithm iteratively updates the changes in warning thresholds and weight coefficients directly related to the current warning judgment in the dynamic diagnostic decision matrix along the negative gradient direction of the loss function with a preset learning rate. For example, if multiple verifications confirm that the threshold change of a certain index at a certain leaf position under a specific growth period is set too sensitively, leading to false alarms, the algorithm will gradually increase the threshold change. Through continuous feedback and iteration, the model optimization module enables the parameters in the dynamic diagnostic decision matrix to adapt to changes in different field environments, variety characteristics, and interannual climate differences, thereby forming a closed-loop diagnostic system with self-correction and improvement capabilities.

[0039] The selection of the first, second, and third characteristic wavelength combinations is completed using a continuous projection algorithm or a competitive adaptive reweighted sampling method to ensure that the wavelengths within each combination are highly sensitive to the target element and that the spectral interference between combinations is low.

[0040] It should be further explained that when screening characteristic wavelength combinations that have specific responses to changes in nitrogen, phosphorus, and potassium element concentrations, the continuous projection algorithm or competitive adaptive reweighted sampling method is implemented as follows: When using the continuous projection algorithm, the laboratory measurement value of the nitrogen element concentration of the leaf to be tested is used as the reference vector, and the reflectance of all hyperspectral bands is used as the initial projection matrix. The algorithm starts from a certain initial wavelength and, through iterative calculation, finds the wavelength with the largest projection residual with the selected wavelength set among the unselected wavelengths and includes it in the characteristic combination. This process is repeated until the preset number of wavelengths or the change in projection residual is lower than a set limit. The wavelengths in the first characteristic wavelength combination obtained are strongly correlated with the nitrogen concentration and have low information redundancy among them. Using the measured values ​​of phosphorus and potassium concentrations as reference vectors, the above process is repeated to independently obtain the second and third characteristic wavelength combinations.

[0041] If a competitive adaptive reweighted sampling method is adopted, the nitrogen concentration measurement value is used as the dependent variable, and the stability of resampling and model regression coefficients is used as the screening criteria: a partial least squares regression model is established through multiple Monte Carlo samplings. After each sampling, the wavelengths are weighted according to the absolute value of the regression coefficients, and wavelengths with smaller weights are eliminated. The retained wavelengths enter the next iteration. After multiple rounds of competition and screening, wavelengths with stable regression coefficients and larger absolute values ​​are retained and constitute the first characteristic wavelength combination. If the dependent variables are changed to phosphorus concentration and potassium concentration and the same resampling, modeling and competitive screening process is performed independently, different second and third characteristic wavelength combinations are obtained respectively.

[0042] By using any of the above methods, it can be ensured that the three wavelength combinations obtained at the end each maintain high diagnostic sensitivity for their target elements, while minimizing mutual interference between diagnostic signals of different elements caused by overlapping or collinearity of spectral bands.

[0043] In the dynamic diagnostic decision matrix, for early warning of potential potassium deficiency, the monitoring priority of lower leaves is set higher than that of middle and upper leaves in multiple growth stages; for early warning of potential nitrogen deficiency, the monitoring priority of upper leaves is set relatively high in the vegetative growth stage.

[0044] It should be further explained that the specific rules for setting the priority of monitoring leaf positions for different nutrient elements in the dynamic diagnostic decision matrix were established and implemented based on the analysis and summarization of historical field trial data: For the early warning scenario of potential potassium deficiency, through continuous observation of summer maize plants under different potassium application levels, it was found that during multiple preset growth stages, when the plants began to experience potassium stress, the spectral characteristics of the lower leaves (such as the reflectance ratio of specific bands or specific sub-items of the constructed synergistic diagnostic index) usually showed earlier and more obvious changes compared to the middle and upper leaves. Therefore, when constructing the matrix, for the row records involving potassium deficiency, the "lower leaves" were systematically set as the first monitoring leaf position, and their corresponding weight coefficients were assigned relatively higher values, while the middle and upper leaves were set as subsequent or auxiliary monitoring leaf positions and assigned relatively lower weight coefficients.

[0045] For early warning of potential nitrogen deficiency, especially during the growth stages (such as seedling and jointing stages), historical data analysis shows that the photosynthetic activity of the upper leaves is more closely related to nitrogen supply, and their spectral response may be more sensitive than that of the middle and lower leaves. Accordingly, in the configuration of the matrix corresponding to these growth stages and nitrogen deficiency types, the monitoring priority of the "upper leaves" is set higher than that of the middle and lower leaves, which is reflected in the allocation of a higher weight coefficient to guide the system to focus more on the spectral information of the upper leaves when calculating the early risk index. These specific priority rules and corresponding weight values ​​are determined based on historical datasets during model initialization and stored in the corresponding fields of the matrix. They are called during the actual diagnosis process to guide leaf position selection and index weighting calculation.

[0046] Alternative spectral assessment pathways include: using vegetation index calculation formulas different from the main assessment pathway, or constructing spectral indices using wavelength combinations obtained from different characteristic wavelength screening algorithms.

[0047] It should be further explained that the specific implementation methods of the backup spectral evaluation path in the first-level verification process include the following two types: The first type uses a vegetation index calculation formula different from the main evaluation path. For example, if the main evaluation path uses a simple ratio index composed of reflectance of two specific wavelengths (such as R780 / R550), the backup path can use normalized difference vegetation index (such as (R780-R550) / (R780+R550)), triangular vegetation index, or red edge position index, etc., based on different mathematical forms or physical meanings for calculation; The second type constructs the spectral index based on wavelength combinations obtained from different characteristic wavelength screening algorithms. That is, even for the same suspected imbalance element, the backup path does not use... The wavelength combination selected by the continuous projection algorithm, which is relied upon by the main path, is replaced by another set of characteristic wavelengths independently selected by the competitive adaptive reweighted sampling method, or a set of differentiated wavelengths obtained by other feature selection methods such as genetic algorithms and stepwise regression. The spectral index is then reconstructed based on this new set of wavelengths (such as recalculating the ratio or normalizing the difference). The core of these two types of backup path designs is that the extraction and synthesis of spectral information are independent of the main evaluation path in terms of mathematical basis or data source. This allows for cross-validation of the nutritional status of the same leaf sample from different perspectives, effectively reducing the possibility of misjudgment due to single algorithm bias, interference in specific bands, or random noise, and improving the reliability of the early warning conclusion.

[0048] Non-destructive or minimally destructive rapid field diagnostics include: measuring the relative chlorophyll content of leaves using a portable chlorophyll meter, and / or acquiring canopy multispectral data using a canopy multispectral sensor, and / or collecting a small number of leaf samples for rapid determination of nutrient elements.

[0049] It should be further explained that non-destructive or minimally destructive rapid field diagnosis is implemented in the following specific manner: In the area where the initial warning signal is generated, the operator uses a portable chlorophyll meter to select the leaf position of concern for the warning on the plant, avoiding the main vein, to measure the SPAD value at a specific location on the leaf, and selects multiple representative points in this area to obtain the average value as one of the verification data; simultaneously, or at a selected location, a handheld or portable canopy multispectral sensor is used to collect canopy reflectance spectral data covering several discrete bands of visible and near-infrared light at a fixed height and angle above the warning field, and... The built-in algorithm calculates vegetation indices related to nitrogen, phosphorus, and potassium nutrient status in real time. In addition, a sampler can be used to collect a small number of leaf samples in the warning area, a leaf perforator can be used to obtain a leaf disc of a fixed area, or a small number of leaves can be cut into pieces and mixed. The sample extract can then be measured using a portable leaf nutrient rapid tester based on the principle of colorimetric reaction, following its operating procedure. This allows for the acquisition of approximate or semi-quantitative results of nitrogen, phosphorus, and potassium content in a short time. These verification data obtained through different independent means are recorded and linked to the warning information, and are used together to verify the accuracy of the spectral warning and provide a real data source for model optimization.

[0050] The critical growth period includes at least the seedling stage, the jointing stage, and the trumpet stage; different leaf positions include at least the upper leaves, the middle leaves, and the lower leaves.

[0051] It should be further explained that the specific definitions and operations for key growth periods and different leaf positions are as follows: The key growth periods of seedling stage, jointing stage, and trumpet stage are determined according to the standard phenological observation guidelines for summer maize fields. For example, the seedling stage is marked by the emergence of the third complete leaf, the jointing stage is marked by the obvious elongation of the basal internodes and their palpable texture, and the trumpet stage is marked by the emergence of clustered central leaves and the unfolding of the upper leaves resembling a trumpet. Data is collected during each designated key growth period.

[0052] The division of leaves into upper, middle, and lower layers is determined based on their position on the main stem of the plant: upper leaves are defined as the leaves from the newly fully unfolded leaves to the leaves at the top that are not yet fully unfolded; middle leaves are defined as several functional leaves located in the middle of the plant, between the upper and lower leaves; and lower leaves are defined as several leaves at the base of the plant that are the first to age. In actual measurements, representative, healthy, and disease-free leaves from these three leaf positions are selected from each selected plant as the objects of spectral measurement and subsequent analysis.

[0053] By independently and systematically acquiring spectral information from leaves with physiological differences at these three spatial levels within the clearly defined reproductive stages, a structured basic data input was provided for the subsequent construction of a diagnostic model that can reflect the spatiotemporal migration and distribution patterns of nutrients.

[0054] By constructing a synergistic diagnostic spectral index integrating information from multiple elements such as nitrogen, phosphorus, and potassium, and introducing a dynamic decision matrix based on the temporal patterns of leaf position response, a comprehensive assessment and early risk quantification of nitrogen, phosphorus, and potassium nutrient imbalances in summer maize were achieved. This method overcomes the limitations of traditional single-element diagnosis, enabling the capture of weak spectral abnormal signals at different leaf positions before the appearance of visible symptoms, thus allowing for earlier warnings.

[0055] This method effectively improves the reliability and stability of early warning results through a built-in two-level independent verification process and a model adaptive optimization mechanism based on field verification. The entire solution forms a complete closed loop from spectral monitoring, risk early warning, field verification to model self-correction, enhancing the applicability and practicality of the technology in different field environments and providing a directly operable tool for the refined nutrient management of summer maize.

[0056] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0057] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A spectral recognition method for early warning of nitrogen, phosphorus and potassium nutrient imbalance of summer corn, characterized in that, The method comprises the following steps: S1: obtaining hyperspectral reflectance data of summer maize target plants at key growth stages and different leaf positions; S2: calculating a nitrogen-phosphorus-potassium collaborative diagnosis spectral index based on the hyperspectral reflectance data; S3: according to a preset dynamic diagnosis decision matrix related to leaf position, growth stage and nutrient element, screening the specified leaf position and the corresponding diagnosis spectral index required for monitoring different potential nutrient imbalance types at the current growth stage, and calculating an early risk index based on the deviation of the actual spectral index value of the selected leaf position from the corresponding normal growth reference value; S4: when the early risk index exceeds the primary warning threshold, starting a first-level verification process, and using a different spectral index or wavelength combination to calculate and judge the consistency of the same sample again; S5: when passing the first-level verification, generating a preliminary warning signal and starting a second-level verification process, conducting a field rapid verification on the area triggering the warning, and feeding back the verification result to the model optimization module for adjusting the parameters in the dynamic diagnosis decision matrix.

2. The spectral identification method for early warning of nitrogen, phosphorus and potassium nutrient imbalance of summer maize according to claim 1, characterized in that: In step S2, the construction process of the nitrogen-phosphorus-potassium collaborative diagnosis spectral index comprises: selecting a first characteristic wavelength combination, a second characteristic wavelength combination and a third characteristic wavelength combination respectively from the hyperspectral reflectance data, which have specific responses to the changes in the concentrations of nitrogen, phosphorus and potassium elements; constructing a first primary spectral index based on the reflectance data of the first characteristic wavelength combination, a second primary spectral index based on the reflectance data of the second characteristic wavelength combination, and a third primary spectral index based on the reflectance data of the third characteristic wavelength combination; coupling the first primary spectral index, the second primary spectral index and the third primary spectral index through a nonlinear function reflecting the physiological coupling relationship among nitrogen, phosphorus and potassium elements to form the nitrogen-phosphorus-potassium collaborative diagnosis spectral index.

3. The spectral identification method for early warning of nitrogen, phosphorus and potassium nutrient imbalance of summer maize according to claim 1, characterized in that: In step S3, the dynamic diagnosis decision matrix is constructed based on historical test data, and the matrix defines: for each potential nutrient imbalance type, the leaf position level that should be monitored first at each preset summer maize growth stage; for the determined leaf position level, one or more target spectral indices for evaluating the nutrient imbalance condition are recommended; the warning threshold variation and the weight coefficient corresponding to the target spectral index; The calculation method of the early risk index is as follows: for one or more leaf position levels and their corresponding target spectral indices specified by the dynamic diagnosis decision matrix at the current growth stage and the suspected imbalance type, the difference between the actual value of each index and the normal growth reference value is calculated, the difference is divided by the corresponding warning threshold variation, then multiplied by the corresponding weight coefficient, and the sum of all terms is obtained.

4. The spectral identification method for early warning of nitrogen, phosphorus and potassium nutrient imbalance of summer maize according to claim 1, characterized in that: In step S4, the first-level verification process is as follows: when the early risk index first exceeds the primary warning threshold, at least one backup spectral evaluation path different from the spectral index or wavelength combination used in step S3 for calculating the early risk index is automatically selected according to the dynamic diagnosis decision matrix or a preset rule, and the consistency of the early risk index calculated by the backup spectral evaluation path is judged. Spectrum index calculation is performed on the same leaf position of the target plant by using the backup spectrum evaluation path, and a verification index or a verification conclusion is generated based on the calculation result; It is judged whether the verification index or the verification conclusion is consistent with the risk direction indicated by the early risk index; Only when the judgment result is consistent, the first level verification is confirmed.

5. The spectral identification method for early warning of nitrogen, phosphorus and potassium nutrient imbalance of summer maize according to claim 1, characterized in that: In step S5, the second level verification process and model optimization process include: Non-destructive or micro-destructive field rapid diagnosis is performed on the area where the preliminary warning signal is generated, and verification data about nitrogen, phosphorus and potassium nutrient status are obtained; The verification data are associated with the spectrum data corresponding to the time when the preliminary warning signal is triggered, the early risk index, and the parameters used in the dynamic diagnosis decision matrix at that time to form a feedback data pair; By using an incremental learning algorithm, the early warning threshold change and the weight coefficient in the dynamic diagnosis decision matrix are dynamically adjusted and optimized based on the feedback data pair.

6. The spectral identification method for early warning of nitrogen, phosphorus and potassium nutrient imbalance of summer maize according to claim 2, characterized in that: The screening of the first, second and third characteristic wavelength combinations is completed by using a successive projections algorithm or a competitive adaptive reweighted sampling method to ensure that the sensitivity of each combination to the target element is high, and the spectral interference degree between the combinations is low.

7. The spectral identification method for early warning of nitrogen, phosphorus and potassium nutrient imbalance of summer maize according to claim 3, characterized in that: In the dynamic diagnosis decision matrix, for the early warning of potential potassium deficiency, the monitoring priority of the lower leaves is set to be higher than that of the middle and upper leaves in multiple growth periods; for the early warning of potential nitrogen deficiency, the monitoring priority of the upper leaves is set to be relatively high in the growth period of the vegetative growth stage.

8. The spectral identification method for early warning of nitrogen, phosphorus and potassium nutrient imbalance of summer maize according to claim 4, characterized in that: The backup spectrum evaluation path includes using a different vegetation index calculation formula from the main evaluation path, or using a wavelength combination obtained based on a different characteristic wavelength screening algorithm to construct a spectrum index.

9. The spectral identification method for early warning of nitrogen, phosphorus and potassium nutrient imbalance of summer maize according to claim 5, characterized in that: The non-destructive or micro-destructive field rapid diagnosis includes using a portable chlorophyll meter to measure the relative content of leaf chlorophyll, and / or using a canopy multispectral sensor to obtain canopy multispectral data, and / or collecting a small amount of leaf samples for rapid determination of nutrient elements.

10. The spectral identification method for early warning of nitrogen, phosphorus and potassium nutrient imbalance of summer maize according to claim 1, characterized in that: The key growth periods include at least the seedling stage, the jointing stage and the trumpet mouth stage, and the different leaf positions include at least the upper leaves, the middle leaves and the lower leaves.

Images