A method and system for evaluating nitrogen content in corn leaves based on hyperspectral remote sensing

Abstract

The present invention provides a method and system for evaluating the nitrogen content of corn leaves based on hyperspectral remote sensing, which relates to the technical field of crop remote sensing analysis. The present invention collects hyperspectral data with a spectral range of 400-1000 nm, performs atmospheric correction, divides sub-bands, and generates environmental correction coefficients for different sub-bands. Based on the least squares method, a functional relationship between spectral reflectance and nitrogen content is constructed for each sub-band, the spectral sensitivity of each sub-band is determined, and the five bands with the highest sensitivity are selected as sensitive bands. The weight coefficients of each sensitive band are corrected by the environmental correction coefficient to generate an improved weighted spectral index. Finally, a deep learning network is used to take the improved weighted spectral index as input to train a nitrogen content prediction model, and the nitrogen content prediction value of corn leaves is generated through hyperspectral data and related environmental information.

Description

Technical Field

[0001] The present invention relates to the technical field of crop remote sensing analysis, and specifically to a method and system for evaluating the nitrogen content of corn leaves based on hyperspectral remote sensing. Background Art

[0002] In modern agricultural production, the application of nitrogen fertilizer is crucial for crop growth. Excessive application of nitrogen fertilizer not only increases production costs but may also have negative impacts on the environment such as water eutrophication. Therefore, accurately evaluating the nitrogen content of crops and then achieving precise fertilization is an important topic for improving agricultural production efficiency, reducing agricultural costs, and achieving sustainable development. Traditional nitrogen content detection methods mostly rely on laboratory chemical analysis, which is not only time-consuming and laborious but also difficult to reflect the growth status of crops and actual nitrogen requirements in real time. In recent years, hyperspectral remote sensing technology has gradually become an effective means for evaluating crop physiological characteristics, but how to improve the accuracy of nitrogen content prediction remains the focus of research. To solve this problem, this solution proposes a method for evaluating the nitrogen content of corn leaves based on hyperspectral remote sensing, aiming to improve the prediction accuracy of nitrogen content while achieving rapid and accurate implementation monitoring, providing technical support for precision agriculture.

[0003] In the prior art, the published patent number CN116026772B discloses a method for predicting the nitrogen content of corn leaves based on hyperspectral remote sensing. It performs spectral measurement and reflection extraction on corn leaves to determine sensitive bands and specific sensitive wavelengths, establishes basic functional relationships of the nitrogen content of corn leaves based on planting density and soil moisture parameters, and planting density and soil nitrogen content parameters respectively, further determines a multi-factor discriminant function, and then establishes a unified multi-weight multi-factor discriminant function to predict the nitrogen content of corn leaves.

[0004] The main problems of the above solution are as follows: It is necessary to collect several environmental parameters including planting density, soil moisture, and soil nitrogen content, and convert these parameters into data for establishing basic functional relationships. This process takes a long time, is difficult to meet the need for dynamic data update during real-time monitoring, and increases the complexity of data processing; moreover, collecting environmental parameters under different conditions is not accurate enough, resulting in a decrease in the accuracy of the overall model, and the combination of multiple functions will introduce more noise and deviation, affecting the accuracy of the prediction results.

[0005] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, so it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and system for evaluating the nitrogen content of corn leaves based on hyperspectral remote sensing, so as to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for evaluating the nitrogen content of corn leaves based on hyperspectral remote sensing, the specific steps include:

[0009] Step 1: Select multiple corn leaf samples with known nitrogen content, collect the hyperspectral data of the corn leaf samples in the spectral range of 400-1000 nm, perform atmospheric correction on the hyperspectral data to obtain the corrected spectral reflectance value, divide the sub-bands, and collect the environmental temperature, soil background reflectance and light intensity to calculate the environmental correction coefficient of different sub-bands;

[0010] Step 2: Based on the least squares method, construct a functional relationship between the spectral reflectance and the nitrogen content in each sub-band, and determine the spectral sensitivity of each sub-band based on the functional relationship;

[0011] Step 3: Select the five bands with the highest spectral sensitivity as the sensitive bands, determine the weight coefficients of each sensitive band, correct the weight coefficients based on the environmental correction coefficient, and generate an improved weighted spectral index by combining the corrected weight coefficients and the corrected spectral reflectance value;

[0012] Step 4: Based on the deep learning network, use the improved weighted spectral index as the input and the nitrogen content of the corn leaves as the label to train the nitrogen content prediction model;

[0013] Step 5: Based on the hyperspectral data, environmental temperature, soil background reflectance and light intensity of the corn leaves, generate the improved weighted spectral index according to the above steps and input it into the nitrogen content prediction model to generate the predicted value of the nitrogen content of the corn leaves.

[0014] Further, the formula for generating the environmental correction coefficient is:

[0015] ;

[0016] Wherein, represents the environmental correction coefficient of the th sub-band, represents the index of the sub-band, represents the environmental temperature, represents the th standard environmental temperature of the sub-band, represents the soil background reflectance, represents the th standard soil background reflectance of the sub-band, represents the light intensity, represents the standard light intensity of the th sub-band, respectively represent the weight coefficients of the ambient temperature, soil background reflectance, and light intensity, , and .

[0017] Furthermore, the principle for constructing the functional relationship between the spectral reflectance and nitrogen content in each sub-band is as follows:

[0018] Select the spectral reflectance at the center wavelength of each sub-band as the spectral reflectance of the sub-band. The functional relationship between the spectral reflectance and nitrogen content in each sub-band is:

[0019] ;

[0020] where represents the nitrogen content in the th sub-band, represents the spectral reflectance of the th sub-band, represents the index of the sub-band, respectively represent the fitting parameters to be determined for the th sub-band;

[0021] Construct an error function for each corn leaf sample in different sub-bands. The error function is:

[0022] ;

[0023] where represents the nitrogen content deviation of all corn leaf samples in the th sub-band, represents the index of the corn sample, and , represents the number of corn leaf samples, represents the nitrogen content of the th corn leaf sample in the th sub-band, represents the spectral reflectance of the th corn leaf sample in the th sub-band;

[0024] For respectively take the partial derivatives with respect to and set the partial derivatives to zero:

[0025] ;

[0026] ;

[0027] ;

[0028] Among them, respectively represent the partial derivative with respect to ;

[0029] After arrangement, the normal equations are obtained:

[0030] ;

[0031] ;

[0032] ;

[0033] Let , , , , , , ;

[0034] Solve the normal equations:

[0035] ;

[0036] ;

[0037] ;

[0038] Substitute the solved into the functional relationship between the spectral reflectance and nitrogen content under each sub - band.

[0039] Furthermore, the principle for determining the spectral sensitivity of each sub - band is:

[0040] Based on the functional relationship between the spectral reflectance and nitrogen content under each sub - band, take the derivative of the spectral reflectance:

[0041] ;

[0042] ;

[0043] Among them, represents the derivative of the nitrogen content with respect to the spectral reflectance;

[0044] Based on the mean value of all corn leaf samples, the formula for generating the spectral sensitivity of this sub - band is:

[0045] ;

[0046] Among them, represents the Spectral sensitivity of each sub-band.

[0047] Furthermore, the principle for determining the weight coefficients of each sensitive band is as follows:

[0048] ;

[0049] Wherein, represents the weight coefficient of the th sensitive band, represents the index of the sensitive band, and , represents the th spectral sensitivity of the sensitive band.

[0050] Furthermore, the principle for generating the improved weighted spectral index is as follows:

[0051] The weight coefficient of the calibrated sensitive band is:

[0052] ;

[0053] Wherein, represents the weight coefficient of the th calibrated sensitive band, represents the environmental correction coefficient of the th sensitive band;

[0054] Perform equal-proportion scaling on to make ;

[0055] The formula for generating the improved weighted spectral index is:

[0056] ;

[0057] Wherein, represents the improved weighted spectral index of the th corn leaf sample.

[0058] The present invention also provides a corn leaf nitrogen content evaluation system based on hyperspectral remote sensing. The system is used to implement the above-mentioned corn leaf nitrogen content evaluation method based on hyperspectral remote sensing, and specifically includes:

[0059] A data acquisition module, which is used to select multiple corn leaf samples with known nitrogen contents, collect hyperspectral data of the corn leaf samples in the spectral range of 400~1000nm, perform atmospheric correction on the hyperspectral data to obtain the spectral reflectance correction value, divide sub-bands, and collect the environmental temperature, soil background reflectance, and light intensity to calculate the environmental correction coefficients of different sub-bands;

[0060] A spectral calculation module, which is used to construct a functional relationship between spectral reflectance and nitrogen content in each sub-band based on the least squares method, and determine the spectral sensitivity of each sub-band based on the functional relationship;

[0061] A spectral optimization module, which is used to select the five bands with the highest spectral sensitivity as sensitive bands, determine the weight coefficients of each sensitive band, correct the weight coefficients based on the environmental correction coefficient, and generate an improved weighted spectral index by combining the corrected weight coefficients and the spectral reflectance correction value;

[0062] A model training module, which is used to construct a deep learning network, use the improved weighted spectral index as the input and the nitrogen content of corn leaves as the label to train a nitrogen content prediction model;

[0063] A comprehensive prediction module, which is used to collect the hyperspectral data, environmental temperature, soil background reflectance and light intensity of corn leaves in real time, generate an improved weighted spectral index according to the steps of the above modules and input it into the nitrogen content prediction model to generate a predicted value of the nitrogen content of corn leaves.

[0064] Compared with the prior art, the beneficial effects of the present invention are:

[0065] The present invention determines the spectral range most significantly affected by nitrogen content on corn leaves by collecting hyperspectral data of corn in the spectral range of 400 - 1000 nm, determining the basic monitoring range; dividing the spectral range into sub-bands according to a fixed interval to provide more refined spectral data, making the sensitivity analysis of different bands more accurate, finding the band most sensitive to nitrogen content changes, and selecting the five sub-bands with the highest spectral sensitivity, improving the efficiency and accuracy of subsequent analysis, and reducing unnecessary data noise by focusing on specific sensitive bands.

[0066] The present invention also adjusts the weight coefficients of different sensitive bands by generating an environmental correction coefficient, sets the environmental correction coefficient according to different bands, while considering the band differences, can effectively reduce the spectral data fluctuations caused by temperature, soil background reflectance and light intensity. The uncorrected weighted spectral index will be affected by environmental changes, resulting in inconsistent performance of the model under different environmental conditions. By introducing the environmental correction coefficient to generate an improved weighted spectral index, the model can more truly reflect the actual nitrogen content, enhance the prediction stability of the model under diverse environmental conditions, and reduce the errors caused by external factors; and training the model based on a deep learning network, in actual use, collecting the real-time spectral data and environmental data of corn leaves can achieve nitrogen content prediction, improving the applicability of the model. Description of the Drawings

[0067] Figure 1 It is a schematic flowchart of the method of the embodiment of the present invention;

[0068] Figure 2 This is a schematic diagram of the system module in an embodiment of the present invention. Detailed implementation manners

[0069] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with specific embodiments.

[0070] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should have the ordinary meanings understood by those with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar terms used in the present invention do not denote any order, quantity or importance, but are only used to distinguish different components. The terms such as "including" or "comprising" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. The terms such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left" and "right" are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0071] Embodiment:

[0072] Please refer to Figure 1 , the present invention provides a technical solution:

[0073] A method for evaluating the nitrogen content of corn leaves based on hyperspectral remote sensing, the specific steps include:

[0074] Step 1: Select multiple corn leaf samples with known nitrogen contents, collect hyperspectral data of the corn leaf samples in the spectral range of 400-1000 nm, perform atmospheric correction on the hyperspectral data to obtain the spectral reflectance correction value, divide the sub-bands, and collect the ambient temperature, soil background reflectance and light intensity to calculate the environmental correction coefficients of different sub-bands;

[0075] In this embodiment, the spectral range of 400 - 1000 nm includes blue - violet light (380 - 490 nm), green light (490 - 575 nm), yellow - orange light (575 - 620 nm), red light (620 - 750 nm), and near - infrared light (750 - 1000 nm). Changes in nitrogen content will cause significant changes in the spectral reflectance of plants in the range of 400 - 1000 nm. In the visible light range, an increase in nitrogen content leads to an increase in chlorophyll content, thus changing the absorption characteristics of blue and red light. In the near - infrared range, changes in nitrogen content will affect the cell structure and water distribution of leaves, thus changing the near - infrared reflectance. Therefore, hyperspectral data needs to be collected in the range of 400 - 1000 nm. When dividing sub - bands, starting from 400 nm, each sub - band is divided every 10 nm until 1000 nm;

[0076] The formula for generating the environmental correction coefficient is:

[0077] ;

[0078] Wherein, represents the environmental correction coefficient of the th sub - band, represents the index of the sub - band, represents the environmental temperature, represents the th standard environmental temperature of the sub - band, represents the soil background reflectance, represents the th standard soil background reflectance of the sub - band, represents the light intensity, represents the th standard light intensity of the sub - band, respectively represent the weight coefficients of environmental temperature, soil background reflectance, and light intensity, , and .

[0079] Standard environmental parameters Represents the environmental parameter values measured under standard conditions, used to correct the deviation between the current environmental parameters and the standard conditions. Different spectral ranges are affected by the environment to different degrees. The near-infrared band is more affected by the thermal radiation of the leaves. The standard environmental temperature in this band increases with the increase of wavelength. The soil has a greater impact on the red and green light bands, and the corresponding standard soil background reflectance is higher, while the standard soil background reflectance in the near-infrared band is lower; the blue and red light bands are more sensitive to photosynthesis, and the corresponding standard light intensity is higher, while a lower standard light intensity is set for the near-infrared band. Specific standard environmental parameters including standard environmental temperature, standard soil background reflectance, and standard light intensity are determined based on expert evaluation; the soil background reflectance has the most significant impact on hyperspectral data and will directly affect the received spectral signal. Therefore, the weight of the soil background reflectance is the highest. The light intensity directly affects the photosynthesis and spectral reflectance of the leaves, so the weight is the second. The environmental temperature indirectly affects the thermal radiation effect of the leaves and has a weak direct impact on the reflectance. Therefore, the weight is the lowest. Take 。

[0080] Step 2: Based on the least squares method, construct the functional relationship between the spectral reflectance and nitrogen content in each sub-band, and determine the spectral sensitivity of each sub-band based on the functional relationship.

[0081] In this embodiment, the principle for constructing the functional relationship between the spectral reflectance and nitrogen content in each sub-band is as follows:

[0082] Select the spectral reflectance at the center wavelength of each sub-band as the spectral reflectance of the sub-band. The functional relationship between the spectral reflectance and nitrogen content in each sub-band is:

[0083] ;

[0084] Among them, represents the nitrogen content in the th sub-band, represents the spectral reflectance of the th sub-band, represents the index of the sub-band, respectively represent the parameters to be fitted in the th sub-band;

[0085] Construct an error function for each corn leaf sample in different sub-bands. The error function is:

[0086] ;

[0087] Among them, represents the nitrogen content deviation of all corn leaf samples in the th sub-band, represents the index of the corn sample, and , represents the number of corn leaf samples, represents the th corn leaf sample's nitrogen content in the th sub - band, represents the th corn leaf sample's spectral reflectance in the th sub - band;

[0088] Take the partial derivatives of with respect to respectively, and set the partial derivatives to zero:

[0089] ;

[0090] ;

[0091] ;

[0092] where respectively represent the partial derivatives of with respect to ;

[0093] After rearrangement, the normal equations are obtained:

[0094] ;

[0095] ;

[0096] ;

[0097] Let , , , , , , ;

[0098] Solve the normal equations:

[0099] ;

[0100] ;

[0101] ;

[0102] Substitute the solved into the functional relationship between the spectral reflectance and the nitrogen content in each sub - band.

[0103] The principle for determining the spectral sensitivity of each sub - band is:

[0104] Based on the functional relationship between the spectral reflectance and nitrogen content in each sub - band, the spectral reflectance is differentiated:

[0105] ;

[0106] ;

[0107] Among them, represents the derivative of the nitrogen content with respect to the spectral reflectance;

[0108] Based on the mean value of all corn leaf samples, the spectral sensitivity of this sub - band is generated, and the formula used is:

[0109] ;

[0110] Among them, represents the spectral sensitivity of the

[0111] reflects the rate of change of the reflectance with respect to the nitrogen content in the th sub - band. The higher the value, the faster the rate of change of the reflectance with respect to the nitrogen content, indicating higher sensitivity. Based on the rate of change of the reflectance with respect to the nitrogen content of different corn leaf samples in the same band, the spectral sensitivities corresponding to all corn samples in different bands are generated.

[0112] The spectral sensitivity of the sub - band reflects the response degree of the spectral reflectance of this band to the change of nitrogen content in corn leaves, and is used to dynamically select the band that is most sensitive to the change of nitrogen content. The higher the spectral sensitivity of the band, the more sensitive the spectral reflectance of this band is to the change of nitrogen content, that is, a small change in nitrogen content will also cause a significant change in reflectance.

[0113] Step 3: Select the five bands with the highest spectral sensitivities as sensitive bands, determine the weight coefficients of each sensitive band, correct the weight coefficients based on the environmental correction coefficient, and generate an improved weighted spectral index by combining the corrected weight coefficients and the spectral reflectance correction values;

[0114] In this embodiment, the principle for determining the weight coefficients of each sensitive band is:

[0115] ;

[0116] Among them, represents the weight coefficient of the th sensitive band, represents the index of the sensitive band, and represents Spectral sensitivity of sensitive bands.

[0117] Based on the spectral sensitivity, the weight coefficients of different sensitive bands are determined, which directly reflect the quantitative relationship between the spectral reflectance and the nitrogen content of corn leaves. The higher the sensitivity of the band, the more obvious the influence on the nitrogen content, and the higher the weight coefficient.

[0118] The principle for generating the improved weighted spectral index is as follows:

[0119] The weight coefficient of the calibrated sensitive band is:

[0120] ;

[0121] Where represents the weight coefficient of the th calibrated sensitive band, represents the environmental correction coefficient of the th sensitive band;

[0122] Perform equal-proportion scaling on to make ;

[0123] represents the comprehensive influence weight of different sensitive bands on the spectral index under the influence of the environment. Multiply by to reflect the influence of environmental factors on the spectral sensitivity under different sensitive bands. The result of may not be 1. Therefore, after calculating , perform equal-proportion scaling to make the sum of all equal to 1. represents the environmental influence coefficient corresponding to the

[0124] band selected from

[0125] The formula for generating the improved weighted spectral index is: ;

[0126] Where represents the improved weighted spectral index of the th corn leaf sample.

[0127] The improved weighted spectral index takes into account the influence of environmental temperature, soil background reflectance, and light intensity on the nitrogen content. It is adjusted through the environmental correction coefficient and considers the spectral sensitivity of the same corn leaf sample under all sensitive bands, which can reflect the comprehensive relationship between the spectral sensitivity of all bands, environmental factors, and nitrogen content, providing a more accurate estimate of the nitrogen content.

[0128] Step 4: Construct a deep learning network, using the improved weighted spectral index as the input and the nitrogen content of corn leaves as the label to train the nitrogen content prediction model;

[0129] In this embodiment, the input value of the deep learning network is the improved weighted spectral index;

[0130] The structure of the deep learning network is as follows:

[0131] Input layer: It contains 1 node and is used to input the improved weighted spectral index of corn leaves;

[0132] The first hidden layer: It contains 64 neurons and uses ReLU as the activation function;

[0133] The second hidden layer: It contains 32 neurons and uses ReLU as the activation function;

[0134] Output layer: It contains 1 stage and is used to output the nitrogen content of corn leaves.

[0135] Step 5: Based on the hyperspectral data, environmental temperature, soil background reflectance, and light intensity of corn leaves, generate the improved weighted spectral index according to the above steps and input it into the nitrogen content prediction model to generate the predicted value of the nitrogen content of corn leaves.

[0136] Please refer to Figure 2 , the present invention also provides a corn leaf nitrogen content evaluation system based on hyperspectral remote sensing. The system is used to implement the above-mentioned corn leaf nitrogen content evaluation method based on hyperspectral remote sensing, and specifically includes:

[0137] Data acquisition module, which is used to select multiple corn leaf samples with known nitrogen content, collect the hyperspectral data of the corn leaf samples in the spectral range of 400 - 1000 nm, perform atmospheric correction on the hyperspectral data to obtain the spectral reflectance correction value, divide sub-bands, collect the environmental temperature, soil background reflectance, and light intensity to calculate the environmental correction coefficient of different sub-bands;

[0138] Spectral calculation module, which is used to construct a functional relationship between the spectral reflectance and the nitrogen content in each sub-band based on the least squares method, and determine the spectral sensitivity of each sub-band based on the functional relationship;

[0139] Spectral optimization module, which is used to select the five bands with the highest spectral sensitivity as the sensitive bands, determine the weight coefficients of each sensitive band, correct the weight coefficients based on the environmental correction coefficient, and generate the improved weighted spectral index by combining the corrected weight coefficients and the spectral reflectance correction value;

[0140] Model training module, which is used to construct a deep learning network, use the improved weighted spectral index as the input and the nitrogen content of corn leaves as the label to train the nitrogen content prediction model;

[0141] The comprehensive prediction module is used to collect the hyperspectral data, ambient temperature, soil background reflectance, and light intensity of corn leaves in real time, generate an improved weighted spectral index according to the steps of the above module, and input it into the nitrogen content prediction model to generate the predicted value of the nitrogen content of corn leaves.

[0142] All the above formulas are dimensionless and take their numerical values for calculation. The formula is a formula obtained by software simulation of a large amount of collected data to approximate the real situation as closely as possible. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0143] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0144] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units. They can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0145] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all should be covered within the protection scope of this application.

Claims

1. A method for evaluating the nitrogen content of corn leaves based on hyperspectral remote sensing, characterized in that, The specific steps include: Step 1: Select multiple corn leaf samples with known nitrogen contents, collect hyperspectral data of the corn leaf samples in the spectral range of 400 - 1000 nm, perform atmospheric correction on the hyperspectral data to obtain the corrected spectral reflectance value, divide sub - bands, and collect the environmental temperature, soil background reflectance, and light intensity to calculate the environmental correction coefficients for different sub - bands; Step 2: Based on the least - squares method, construct a functional relationship between the spectral reflectance and nitrogen content for each sub - band, and determine the spectral sensitivity of each sub - band based on the functional relationship; Step 3: Select the five bands with the highest spectral sensitivity as sensitive bands, determine the weight coefficients of each sensitive band, correct the weight coefficients based on the environmental correction coefficients, and generate an improved weighted spectral index by combining the corrected weight coefficients and the corrected spectral reflectance value; The formula for generating the improved weighted spectral index is: Among them, W j represents the improved weighted spectral index of the j-th corn leaf sample; Step 4: Construct a deep - learning network, use the improved weighted spectral index as the input and the nitrogen content of corn leaves as the label to train the nitrogen content prediction model; Step 5: Real - time collect the hyperspectral data, environmental temperature, soil background reflectance, and light intensity of corn leaves, generate the improved weighted spectral index according to the above steps and input it into the nitrogen content prediction model to generate the predicted value of the nitrogen content of corn leaves; The formula for generating the environmental correction coefficient is: Among them, EFC i represents the environmental correction coefficient of the i-th sub-band, i represents the index of the sub-band, T represents the environmental temperature, T i,0 represents the standard environmental temperature of the i-th sub-band, S represents the soil background reflectance, S i,0 represents the standard soil background reflectance of the i-th sub-band, L represents the light intensity, L i,0 represents the standard light intensity of the i-th sub-band, α, β, and γ respectively represent the weight coefficients of environmental temperature, soil background reflectance, and light intensity, α + β + γ = 1, and β > γ > α; Based on the mean of all corn leaf samples, the formula for generating the spectral sensitivity of this sub - band is: Among them, Q i represents the spectral sensitivity of the i-th sub-band, and a i , b i respectively represent the parameters to be fitted for the i-th sub-band. R i,j represents the spectral reflectance of the j-th corn leaf sample in the i-th sub-band; The principle for determining the weight coefficients of each sensitive band is: Among them, w n represents the weight coefficient of the nth sensitive band, n represents the index of the sensitive band, and n ∈ [1, 5], Q n represents the spectral sensitivity of the nth sensitive band.

2. The method for evaluating the nitrogen content of corn leaves based on hyperspectral remote sensing according to claim 1, wherein: The principle for constructing the functional relationship between the spectral reflectance and nitrogen content for each sub - band in step 2 is: Select the spectral reflectance at the central wavelength of each sub - band as the spectral reflectance of the sub - band. The functional relationship between the spectral reflectance and nitrogen content for each sub - band is: Among them, N i represents the nitrogen content in the i-th sub-band, and R i represents the spectral reflectance of the i-th sub-band. i represents the index of the sub-band, and a i , b i , c i respectively represent the parameters to be fitted for the i-th sub-band; Construct an error function for each corn leaf sample in different sub - bands. The error function is: Among them, E i represents the nitrogen content deviation of all corn leaf samples in the i-th sub-band, j represents the index of the corn sample, and j ∈ [1, J], where J represents the number of corn leaf samples, N i,j represents the nitrogen content of the j-th corn leaf sample in the i-th sub-band, R i,j represents the spectral reflectance of the j-th corn leaf sample in the i-th sub-band; For E i Take the partial derivatives with respect to a i , b i , c i respectively, and set the partial derivatives equal to zero: Among them, respectively represent E i partial derivatives with respect to a i , b i , c i ; Arrange to obtain the normal equations: Let Solve the normal equations: Substitute the obtained a i , b i , c i into the functional relationship between the spectral reflectance and nitrogen content under each sub-band.

3. The method for evaluating nitrogen content in corn leaves based on hyperspectral remote sensing according to claim 2, wherein: The principle for determining the spectral sensitivity of each sub - band in step 2 is: Based on the functional relationship between the spectral reflectance and nitrogen content for each sub - band, take the derivative of the spectral reflectance: Among them, represents the derivative of the spectral reflectance with respect to the nitrogen content.

4. A method for evaluating the nitrogen content of corn leaves based on hyperspectral remote sensing according to claim 1, characterized in that: The principle for generating the improved weighted spectral index in step 3 is: The weight coefficient of the corrected sensitive band is: w′ n = EFC n × w n where, w′ n represents the weight coefficient of the nth sensitive band after calibration, and EFC n represents the environmental correction coefficient of the nth sensitive band; Scale w' n proportionally so that 5. A corn leaf nitrogen content evaluation system based on hyperspectral remote sensing, which is used to implement the corn leaf nitrogen content evaluation method based on hyperspectral remote sensing according to any one of claims 1 - 4. Specifically, it includes: A data acquisition module, which is used to select multiple corn leaf samples with known nitrogen contents, collect hyperspectral data of the corn leaf samples in the spectral range of 400 - 1000 nm, perform atmospheric correction on the hyperspectral data to obtain the corrected spectral reflectance value, divide sub - bands, and collect the environmental temperature, soil background reflectance, and light intensity to calculate the environmental correction coefficients for different sub - bands; A spectral calculation module, which is used to construct a functional relationship between the spectral reflectance and nitrogen content for each sub - band based on the least - squares method, and determine the spectral sensitivity of each sub - band based on the functional relationship; A spectral optimization module, which is used to select the five bands with the highest spectral sensitivity as sensitive bands, determine the weight coefficients of each sensitive band, correct the weight coefficients based on the environmental correction coefficient, and generate an improved weighted spectral index by combining the corrected weight coefficients and the spectral reflectance correction value; A model training module, which is used to construct a deep learning network, use the improved weighted spectral index as the input and the nitrogen content of corn leaves as the label to train the nitrogen content prediction model; A comprehensive prediction module, which is used to collect the hyperspectral data, environmental temperature, soil background reflectance and light intensity of corn leaves in real time, generate the improved weighted spectral index according to the steps of the above modules and input it into the nitrogen content prediction model to generate the predicted value of the nitrogen content of corn leaves.

Images