A method and system for constructing a monitoring model for early wheat diseases

Abstract

This invention provides a method and system for constructing an early wheat disease monitoring model, relating to the field of remote sensing monitoring technology for crop diseases. The method includes: acquiring spectral reflectance data of wheat ear samples and the severity of disease based on the spectral reflectance data; dividing the spectral reflectance data into a source domain dataset and a target domain dataset according to the severity of disease; wherein the source domain dataset includes first spectral reflectance data corresponding to healthy wheat and the first degree of disease, and the target domain dataset includes second spectral reflectance data corresponding to healthy wheat and the second degree of disease; and filtering the sensitive band ranges for both the source and target domain datasets. The model construction method and system provided by this invention, through feature transfer and intelligent dynamic parameter tuning, can improve the accuracy of early disease monitoring models to an ideal level.

Description

Technical Field

[0001] This invention relates to the field of remote sensing monitoring technology for crop diseases, and in particular to a method and system for constructing a monitoring model for early wheat diseases. Background Technology

[0002] In the later stages of wheat scab, when the infection is severe, the symptoms are particularly pronounced. Therefore, researchers typically conduct monitoring studies during a period of wheat scab outbreak (usually the late grain-filling stage) to capture more significant specific characteristics and improve the accuracy of monitoring models. However, in reality, once a severe outbreak of wheat scab occurs, it is very difficult to control and can cause irreversible and serious damage to the grains.

[0003] However, research on early monitoring of wheat scab is still relatively lacking. Although current disease monitoring models and methods have made some progress, their application in early wheat scab monitoring is limited due to weak generalization ability and low transferability. For example, when there is a large difference in the data distribution between the source and target domains, the model may exhibit negative transfer, leading to a significant decrease in the performance of early scab prediction. Summary of the Invention

[0004] In view of the shortcomings of the prior art described above, the purpose of this invention is to provide a method and system for constructing a monitoring model for early wheat diseases, in order to solve the problem that the model may exhibit negative migration due to the large difference in data distribution between the source domain and the target domain, which leads to a significant decrease in the performance of early wheat scab prediction.

[0005] To achieve the above and other related objectives, this invention provides a method for constructing a wheat early disease monitoring model, comprising: acquiring spectral reflectance data of wheat ear samples and the disease severity of the spectral reflectance data; dividing the spectral reflectance data into a source domain dataset and a target domain dataset according to the disease severity; wherein, the source domain dataset includes first spectral reflectance data corresponding to healthy conditions and the first degree of disease, and the target domain dataset includes second spectral reflectance data corresponding to healthy conditions and the second degree of disease; filtering the sensitive wavelength range of the source domain dataset and the target domain dataset respectively to obtain the first sensitive wavelength reflectance corresponding to the source domain dataset and the second sensitive wavelength reflectance corresponding to the target domain dataset; and constructing a source domain disease monitoring model based on the first sensitive wavelength reflectance, the second sensitive wavelength reflectance, and the disease severity label.

[0006] In one embodiment of the present invention, sensitive band ranges are filtered for source domain datasets and target domain datasets respectively to obtain a first sensitive wavelength reflectance corresponding to the source domain dataset and a second sensitive wavelength reflectance corresponding to the target domain dataset. This includes: establishing a first partial least squares regression model based on the source domain dataset with a set band range; predicting a first disease severity value using the first partial least squares regression model based on the first spectral reflectance data; obtaining a first maximum determination coefficient based on the actual value and predicted value of the first disease severity corresponding to the first spectral reflectance data; establishing a second partial least squares regression model based on the target domain dataset with a set band range; predicting a second disease severity value using the second partial least squares regression model based on the second spectral reflectance data; obtaining a second maximum determination coefficient based on the actual value and predicted value of the second disease severity corresponding to the second spectral reflectance data; and filtering the sensitive band ranges for the source domain dataset and target domain dataset respectively based on the first and second maximum determination coefficients to obtain the first and second sensitive wavelength reflectances.

[0007] In one embodiment of the present invention, sensitive band ranges are filtered for the source domain dataset and the target domain dataset according to a first maximum determination coefficient and a second maximum determination coefficient to obtain a first sensitive wavelength reflectance and a second sensitive wavelength reflectance. This includes: filtering the sensitive band range within a set band range according to the first maximum determination coefficient and the second maximum determination coefficient to obtain a target sensitive band range; wherein the target sensitive band range is the same band range for both the source domain dataset and the target domain dataset; and selecting the sensitive wavelength reflectance based on the target sensitive band range to obtain the first sensitive wavelength reflectance and the second sensitive wavelength reflectance.

[0008] In one embodiment of the present invention, a source domain disease monitoring model is constructed based on a first sensitive wavelength reflectance, a second sensitive wavelength reflectance, and a disease severity label. The method includes: standardizing the first sensitive wavelength reflectance and the second sensitive wavelength reflectance; mapping the standardized first sensitive wavelength reflectance and the second sensitive wavelength reflectance to the same high-dimensional regenerative kernel Hilbert space for feature space alignment; and obtaining a target source domain dataset after feature space alignment based on the disease severity label; and optimizing the regularization parameters of the regression model using the target source domain dataset to construct the source domain disease monitoring model.

[0009] In one embodiment of the present invention, the standardization processing of the first sensitive wavelength reflectance and the second sensitive wavelength reflectance includes: standardizing the first sensitive wavelength reflectance using the z-score method based on the first mean and the first standard deviation of all first sensitive wavelength reflectances to obtain the standardized first sensitive wavelength reflectance; and standardizing the second sensitive wavelength reflectance using the z-score method based on the second mean and the second standard deviation of all second sensitive wavelength reflectances to obtain the standardized second sensitive wavelength reflectance.

[0010] In one embodiment of the present invention, the standardized first and second sensitive wavelength reflectances are mapped to the same high-dimensional regenerative kernel Hilbert space for feature space alignment, and a target source domain dataset after feature space alignment is obtained based on the disease severity label. This includes: merging the standardized first and second sensitive wavelength reflectances row-wise to obtain a joint matrix; obtaining the dimension information of the joint matrix; obtaining a symmetric matrix based on the dimension information and preset weight coefficients; obtaining a centered matrix based on the dimension information, an identity matrix, and a matrix of all ones; obtaining a kernel matrix based on the standardized first and second sensitive wavelength reflectances, a preset linear kernel function type, and matrix transpose; obtaining an optimal projection matrix based on the symmetric matrix, kernel matrix, centered matrix, identity matrix, and regularization parameters; obtaining a target joint matrix mapped from the joint matrix to the regenerative kernel Hilbert space based on the optimal projection matrix, matrix transpose, and kernel matrix; and reconstructing the target joint matrix by segmenting and reconstructing the source and target domains based on the disease severity label to obtain a target source domain dataset after feature space alignment.

[0011] In one embodiment of the present invention, the regularization parameters of a regression model are optimized using a target source domain dataset to construct a source domain disease monitoring model. This includes: obtaining a regression model based on the feature matrix, disease severity label, and error of the source domain training data in the target source domain dataset; wherein the regression model includes a regression parameter vector to be solved; obtaining a loss function of the regression model based on the regression model, the L2 regularization term, and the number of samples in the source domain training data; wherein the loss function includes regularization parameters; and dynamically adjusting the regularization parameters based on the initial search range of the regularization parameters, the random number generation function, and the energy factor to construct the source domain disease monitoring model.

[0012] In one embodiment of the present invention, the regularization parameter is dynamically adjusted based on the initial search range of the regularization parameter, the random number generation function, and the energy factor to construct a source domain disease monitoring model. This includes: generating initial candidate values ​​for the regularization parameter based on the initial search range; obtaining a regression model loss value based on the initial candidate values ​​and a loss function; when the energy factor is greater than a first set value and the generated value of the random number generation function is less than a second set value, performing a global migration perturbation exploration based on a preset number of iterations to obtain the first regularization parameter; and when the energy factor is greater than the first set value and the generated value of the random number generation function is greater than or equal to the second set value, performing a local fine-grained search based on a preset number of iterations. To obtain a second regularization parameter; when the energy factor is less than a first set value and the value generated by the random number generation function is less than a third set value, a spiral path density search is performed according to a preset number of iterations to obtain a third regularization parameter; when the energy factor is less than a first set value and the value generated by the random number generation function is greater than or equal to the third set value, a Levy flight hybrid search is performed according to a preset number of iterations to obtain a fourth regularization parameter; based on the first, second, third, and fourth regularization parameters and the loss function of the regression model, a target regularization parameter corresponding to the minimum regression model loss value under a preset number of iterations is obtained; based on the target regularization parameter, a source domain disease monitoring model is constructed.

[0013] In one embodiment of the present invention, the method further includes: inputting source domain validation data from the target source domain dataset into a source domain disease monitoring model to obtain model prediction results; fitting the model prediction results with disease severity labels to obtain a first coefficient of determination; inputting source domain validation data from the target source domain dataset into a comparison model to obtain a second coefficient of determination; wherein the comparison model includes a random forest model and a support vector regression model; comparing the first coefficient of determination and the second coefficient of determination; when the first coefficient of determination is greater than the second coefficient of determination, the source domain disease monitoring model is valid; when the first coefficient of determination is less than the second coefficient of determination, the source domain disease monitoring model is invalid.

[0014] To achieve the above and other related objectives, the present invention also provides a wheat early disease monitoring model construction system, comprising: an acquisition unit for acquiring spectral reflectance data of wheat ear samples and the disease severity of the spectral reflectance data; a division unit for dividing the spectral reflectance data into a source domain dataset and a target domain dataset according to the disease severity; wherein the source domain dataset includes first spectral reflectance data corresponding to healthy conditions and the first degree of disease, and the target domain dataset includes second spectral reflectance data corresponding to healthy conditions and the second degree of disease; a selection unit for filtering the sensitive band range of the source domain dataset and the target domain dataset respectively to obtain the first sensitive wavelength reflectance corresponding to the source domain dataset and the second sensitive wavelength reflectance corresponding to the target domain dataset; and a modeling unit for constructing a source domain disease monitoring model based on the first sensitive wavelength reflectance, the second sensitive wavelength reflectance, and the disease severity label.

[0015] To achieve the above and other related objectives, the present invention provides an electronic device comprising: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the electronic device enables the aforementioned method for constructing a wheat early disease monitoring model.

[0016] To achieve the above and other related objectives, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a computer processor, causes the computer to execute the aforementioned method for constructing a monitoring model for early wheat diseases.

[0017] As described above, the wheat early disease monitoring model construction method and system of the present invention has the following beneficial effects: By dividing spectral reflectance data into source domain datasets and target domain datasets according to disease severity, and then selecting sensitive wavelength reflectance, the TCA method is used to map the features of source domain and target domain samples with different disease severity to a high-dimensional space, achieving alignment of feature distributions between the source domain and target domain. Simultaneously, the ALA algorithm is used to achieve intelligent parameter optimization, and the Ridge regression algorithm is combined to construct the source domain disease monitoring model. This method integrates the domain adaptability of TCA, the adaptive optimization of ALA, and the regularization advantages of Ridge regression, making it suitable for early wheat scab monitoring modeling scenarios. It improves the early disease monitoring capability of the source domain disease monitoring model and solves the technical problems of poor model generalization ability, low transferability, and low accuracy in early wheat scab monitoring scenarios using traditional methods. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the method for constructing a wheat early disease monitoring model according to an embodiment of the present invention.

[0019] Figure 2The image shown is a health sample diagram provided in an embodiment of the present invention.

[0020] Figure 3 The graph shown represents the relationship between the number of healthy samples and the severity of the disease, as provided in an embodiment of the present invention.

[0021] Figure 4 The image shown is a mild sample diagram provided in an embodiment of the present invention.

[0022] Figure 5 The graph shown represents the relationship between the number of mild cases and the severity of the disease, as provided in an embodiment of the present invention.

[0023] Figure 6 The image shown is a heavy sample image provided in an embodiment of the present invention.

[0024] Figure 7 The graph shown is a curve illustrating the relationship between the number of severe samples and the severity of the disease, provided in an embodiment of the present invention.

[0025] Figure 8 The figure shown is a PLSR fitting result of wavelength spectral reflectance and disease severity in any band of source domain 350-1050nm provided by an embodiment of the present invention.

[0026] Figure 9 The figure shown is a PLSR fitting result of wavelength spectral reflectance and disease severity in any band range of 350-1050nm provided by an embodiment of the present invention.

[0027] Figure 10 The diagram shown is a VIP weight distribution diagram of the source-domain sensitive wavelength provided in an embodiment of the present invention.

[0028] Figure 11 The diagram shown is a VIP weight distribution diagram of the target domain sensitive wavelength provided in an embodiment of the present invention.

[0029] Figure 12 The image shown is a source domain monitoring result based on the TCA-ALA-Ridge regression model provided in an embodiment of the present invention.

[0030] Figure 13 The image shown is a source domain monitoring result based on an SVR regression model provided in an embodiment of the present invention.

[0031] Figure 14 The image shown is a source domain monitoring result based on an RF regression model provided in an embodiment of the present invention.

[0032] Figure 15 The diagram shown is a structural block diagram of a wheat early disease monitoring system provided in an embodiment of the present invention.

[0033] Figure 16 The diagram shown is a structural schematic of an electronic device according to an embodiment of the present invention.

[0034] Component designation explanation

[0035] Electronic device 1; Wheat early disease monitoring model construction system 11; Memory 12; Processor 13; Acquisition unit 111; Division unit 112; Selection unit 113; Modeling unit 114. Detailed Implementation

[0036] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0037] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0038] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0039] Please see Figure 1This invention provides a method for constructing an early wheat disease monitoring model. By utilizing the spectral reflectance data of wheat ear samples and the severity of disease based on the spectral reflectance data, a source domain dataset and a target domain dataset can be divided according to the severity of the disease. This improves the monitoring accuracy of the source domain dataset to the level of the target domain dataset. After dividing the source and target domain datasets, the sensitive band range is further filtered to obtain the first sensitive wavelength reflectance corresponding to the source domain dataset and the second sensitive wavelength reflectance corresponding to the target domain dataset. The model is then trained based on the first sensitive wavelength reflectance of the source domain dataset and the second sensitive wavelength reflectance of the target domain dataset, thereby constructing a source domain disease monitoring model. This improves the monitoring accuracy of the source domain disease monitoring model to the level of the target domain disease monitoring model, thus raising the accuracy of the early disease monitoring model to an ideal level. This solves the technical problems of poor model generalization ability, low transferability, and low accuracy in the early wheat scab monitoring scenario of traditional methods.

[0040] Figure 1 A flowchart illustrating a method for constructing a wheat early disease monitoring model according to an exemplary embodiment of this application is shown. This method is applied to a wheat early disease monitoring model construction system and includes steps S10-S40. The following will be combined with… Figure 1 The technical solution of this application will be described in detail below. It is worth noting that the reflectivity of the sensitive wavelength can also be directly referred to as the sensitive wavelength in the following text.

[0041] First, perform step S10 to obtain the spectral reflectance data of the wheat ear samples and the severity of the disease based on the spectral reflectance data.

[0042] Before acquiring the spectral reflectance data and disease severity of the wheat ear samples, disease severity data at different infection levels are first obtained to calculate and statistically analyze the disease severity of each sample. Specifically, the disease severity of each sample corresponds to each wheat ear. Then, the proportion of infected grains on each wheat ear to the total number of grains is used to determine the severity. The disease severity can be set to a range of 0-100%, with a smaller ratio indicating a milder infection. Simultaneously, the spectral reflectance data of each wheat ear sample is measured using a spectral data measurement system, such as a SOC710E imaging spectrometer (Surface Optics Corporation, San Diego, USA), in a completely sealed black box environment to reduce interference from external light during data acquisition. Furthermore, during the measurement of spectral reflectance data, parameters such as the height, focal length, and exposure of the spectrometer can be adjusted to capture hyperspectral imaging data of all samples with an exposure time of 18ms. A thresholding method is used to segment the wheat ear samples from the background, and morphological details are eliminated. The average spectral reflectance of all pixels within the Region of Interest (ROI) of each sample is used as the final spectral data for each sample, i.e., the spectral reflectance data for each wheat ear sample. Then, the spectral reflectance data and disease severity of the corresponding wheat ear samples are obtained through a wheat early disease monitoring model construction system.

[0043] Next, step S20 is executed, dividing the spectral reflectance data into a source domain dataset and a target domain dataset according to the severity of the disease; wherein, the source domain dataset includes the first spectral reflectance data corresponding to health and the first degree of disease, and the target domain dataset includes the second spectral reflectance data corresponding to health and the second degree of disease.

[0044] The wheat early disease monitoring model construction system primarily uses spectral reflectance data as a sample feature and the corresponding disease severity as a label when dividing the source and target domain datasets. This divides the spectral reflectance data into source and target domain datasets based on disease severity. Furthermore, to ensure the monitoring accuracy of the source domain dataset is improved to match that of the target domain dataset, the disease severity label for the source domain dataset can include both "healthy" and "first degree of infection," while the disease severity label for the target domain dataset can include both "healthy" and "second degree of infection." This results in the source domain dataset containing the first spectral reflectance data corresponding to both "healthy" and "first degree of infection," and the target domain dataset containing the second spectral reflectance data corresponding to both "healthy" and "second degree of infection." Figure 2-7 The given samples and disease severity distribution under different infection levels show that, Figure 2-3The wheat sample data obtained were healthy samples (infection rate was 0); Figure 4-5 The wheat sample data obtained were mild samples (0 < infection rate < 15%). Figure 6-7 The wheat ear sample data obtained were severe samples (infection rate >15%). By using the above-mentioned classification method, the monitoring accuracy of the source domain dataset can be effectively improved to the level of the target domain dataset, thus providing high-precision technical guidance for early monitoring of wheat scab.

[0045] Next, step S30 is executed to filter the sensitive band ranges of the source domain dataset and the target domain dataset respectively, so as to obtain the first sensitive wavelength reflectance corresponding to the source domain dataset and the second sensitive wavelength reflectance corresponding to the target domain dataset.

[0046] After the source domain dataset and target domain dataset are divided, the sensitive band range is further filtered by the wheat early disease monitoring model construction system, so as to obtain the first sensitive wavelength reflectance corresponding to the source domain dataset and the second sensitive wavelength reflectance corresponding to the target domain dataset.

[0047] In step S30, the sensitive band ranges of the source domain dataset and the target domain dataset are filtered respectively to obtain the first sensitive wavelength reflectance corresponding to the source domain dataset and the second sensitive wavelength reflectance corresponding to the target domain dataset. This may further include:

[0048] Based on the source domain dataset with a defined band range, a first partial least squares regression model is established.

[0049] Based on the first spectral reflectance data, the predicted value of the first disease severity is obtained by the first partial least squares regression model.

[0050] The first maximum determination coefficient is obtained based on the actual value and predicted value of the first disease severity corresponding to the first spectral reflectance data.

[0051] Based on the target domain dataset with a defined band range, a second partial least squares regression model is established.

[0052] Based on the second spectral reflectance data, the predicted value of the second disease severity is obtained by using the second partial least squares regression model.

[0053] The second maximum coefficient of determination is obtained based on the actual value of the second disease severity and the predicted value of the second disease severity corresponding to the second spectral reflectance data.

[0054] Based on the first and second maximum determination coefficients, the sensitive band ranges of the source domain dataset and the target domain dataset are filtered to obtain the first and second sensitive wavelength reflectances.

[0055] In the process of extracting the first and second sensitive wavelength reflectances through the wheat early disease monitoring model, a first partial least squares regression model and a second partial least squares regression model are established based on a set wavelength range, the source domain dataset, and the target domain dataset within that set wavelength range; of course, other wavelength ranges can also be used depending on the specific situation. Then, based on the first spectral reflectance data from the source domain dataset, the first partial least squares regression model can predict the first disease severity value. Then, based on the second spectral reflectance data from the target domain dataset, the second partial least squares regression model can predict the second disease severity value. Finally, based on the actual first disease severity value corresponding to the first spectral reflectance data (i.e., the disease severity label and the predicted first disease severity value), the first maximum coefficient of determination can be determined. And based on the actual second disease severity value corresponding to the second spectral reflectance data (i.e., the disease severity label and the predicted second disease severity value), the second maximum coefficient of determination can be determined. Then, based on the first and second maximum coefficients of determination, the sensitive band ranges of the source and target domain datasets are determined. These sensitive band ranges can be the same in both datasets. Based on these same sensitive band ranges, the corresponding first sensitive wavelength reflectance in the source dataset and the second sensitive wavelength reflectance in the target dataset can be determined. This ensures that the sensitive wavelength reflectances in the source and target datasets can effectively reflect the physiological differences in wheat, providing biochemically interpretable spectral characteristic parameters for the subsequent construction of the TCA-ALA-Ridge model (i.e., the source domain disease monitoring model).

[0056] Specifically, in the process of feature extraction of the first and second sensitive wavelength reflectance, for example, a partial least squares regression (PLSR) model can be established within the 350-1050nm wavelength range to correlate spectral reflectance data with the degree of disease verification. This could include a first partial least squares regression model corresponding to the source domain dataset and a second partial least squares regression model corresponding to the target domain dataset. The PLSR algorithm corresponding to the partial least squares regression (PLSR) model can handle the multicollinearity problem between wavelength reflectances of samples at different wavelengths. It projects the original variables into a new space through projection, thereby generating new, orthogonal but uncorrelated latent variables. The partial least squares regression (PLSR) model (which can be either the first or second partial least squares regression model) can be expressed as:

[0057] X = T × P t +E

[0058] Y = T × Q t +F

[0059] Where X and Y represent the wavelength reflectance of the sample and the corresponding severity of the disease, respectively. The sample includes samples from the source domain dataset or samples from the target domain dataset. E and F represent the projection residuals of X and Y, respectively. T represents the common latent variable matrix connecting X and Y, which is essentially a low-dimensional comprehensive feature matrix extracted from the high-dimensional spectral data X and directly associated with the prediction target Y. t represents a single latent variable (a single column vector of T). P and Q represent the weight matrices of X and Y, respectively, used to reveal the contribution of each band to the latent variable and to establish the quantitative relationship between the latent variable and the severity of the disease.

[0060] Then, the maximum coefficient of determination R is determined using the formula for calculating the maximum coefficient of determination. 2 :

[0061]

[0062] Among them, y i For each sample, the actual severity value of the disease is given. The samples include those from the source domain dataset or the target domain dataset. i This is the predicted severity value of the disease after processing with the PLSR algorithm, which is the predicted target Y obtained after processing with the PLSR algorithm. The maximum coefficient of determination R is the actual average severity of the disease across all samples, where n is the sample size. 2 This includes the first and second largest coefficients of determination.

[0063] Furthermore, based on the first and second maximum determination coefficients, the source domain dataset and the target domain dataset are respectively screened for sensitive band ranges to obtain the first and second sensitive wavelength reflectances, which may include:

[0064] Based on the first and second maximum determination coefficients, sensitive band ranges are filtered within a set band range to obtain the target sensitive band range; wherein, the target sensitive band range is the same band range of the source domain dataset and the target domain dataset;

[0065] The reflectivity of the sensitive wavelength is selected based on the target sensitive band range to obtain the first sensitive wavelength reflectivity and the second sensitive wavelength reflectivity.

[0066] After obtaining the first maximum coefficient of determination for the source domain dataset and the second maximum coefficient of determination for the target domain dataset, the wheat early disease monitoring model construction system can filter the sensitive band range according to the source and target domain datasets to obtain the same target sensitive band range for both datasets. Then, using this target sensitive band range, sensitive wavelength reflectance is selected in both the source and target domain datasets to obtain the first and second sensitive wavelength reflectances.

[0067] For example, when selecting the sensitive band range for the source and target domains, the source and target domains can be the sensitive wavelengths and fitting results shown in the table below:

[0068]

[0069] As shown in the table above and Figure 8 and Figure 9 The provided example shows the PLSR fitting results for wavelength spectral reflectance and disease severity within any wavelength range of 350-1050nm in both the source and target domains. It can be seen that within the set wavelength range of 350-1050nm, the PLSR fitting results for spectral reflectance and disease severity of the source and target domain samples are displayed. Furthermore, overall, the maximum coefficient of determination R for the target domain is... 2 The accuracy is higher than that of the source domain, and the maximum coefficient of determination R is higher in most bands. 2 The accuracy exceeds 0.8, and the source domain accuracy exceeding 0.8 is mainly concentrated in the 400-700nm and 700-1000nm ranges. The maximum coefficient of determination R in the source domain is... 2 The highest accuracy reached 0.83, corresponding to a sensitive wavelength range of 648-681nm, and the maximum coefficient of determination R in the target domain was [value missing]. 2 The highest accuracy reached 0.88, corresponding to a sensitive wavelength range of 648-681 nm. Therefore, the maximum coefficient of determination R between the source and target domains was [value missing]. 2 The sensitive band range corresponding to the accuracy can be selected as 648-681nm. Within the sensitive band range, the first 14 sensitive wavelengths can be selected as subsequent model input variables, namely the first sensitive wavelength reflectance and the second sensitive wavelength reflectance.

[0070] Furthermore, after determining the reflectance at the first and second sensitive wavelengths, the relative contribution of each wavelength to the PLSR model can be further evaluated using variable importance in projection (VIP). Of course, other methods can also be used to assess contributions, such as significance multivariate correlation (SMC), selectivity ratio (SR), and regression coefficients (RC). Figure 10 and Figure 11 The provided example of VIP weight distribution for sensitive wavelengths in the source and target domains shows a significant correlation (p<0.01) in the spectral characteristic distributions of the top 14 key sensitive wavelengths in both domains through Importance Projection (VIP) analysis. Within the 648-681 nm spectral range, the VIP scores for both domains exceed 0.25, with the relative contributions at wavelengths of 655 and 648 nm reaching peak values ​​(source domain VIP = 0.66, target domain VIP = 0.67). This wavelength range closely matches the chlorophyll a / b absorption peak (640-660 nm) in plants, indicating that the extracted sensitive wavelengths can effectively characterize physiological differences in wheat and provide biochemically explanatory spectral characteristic parameters for the subsequent construction of the TCA-ALA-Ridge model.

[0071] Next, step S40 is executed: a source domain disease monitoring model is constructed based on the reflectance of the first sensitive wavelength, the reflectance of the second sensitive wavelength, and the disease severity label.

[0072] After determining the first sensitive wavelength reflectance of the source domain dataset and the second sensitive wavelength reflectance of the target domain dataset, a disease severity label can be combined to construct a disease monitoring model for the source domain. The constructed disease monitoring model can accurately predict the severity of early-stage diseases.

[0073] In step S40, a source-domain disease monitoring model is constructed based on the reflectance of the first sensitive wavelength, the reflectance of the second sensitive wavelength, and the disease severity label, including:

[0074] The reflectance at the first and second sensitive wavelengths is standardized.

[0075] The standardized first and second sensitive wavelength reflectances are mapped to the same high-dimensional regenerative kernel Hilbert space for feature space alignment, and the target source domain dataset after feature space alignment is obtained according to the disease severity label.

[0076] The regularization parameters of the regression model are optimized using the target source domain dataset to construct a source domain disease monitoring model.

[0077] In constructing the source-domain disease monitoring model, the reflectance of the first and second sensitive wavelengths is first standardized to ensure the data are on the same scale. Then, the standardized first and second sensitive wavelength reflectances are mapped to the same high-dimensional regenerative kernel Hilbert space for feature space alignment to reduce the distributional differences between them. After feature space alignment, the target source-domain dataset is obtained by combining disease severity labels. Subsequently, the regularization parameters of the regression model are optimized during model training using the target source-domain dataset to achieve the construction of the source-domain disease monitoring model.

[0078] The standardization process for the reflectivity of the first and second sensitive wavelengths includes:

[0079] Based on the first mean and the first standard deviation of the reflectance of all first sensitive wavelengths, the reflectance of the first sensitive wavelengths is standardized by the z-score method to obtain the standardized reflectance of the first sensitive wavelengths.

[0080] Based on the second mean and the second standard deviation of the reflectance of all second sensitive wavelengths, the reflectance of the second sensitive wavelengths is standardized using the z-score method to obtain the standardized reflectance of the second sensitive wavelengths.

[0081] In the preprocessing of standardizing the reflectance of the first and second sensitive wavelengths, the first mean and first standard deviation of the first sensitive wavelength reflectance, and the second mean and second standard deviation of the second sensitive wavelength reflectance, can be used to standardize the first and second sensitive wavelength reflectances respectively using the z-score method. This yields standardized first and second sensitive wavelength reflectances. Standardizing the sensitive wavelengths ensures that the first and second sensitive wavelength reflectances are on the same scale. The z-score method is a fundamental and powerful data standardization tool; its core value lies in eliminating incomparability between data and providing a unified benchmark for subsequent analysis. Of course, other methods can also be used to standardize the first and second sensitive wavelength reflectances.

[0082] Specifically, in the data standardization preprocessing, the source domain dataset and the target domain dataset can be defined first, and can be represented as follows:

[0083] data1 = [x s1 ,x s2 ,x s3 ,…x sns ,y1]

[0084] data2 = [x t1 ,x t2 ,x t3 ,…x tnt ,y2]

[0085] Where data1 represents the source domain dataset, data2 represents the target domain dataset, and X1 = [x s1 ,x s2 ,x s3 ,…x sns ]∈R ns×d Representing the source-domain sensitive wavelength, X2 = [x t1 ,x t2 ,x t3 ,…x tnt ]∈R nt×d y1 represents the target domain sensitive wavelength, and y2 represents the disease severity labels in the source domain and target domain, respectively.

[0086] Then, the source domain sensitive wavelength X1 and target domain sensitive wavelength X2 data can be read separately using software such as MATLAB, and the data can be standardized using the z-score method. The corresponding formula for standardization can be expressed as:

[0087]

[0088] Among them, X src and X tar σ1 and σ2 represent the sensitive wavelength data after z-score processing, namely the first and second sensitive wavelength reflectances after standardization. μ1 and μ2 represent the mean of the original sensitive wavelength reflectances of all samples of X1 and X2, respectively. σ1 and σ2 represent the standard deviation of the original sensitive wavelength reflectances of all samples of X1 and X2, respectively.

[0089] In addition, the standardized first and second sensitive wavelength reflectances are mapped to the same high-dimensional regenerative kernel Hilbert space for feature space alignment. Based on the disease severity label, the target source domain dataset after feature space alignment is obtained, including:

[0090] The standardized reflectance of the first and second sensitive wavelengths are merged row by row to obtain the joint matrix;

[0091] Based on the joint matrix, obtain the dimension information of the joint matrix;

[0092] Based on the dimensional information and preset weight coefficients, obtain a symmetric matrix;

[0093] Obtain the centered matrix based on the dimension information, the identity matrix, and the matrix of all ones;

[0094] The kernel matrix is ​​obtained based on the standardized reflectance of the first and second sensitive wavelengths, the preset linear kernel function type, and the matrix transpose.

[0095] The optimal projection matrix is ​​obtained based on the symmetric matrix, kernel matrix, centering matrix, identity matrix, and regularization parameter.

[0096] Based on the optimal projection matrix, matrix transpose, and kernel matrix, obtain the target joint matrix that maps the joint matrix to the reproducing kernel Hilbert space;

[0097] Based on the severity labels, the target joint matrix is ​​segmented and reconstructed into source and target domains to obtain a target-source domain dataset aligned with the feature space.

[0098] In the process of aligning the standardized first and second sensitive wavelength reflectances in the high-dimensional regenerative kernel Hilbert space and establishing the feature-space aligned target source domain dataset, the standardized first and second sensitive wavelength reflectances are first merged row-wise to obtain a joint matrix. Then, based on the dimension information, preset weight coefficients, identity matrix, all-one matrix, preset linear kernel function type, matrix transpose identity matrix, and regularization parameters of the joint matrix, the symmetric matrix, centered matrix, kernel matrix, and optimal projection matrix can be obtained respectively. Using the optimal projection matrix, matrix transpose, and kernel matrix, the target joint matrix mapped to the regenerative kernel Hilbert space can be obtained. Then, the target joint matrix is ​​segmented and reconstructed into source and target domains. Combined with the disease severity label, the feature-space aligned target source domain dataset can be obtained, thus realizing the feature-space alignment of the standardized first and second sensitive wavelength reflectances to reduce the distribution difference between the first and second sensitive wavelength reflectances.

[0099] Specifically, the normalized reflectance X of the first sensitive wavelength can be obtained by using the Transfer Component Analysis (TCA) method. src Second sensitive wavelength reflectivity X tarMapping to a common high-dimensional subspace, such as the regenerating kernel Hilbert space, reduces the distributional differences between the two, providing a more consistent feature space for model construction. Furthermore, when performing preprocessing to normalize the reflectance of the first and second sensitive wavelengths, improved TCA methods (such as TCADS), Joint Distribution Adaptation (JDA), and Deep Domain Adaptation Networks (DDAN) can be used as alternatives.

[0100] When processing using the Transfer Component Analysis (TCA) method, the standardized reflectance of the first and second sensitive wavelengths is first merged row-wise to obtain a joint matrix. Specifically, the TCA parameters are first set: the kernel function type is selected as "linear," the regularization parameter is set to 1.0, and the kernel parameter is set to 2.0. X... src and X tar The rows are merged to form a new joint matrix X. Based on the new joint matrix X, the dimensional information of X is calculated, expressed by the formula:

[0101] X = [X] src ;X tar ]∈R n×d

[0102] n = n s +n t

[0103] Where, n s n t denoted by , and d represents the number of samples in the source and target domains, respectively, where d is the feature dimension.

[0104] Then, based on the dimensional information and preset weight coefficients, a symmetric matrix is ​​obtained. A symmetric matrix L∈R is constructed using the preset weight coefficients. n×n To reflect the distributional differences between samples across domains, the L matrix is ​​normalized using the Frobenius norm to enhance numerical stability. The formula for calculating the symmetric matrix L is:

[0105]

[0106] Next, based on the dimensional information, the identity matrix, and the matrix of all ones, the centered matrix is ​​obtained. By constructing the centered matrix H, it can be used to eliminate the mean bias of X and preserve the relative relationships between samples. The formula for calculating the centered matrix H is:

[0107]

[0108] Where l is an n×n identity matrix and P is an n×n matrix of all ones.

[0109] Furthermore, based on the standardized reflectance of the first and second sensitive wavelengths, the preset linear kernel function type, and the matrix transpose, the kernel matrix is ​​obtained. That is, the kernel matrix K can be calculated according to the preset linear kernel function type, implicitly mapping the data to a high-dimensional reproducing kernel Hilbert space (RKHS). For the linear kernel function "linear", K... i,j =X i T X j Therefore, the kernel matrix K can be represented as:

[0110]

[0111] Among them, K src,src K src,tar K tar,src and K tar,tar It is cross-domain data defined in the Hilbert space embedded in the source and target domains, where T represents the matrix transpose.

[0112] Next, the optimal projection matrix can be obtained based on the symmetry matrix, kernel matrix, centering matrix, identity matrix, and regularization parameter. That is, based on the given symmetry matrix L, centering matrix H, and kernel matrix K, eigenvalue decomposition and matrix projection decomposition are performed to solve for the optimal projection matrix A. The formula can be expressed as:

[0113] A=(KLK+λl) -1 KHK

[0114] Where λ is the regularization parameter and l is the identity matrix.

[0115] Subsequently, based on the optimal projection matrix, matrix transpose, and kernel matrix, the target joint matrix mapped to the reproducing kernel Hilbert space is obtained. That is, data mapping and output are performed to realize the mapping calculation from the joint matrix X to the subspace, which can be expressed by the formula:

[0116] X new =A T K

[0117] Among them, X new The joint matrix formed by mapping X to the new subspace is the target joint matrix.

[0118] Then, based on the severity labels, the target joint matrix is ​​segmented and reconstructed into source and target domains to obtain the target-source domain dataset aligned with the feature space. That is, X... newThe source and target domain data are reconstructed through segmentation, ultimately yielding new source and target domain feature parameters. This aligns the feature distributions, resulting in new sensitive wavelengths for the source and target domains. The formula can be expressed as:

[0119] X src_new =X new [:,1:n s ] T

[0120] X tar_new =X new [:,n s+1 :n] T

[0121] Among them, X src_new and X tar_new These represent the new source and target domain sensitive wavelengths obtained after TCA processing, respectively.

[0122] Furthermore, based on the new source-domain sensitive wavelength X obtained after TCA processing... src_new and target domain sensitive wavelength X tar_new This redefines the source and target domain data, namely:

[0123] data 1_new =[X src_new ,y1]

[0124] data 2_new =[X src_new ,y2]

[0125] Among them, data 1_new =[X src_new ,y1] represents the target source domain dataset after feature space alignment.

[0126] Next, the regularization parameters of the regression model are optimized using the target source domain dataset to construct a source domain disease monitoring model, which may further include:

[0127] Based on the feature matrix, disease severity label, and error of the source domain training data in the target source domain dataset, a regression model is obtained; the regression model includes the regression parameter vector to be solved.

[0128] Based on the regression model, the L2 regularization term, and the number of samples in the source domain training data, obtain the loss function of the regression model; where the loss function includes the regularization parameter;

[0129] The regularization parameters are dynamically adjusted based on the initial search range, random number generation function, and energy factor to construct a source domain disease monitoring model.

[0130] In constructing a source domain disease monitoring model, model training can be performed based on the target source domain dataset aligned with the feature space. Furthermore, during model training, the Artificial Lemming Algorithm (ALA) can be applied to optimize the regularization parameters of the regression model (e.g., a Ridge regression model) to ensure the model's stability and accuracy in complex data environments. The regularization parameter optimization can utilize the Artificial Lemming Algorithm (ALA), but other algorithms such as Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and Differential Evolution (DE) can also be used. Specifically, an initial regression model can be established using the feature matrix, disease severity labels, and errors of the source domain training data in the target source domain dataset. This could be a Ridge regression model, or other types of regression models. Then, based on the regression model, the L2 regularization term, and the number of samples in the source domain training data, the loss function of the regression model is further determined. Furthermore, based on the regularization parameter to be optimized in the loss function, the initial search range of the regularization parameter, the random number generation function, and the energy factor, the regularization parameter is dynamically adjusted to obtain the optimized regularization parameter for constructing the source domain disease monitoring model, thereby ensuring the stability and accuracy of the source domain disease monitoring model in complex data environments.

[0131] The Ridge regression algorithm is an improvement on the least squares estimation algorithm, but it is a biased estimate. It abandons the unbiasedness and partial accuracy of the least squares algorithm, seeking a regression process that better reflects reality. In the Ridge regression model, the larger the value of the regularization parameter λ, the greater the model bias; therefore, determining an appropriate value for λ is crucial for the Ridge regression method. The ALA algorithm simulates the natural behaviors of lemmings, such as migration, burrowing, foraging, and predator evasion. Through mathematical modeling, it transforms lemming behavior into four cooperative search operators and introduces an energy reduction mechanism to dynamically adjust the selection probability of the operators, thus gradually transitioning from global coarse-grained search to local fine-grained optimization during the iteration process.

[0132] When constructing a Ridge regression model, for example, 2 / 3 of the target source domain dataset can be randomly selected as the source domain training data to construct the source domain Ridge regression model. Then, when optimizing the regularization parameter λ of the Ridge regression model using the Artificial Lemming Algorithm (ALA), the feature matrix of the source domain training data is first set to X∈R.m×d , target vector y∈R m The Ridge regression model for the source domain can then be defined as:

[0133] y = Xw + ε

[0134] Where X and y represent the sensitive wavelength and disease severity label of the training data, respectively, w is the regression parameter vector to be solved, m is the number of samples in the training data, and ε represents the error.

[0135] Then, since the goal of Ridge regression is to find w, which requires Xw to be infinitely close to y, Ridge regression introduces an L2 regularization term on top of ordinary least squares (OLS), and the loss function is:

[0136]

[0137] Where, min w This indicates the desire to find the weight vector w that minimizes the loss of the objective function. It is the L2 regularization term, λ is used to constrain the magnitude of the weight vector w to prevent the model from overfitting or underfitting.

[0138] Finally, based on the initial search range of λ set to

[10] , -5 10 3 The λ parameter is dynamically adjusted using the ALA algorithm to construct a source domain disease monitoring model using the optimized λ parameter.

[0139] Specifically, the regularization parameters are dynamically adjusted based on the initial search range, the random number generation function, and the energy factor to construct a source domain disease monitoring model, which may further include:

[0140] Based on the initial search range of the regularization parameter, generate initial candidate values ​​for the regularization parameter;

[0141] Based on the initial candidate values ​​and the loss function, obtain the regression model loss value;

[0142] When the energy factor is greater than the first set value and the value generated by the random number generation function is less than the second set value, a global exploration of migration perturbation is performed according to the preset number of iterations to obtain the first regularization parameter;

[0143] When the energy factor is greater than the first set value and the value generated by the random number generation function is greater than or equal to the second set value, a local fine search is performed according to the preset number of iterations to obtain the second regularization parameter.

[0144] When the energy factor is less than the first set value and the value generated by the random number generation function is less than the third set value, a spiral path density search is performed according to the preset number of iterations to obtain the third regularization parameter.

[0145] When the energy factor is less than the first set value and the value generated by the random number generation function is greater than or equal to the third set value, a Levy flight hybrid search is performed according to the preset number of iterations to obtain the fourth regularization parameter.

[0146] Based on the first regularization parameter, the second regularization parameter, the third regularization parameter, the fourth regularization parameter, and the loss function of the regression model, the target regularization parameter corresponding to the minimum regression model loss value under the preset number of iterations is obtained.

[0147] Based on the target regularization parameters, a source domain disease monitoring model is constructed.

[0148] When dynamically adjusting the regularization parameters, the initial candidate values ​​of the regularization parameters and the loss function are obtained using the initial search range of the regularization parameters. This yields the regression model loss value. Then, based on the relationship between the energy factor and the first set value, and the relationship between the generated value of the random number generation function and the second or third set value, different regularization parameters can be obtained: the first, second, third, and fourth regularization parameters. Using these parameters and the regression model's loss function, the target regularization parameter corresponding to the minimum regression model loss value under a preset number of iterations can be obtained. This target regularization parameter is then used to construct a source domain disease monitoring model, ensuring that the constructed model maintains stability and accuracy even in complex data environments.

[0149] When dynamically adjusting the regularization parameter, the ALA algorithm can be used to dynamically adjust the λ parameter. Before dynamically adjusting the λ parameter using the ALA algorithm, the parameter can be initialized; that is, the population size and the maximum number of iterations T can be set. max Set the value to 100 to initialize the population position and generate initial candidate values ​​for λ.

[0150] Then, differentiate the loss function J(w) and set the derivative to 0, then solve for the current w using a closed-form solution:

[0151] w = (X T X+λP) -1 X T y

[0152] Where P is an m×m identity matrix.

[0153] Then, the deviation between the predicted and actual values ​​is measured by calculating the mean squared error (EMS).

[0154]

[0155] Calculate the model loss at this moment based on the mean squared error:

[0156]

[0157] Subsequently, the current λ parameter is dynamically updated using the rand() function and the energy factor E. Here, rand is a function that can randomly generate numbers uniformly distributed in the interval [0,1], and the energy factor E... Where the angle attenuation term θ = 2arctan(1-t / T) max) It decreases dynamically with the number of iterations.

[0158] Furthermore, when E>1 and rand()<0.3, a global exploration of λ is performed using migration perturbation. By simulating mouse migration behavior, and combining Gaussian noise with random direction perturbation, potential high-quality λ regions are explored globally, as shown in the formula:

[0159]

[0160] Where F∈{-1,1} is the random direction indicator, BM~N(0,1) is the Brownian motion perturbation, the random vector R∈[-1,1] is the mixed weight, t represents the iteration number, and i represents the population individual. Let represent the position vector of the i-th individual in the t-th iteration. Let represent the updated position vector of the i-th individual at the (t+1)-th iteration. This represents the position of the optimal individual in the population at the t-th iteration. Let represent the average position vector of the population at the t-th iteration. Furthermore, in the above formula, the position vector of each individual... For a corresponding candidate value of λ, and based on the current optimal value... Perturbation to explore new λ values.

[0161] When E>1 and rand()≥0.3, a local fine-grained search for λ is performed. The local search step size L is adjusted based on the dynamic sine coefficients. By controlling L to perform a fine-grained search near the optimal solution, the high-quality λ region is avoided due to an excessively large step size. The formula for L is:

[0162]

[0163] Where L represents the search step size. In this case, λ is calculated using the following formula:

[0164]

[0165] When E < 1 and rand() < 0.5, perform a spiral path density search for λ. Generate a spiral path around the current optimal λ, and conduct a circular exploration with controllable density, using a random phase angle to control the search density.

[0166] spird=radius×(sin(2×π×rand)+cos(2×π×rand))

[0167] Where sprial is the coordinate value of a point on the spiral path, and radius is the radius of the spiral, used to control the search range. radius can be defined as:

[0168]

[0169] Where Dim represents the dimension of the optimization problem, which directly affects the search space complexity, z best,j (t) represents the optimal position of the population on the j-th dimension at the t-th iteration, z i,j (t) represents the current position of the i-th particle in the j-th dimension at the t-th iteration. At this point, λ is calculated using the following formula:

[0170]

[0171] When E < 1 and rand() ≥ 0.5, perform a hybrid Levy flight search λ. Levy flight is introduced to implement a hybrid search combining long-step (global) and short-step (local) approaches:

[0172]

[0173] Wherein, levy_step is a random step size based on Lévy flight, following a heavy-tailed distribution levy~t. -λ (1 < λ ≤ 3), sign is a function that returns -1, 0, or 1, and its value is determined by rand(). G is the dynamic coefficient that balances exploration and development, where the dynamic coefficient G can be defined as:

[0174]

[0175] Among them, T max This represents the maximum number of iterations.

[0176] Based on the above, the model loss is repeatedly calculated for each searched λ, and the calculated model loss is compared with the previous model loss. This process continues until the maximum number of iterations T is achieved. max During the entire iteration process, the individual positions that minimize the model loss are retained. The λ corresponding to this position is the optimal solution.

[0177] After determining the optimal solution λ, the formula w = (X) is used. T X+λP) -1 X Ty, solve for w in the closed form, and then substitute it into the Ridge regression model y=Xw+ε to optimize and obtain the source domain disease monitoring model.

[0178] In addition, the validation and evaluation of early wheat disease monitoring models also includes:

[0179] Input the source domain validation data from the target source domain dataset into the source domain disease monitoring model to obtain the model prediction results;

[0180] The model predictions were fitted to disease severity labels to obtain the first coefficient of determination.

[0181] The source domain validation data from the target source domain dataset is input into the comparison model to obtain the second coefficient of determination; the comparison model includes a random forest model and a support vector regression model.

[0182] Compare the first and second coefficients of determination;

[0183] When the first coefficient of determination is greater than the second coefficient of determination, the source domain disease monitoring model is effective.

[0184] When the first coefficient of determination is less than the second coefficient of determination, the source domain disease monitoring model is invalid.

[0185] Specifically, the remaining one-third of the target source domain dataset, excluding the source domain training data, is used as source domain validation data. This data is then input into the source domain disease monitoring model optimized with regularization parameters to obtain the corresponding model prediction results. The model prediction results are then fitted with the actual disease severity labels to obtain the first coefficient of determination. The validation data is then similarly input into a comparative model, such as a random forest (RF) or support vector regression (SVR) model, to obtain the second coefficient of determination. Finally, the first and second coefficients of determination are compared. If the first coefficient of determination is greater than the second coefficient, the source domain disease monitoring model is considered effective; otherwise, it is considered ineffective. Figure 12 Example of source domain monitoring results based on the TCA-ALA-Ridge regression model. Figure 13 and Figure 14 The provided examples of source domain monitoring results based on SVR and RF regression models show that the monitoring accuracy R of the model optimized by TCA-ALA-Ridge for source domain data is [missing information]. 2 The accuracy was improved to 88%, achieving the target domain monitoring accuracy shown in the table above. This invention compares the TCA-ALA-Ridge regression algorithm with conventional SVR and RF algorithms. Figure 13 , Figure 14It is known that conventional algorithms do not perform well in model transfer evaluation across different datasets, with SVR achieving an accuracy of only 74% and RF only reaching 61%. Therefore, the technical method proposed in this invention improves upon traditional SVR and RF algorithms by 14% and 27% respectively, demonstrating the effectiveness of the source domain disease monitoring model. The optimized source domain disease monitoring model can enhance early disease monitoring capabilities.

[0186] Please see Figure 15 The present invention also provides a wheat early disease monitoring model construction system 11, comprising: an acquisition unit 111 for acquiring spectral reflectance data of wheat ear samples and the disease severity of the spectral reflectance data; a division unit 112 for dividing the spectral reflectance data into a source domain dataset and a target domain dataset according to the disease severity; wherein the source domain dataset includes first spectral reflectance data corresponding to healthy and first disease severity, and the target domain dataset includes second spectral reflectance data corresponding to healthy and second disease severity; a selection unit 113 for filtering the sensitive band range of the source domain dataset and the target domain dataset respectively to obtain the first sensitive wavelength reflectance corresponding to the source domain dataset and the second sensitive wavelength reflectance corresponding to the target domain dataset; and a modeling unit 114 for constructing a source domain disease monitoring model based on the first sensitive wavelength reflectance, the second sensitive wavelength reflectance, and the disease severity label.

[0187] It should be noted that the wheat early disease monitoring model construction system 11 provided in the above embodiments and the wheat early disease monitoring model construction method provided in the above embodiments belong to the same concept. The specific operation methods of each module and unit have been described in detail in the method embodiments and will not be repeated here. In practical applications, the wheat early disease monitoring system 11 provided in the above embodiments can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. This is not a limitation here.

[0188] Please see Figure 16 The electronic device 1 may include a memory 12, a processor 13 and a bus, and may also include a computer program stored in the memory 12 and capable of running on the processor 13, such as a wheat early disease monitoring model construction program.

[0189] The memory 12 includes at least one type of readable storage medium, such as flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 12 can be an internal storage unit of the electronic device 1, such as the portable hard drive of the electronic device 1. In other embodiments, the memory 12 can be an external storage device of the electronic device 1, such as a plug-in portable hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 1. Furthermore, the memory 12 can include both internal and external storage units of the electronic device 1. The memory 12 can be used not only to store application software and various types of data installed on the electronic device 1, such as code for constructing early wheat disease monitoring models, but also to temporarily store data that has been output or will be output.

[0190] In some embodiments, the processor 13 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 13 is the control unit of the electronic device 1, connecting various components of the electronic device 1 via various interfaces and lines. It executes programs or modules stored in the memory 12 (e.g., a wheat early disease monitoring model construction program) and calls data stored in the memory 12 to perform various functions of the electronic device 1 and process data.

[0191] The processor 13 executes the operating system of the electronic device 1 and various installed applications. The processor 13 executes the applications to implement the steps in the above-described method for constructing a wheat early disease monitoring model.

[0192] For example, the computer program may be divided into one or more modules, which are stored in the memory 12 and executed by the processor 13 to complete this application. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the electronic device 1. For example, the computer program may be divided into unit modules of a wheat early disease monitoring system 11.

[0193] The integrated unit implemented as a software functional module can be stored in a computer-readable storage medium, which can be non-volatile or volatile. The software functional module, stored in the storage medium, includes several instructions to cause a computer device (which may be a personal computer, computer equipment, or network device, etc.) or processor to execute some functions of the wheat early disease monitoring model construction method described in the various embodiments of this application.

[0194] In summary, the present invention discloses a method and system for constructing an early wheat disease monitoring model. This method divides spectral reflectance data into source domain and target domain datasets according to disease severity, selects sensitive wavelength reflectance, and then uses the TCA method to map the features of source and target domain samples with different disease severity levels into a high-dimensional space, achieving alignment of feature distributions between the source and target domains. Simultaneously, the ALA algorithm enables intelligent parameter optimization, and the Ridge regression algorithm is combined to construct the source domain disease monitoring model. This method integrates the domain adaptability of TCA, the adaptive optimization of ALA, and the regularization advantages of Ridge regression. It is suitable for early wheat scab disease monitoring and modeling scenarios, improving the early disease monitoring capability of the source domain disease monitoring model and solving the technical problems of poor model generalization, low transferability, and low accuracy in early wheat scab monitoring scenarios using traditional methods. Therefore, this invention effectively overcomes the various shortcomings of existing technologies and has high industrial application value.

[0195] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A method for constructing a monitoring model for early wheat diseases, characterized in that, include: Obtain spectral reflectance data of wheat ear samples and the severity of the disease based on the spectral reflectance data; The spectral reflectance data is divided into a source domain dataset and a target domain dataset according to the severity of the disease; wherein, the source domain dataset includes first spectral reflectance data corresponding to health and the first degree of disease, and the target domain dataset includes second spectral reflectance data corresponding to health and the second degree of disease. The source domain dataset and the target domain dataset are respectively filtered for sensitive band ranges to obtain the first sensitive wavelength reflectance corresponding to the source domain dataset and the second sensitive wavelength reflectance corresponding to the target domain dataset. A source domain disease monitoring model is constructed based on the reflectance of the first sensitive wavelength, the reflectance of the second sensitive wavelength, and the disease severity label. Specifically, a source-domain disease monitoring model is constructed based on the first sensitive wavelength reflectance, the second sensitive wavelength reflectance, and the disease severity label, including: The reflectance of the first sensitive wavelength and the reflectance of the second sensitive wavelength are standardized. The standardized reflectance of the first and second sensitive wavelengths are merged row by row to obtain a joint matrix; Based on the joint matrix, obtain the dimension information of the joint matrix; Based on the dimensional information and preset weight coefficients, a symmetric matrix is ​​obtained; Based on the dimensional information, the identity matrix, and the all-ones matrix, obtain the centered matrix; Based on the standardized reflectance of the first sensitive wavelength, the reflectance of the second sensitive wavelength, the preset linear kernel function type, and the matrix transpose, obtain the kernel matrix; The optimal projection matrix is ​​obtained based on the symmetric matrix, the kernel matrix, the centering matrix, the identity matrix, and the regularization parameter. Based on the optimal projection matrix, the matrix transpose, and the kernel matrix, obtain the target joint matrix that maps the joint matrix to the reproducing kernel Hilbert space; Based on the severity labels, the target joint matrix is ​​segmented and reconstructed into source and target domains to obtain a target source domain dataset aligned with the feature space. The regularization parameters of the regression model are optimized using the target source domain dataset to construct the source domain disease monitoring model.

2. The method for constructing a wheat early disease monitoring model according to claim 1, characterized in that, Sensitive band range filtering is performed on the source domain dataset and the target domain dataset respectively to obtain the first sensitive wavelength reflectance corresponding to the source domain dataset and the second sensitive wavelength reflectance corresponding to the target domain dataset, including: Based on the source domain dataset within the defined band range, a first partial least squares regression model is established; Based on the first spectral reflectance data, the first disease severity prediction value is obtained by the first partial least squares regression model. Based on the actual value of the first disease severity corresponding to the first spectral reflectance data and the predicted value of the first disease severity, the first maximum determination coefficient is obtained; Based on the target domain dataset within the defined band range, a second partial least squares regression model is established; Based on the second spectral reflectance data, the second predicted value of the severity of the condition is obtained by the second partial least squares regression model. The second maximum determination coefficient is obtained based on the actual value of the second disease severity and the predicted value of the second disease severity corresponding to the second spectral reflectance data. Based on the first maximum determination coefficient and the second maximum determination coefficient, the source domain dataset and the target domain dataset are respectively filtered for sensitive band ranges to obtain the first sensitive wavelength reflectance and the second sensitive wavelength reflectance.

3. The method for constructing a wheat early disease monitoring model according to claim 2, characterized in that, Based on the first maximum determination coefficient and the second maximum determination coefficient, the source domain dataset and the target domain dataset are respectively filtered for sensitive band ranges to obtain the first sensitive wavelength reflectance and the second sensitive wavelength reflectance, including: Based on the first maximum determination coefficient and the second maximum determination coefficient, a sensitive band range is filtered within the set band range to obtain a target sensitive band range; wherein, the target sensitive band range is the same band range of the source domain dataset and the target domain dataset; The reflectivity of the sensitive wavelength is selected according to the target sensitive band range to obtain the first sensitive wavelength reflectivity and the second sensitive wavelength reflectivity.

4. The method for constructing a wheat early disease monitoring model according to claim 1, characterized in that, The reflectivity of the first sensitive wavelength and the reflectivity of the second sensitive wavelength are standardized, including: Based on the first mean and the first standard deviation of the reflectance of all first sensitive wavelengths, the reflectance of the first sensitive wavelengths is standardized by the z-score method to obtain the standardized reflectance of the first sensitive wavelengths. Based on the second mean and the second standard deviation of the reflectance of all second sensitive wavelengths, the reflectance of the second sensitive wavelengths is standardized using the z-score method to obtain the standardized reflectance of the second sensitive wavelengths.

5. The method for constructing a wheat early disease monitoring model according to claim 1, characterized in that, The regularization parameters of the regression model are optimized using the target source domain dataset to construct the source domain disease monitoring model, including: Based on the feature matrix of the source domain training data in the target source domain dataset, the disease severity label, and the error, a regression model is obtained; wherein, the regression model includes a regression parameter vector to be solved; Based on the regression model, the L2 regularization term, and the number of samples in the source domain training data, the loss function of the regression model is obtained; wherein, the loss function includes the regularization parameter; The regularization parameters are dynamically adjusted based on the initial search range, random number generation function, and energy factor to construct the source domain disease monitoring model.

6. The method for constructing a wheat early disease monitoring model according to claim 5, characterized in that, The regularization parameters are dynamically adjusted based on the initial search range, random number generation function, and energy factor to construct the source domain disease monitoring model, including: Based on the initial search range of the regularization parameter, generate initial candidate values ​​for the regularization parameter; Based on the initial candidate values ​​and the loss function, the regression model loss value is obtained; When the energy factor is greater than the first set value and the value generated by the random number generation function is less than the second set value, then a global exploration of migration perturbation is performed according to a preset number of iterations to obtain the first regularization parameter; When the energy factor is greater than the first set value and the value generated by the random number generation function is greater than or equal to the second set value, a local fine search is performed according to the preset number of iterations to obtain the second regularization parameter; When the energy factor is less than the first set value and the generated value of the random number generation function is less than the third set value, a spiral path density search is performed according to the preset number of iterations to obtain the third regularization parameter. When the energy factor is less than the first set value and the generated value of the random number generation function is greater than or equal to the third set value, then a Levi flight hybrid search is performed according to the preset number of iterations to obtain the fourth regularization parameter; Based on the first regularization parameter, the second regularization parameter, the third regularization parameter, the fourth regularization parameter, and the loss function of the regression model, the target regularization parameter corresponding to the minimum regression model loss value under a preset number of iterations is obtained; The source domain disease monitoring model is constructed based on the target regularization parameters.

7. The method for constructing a wheat early disease monitoring model according to claim 1, characterized in that, Also includes: Input the source domain validation data from the target source domain dataset into the source domain disease monitoring model to obtain the model prediction results; The model prediction results are fitted to the disease severity labels to obtain the first coefficient of determination; The source domain validation data from the target source domain dataset is input into the comparison model to obtain the second coefficient of determination; wherein, the comparison model includes a random forest model and a support vector regression model; Compare the first coefficient of determination and the second coefficient of determination; When the first determination coefficient is greater than the second determination coefficient, the source domain disease monitoring model is effective; When the first determination coefficient is less than the second determination coefficient, the source domain disease monitoring model is invalid.

8. A system for constructing a monitoring model for early wheat diseases, characterized in that, include: The acquisition unit is used to acquire the spectral reflectance data of wheat ear samples and the severity of the disease based on the spectral reflectance data. A partitioning unit is used to divide the spectral reflectance data into a source domain dataset and a target domain dataset according to the severity of the disease; wherein, the source domain dataset includes first spectral reflectance data corresponding to health and the first degree of disease, and the target domain dataset includes second spectral reflectance data corresponding to health and the second degree of disease; The selection unit is used to filter the sensitive band range of the source domain dataset and the target domain dataset respectively, to obtain the first sensitive wavelength reflectance corresponding to the source domain dataset and the second sensitive wavelength reflectance corresponding to the target domain dataset; and The modeling unit is used to construct a source domain disease monitoring model based on the first sensitive wavelength reflectance, the second sensitive wavelength reflectance, and the disease severity label. Specifically, a source-domain disease monitoring model is constructed based on the first sensitive wavelength reflectance, the second sensitive wavelength reflectance, and the disease severity label, including: The reflectance of the first sensitive wavelength and the reflectance of the second sensitive wavelength are standardized. The standardized reflectance of the first and second sensitive wavelengths are merged row by row to obtain a joint matrix; Based on the joint matrix, obtain the dimension information of the joint matrix; Based on the dimensional information and preset weight coefficients, a symmetric matrix is ​​obtained; Based on the dimensional information, the identity matrix, and the all-ones matrix, obtain the centered matrix; Based on the standardized reflectance of the first sensitive wavelength, the reflectance of the second sensitive wavelength, the preset linear kernel function type, and the matrix transpose, obtain the kernel matrix; The optimal projection matrix is ​​obtained based on the symmetric matrix, the kernel matrix, the centering matrix, the identity matrix, and the regularization parameter. Based on the optimal projection matrix, the matrix transpose, and the kernel matrix, obtain the target joint matrix that maps the joint matrix to the reproducing kernel Hilbert space; Based on the severity labels, the target joint matrix is ​​segmented and reconstructed into source and target domains to obtain a target source domain dataset aligned with the feature space. The regularization parameters of the regression model are optimized using the target source domain dataset to construct the source domain disease monitoring model.

Images