A wheat stripe rust identification method, device, equipment and storage medium

Abstract

The application discloses a wheat stripe rust identification method and device, equipment and a storage medium. The sunlight-induced chlorophyll fluorescence data and wavelet features of a target wheat to be measured are input into a wheat stripe rust identification model for processing to obtain a target disease index of the target wheat to be measured. The wheat stripe rust identification model is a prediction model trained by taking the continuous wavelet features of wheat samples in cooperation with the sunlight-induced chlorophyll fluorescence data as characteristic values and taking the disease indexes of the wheat samples as target values. The continuous wavelet features can highlight the features sensitive to wheat stripe rust in the wheat canopy spectrum, and the sunlight-induced chlorophyll fluorescence can be effectively and intuitively applied to the detection of changes in the physiological state of plants to represent the degree of stress on the wheat caused by diseases. The embodiments of the application analyze the full-band data of the wheat to obtain key information for identifying the wheat stripe rust, prevent the loss of disease-specific factors of the wheat stripe rust, and improve the accuracy of identifying the wheat stripe rust.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of identification, more particularly, to a wheat stripe rust identification method, device, equipment and storage medium. BACKGROUND

[0002] Wheat stripe rust is a serious disease that has a wide distribution, fast spread and large damage area in wheat production, which seriously affects the yield and quality of wheat. It is very important to detect whether the planted wheat is infected with wheat stripe rust.

[0003] In the prior art, the detection method of wheat stripe rust usually uses the narrow band information provided by hyperspectral technology to reflect its stress state, calculates specific vegetation indices, reflectance differential indices or absorption characteristics using reflectance data, and then analyzes the relationship between the above indices and the severity of crop diseases to screen out vegetation indices that can represent the growth status of vegetation under disease stress, and constructs a monitoring model for different diseases. However, the above monitoring method mainly detects diseases through a small amount of spectral information or absorption characteristics of specific wave bands, and improper selection of disease-specific factors may result in the loss of some key information related to disease detection, which cannot obtain more sensitive characteristic information about wheat stripe rust, and cannot guarantee the accuracy of the obtained wheat stripe rust detection results. SUMMARY

[0004] Therefore, the present application provides a wheat stripe rust identification method, device, equipment and storage medium to solve the problem of inaccurate prediction caused by insufficient use of spectral data, improper selection of disease-specific factors, etc. in the existing wheat stripe rust prediction method.

[0005] In order to achieve the above purpose, the present scheme is as follows:

[0006] A wheat stripe rust identification method, comprising:

[0007] obtaining daylight-induced chlorophyll fluorescence data and continuous wavelet features of a target wheat to be tested;

[0008] inputting the daylight-induced chlorophyll fluorescence data and the continuous wavelet features into a wheat stripe rust identification model for processing to obtain a target disease index matched with the target wheat to be tested output by the wheat stripe rust identification model;

[0009] The wheat stripe rust identification model is a model trained with the continuous wavelet features of the wheat sample in cooperation with the daylight-induced chlorophyll fluorescence data as feature values, and the disease index corresponding to the wheat sample as a target value, and the wheat stripe rust identification model is a regression prediction model determined based on iterative training.

[0010] Preferably, the method further comprises:

[0011] Obtaining spectral data and disease index corresponding to wheat samples at different growth stages;

[0012] Processing based on the spectral data and the disease index to obtain effective data, the effective data including one or more of daylight-induced chlorophyll fluorescence data, continuous wavelet features, fractional differential spectral features and vegetation index, the daylight-induced chlorophyll fluorescence data including daylight-induced chlorophyll fluorescence relative intensity and fluorescence index;

[0013] Generating a sample data set based on the effective data and the disease index, each sample data in the sample data set including at least one effective data feature and data labeled with the disease index;

[0014] Model training based on each preset model structure and the sample data set to obtain an initial identification model matched with each preset model and each sample data, wherein the model training process takes at least one effective data in each sample data as a feature value and the disease index as a target value, and an initial identification model is obtained after training;

[0015] Model evaluation of the initial identification model to obtain an evaluation result corresponding to each initial identification model;

[0016] Comparing the evaluation results, the initial identification model corresponding to the evaluation result reaching a preset condition is determined as a wheat stripe rust identification model.

[0017] Preferably, the effective data includes continuous wavelet features, wherein the processing based on the spectral data and the disease index to obtain effective data includes:

[0018] Processing the spectral data to obtain continuous wavelet features at different decomposition scales;

[0019] Statistically analyzing the correlation between the continuous wavelet features and the disease index to obtain correlation coefficient absolute values corresponding to the continuous wavelet features at different decomposition scales;

[0020] Determining sensitive features based on the correlation coefficient absolute values;

[0021] Calculating the sensitive features and the disease index to obtain a projection importance index corresponding to the sensitive features;

[0022] Based on the projection importance index, determining the continuous wavelet features corresponding to the projection importance index reaching a preset condition as effective data.

[0023] Preferably, the effective data comprises sunlight-induced chlorophyll fluorescence data, and the sunlight-induced chlorophyll fluorescence data comprises relative intensity of sunlight-induced chlorophyll fluorescence, wherein the processing based on the spectral data and the disease index to obtain effective data comprises:

[0024] Based on the spectral data, the first sunlight-induced chlorophyll fluorescence absolute intensity corresponding to the first oxygen absorption wave band and the second sunlight-induced chlorophyll fluorescence absolute intensity corresponding to the second oxygen absorption wave band are calculated by using irradiance and radiance data respectively;

[0025] Based on the first sunlight-induced chlorophyll fluorescence absolute intensity and the second sunlight-induced chlorophyll fluorescence absolute intensity, the first sunlight-induced chlorophyll fluorescence relative intensity corresponding to the first sunlight-induced chlorophyll fluorescence absolute intensity and the second sunlight-induced chlorophyll fluorescence relative intensity corresponding to the second sunlight-induced chlorophyll fluorescence absolute intensity are obtained by processing;

[0026] The first sunlight-induced chlorophyll fluorescence relative intensity and the second sunlight-induced chlorophyll fluorescence relative intensity are respectively correlated with the disease index, and the correlation coefficients corresponding to the first sunlight-induced chlorophyll fluorescence relative intensity and the second sunlight-induced chlorophyll fluorescence relative intensity are obtained respectively;

[0027] By comparing the correlation coefficients, the sunlight-induced chlorophyll fluorescence relative intensity corresponding to the correlation coefficient reaching a preset condition is determined as effective data.

[0028] Preferably, the effective data comprises sunlight-induced chlorophyll fluorescence data, and the sunlight-induced chlorophyll fluorescence data further comprises a fluorescence index, and further comprises:

[0029] Based on the spectral data, the reflectance index and the first derivative index of reflectance are calculated to obtain the fluorescence index based on reflectance, and the fluorescence index reflects the intensity of sunlight-induced chlorophyll fluorescence;

[0030] Based on the fluorescence index and the disease index, the correlation analysis is performed to obtain the correlation coefficient corresponding to the fluorescence index;

[0031] By comparing the correlation coefficients, the fluorescence index corresponding to the correlation coefficient reaching a preset condition is determined as effective fluorescence index;

[0032] The effective fluorescence index is determined as effective data.

[0033] Preferably, the effective data comprises vegetation index, and the processing based on the spectral data and the disease index to obtain effective data comprises:

[0034] characterize a growth condition of the wheat sample under the stripe rust stress based on the spectral data, to obtain a plurality of vegetation indexes of the wheat sample;

[0035] perform correlation analysis on the vegetation indexes and the disease indexes, to obtain a correlation coefficient corresponding to each vegetation index;

[0036] compare the correlation coefficients corresponding to each vegetation index, and determine the vegetation index corresponding to the correlation coefficient meeting a preset condition as effective data.

[0037] Preferably, the effective data includes a fractional derivative spectral feature, wherein the processing based on the spectral data and the disease indexes to obtain effective data includes:

[0038] performing derivative processing of different orders on the spectral data to obtain derivative spectral features of different orders;

[0039] performing correlation analysis on the derivative spectral features of each order and the disease indexes to obtain a correlation coefficient corresponding to the derivative spectral features of each order;

[0040] determining a sensitive feature based on an absolute value of the correlation coefficient;

[0041] calculating the sensitive feature and the disease index to obtain a projection importance index corresponding to the sensitive feature;

[0042] determining a derivative spectral feature corresponding to a projection importance index meeting a preset condition based on the projection importance index;

[0043] combining the derivative spectral features to obtain a fractional derivative spectral feature.

[0044] A wheat stripe rust identification device includes:

[0045] a wheat information acquisition unit configured to acquire sunlight-induced chlorophyll fluorescence data and continuous wavelet features of a target wheat to be tested;

[0046] a wheat information processing unit configured to input the sunlight-induced chlorophyll fluorescence data and the continuous wavelet features into a wheat stripe rust identification model for processing, to obtain a target disease index matching the target wheat to be tested output by the wheat stripe rust identification model;

[0047] The wheat stripe rust identification model is a model trained by taking the continuous wavelet features and the sunlight-induced chlorophyll fluorescence data of a wheat sample as feature values and taking a disease index corresponding to the wheat sample as a target value, and the wheat stripe rust identification model is a regression prediction model determined based on iterative training.

[0048] A wheat stripe rust identification device, comprising a processor and a memory;

[0049] The processor is configured to execute programs stored in the memory.

[0050] The memory is configured to store programs, which implement at least each step of the wheat stripe rust identification method.

[0051] A storage medium, in which computer executable instructions are stored, the computer executable instructions are loaded and executed by a processor to implement each step of the wheat stripe rust identification method.

[0052] As can be seen from the above technical solutions, the embodiments of the present application provide a wheat stripe rust identification method, device, equipment and storage medium. The continuous wavelet features used in the above scheme can highlight the features in the wheat canopy spectrum that are sensitive to wheat stripe rust. Sunlight-induced chlorophyll fluorescence has advantages in photosynthesis detection, can be effectively and intuitively applied to plant physiological state change detection, and can directly quantify the actual photosynthesis and represent the degree of wheat stress caused by diseases. The embodiments of the present application analyze the full-band data of wheat, obtain disease-specific factors for identifying the above wheat stripe rust, prevent the loss of more sensitive feature information of wheat stripe rust, and improve the accuracy of identifying wheat stripe rust. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only embodiments of the present application, and those skilled in the art can obtain other drawings according to the provided drawings without creative labor.

[0054] Figure 1 A flowchart for implementing the wheat stripe rust identification method provided by the embodiments of the present application is provided.

[0055] Figure 2 A flowchart for constructing the wheat stripe rust identification model provided by the embodiments of the present application is provided.

[0056] Figure 3 A structural diagram of the wheat stripe rust identification device provided by the embodiments of the present application is provided.

[0057] Figure 4 A structural diagram of the wheat stripe rust identification device provided by the embodiments of the present application is provided. DETAILED DESCRIPTION

[0058] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0059] Figure 1 A flowchart of a method for identifying wheat stripe rust is shown, and the method is provided in the embodiments of the present application, as shown in Figure 1 The flowchart can include the following steps.

[0060] In step S110, the sunlight-induced chlorophyll fluorescence data and continuous wavelet features of the target wheat to be tested are obtained.

[0061] The sunlight-induced chlorophyll fluorescence data and the continuous wavelet features can be obtained by analyzing the full-band spectral data of the target wheat to be tested. The spectral data of the target wheat to be tested can be obtained by various methods, such as ground measurement by a ground object spectrometer and the like, and further analysis based on the obtained spectral data of the wheat to obtain key information for identifying wheat diseases.

[0062] The sunlight-induced chlorophyll fluorescence data can represent the degree of stress of the wheat caused by diseases. The sunlight-induced chlorophyll fluorescence data corresponding to the target wheat to be tested is obtained by performing relevant calculations based on the spectral data of the target wheat to be tested. The continuous wavelet features are obtained by performing continuous wavelet transform on the spectral data of the target wheat to be tested, and the continuous wavelet features highlight the sensitive features of the wheat stripe rust in the spectral data. The above feature information reflecting the wheat stripe rust is input into a wheat stripe rust identification model for identification.

[0063] In step S120, the sunlight-induced chlorophyll fluorescence data and the continuous wavelet features are input into the wheat stripe rust identification model.

[0064] In step S130, the target disease index of the target wheat to be tested is obtained.

[0065] The wheat stripe rust identification model is a model trained by taking the continuous wavelet features and the sunlight-induced chlorophyll fluorescence data of the wheat samples as feature values, and taking the disease indexes corresponding to the wheat samples as target values. The wheat stripe rust identification model is a regression prediction model determined based on iterative training.

[0066] The wheat stripe rust identification model can be based on the same sample data, and an optimal identification model is obtained by using multiple different initial model structures for multiple modeling, training, detection and evaluation. The optimal identification model can also be obtained by step-by-step optimization of the initial model by changing the structure, algorithm and data structure of the same initial model, so as to ensure that the final wheat stripe rust identification model can obtain the optimal identification result. The wheat stripe rust identification model can output the target disease index corresponding to the target wheat to be tested according to the input sunlight-induced chlorophyll fluorescence data and continuous wavelet features.

[0067] The embodiments of the present application can input sunlight-induced chlorophyll fluorescence data and continuous wavelet features that can better reflect the sensitive characteristics of the target wheat to be tested into the wheat stripe rust identification model, and the wheat stripe rust identification model can output the target disease index corresponding to the target wheat to be tested. The continuous wavelet feature can highlight the characteristics sensitive to wheat stripe rust in the wheat canopy spectrum. Sunlight-induced chlorophyll fluorescence has an advantage in photosynthesis detection, can be effectively and intuitively applied to plant physiological state change detection, and can directly quantify the actual photosynthesis and represent the degree of wheat stress caused by diseases. The embodiments of the present application can analyze the full-band data of wheat, obtain disease-specific factors for identifying wheat stripe rust, prevent the loss of more sensitive feature information of wheat stripe rust, and improve the accuracy of identifying wheat stripe rust.

[0068] Next, the embodiments of the present application will further introduce the above-mentioned wheat stripe rust identification method.

[0069] The above-mentioned wheat stripe rust identification model needs to process the data of the target wheat to be tested based on the input data to obtain the target disease index corresponding to the target wheat to be tested. The wheat stripe rust identification model is not unique, and an initial model can be established based on different sample sets, different model structures and different model algorithms, and then an optimal model is determined from multiple initial models as the final wheat stripe rust identification model.

[0070] Reference Figure 2 which shows a flowchart of constructing a wheat stripe rust identification model, which can include the following steps:

[0071] In step S121, the spectral data and disease index corresponding to the wheat sample at different growth stages are obtained.

[0072] The spectral data can be obtained by detecting the test field of the wheat sample on the ground by an instrument. The wheat variety of the test field can be selected as the conventional planting variety in the test area, or the specific variety of wheat for test research can be planted, and there is no special requirement for the variety of the wheat sample.

[0073] The wheat samples need to be divided into diseased samples and healthy samples. The growth conditions of the wheat in the diseased samples and the healthy samples are detected by a spectral detection instrument to obtain spectral data and disease index of the wheat in different growth periods. The spectral data and the growth conditions of the diseased samples need to be monitored in different disease stages, and the spectral data and the growth conditions of the healthy samples need to be monitored in different growth periods. The growth conditions of the wheat are divided into multiple gradient levels according to the severity of the wheat yellow rust, and the disease index of the wheat samples in different gradient levels is calculated. The gradient level can be calculated from level 0, which represents that the wheat in this level is healthy, and the disease index is relatively 0, indicating that the wheat does not have wheat yellow rust. The higher the level is, the more serious the disease is, and the higher the disease index is.

[0074] Specifically, the preferred wheat sample data can be obtained by screening all the spectral data and the disease index of the wheat. The sample data includes spectral data and disease index of the wheat samples. In the embodiments of the present application, the wheat sample data can include 51 wheat sample data, including 45 diseased sample data and 6 healthy sample data.

[0075] In step S122, the spectral data and the disease index are processed to obtain effective data.

[0076] The effective data can include one or more of daylight-induced chlorophyll fluorescence data, continuous wavelet features, fractional differential spectral features, and vegetation indices. The daylight-induced chlorophyll fluorescence data can include, for example, relative intensity of daylight-induced chlorophyll fluorescence and fluorescence index, which is obtained based on reflectance processing. Different effective data represent different characteristics of the wheat, but different effective data can reflect and analyze the disease degree of the wheat to a certain extent.

[0077] When the continuous wavelet features are obtained based on the spectral data and the disease index, the specific process can include the following steps: processing the spectral data to obtain continuous wavelet features at different decomposition scales; calculating the correlation between the continuous wavelet features and the disease index to obtain correlation coefficient absolute values corresponding to the continuous wavelet features at different decomposition scales; determining sensitive features based on the correlation coefficient absolute values, and calculating the sensitive features and the disease index to obtain projection importance indices corresponding to the sensitive features; and determining, based on the projection importance indices, that the continuous wavelet features corresponding to the projection importance indices that meet a preset condition are effective data.

[0078] The continuous wavelet transform is performed on the spectral data of the wheat sample, and the wavelet transform is a signal analysis method that can effectively extract characteristic information sensitive to the wheat stripe rust disease degree from the spectral reflectance, i.e., the final continuous wavelet feature.

[0079] In the embodiments of the present application, the 51 sample data obtained above can be processed. The wheat canopy spectral data of the 51 sample data in the range of 350-1800 nm is taken as the processing object, and the canopy spectral data is subjected to continuous wavelet transform of 1-10 decomposition scales by using the base functions db5 and mexh respectively. Correlation analysis is performed on the continuous wavelet features of 10 different decomposition scales and the disease index respectively, and the correlation coefficients corresponding to the continuous wavelet features of each decomposition scale are obtained. The sensitive features are determined based on the absolute values of the correlation coefficients, and variable projection importance calculation is performed on the sensitive features and the disease index, so as to screen out suitable and effective wavelet features.

[0080] Specifically, the correlation coefficients can be arranged in descending order based on the size, and the variable projection importance is screened based on the absolute values of the correlation coefficients. Calculation is established between the primary screening sensitive features obtained in the above variable screening and the disease index, and the projection importance index corresponding to the disease index of the primary screening sensitive features is obtained. When the variables have strong correlation, the projection importance index can quantify the explanation degree of each input variable to the dependent variable. Then, the projection importance index that meets the preset condition is screened from the projection importance index, and the continuous wavelet feature corresponding to the projection importance index that meets the preset condition is taken as the effective data. The preset condition can be that the projection importance indexes are arranged in descending order according to the size, and the projection importance indexes in the top 15 positions are the preset condition, which can ensure that the continuous wavelet features corresponding to the projection importance indexes are highly correlated with the disease index.

[0081] In addition, the number of wavelet transform bands that pass the P=0.1% extremely significant test under different decomposition scales of the two base functions db5 and mexh can be counted, as shown in Table 1. When the base function is db5, the number of characteristic bands decreases sharply and then increases slowly with the increase of the scale. When the decomposition scale is 1, the number of characteristic bands is the largest, reaching 842. When the base function is mexh, the number of characteristic bands increases with the increase of the scale. When the decomposition scale is 10, the number of characteristic bands is 1046. The number of bands that pass the P=0.1% extremely significant test without wavelet transform is 795. The wavelet transform under the two base functions can increase the number of sensitive bands, and the wavelet transform is more helpful to extract the characteristic data sensitive to the stripe rust disease from the spectral data.

[0082] Table 1 Number of bands that pass 0.1% significance test of continuous wavelet transform

[0083]

[0084] Further statistics of the maximum correlation coefficient absolute value and the corresponding wavelength of the canopy spectrum and disease index under 10 different decomposition scales, the maximum correlation coefficient appears at scale 6 when the basis function is db5, the value is 0.889, and the corresponding wavelength is 701 nm. When the basis function is mexh, the maximum correlation coefficient appears at scale 1, the value is 0.877, and the corresponding wavelength is 714 nm. The maximum correlation coefficient between the spectral data without wavelet transform and the disease index is only 0.709. Through the comparison of the above correlation coefficients, it can be clear that the role of continuous wavelet transform in mining disease-specific factors is obvious. The continuous wavelet features obtained by wavelet transform can better reflect the wheat stripe rust and highlight the features sensitive to wheat stripe rust in the wheat canopy spectrum. The band information at a specific scale is highly correlated with the disease index. Based on this, the training data of the model can better improve the accuracy of the model in identifying wheat stripe rust.

[0085] In addition to the continuous wavelet feature, the effective data also includes the sunlight-induced chlorophyll fluorescence data, which can include the relative intensity of sunlight-induced chlorophyll fluorescence. By processing the spectral data, the sunlight-induced chlorophyll fluorescence data can be obtained as effective data. The process of obtaining the relative intensity of sunlight-induced chlorophyll fluorescence can include: based on the spectral data, using irradiance and radiance data to calculate, respectively, to obtain the first sunlight-induced chlorophyll fluorescence absolute intensity corresponding to the first oxygen absorption waveband, and the second sunlight-induced chlorophyll fluorescence absolute intensity corresponding to the second oxygen absorption waveband; based on the first sunlight-induced chlorophyll fluorescence absolute intensity and the second sunlight-induced chlorophyll fluorescence absolute intensity, processing to obtain the first sunlight-induced chlorophyll fluorescence relative intensity corresponding to the first sunlight-induced chlorophyll fluorescence absolute intensity, and the second sunlight-induced chlorophyll fluorescence relative intensity corresponding to the second sunlight-induced chlorophyll fluorescence absolute intensity; respectively, the first sunlight-induced chlorophyll fluorescence relative intensity and the second sunlight-induced chlorophyll fluorescence relative intensity are correlated with the disease index to obtain the correlation coefficient corresponding to the first sunlight-induced chlorophyll fluorescence relative intensity and the second sunlight-induced chlorophyll fluorescence relative intensity, respectively; comparing the correlation coefficients, the sunlight-induced chlorophyll fluorescence relative intensity corresponding to the correlation coefficient that meets the preset condition is determined as the effective data.

[0086] Wherein, the embodiment of the application can extract the principle of Fraunhofer dark line and apply 3FLD algorithm to estimate the sunlight-induced chlorophyll fluorescence intensity in the two oxygen absorption bands O2-A, O2-B of the spectral data of the wheat sample. First, the absolute intensity of the sunlight-induced chlorophyll fluorescence (SIF) is calculated by using the irradiance and radiance. At the same time, in order to weaken the interference of external factors such as sunlight intensity on the estimation of chlorophyll fluorescence, the calculated absolute intensity of SIF is divided by the incident solar irradiance in the Fraunhofer absorption dark line to obtain the relative intensity of SIF at the absorption line.

[0087] In addition, in addition to the above-mentioned sunlight-induced chlorophyll fluorescence intensity extracted by the Fraunhofer dark line, the data reflecting the sunlight-induced chlorophyll fluorescence intensity can also be obtained based on the method of processing reflectance. Specifically, the data reflecting the sunlight-induced chlorophyll fluorescence intensity can also include a fluorescence index, which can be obtained based on reflectance processing. Then, the spectral data and the disease index based on the sample data are processed to obtain effective data, and the process can further include:

[0088] Based on the spectral data, the reflectance index and the reflectance first derivative index are calculated to obtain the fluorescence index, which reflects the sunlight-induced chlorophyll fluorescence intensity. Correlation analysis is performed between the fluorescence index and the disease index to obtain the correlation coefficient corresponding to the fluorescence index. The fluorescence index corresponding to the correlation coefficient that meets the preset condition is determined as the effective fluorescence index. The effective fluorescence index is determined as the effective data.

[0089] In the embodiment of the application, the influence of chlorophyll fluorescence on the reflectance of the red edge region (650-800nm) can be analyzed by calculating the reflectance ratio index and the reflectance first derivative index to obtain the fluorescence index reflecting the sunlight-induced chlorophyll fluorescence intensity. Preferably, the following eight indexes can be selected for calculation, including: 730 / D 706 , R 740 / R 720 , D 705 / D 722 , R 740 / R 800 , R 685 / R 655 Based on the above eight formulas, the fluorescence index can be obtained, and the fluorescence index and the disease index are subjected to correlation analysis, and the results are shown in Table 2. The chlorophyll fluorescence index is the fluorescence index described above, and the reflectance index and the reflectance first derivative index representing the sunlight-induced chlorophyll fluorescence intensity.

[0090] Table 2 Relationship between chlorophyll fluorescence index and disease index

[0091]

[0092] Based on the chlorophyll fluorescence index and the corresponding Pearson correlation coefficient shown in Table 2, it can be determined that the fluorescence index and the disease index are significantly correlated, wherein SIF(O2-A), D 730 / D 706 The correlation coefficients with the disease index are -0.819 and -0.801, respectively, and the absolute values are greater than 0.8, indicating that they have strong correlation with the disease index. Therefore, the wheat stripe rust recognition model trained based on the data can have high recognition accuracy.

[0093] In addition, based on the above method, the spectral data and the disease index of the wheat sample are processed, and the fluorescence index corresponding to the correlation coefficient that meets the preset condition is determined as effective data, wherein the preset condition can be that all correlation coefficients meet the preset threshold and have high correlation with the disease index of the wheat, so that the feature value used in the model training can accurately reflect the feature information of the wheat.

[0094] In addition, in the embodiments of the present application, the effective data also includes vegetation index and fractional order differential spectral features. Among them, there are multiple vegetation indexes that can be used to observe wheat stripe rust, but not all vegetation indexes can accurately or approximately reflect the disease degree of wheat. Therefore, it is necessary to screen the above vegetation indexes to determine the vegetation indexes that can best reflect the disease degree of wheat. Specifically, the growth status of the wheat sample under the stress of wheat stripe rust is characterized based on the spectral data of the wheat sample, and multiple vegetation indexes of the wheat sample are obtained. The correlation between each vegetation index and the disease index is analyzed to obtain the correlation coefficient corresponding to each vegetation index. The vegetation index corresponding to the correlation coefficient that meets the preset condition is determined as effective data.

[0095] Specifically, in the embodiments of the present application, the vegetation index that can characterize the growth status of wheat under the stress of stripe rust can be obtained by analyzing the spectral data. For example, Table 3 shows the vegetation indexes to be selected for wheat stripe rust prediction.

[0096] Table 3 Vegetation index for stripe rust prediction

[0097]

[0098]

[0099] Since different vegetation indices represent different physical and chemical components, the correlation with the disease index of wheat stripe rust is also significantly different. Since the correlation between the effective data in the sample data and the disease index of wheat can affect the recognition accuracy of the finally trained model, the above to-be-selected vegetation indices need to be screened.

[0100] Specifically, the correlation between the to-be-selected vegetation indices and the disease index can be analyzed to obtain a correlation coefficient corresponding to each to-be-selected vegetation index. For details, refer to Table 4, which shows the correlation coefficient corresponding to each to-be-selected vegetation index.

[0101] Table 4 Relationship between disease index and vegetation index

[0102]

[0103] Note: ** indicates 0.001 level extremely significant correlation, R 0.001

[51] = 0447.

[0104] Among them, except that the correlation between MCARI, ARI, PhRI, RVSI and the disease index of wheat stripe rust is not significant, the other 13 to-be-selected vegetation indices are extremely significantly correlated with the disease index. Among them, NPCI, as a normalized index for estimating chlorophyll concentration, has the highest correlation coefficient with the disease index, reaching 0.862. When the wheat leaves are infected by the stripe rust fungus, the nutrients and water in the plant will be consumed, the chlorophyll tissue will be damaged, and the photosynthesis function will decline. By monitoring the NPCI and other indicators, the stress state of the plant and the disease information can be effectively reflected. That is, the above to-be-selected vegetation indices that meet the preset conditions can be determined as effective data, and the preset conditions can be that the correlation coefficient corresponding to the vegetation index reaches a preset threshold value or the like.

[0105] In addition, the embodiment of the present application adopts the method of fractional differential to process the spectral data of the wheat samples, and screens the fractional differential spectral features sensitive to wheat stripe rust as effective data. Among them, the processing based on the spectral data and the disease index to obtain effective data can include:

[0106] Different order differential processing is performed on the spectral data to obtain differential spectral features of different orders; correlation analysis is performed on each order differential spectral feature and the disease index to obtain a correlation coefficient corresponding to each order differential spectral feature; based on the absolute value of the correlation coefficient, a sensitive feature is determined, and the sensitive feature and the disease index are calculated to obtain a projection importance index corresponding to the sensitive feature; based on the projection importance index, a differential spectral feature corresponding to a projection importance index that meets a preset condition is determined; and the differential spectral features are combined to obtain fractional differential spectral features.

[0107] Specifically, in the embodiments of the present application, the 0-2 order fractional order differential processing can be performed on the spectral data of 51 samples in the range of 350-1800 nm, the order interval can be set to 0.1, the differential spectral data of different orders is obtained, and the correlation analysis of the differential spectral data of different orders and the disease index of wheat stripe rust is performed. Further, the maximum correlation coefficient of the differential spectral data under different differential orders and the corresponding wavelength are statistically analyzed. With the increase of the differential order, the maximum correlation coefficient of each order generally shows an increasing trend. For example, when the differential order is 1.2 order, the maximum correlation coefficient reaches a peak value of 0.888, which is increased by 25.2% compared with the maximum correlation coefficient 0.709 of the original spectral data without any processing, and is increased by 6.3% compared with the maximum correlation coefficient 0.835 of the 1.0 order differential spectral feature. This further indicates that the original spectral data without processing and the integer order differential spectral feature cannot deeply mine the effective information related to the disease degree, the fractional order differential spectral feature can extract the disease-sensitive characteristic factor, and the projection importance index between the disease-sensitive characteristic factor and the disease index of wheat stripe rust can be obtained based on the calculation of the disease-sensitive characteristic factor and the disease index. Further, based on the projection importance index, the differential spectral features corresponding to the projection importance indexes reaching the preset threshold in the projection importance index are combined and determined as effective data, which can make the recognition model have better disease index estimation potential.

[0108] In step 123, training is performed based on a preset model structure and the effective data to obtain a plurality of initial recognition models.

[0109] The effective data obtained by the above steps have a high correlation with the disease index and can significantly represent the growth state of wheat. Therefore, the above effective data are used as characteristic values for model training, the disease index is used as a target value, and different model algorithms and model structures are modeled and trained to obtain a plurality of initial recognition models.

[0110] Specifically, one or more of the SIF data, the continuous wavelet feature, the vegetation index, and the fractional order differential spectral feature can be used as characteristic values, the disease index can be used as a target value, the PLSR, BPNN, RF, and XGBoost algorithms selected in the embodiments of the present application can be used for modeling and training, and after the effective data of the target wheat to be measured are input, the target disease index corresponding to the target wheat to be measured can be output.

[0111] In step S124, the initial recognition models are evaluated to obtain an evaluation result corresponding to each initial recognition model.

[0112] All the initial identification models trained based on the above method are evaluated. Optionally, the embodiments of the present application evaluate the above models by K-fold cross-validation. As shown in Table 5, the evaluation results of each initial identification model.

[0113] Based on the evaluation results, it can be determined that no matter which model algorithm and structure is used, the model constructed by using the continuous wavelet feature as the characteristic value is overall superior in accuracy to the model constructed by using the fractional order differential spectral feature and the vegetation index, and the model constructed by using the vegetation index as the characteristic value has the worst accuracy. After the characteristic values are respectively combined with the SIF data, the model constructed by using the continuous wavelet feature still has the highest accuracy. Among the four model algorithms PLSR, BPNN, RF and XGBoost used in the embodiments of the present application, the model constructed by using the PLSR algorithm has the worst recognition accuracy.

[0114] In addition, different combinations of characteristic values correspond to different optimal model algorithms. For example, when the input characteristic is the vegetation index or the vegetation index combined with the SIF data, the model constructed by using the BPNN algorithm is optimal. The model constructed by using the BPNN algorithm and the vegetation index combined with the SIF data has an R 2 which is improved by 9.9% and the RMSE is reduced by 18.1%. When the input characteristic is the fractional order differential spectrum or the fractional order differential spectrum combined with the SIF data, the model constructed by using the RF algorithm is optimal. The model constructed by using the RF algorithm and the fractional order differential spectrum combined with the SIF data has an R 2 which is improved by 4.5% and the RMSE is reduced by 13.4%. When the input characteristic is the continuous wavelet feature or the continuous wavelet feature combined with the SIF data, the model constructed by using the XGBoost algorithm is optimal, and the R 2 are 0.847 and 0.867 respectively, and the RMSEs are 0.114 and 0.104 respectively. The model constructed by using the XGBoost algorithm and the continuous wavelet feature combined with the SIF data has an R 2 which is improved by 7.8% and the RMSE is reduced by 19.3%.

[0115] Table 5 Evaluation results of initial identification models

[0116]

[0117] Step S125: The initial identification model corresponding to the evaluation result that meets the preset condition is determined as the wheat stripe rust identification model.

[0118] Based on the above steps, the evaluation results of different initial identification models can be obtained, and the differences between the initial identification models can be distinguished based on the evaluation results. Then, based on the above evaluation results, it can be judged that the initial identification model corresponding to the evaluation result reaching the preset condition is determined as the final wheat stripe rust identification model.

[0119] In the embodiment of the present application, based on the evaluation results of the above initial identification models, the initial identification model trained by the XGBoost algorithm modeling with continuous wavelet features and SIF data as characteristic values and the disease index of the wheat sample as the target value is the optimal model. Compared with the lowest precision model trained by the PLSR algorithm modeling with only the vegetation index as the characteristic value, R 2 The accuracy of the model is improved by 16.6% and the RMSE is reduced by 32.4%, which better reflects the high accuracy of the identification model.

[0120] Through the screening process of the wheat stripe rust identification model, the optimal identification model trained by the above four models is screened out, and the model with the highest identification accuracy is used to identify the disease of the target wheat to be tested to obtain the target disease index corresponding to the target wheat to be tested. And the data input into the identification model is obtained by analyzing the full-band spectral data of the target wheat to be tested, which can better reflect the disease level and sensitivity to physiological state of the target wheat to be tested, and avoid missing disease-specific factors, thereby improving the accuracy of disease prediction of the target wheat to be tested.

[0121] The following is a practical example proposed in the embodiment of the present application. A wheat field needs to be detected whether it is suffering from wheat stripe rust, so as to arrange the subsequent planting plan for the wheat field. Specifically, ASD Field Pro FR spectrometer and QE 65pro spectrometer can be used to measure the hyperspectral data of the wheat canopy to obtain ASD Field Pro FR spectral data and QE 65pro spectral data, by analyzing and calculating the above two kinds of spectral data, the sun-induced chlorophyll fluorescence data and continuous wavelet features of the wheat field are obtained. Then, the sun-induced chlorophyll fluorescence data and the continuous wavelet features are input into the wheat stripe rust identification model for processing to obtain the disease index of the target to be detected of the wheat field, and determine whether the wheat field is suffering from wheat stripe rust.

[0122] The wheat stripe rust identification device provided in the embodiment of the present application is described below. The wheat stripe rust identification device described below corresponds to the wheat stripe rust identification method described above.

[0123] First, the wheat stripe rust identification device is combined with the wheat stripe rust identification method described above. Figure 3 The wheat stripe rust identification device is introduced, such as Figure 3The wheat stripe rust identification device can include:

[0124] The wheat information acquisition unit 100 is configured to acquire sunlight-induced chlorophyll fluorescence data and continuous wavelet features of a target wheat to be tested;

[0125] The wheat information processing unit 200 is configured to input the sunlight-induced chlorophyll fluorescence data and the continuous wavelet features into a wheat stripe rust identification model for processing, so as to obtain a target disease index matched with the target wheat to be tested output by the wheat stripe rust identification model;

[0126] The wheat stripe rust identification model is a model trained by taking the continuous wavelet features and the sunlight-induced chlorophyll fluorescence data of a wheat sample as characteristic values and taking a disease index corresponding to the wheat sample as a target value, and the wheat stripe rust identification model is a regression prediction model determined based on iterative training.

[0127] Optionally, the device further includes:

[0128] The sample data acquisition unit is configured to acquire spectral data and disease indexes corresponding to a wheat sample at different growth stages;

[0129] The effective data acquisition unit is configured to process the spectral data and the disease indexes to obtain effective data, the effective data including one or more of sunlight-induced chlorophyll fluorescence data, continuous wavelet features, fractional differential spectral features, and vegetation indexes, and the sunlight-induced chlorophyll fluorescence data including sunlight-induced chlorophyll fluorescence relative intensity and fluorescence index;

[0130] The sample data set acquisition unit is configured to generate a sample data set based on the effective data and the disease indexes, each sample data in the sample data set including at least one effective data feature and data labeled with the disease index;

[0131] The model training unit is configured to train a model based on each preset model structure and the sample data set to obtain an initial identification model matched with each preset model and each sample data, wherein the model training process takes at least one effective data in each sample data as a characteristic value and takes the disease index as a target value, and an initial identification model is obtained after training;

[0132] The model evaluation unit is configured to evaluate the initial identification model to obtain an evaluation result corresponding to each initial identification model;

[0133] The target model determination unit is configured to compare the evaluation results and determine an initial identification model corresponding to an evaluation result meeting a preset condition as a wheat stripe rust identification model.

[0134] Optionally, the effective data comprises continuous wavelet features, wherein the effective data acquisition unit comprises:

[0135] a spectral decomposition subunit configured to process the spectral data to obtain continuous wavelet features at different decomposition scales;

[0136] a wavelet feature correlation analysis subunit configured to statistically analyze a correlation between the continuous wavelet features and the disease index to obtain absolute values of correlation coefficients corresponding to the continuous wavelet features at different decomposition scales;

[0137] a sensitive feature determination first subunit configured to determine sensitive features based on the absolute values of correlation coefficients;

[0138] a projection importance analysis subunit configured to calculate the sensitive features and the disease index to obtain a projection importance index corresponding to the sensitive features;

[0139] an effective wavelet feature determination subunit configured to determine, based on the projection importance index, a continuous wavelet feature corresponding to a projection importance index that meets a preset condition as effective data.

[0140] Preferably, the effective data comprises sunlight-induced chlorophyll fluorescence data, and the sunlight-induced chlorophyll fluorescence data comprises relative intensities of sunlight-induced chlorophyll fluorescence, wherein the effective data acquisition unit comprises:

[0141] an absolute intensity acquisition subunit configured to calculate, based on the spectral data, a first absolute intensity of sunlight-induced chlorophyll fluorescence corresponding to a first oxygen absorption waveband and a second absolute intensity of sunlight-induced chlorophyll fluorescence corresponding to a second oxygen absorption waveband using irradiance and radiance data;

[0142] a relative intensity acquisition subunit configured to process the first absolute intensity of sunlight-induced chlorophyll fluorescence and the second absolute intensity of sunlight-induced chlorophyll fluorescence to obtain a first relative intensity of sunlight-induced chlorophyll fluorescence corresponding to the first absolute intensity of sunlight-induced chlorophyll fluorescence and a second relative intensity of sunlight-induced chlorophyll fluorescence corresponding to the second absolute intensity of sunlight-induced chlorophyll fluorescence;

[0143] a chlorophyll correlation calculation subunit configured to respectively analyze correlations between the first relative intensity of sunlight-induced chlorophyll fluorescence and the second relative intensity of sunlight-induced chlorophyll fluorescence and the disease index to obtain correlation coefficients corresponding to the first relative intensity of sunlight-induced chlorophyll fluorescence and the second relative intensity of sunlight-induced chlorophyll fluorescence, respectively;

[0144] The effective chlorophyll fluorescence data determination subunit is configured to determine, by comparing the correlation coefficients, a relative intensity of sunlight-induced chlorophyll fluorescence corresponding to a correlation coefficient meeting a preset condition as effective data.

[0145] Optionally, the effective data includes sunlight-induced chlorophyll fluorescence data, and the sunlight-induced chlorophyll data includes a fluorescence index, and the method further includes:

[0146] The fluorescence index acquisition subunit is configured to calculate a reflectance index and a reflectance first derivative index based on the spectral data to obtain a fluorescence index reflecting a sunlight-induced chlorophyll fluorescence intensity.

[0147] The fluorescence index correlation analysis subunit is configured to perform correlation analysis on the fluorescence index and the disease index to obtain a correlation coefficient corresponding to the fluorescence index.

[0148] The effective fluorescence index determination subunit is configured to determine, by comparing the correlation coefficients, a fluorescence index corresponding to a correlation coefficient meeting a preset condition as an effective fluorescence index.

[0149] Optionally, the effective data includes a vegetation index, and the effective data acquisition unit includes:

[0150] The vegetation index acquisition subunit is configured to obtain a plurality of vegetation indexes of the wheat sample based on the growth status of the wheat sample under the stress of the wheat stripe rust represented by the spectral data.

[0151] The vegetation index correlation analysis subunit is configured to perform correlation analysis on the vegetation index and the disease index to obtain a correlation coefficient corresponding to each vegetation index.

[0152] The effective vegetation index determination subunit is configured to determine, by comparing the correlation coefficients corresponding to each vegetation index, a vegetation index corresponding to a correlation coefficient meeting a preset condition as effective data.

[0153] Optionally, the effective data includes a fractional derivative spectral feature, and the effective data acquisition unit includes:

[0154] The derivative spectral data acquisition subunit is configured to perform derivative processing of different orders on the spectral data to obtain derivative spectral features of different orders.

[0155] The derivative spectral data correlation analysis subunit is configured to perform correlation analysis on each order of the derivative spectral feature and the disease index to obtain a correlation coefficient corresponding to each order of the derivative spectral feature.

[0156] The sensitive feature determination second subunit is configured to determine a sensitive feature based on an absolute value of the correlation coefficient.

[0157] a differential spectral data importance analysis subunit configured to calculate the sensitive features and the disease index to obtain a projection importance index corresponding to the sensitive features;

[0158] an effective differential spectral data determination subunit configured to determine a differential spectral feature corresponding to a projection importance index that meets a preset condition based on the projection importance index;

[0159] an effective data acquisition subunit configured to combine the differential spectral features to obtain fractional differential spectral features.

[0160] The embodiments of the present application filter the optimal recognition model through the filtering process of the wheat stripe rust recognition model, and the model with the highest recognition accuracy is used to recognize the disease of the target wheat to be tested to obtain a target disease index corresponding to the target wheat to be tested. The data input into the recognition model is obtained through analysis of the full-band spectral data of the target wheat to be tested, which can better reflect the disease degree of the target wheat to be tested and the feature information sensitive to the physiological state, avoid missing disease-specific factors of the wheat stripe rust, and improve the accuracy of disease prediction of the target wheat to be tested.

[0161] The wheat stripe rust recognition device provided by the embodiments of the present application can be applied to a wheat stripe rust recognition device, which can be a terminal computing device connected to a spectral detection instrument. Figure 4 A structural schematic diagram of the wheat stripe rust recognition device is shown, and the structure of the wheat stripe rust recognition device can include at least one processor 10, at least one memory 20, at least one communication bus 30, and at least one communication interface 40. Figure 4

[0162] In the embodiments of the present application, the number of the processor 10, the memory 20, the communication bus 30, and the communication interface 40 is at least one, and the processor 10, the memory 20, the communication bus 30, and the communication interface 40 complete communication between each other through the communication bus 30.

[0163] The processor 10 can be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application, etc.

[0164] The memory 20 can include a RAM memory and can also include a non-volatile memory such as at least one disk memory.

[0165] ​The memory stores a program, and the processor can invoke the program stored in the memory, and the program is used to implement each processing step in the above-mentioned wheat stripe rust identification method.

[0166] The embodiment of the present application further provides a storage medium, which can store computer executable instructions loaded and executed by a processor. The instructions can be in the form of a computer program, which is used to implement each step in the above-mentioned wheat stripe rust identification method.

[0167] Finally, it needs to be explained that, in this document, the relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment. Without more limitations, the element defined by the statement "including a" does not exclude the presence of other identical elements in the process, method, article or equipment including the element.

[0168] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same or similar parts between various embodiments can be referred to each other.

[0169] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method of identifying wheat stripe rust, characterized by, The method comprises the following steps: obtaining the sun-induced chlorophyll fluorescence data and continuous wavelet features of the target wheat to be tested; inputting the sun-induced chlorophyll fluorescence data and continuous wavelet features into a wheat stripe rust identification model for processing to obtain a target disease index matched with the target wheat to be tested output by the wheat stripe rust identification model; the wheat stripe rust identification model is a model trained by taking the continuous wavelet features and sun-induced chlorophyll fluorescence data of a wheat sample as characteristic values and taking the disease index corresponding to the wheat sample as a target value, and the wheat stripe rust identification model is a regression prediction model determined based on iterative training; the wheat stripe rust identification method further comprises the following steps: obtaining the spectral data and disease index of a wheat sample at different growth stages; processing the spectral data and the disease index to obtain effective data, wherein the effective data comprises sun-induced chlorophyll fluorescence data, continuous wavelet features, fractional differential spectral features, and vegetation indexes, and the sun-induced chlorophyll fluorescence data comprises sun-induced chlorophyll fluorescence relative intensity and fluorescence index; generating a sample data set based on the effective data and the disease index, wherein each sample data in the sample data set comprises at least one effective data feature and data labeled with the disease index; training a model based on each preset model structure and the sample data set to obtain an initial identification model matched with each preset model and each sample data, wherein the model training process takes at least one effective data in each sample data as a characteristic value and takes the disease index as a target value to obtain the initial identification model after training; evaluating the initial identification model to obtain an evaluation result corresponding to each initial identification model; comparing the evaluation results to determine the initial identification model corresponding to the evaluation result that meets a preset condition as the wheat stripe rust identification model; the effective data comprises continuous wavelet features, and the processing of the spectral data and the disease index to obtain effective data comprises the following steps: processing the spectral data to obtain continuous wavelet features at different decomposition scales; statistically analyzing the correlation between the continuous wavelet features and the disease index to obtain correlation coefficient absolute values corresponding to the continuous wavelet features at different decomposition scales; determining sensitive features based on the correlation coefficient absolute values; calculating the sensitive features and the disease index to obtain a projection importance index corresponding to the sensitive features; determining the continuous wavelet features corresponding to the projection importance index that meets the preset condition as the effective data based on the projection importance index; the effective data comprises fractional differential spectral features, and the processing of the spectral data and the disease index to obtain effective data comprises the following steps: performing differential processing on the spectral data to obtain differential spectral features of different orders; analyzing the correlation between the differential spectral features of each order and the disease index to obtain correlation coefficients corresponding to the differential spectral features of each order; determine a sensitive feature based on an absolute value of the correlation coefficient; calculate the sensitive feature and the disease index to obtain a projection importance index corresponding to the sensitive feature; determine a differential spectral feature corresponding to a projection importance index reaching a preset condition based on the projection importance index; combine the differential spectral feature to obtain a fractional differential spectral feature.

2. The method of claim 1, wherein, The effective data includes sunlight-induced chlorophyll fluorescence data, and the sunlight-induced chlorophyll fluorescence data includes sunlight-induced chlorophyll fluorescence relative intensity, wherein the processing based on the spectral data and the disease index to obtain effective data includes: Based on the spectral data, the irradiance and radiance data are used to calculate, respectively, to obtain the first sunlight-induced chlorophyll fluorescence absolute intensity corresponding to the first oxygen absorption wave band, and the second sunlight-induced chlorophyll fluorescence absolute intensity corresponding to the second oxygen absorption wave band; Based on the first sunlight-induced chlorophyll fluorescence absolute intensity and the second sunlight-induced chlorophyll fluorescence absolute intensity, the first sunlight-induced chlorophyll fluorescence relative intensity corresponding to the first sunlight-induced chlorophyll fluorescence absolute intensity and the second sunlight-induced chlorophyll fluorescence relative intensity corresponding to the second sunlight-induced chlorophyll fluorescence absolute intensity are obtained; The first sunlight-induced chlorophyll fluorescence relative intensity and the second sunlight-induced chlorophyll fluorescence relative intensity are respectively analyzed for correlation with the disease index, and the correlation coefficients corresponding to the first sunlight-induced chlorophyll fluorescence relative intensity and the second sunlight-induced chlorophyll fluorescence relative intensity are obtained respectively; By comparing the correlation coefficients, the sunlight-induced chlorophyll fluorescence relative intensity corresponding to the correlation coefficient reaching the preset condition is determined as the effective data.

3. The method of claim 2, wherein, The effective data includes sunlight-induced chlorophyll fluorescence data, and the sunlight-induced chlorophyll fluorescence data further includes a fluorescence index, and further includes: Based on the spectral data, the reflectance index and the reflectance first derivative index are calculated to obtain the fluorescence index, which reflects the sunlight-induced chlorophyll fluorescence intensity; Based on the fluorescence index and the disease index, a correlation analysis is performed to obtain a correlation coefficient corresponding to the fluorescence index; By comparing the correlation coefficients, the fluorescence index corresponding to the correlation coefficient reaching the preset condition is determined as the effective fluorescence index; The effective fluorescence index is determined as the effective data.

4. The method of claim 1, wherein, The effective data includes vegetation index, wherein the processing based on the spectral data and the disease index to obtain effective data includes: Based on the spectral data, the growth status of the wheat sample under the stress of wheat stripe rust is obtained to obtain a plurality of vegetation indexes of the wheat sample; The vegetation index and the disease index are analyzed for correlation to obtain a correlation coefficient corresponding to each vegetation index; By comparing the correlation coefficients corresponding to each vegetation index, the vegetation index corresponding to the correlation coefficient reaching the preset condition is determined as the effective data.

5. A device for identifying wheat stripe rust, characterized by, It includes: A wheat information acquisition unit is configured to acquire sunlight-induced chlorophyll fluorescence data and continuous wavelet features of a target wheat to be tested. The wheat information processing unit is configured to input the sunlight-induced chlorophyll fluorescence data and the continuous wavelet feature into a wheat stripe rust identification model for processing, so as to obtain a target disease index matched with the target wheat to be detected, which is output by the wheat stripe rust identification model. The wheat stripe rust identification model is a regression prediction model determined based on iterative training, and is trained by taking the continuous wavelet feature and the sunlight-induced chlorophyll fluorescence data of the wheat sample as characteristic values and taking the disease index corresponding to the wheat sample as a target value. The wheat stripe rust identification device further comprises: A sample data acquisition unit configured to acquire spectral data and disease indexes corresponding to wheat samples at different growth stages. An effective data acquisition unit configured to process the spectral data and the disease indexes to obtain effective data, wherein the effective data comprises sunlight-induced chlorophyll fluorescence data, continuous wavelet features, fractional differential spectral features, and vegetation indexes, and the sunlight-induced chlorophyll fluorescence data comprises relative intensity of sunlight-induced chlorophyll fluorescence and fluorescence index. A sample data set acquisition unit configured to generate a sample data set based on the effective data and the disease indexes, wherein each sample data in the sample data set comprises at least one effective data feature and data labeled with the disease index. A model training unit configured to train a model based on each preset model structure and the sample data set to obtain an initial identification model matched with each preset model and each sample data, wherein the model training process takes at least one effective data in each sample data as a characteristic value and takes the disease index as a target value to obtain the initial identification model after training. A model evaluation unit configured to evaluate the initial identification model to obtain an evaluation result corresponding to each initial identification model. A target model determination unit configured to compare the evaluation results and determine the initial identification model corresponding to the evaluation result meeting a preset condition as the wheat stripe rust identification model. The effective data comprises continuous wavelet features, and the effective data acquisition unit comprises: A spectral decomposition subunit configured to process the spectral data to obtain continuous wavelet features at different decomposition scales. A wavelet feature correlation analysis subunit configured to statistically analyze the correlation between the continuous wavelet features and the disease indexes to obtain correlation coefficient absolute values corresponding to the continuous wavelet features at different decomposition scales. A sensitive feature determination first subunit configured to determine sensitive features based on the correlation coefficient absolute values. A projection importance analysis subunit configured to calculate the sensitive features and the disease indexes to obtain projection importance indexes corresponding to the sensitive features. An effective wavelet feature determination subunit configured to determine, based on the projection importance indexes, that the continuous wavelet features corresponding to the projection importance indexes meeting a preset condition are effective data. The effective data comprises fractional differential spectral features, and the effective data acquisition unit comprises: The differential spectrum data acquisition subunit is configured to perform differential processing of different orders on the spectrum data to obtain differential spectrum features of different orders. The differential spectrum data correlation analysis subunit is configured to perform correlation analysis on each order of the differential spectrum features and the disease index to obtain a correlation coefficient corresponding to each order of the differential spectrum features. The sensitive feature determination second subunit is configured to determine a sensitive feature based on an absolute value of the correlation coefficient. The differential spectrum data importance analysis subunit is configured to perform calculation on the sensitive feature and the disease index to obtain a projection importance index corresponding to the sensitive feature. The effective differential spectrum data determination subunit is configured to determine a differential spectrum feature corresponding to a projection importance index that meets a preset condition based on the projection importance index. The effective data acquisition subunit is configured to combine the differential spectrum features to obtain a fractional differential spectrum feature.

6. A wheat stripe rust identifying apparatus, characterized by, The processor and the memory are included. The processor is configured to execute a program stored in the memory. The memory is configured to store a program, and the program implements the method for identifying wheat stripe rust according to any one of claims 1-4. The storage medium stores computer executable instructions, and the computer executable instructions are loaded and executed by the processor to implement the method for identifying wheat stripe rust according to any one of claims 1-4.

7. A storage medium, characterized by ​

Images