Cotton verticillium wilt resistance remote sensing accurate evaluation method and system based on disease dynamic development rate
By constructing a remote sensing assessment method for cotton Verticillium wilt resistance based on the dynamic development rate of the disease, and utilizing multi-source datasets and a random forest model, the subjectivity and error problems of traditional assessment methods are solved, and rapid and accurate resistance assessment is achieved.
Patent Information
- Application Number
- CN202511477328.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-16
- Publication Date
- 2026-01-06
AI Technical Summary
Traditional methods for assessing resistance to cotton Verticillium wilt are subjective, inefficient, and difficult to perform accurately and quickly. They also ignore differences in disease development rates, leading to large errors in resistance assessment.
By collecting multi-source datasets, performing hyperspectral data preprocessing, calculating the spectral characteristics of dynamic development rate, and combining Cohen's effect size and SBS method to screen sensitive features, a random forest classification model is constructed for resistance assessment.
This technology enables the precise and rapid screening of cotton wilt-resistant varieties in a short period of time, improving the accuracy and efficiency of the assessment while reducing manpower consumption.
Smart Images
Figure CN121280902A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of disease assessment technology, and more specifically to a remote sensing method and system for accurate assessment of cotton Verticillium wilt resistance based on the dynamic development rate of the disease. Background Technology
[0002] Cotton, as one of the world's most important economic crops, is not only the foundation of the textile industry but also a vital source of livelihood for farmers. From planting to textiles and deep processing, the cotton industry chain spans multiple stages. However, this chain also faces numerous severe challenges. Verticillium wilt (VW), often referred to as "cotton field cancer," is a significant soil-borne vascular disease threatening cotton growth and yield. The VW pathogen infects plants through the roots, affecting the vascular system and causing wilting, premature aging, and in severe cases, substantial yield reduction or even crop failure. Due to the wide host range, long survival time in the soil, and latent infection characteristics of the VW pathogen, traditional control methods such as crop rotation, soil treatment, and chemical control are often ineffective. Developing and applying disease-resistant varieties is widely considered the most economical, effective, and environmentally friendly way to control cotton Verticillium wilt.
[0003] Accurate and rapid assessment of cotton Verticillium wilt resistance is crucial for improving the efficiency of screening high-quality resistant cotton varieties and identifying superior resistance genes. However, current traditional methods for assessing cotton Verticillium wilt resistance still have the following drawbacks: 1) Professional investigators survey the disease incidence and disease index of the plant population in the field, and calculate the disease index based on the severity level (0-4) of each plant in the population during the peak period of cotton Verticillium wilt. However, the disease level classification method is relatively crude, highly subjective, and the evaluation criteria are not accurate and difficult to unify. In addition, multiple surveys are inefficient and consume a lot of manpower, making it difficult to accurately and quickly screen cotton Verticillium wilt resistant varieties. 2) Traditional remote sensing technology for Verticillium wilt resistance assessment mostly screens cotton Verticillium wilt resistant varieties by using the canopy spectral characteristics of a single or a few times during the peak period of Verticillium wilt. However, there is no unified standard for the time window for obtaining the canopy spectrum. The timing of disease onset and the rate of disease growth vary among different resistant varieties. Therefore, this resistance assessment method has a certain identification error. 3) Different resistant varieties have significant differences in disease onset time and disease rate. Therefore, the disease development rate is of great significance for accurate and rapid resistance assessment. At present, cotton Verticillium wilt resistance assessment based on time-series remote sensing data effectively avoids the error caused by a single assessment, but ignores the role of the difference in disease development rate in accurate assessment, making it difficult to complete accurate and rapid resistance assessment in a short period of time.
[0004] Therefore, exploring the differences in disease development rates among different resistant varieties by combining the dynamic development rate of the disease in the early stage of disease infection, and screening and determining the spectral temporal characteristics of sensitive dynamic development rates, is an effective method to achieve accurate and rapid resistance assessment for specific biological stresses, and is also a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0005] In view of this, the present invention provides a method and system for accurate remote sensing assessment of cotton Verticillium wilt resistance based on the dynamic development rate of the disease, which solves the problems existing in the background technology.
[0006] To achieve the above objectives, the present invention provides the following technical solution: A remote sensing method for precise assessment of cotton Verticillium wilt resistance based on the dynamic development rate of the disease includes the following steps: S1. Collect ground data, remote sensing data, and resistance assessment data of different resistant varieties within the test area to construct a multi-source dataset; S2. Preprocess the hyperspectral data in the multi-source dataset to obtain time-series spectral reflectance and vegetation index. Calculate the dynamic development rate spectral characteristics based on the spectral reflectance and spectral vegetation index at adjacent time points. S3. Based on Cohen's effect size, screen for significant differences in dynamic development rate characteristics among different resistant varieties; S4. Combining Cohen's effect size confidence intervals with the SBS method, further screen for sensitive dynamic development rate characteristics; S5. Using the sensitive dynamic development rate characteristics as input and the resistance level as output, construct a random forest classification model and train the model. Use the trained cotton Verticillium wilt resistance assessment model to screen resistant varieties.
[0007] Optionally, the multi-source dataset in S1 is specifically as follows: Ground data includes physiological data and canopy structure data of leaves of different genotypes. Physiological data includes chlorophyll content, carotenoid content, anthocyanin, and water content, while canopy structure data includes leaf area index. The remote sensing data consists of canopy-scale hyperspectral data collected periodically by drones in the early stages of disease infection; The resistance assessment data includes cotton plant disease grading, relative disease index of the plant population, and resistance assessment results.
[0008] Optionally, the specific methods for obtaining resistance assessment data are as follows: Before and after the acquisition of hyperspectral data at the canopy scale, a resistance assessment survey was conducted on the experimental area. Based on the resistance survey standards, the severity of cotton Verticillium wilt was divided into 5 levels, of which level 0 is healthy, level 1 is mild disease, level 2 is moderate disease, level 3 is severe disease, and level 4 is death. The formula for calculating the disease index is:
[0009] In the formula: This indicates the disease index of the identified variety. Indicates the corresponding disease level. This indicates the number of diseased plants at each disease level. Indicates the total number of plants; When the disease index of the susceptible control variety is between 35.1 and 65.0, the relative disease index is used to measure the resistance of the variety. The calculation formula is as follows:
[0010] In the formula: This indicates a relative disease index. K This represents the correction factor; 50.0 is the standard disease index for the infected control group. This indicates the disease severity index of the control group in this assessment; For each time point The average relative disease index was obtained by taking the average value, and the cotton resistance assessment results of the test varieties were divided according to the resistance survey standards.
[0011] Optionally, in S2, the preprocessing operation specifically includes: The hyperspectral data were preprocessed, and the images were stitched together using ENVI 5.3 software. The DN values were converted into BRF bidirectional reflectance factors using a diffuse reflectance plate to obtain the preprocessed hyperspectral image of the test area. Savitzky-Golay filters were used to remove spectral noise from hyperspectral images to obtain time-series hyperspectral data.
[0012] Optionally, in S2, the formula for calculating the spectral characteristics of the dynamic evolution rate is:
[0013] In the formula: express Spectral reflectance over a time period , Representing time points , wavelength reflectivity, express Spectral vegetation index over a period of time , Representing time points , The vegetation index value, Indicates a time interval.
[0014] Optionally, in S3, Cohen's effect size is calculated as follows:
[0015] filter The significant differences were identified, and the significant differences in wavelength and vegetation index were obtained.
[0016] Optionally, in S5, the classification results of the random forest classification model are evaluated using five metrics: Overall Accuracy, Precision, Recall, F1-score, and Kappa coefficient. Overall Accuracy represents the overall accuracy of the model's classification; Precision measures the accuracy of positive class predictions; Recall reflects the ability to identify positive class samples; F1-score is the harmonic mean of precision and recall; and the Kappa coefficient assesses the consistency between the model's predictions and the true classification.
[0017] A remote sensing precision assessment system for cotton Verticillium wilt resistance based on the dynamic development rate of the disease, comprising executing any of the above-described methods for remote sensing precision assessment of cotton Verticillium wilt resistance based on the dynamic development rate of the disease, including: The data acquisition module is used to collect ground data, remote sensing data, and resistance assessment data of different resistant varieties within the test area to construct a multi-source dataset; The feature extraction module is used to preprocess the hyperspectral data in the multi-source dataset, obtain the time-series spectral reflectance and vegetation index, and calculate the dynamic development rate spectral features based on the spectral reflectance and spectral vegetation index of adjacent time points. The first screening module is used to screen for significant differences in dynamic development rate characteristics among different resistant varieties based on Cohen's effect size. The second screening module is used to combine Cohen's effect size confidence interval with the SBS method to further screen for sensitive dynamic development rate characteristics; The model building and evaluation module is used to construct a random forest classification model with sensitive dynamic development rate characteristics as input and resistance level as output, and to train the model. The trained cotton Verticillium wilt resistance evaluation model is then used to screen resistant varieties.
[0018] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method and system for accurate remote sensing assessment of cotton Verticillium wilt resistance based on the dynamic development rate of the disease. Combining the differences in the dynamic development rate of different resistant varieties during the occurrence and development of cotton Verticillium wilt, and based on the ground-measured physiological and non-physiological parameters, the spectral characteristics of the sensitive dynamic development rate are identified, and a cotton Verticillium wilt resistance assessment model is constructed. This provides a feasible technical support solution for realizing large-scale, rapid and accurate resistance assessment of cotton Verticillium wilt, and is of great significance for exploring high-quality germplasm resources and reducing cotton Verticillium wilt losses economically and efficiently. Attached Figure Description
[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0020] Figure 1 The process flow diagram of the remote sensing precision assessment method for cotton Verticillium wilt resistance based on the dynamic development rate of the disease provided by this invention. Detailed Implementation
[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0022] Traditional resistance assessment methods based on field surveys suffer from drawbacks such as high subjectivity, long timeframes, low efficiency, and high manpower consumption, making it difficult to accurately and quickly assess large-scale cotton Verticillium wilt resistance. Current remote sensing technologies mostly assess resistance at single time points, neglecting the differences in physiological changes during the disease process of different resistant varieties, thus failing to accurately reflect varietal resistance. Furthermore, resistance assessments using time-series remote sensing data still focus on long periods during peak disease occurrence, ignoring differences in the dynamic development rate of the disease and failing to complete resistance assessments quickly enough.
[0023] To address the aforementioned problems, this invention discloses a remote sensing method for accurately assessing cotton Verticillium wilt resistance based on the dynamic development rate of the disease. Figure 1 As shown, it includes the following steps: S1. Collect ground data, remote sensing data, and resistance assessment data of different resistant varieties within the test area to construct a multi-source dataset; S2. Preprocess the hyperspectral data in the multi-source dataset to obtain time-series spectral reflectance and vegetation index. Calculate the dynamic development rate spectral characteristics based on the spectral reflectance and spectral vegetation index at adjacent time points. S3. Based on Cohen's effect size, screen for significant differences in dynamic development rate characteristics among different resistant varieties; S4. Combining Cohen's effect size confidence intervals with the SBS method, further screen for sensitive dynamic development rate characteristics; S5. Using the sensitive dynamic development rate characteristics as input and the resistance level as output, construct a random forest classification model and train the model. Use the trained cotton Verticillium wilt resistance assessment model to screen resistant varieties.
[0024] based on Figure 1 As shown in the flowchart, this embodiment uses a UAV time-series hyperspectral system to investigate the differences in the dynamic change rate of different resistant varieties of Verticillium wilt phenotypes in the early stage of infection. Through significant difference analysis and SBS method screening, the spectral time-series characteristics of sensitive dynamic development rate in the early stage of infection were determined, which can be used to achieve accurate and rapid resistance assessment of different genotypes of cotton Verticillium wilt.
[0025] Next, for Figure 1 The process shown is described in detail to further understand the technical solution to be protected by this invention.
[0026] I. Collecting multi-source data on different resistant varieties
[0027] In this embodiment, the collected data includes ground data, remote sensing data, and resistance assessment data, wherein: Ground data includes physiological data and canopy structure data of leaves of different genotypes. Physiological data includes chlorophyll content, carotenoid content, anthocyanin, and water content. Canopy structure data includes leaf area index. The remote sensing data consists of canopy-scale hyperspectral data collected periodically by drones in the early stages of disease infection; The resistance assessment data includes cotton plant disease grading, relative disease index of the plant population, and resistance assessment results.
[0028] The specific data collection method is as follows: Before and after drone data collection, breeding experts conducted a resistance assessment survey of the experimental area. A total of 80 varieties were tested, with approximately 45 cotton plants per variety. According to the resistance survey standard (NY / T2952-2016), the severity of cotton Verticillium wilt was divided into 5 levels, where level 0 is healthy (0 cases), level 1 is mild disease (0-1 / 3), level 2 is moderate disease (1 / 3-2 / 3), level 3 is severe disease (2 / 3-1), and level 4 is death. The formula for calculating the disease index is:
[0029] In the formula: This indicates the disease index of the identified variety. Indicates the corresponding disease level. This indicates the number of diseased plants at each disease level. Indicates the total number of plants; When the disease index of the susceptible control variety is between 35.1 and 65.0, the relative disease index is used to measure the resistance of the variety. The calculation formula is as follows:
[0030] In the formula: This indicates a relative disease index. K This represents the correction factor; 50.0 is the standard disease index for the infected control group. This indicates the disease severity index of the control group in this assessment; For each time point The mean relative disease index (ARDI) was obtained by taking the average. The cotton resistance assessment results of the test varieties were divided according to the resistance survey standard (NY / T2952-2016), as shown in Table 1.
[0031] Table 1. Criteria for classifying cotton variety resistance levels
[0032] II. Extracting Spectral Characteristics of Dynamic Evolution Rate
[0033] In S2 of this embodiment, the preprocessing operation specifically includes: MegaCube_V2.15.0 software was used for preliminary preprocessing of the hyperspectral data. ENVI 5.3 software was used for image stitching, and the DN values were converted into BRF (Bidirectional Reflectance Factor) using a diffuse reflectance plate to obtain preprocessed hyperspectral images of the experimental area. To smooth the spectral curves, a Savitzky-Golay filter (second-order polynomial, filter window 11) was used to remove spectral noise from the hyperspectral images, obtaining time-series hyperspectral data. Before selecting sensitive bands, envelope processing was performed on the data.
[0034] Furthermore, 12 commonly used spectral vegetation indices related to plant diseases in remote sensing monitoring were selected. These indices are related to plant pigments, structure (e.g., NDVI, RDVI), water status (e.g., WI), red edge, and photosynthetic physiology (e.g., Healthy index, CIRed_edge). A complete list of indices is shown in Table 2. Wavelengths at the peaks, troughs, and red edge positions in the visible light region were obtained for each time point at all experimental sites. Wavelengths with higher frequencies were selected as representative, and these wavelengths were used to replace the relevant original wavelengths in the vegetation indices according to the definition of vegetation index wavelengths. For cases where the required wavelengths were not completely consistent in the data, the nearest wavelength was used to replace the reflectance value of that wavelength, ultimately obtaining the time-series spectral vegetation index data.
[0035] Table 2 Spectral Vegetation Indices and Construction Methods
[0036] The rate of change of spectral reflectance and vegetation index in adjacent time periods is obtained based on the spectral reflectance and vegetation index of each test point at a single time point. In S2, the formula for calculating the spectral characteristics of the dynamic development rate is:
[0037] In the formula: express Spectral reflectance over a time period , Representing time points , wavelength reflectivity, express Spectral vegetation index over a period of time , Representing time points , The vegetation index value, This indicates a time interval. Furthermore, "bands" can be replaced with specific wavelengths in the spectrum; the data acquisition rate calculation method and spectral characteristics are consistent with those used for ground-based data.
[0038] III. Dynamic Development Rate Characteristics of Significant Differences in Cotton Verticillium Wilt Resistance Assessment Based on Cohen's Effect Size
[0039] This embodiment defines an effect threshold d, calculated as follows, to determine whether there are significant differences between different resistance test points:
[0040] filter The significant differences were identified, and the significant differences in wavelength and vegetation index were obtained.
[0041] Table 3 shows the feature set with significant differences selected based on effect size. Significant differences in wavelengths only existed among different resistances in the early stages of disease (71-81 days after emergence). The significant differences were concentrated around 510 nm (the characteristic wavelength of the green peak) and 680 nm (the chlorophyll absorption trough), and the rate of change near 680 nm (chlorophyll absorption trough) was more pronounced in different resistant varieties than near 510 nm (the characteristic wavelength of the green peak). Specifically, as disease severity increased, the spectral reflectance in both wavelength ranges showed an upward trend, and the increase in reflectance in susceptible varieties in the early stages of disease was significantly greater than that in resistant varieties. Therefore, the magnitude of the significant difference was negative, possibly due to the increased canopy scattering caused by the destruction of leaf mesophyll structure due to pathogen infection. Further time-series analysis of vegetation indices revealed that the early differences mainly stemmed from the synergistic response of red edge and photosynthetic physiological parameters (CIRed_edge index, d=1.59) and pigment and structural parameters (PSSRc index, d=1.40). However, in the mid-stage of the disease (81-101 days after emergence), although chlorophyll content showed different trends (chl, d=0.79), the canopy leaf area decline trend converged (LAI, d=-0.01), and this was also the case in the later stages of the disease, leading to a lack of significant differences in the spectral vegetation index response.
[0042] Table 3. Feature sets with significant differences selected based on effect size.
[0043] IV. Further screening of sensitive features based on the SBS (Sequential Backward Selection) method
[0044] Before applying the SBS method, more robust features were selected by combining the confidence interval of Cohen's effect size (excluding 0). SBS is a feature selection method that starts with all features and removes one feature at a time that has the least impact on the model until the optimal feature subset is found. The advantage of this method is that it can remove redundant features and select the most representative set of resistance assessment features. Finally, it combines the dynamic development rate of the disease to construct a resistance assessment model. Table 4 shows the feature set after processing using the SBS method.
[0045] Table 4 Feature sets after processing by the SBS method
[0046] V. Model Construction
[0047] To further determine the effectiveness of the selected features, the classic Random Forest (RF) classification model was chosen, and the final resistance assessment model and hyperspectral features were selected based on the results.
[0048] Random Forest (RF): It achieves classification by integrating the votes of multiple decision trees. Each tree is trained based on different subsets of data and feature subsets, thereby improving the model's generalization ability. Due to its randomness and diversity, Random Forest can effectively avoid overfitting and improve the robustness of the model. It is suitable for binary classification problems with large-scale and high-noise data.
[0049] The training set (n=96) and validation set (n=23) were partitioned in a 4:1 ratio. The classification results of the random forest classification model were evaluated using five metrics: Overall Accuracy (OA), Precision, Recall, F1-score, and Kappa coefficient. Overall Accuracy represents the overall accuracy of the model's classification; Precision measures the accuracy of positive class predictions; Recall reflects the ability to identify positive samples; F1-score is the harmonic mean of precision and recall; and the Kappa coefficient assesses the consistency between the model's predictions and the true classification, effectively eliminating the influence of random consistency. When resistance was evaluated based on the vegetation index dynamic development rate and the random forest (RF) model, the model achieved an accuracy of 0.91 and a Kappa coefficient of 0.83.
[0050] and Figure 1 Corresponding to the method described above, this invention also provides a remote sensing precision assessment system for cotton Verticillium wilt resistance based on the dynamic development rate of the disease, used for... Figure 1 The specific implementation of the method, as provided in this embodiment of the invention, is a remote sensing precision assessment system for cotton Verticillium wilt resistance based on the dynamic development rate of the disease. This system can be applied to computer terminals or various mobile devices, and specifically includes: The data acquisition module is used to collect ground data, remote sensing data, and resistance assessment data of different resistant varieties within the test area to construct a multi-source dataset; The feature extraction module is used to preprocess the hyperspectral data in the multi-source dataset, obtain the time-series spectral reflectance and vegetation index, and calculate the dynamic development rate spectral features based on the spectral reflectance and spectral vegetation index of adjacent time points. The first screening module is used to screen for significant differences in dynamic development rate characteristics among different resistant varieties based on Cohen's effect size. The second screening module is used to combine Cohen's effect size confidence interval with the SBS method to further screen for sensitive dynamic development rate characteristics; The model building and evaluation module is used to construct a random forest classification model with sensitive dynamic development rate characteristics as input and resistance level as output, and to train the model. The trained cotton Verticillium wilt resistance evaluation model is then used to screen resistant varieties.
[0051] In summary, this embodiment, by combining the dynamic development rate of the disease, clarified the spectral characteristics of the sensitive dynamic development rate of different cotton Verticillium wilt resistant varieties based on ground-measured physiological and non-physiological parameters in the early stage of disease infection. Based on the spectral characteristics, resistant varieties can be screened quickly and accurately, providing technical support for resistance assessment and disease monitoring research, and is beneficial to the discovery of high-quality germplasm resources.
[0052] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.
[0053] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for remote sensing precision evaluation of cotton Verticillium wilt resistance based on disease dynamic development rate, characterized in that, The method comprises the following steps: S1, collecting ground data, remote sensing data and resistance evaluation data of different resistant varieties in the test area to construct a multi-source data set; S2, preprocessing the hyperspectral data in the multi-source data set to obtain time-series spectral reflectance and vegetation index, and calculating the dynamic development rate spectral characteristics based on the spectral reflectance and spectral vegetation index of adjacent time points; S3, screening the significant difference dynamic development rate characteristics between different resistant varieties based on Cohen's effect size; S4, further screening sensitive dynamic development rate characteristics by combining Cohen's effect size confidence interval and SBS method; S5, taking the sensitive dynamic development rate characteristics as input and the resistance level as output, constructing a random forest classification model and training the model, and using the trained cotton Verticillium wilt resistance evaluation model to screen resistant varieties.
2. The method according to claim 1, wherein the method is characterized by, The multi-source data set in S1 is specifically: The ground data includes physiological data and canopy structure data of different genotypes of leaves, wherein the physiological data includes chlorophyll content, carotenoid content, anthocyanin, and water content, and the canopy structure data includes leaf area index; The remote sensing data is the canopy scale hyperspectral data collected by the unmanned aerial vehicle at regular intervals in the early stage of disease infection; The resistance evaluation data includes cotton plant disease classification, group plant relative disease index, and resistance evaluation results.
3. The method according to claim 2, wherein the method is characterized by, The resistance evaluation data is obtained in the following manner: Before and after the collection of canopy scale hyperspectral data, resistance evaluation investigation is conducted on the test area; according to the resistance investigation standard, the severity of cotton Verticillium wilt is divided into 5 levels, of which 0 level is healthy, 1 level is mild disease, 2 level is moderate disease, 3 level is severe disease, and 4 level is dead; The formula for calculating the disease index is: wherein: represents the disease index of the identified variety, represents the corresponding disease class, represents the number of diseased plants of each disease class, represents the total number of plants; When the disease index of the susceptible control variety is 35.1-65.0, the relative disease index is used to measure the resistance of the variety, and the formula is: In the formula: represents the relative disease index, K represents the correction coefficient, 50.0 is the disease control standard disease index, represents the disease control disease index of this identification; The relative disease index (RDI) was calculated for each time point The average of the relative disease index (RDI) was calculated, and the resistance of the test varieties was evaluated according to the resistance investigation standard.
4. The method according to claim 1, wherein the method is characterized by, In S2, the preprocessing operation is specifically: Preliminary preprocessing of the hyperspectral data, using ENVI 5.3 software for image stitching, and converting the DN value to BRF bidirectional reflectance factor through a diffuse reflectance plate to obtain the preprocessed hyperspectral image of the test area; Spectral noise in the hyperspectral image is removed by using Savitzky-Golay filter to obtain time-series hyperspectral data.
5. The method according to claim 1, wherein the method is characterized by, In S2, the formula for calculating the dynamic development rate spectral characteristics is: wherein: denotes spectral reflectance of the time period, , denotes the wavelength reflectance at the time point , denotes the wavelength reflectance at the time point denotes spectral vegetation index of the time period, , denotes the vegetation index value at the time point , denotes the vegetation index value at the time point denotes the time interval.
6. The method according to claim 1, wherein the method is characterized by, In S3, the calculation method of Cohen's effect size is: Screening The significant difference features of the difference spectrum are obtained, and the significant difference wavelength and the significant difference vegetation index are obtained.
7. The method according to claim 1, wherein the method is characterized by, In S5, the classification results of the random forest classification model are evaluated by five evaluation indexes of Overall Accuracy, Precision, Recall, F1-score and Kappa coefficient; wherein Overall Accuracy represents the accuracy of the overall classification of the model, Precision measures the accuracy of the positive class prediction, Recall reflects the recognition ability of the positive class samples, F1-score is the harmonic mean of precision and recall, and Kappa coefficient evaluates the consistency between the model prediction and the true classification.
8. A system for remote sensing precision evaluation of cotton Verticillium wilt resistance based on disease dynamic development rate, characterized in that, The method comprises the following steps: The data acquisition module is configured to collect ground data, remote sensing data and resistance evaluation data of different resistant varieties in a test area, and construct a multi-source data set; The feature extraction module is configured to preprocess hyperspectral data in the multi-source data set, obtain time-series spectral reflectance and vegetation index, and calculate a dynamic development rate spectral feature based on spectral reflectance and spectral vegetation index of adjacent time points; The first screening module is configured to screen significant difference dynamic development rate features between different resistant varieties according to Cohen's effect size; The second screening module is configured to further screen sensitive dynamic development rate features by combining a Cohen's effect size confidence interval and an SBS method; The model construction and evaluation module is configured to construct a random forest classification model and perform model training by taking sensitive dynamic development rate features as input and resistance levels as output, and to screen resistant varieties by using the trained cotton verticillium wilt resistance evaluation model.