Large-area drawing method for severity of rice bacterial blight by cooperation of star-sky-ground multi-source data

Abstract

The invention relates to a large-area drawing method for the severity of rice bacterial blight in cooperation with starry sky-ground multi-source data. The large-area drawing method comprises the large-area drawing method for the severity of rice bacterial blight in cooperation with starry sky-ground multi-source data. Various common vegetation indexes are calculated according to the unmanned aerial vehicle image wave bands, and the vegetation index with the highest rice bacterial leaf blight classification precision and a corresponding threshold value are determined; constructing a severity index of rice harmed by bacterial leaf blight; determining an optimal model through multi-model comparison and cross validation; and drawing a spatial distribution diagram of the severity of the rice bacterial blight in the research area and carrying out grade division. The method has the beneficial effects that the key technical problems of multi-source remote sensing data cross-scale fusion, disease feature quantitative extraction, large-range inversion model optimization and the like are innovatively solved, and the method has the characteristics of high monitoring precision, wide space coverage and the like.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing image technology, and more specifically, it relates to a method for large-area mapping of the severity of rice bacterial blight by synergistically using multi-source data from space, air, and ground. Background Art

[0002] Rice bacterial blight (BLB) is one of the three major diseases affecting rice production. Traditional monitoring of rice bacterial blight mainly relies on manual field surveys. This method not only consumes a large amount of manpower and time, but also is difficult to achieve large-scale monitoring, and it is impossible to obtain information on the occurrence range and severity of the disease in a timely and accurate manner.

[0003] With the development of remote sensing technology, it has been widely used in the field of crop disease monitoring. Early research mainly focused on using ground observation equipment to obtain spectral data of crop leaves and canopies, and then using methods such as spectral indices, variance analysis, and convolutional neural networks for disease feature extraction and classification. Although these methods have achieved high accuracy at the leaf / canopy scale, due to the limitations of ground data acquisition, it is difficult to achieve large-area disease monitoring. Unmanned aerial vehicles (UAVs) can obtain remote sensing data with high spatial and spectral resolutions, and have gradually been applied in disease monitoring and achieved good results. However, the monitoring range of UAVs is limited and it is difficult to meet the needs of large-scale monitoring.

[0004] Satellite data has the advantage of wide coverage and has become an important data source for large-scale disease monitoring. However, the spatial / spectral resolution of satellite data is limited, and there are problems with mixed pixels, resulting in relatively low disease recognition accuracy. In addition, when using remote sensing data for quantitative research on disease severity, it is usually necessary to obtain reference disease severity data, such as disease index (DI). However, it is difficult to obtain accurate DI values, and the spatio-temporal resolution of existing satellite data also limits the accurate estimation of disease severity. Therefore, how to use multi-scale remote sensing data to achieve accurate monitoring of the severity of rice bacterial blight has become an urgent problem to be solved in the current field of agricultural pest and disease monitoring. Summary of the Invention

[0005] The purpose of the present invention is to propose a method for large-area mapping of the severity of rice bacterial blight by synergistically using multi-source data from space, air, and ground in view of the deficiencies of the prior art.

[0006] In the first aspect, a method for large-area mapping of the severity of rice bacterial blight by synergistically using multi-source data from space, air, and ground is provided, including:

[0007] S1. Obtain UAV multispectral images and Sentinel-2 images and perform preprocessing;

[0008] S2. Combine the ground survey data with visual interpretation to determine the healthy and bacterial blight-infected rice sample points in the UAV multispectral images; calculate various common vegetation indices based on the UAV image bands, and determine the vegetation index and corresponding threshold with the highest classification accuracy for rice bacterial blight.

[0009] S3. Superimpose the classification results of the UAV multispectral images with the Sentinel-2 pixel space, count the area ratio and vegetation index of the disease area within each Sentinel-2 pixel, and construct an index for the severity of rice infected by bacterial blight.

[0010] S4. Use the reflectance and vegetation index of the Sentinel-2 images as independent variables, and the index for the severity of rice infected by bacterial blight as the dependent variable, and determine the optimal model through multi-model comparison and cross-validation.

[0011] S5. Use the optimal model and the Sentinel-2 images to draw a spatial distribution map of the severity of rice bacterial blight in the study area and conduct a grade division.

[0012] Preferably, in S3, the calculation formula for the index for the severity of rice infected by bacterial blight is:

[0013]

[0014] where BLB_DSI is the index for the severity of rice infected by bacterial blight, N BLB represents the number of UAV pixels marked as rice infected by BLB within the Sentinel-2 pixel, N total represents the total number of UAV pixels within the Sentinel-2 pixel, and VI all represents the normalized vegetation index value of all UAV pixels within a Sentinel-2 pixel.

[0015] Preferably, S2 includes:

[0016] S201. Select the sample areas of healthy rice and rice infected by bacterial blight from the UAV multispectral images according to the ground survey data and visual interpretation results.

[0017] S202. Use the spectral index threshold method and combine various common vegetation indices to classify the healthy rice and rice infected by bacterial blight in the UAV multispectral images.

[0018] S203. Evaluate the classification effects of various common vegetation indices, and determine the vegetation index and corresponding threshold with the highest classification accuracy for rice bacterial blight.

[0019] Preferably, S3 includes:

[0020] S301. Spatially overlay the UAV classification results with the pixels of the Sentinel-2 image to ensure that each Sentinel-2 pixel contains several UAV pixels;

[0021] S302. Count the number of UAV pixels affected by bacterial blight in each Sentinel-2 pixel and compare it with the number of all rice pixels in this pixel, so as to calculate the proportion of the area damaged by bacterial blight; At the same time, calculate the average normalized difference vegetation index value of all UAV pixels in each Sentinel-2 pixel to construct an index of the severity of bacterial blight on rice.

[0022] Preferably, S4 includes:

[0023] S401. Extract the band reflectance and vegetation index of the Sentinel-2 pixel as independent variables and use the index of the severity of bacterial blight on rice as the dependent variable to construct a sample set;

[0024] S402. Use ten-fold cross-validation and grid search to determine the optimal model among the RF model, SVM model and MLR model.

[0025] In a second aspect, a multi-source data collaborative large-area mapping system for the severity of bacterial blight in rice using space-air-ground is provided, which is used to execute any of the methods in the first aspect, including:

[0026] An acquisition module, which is used to acquire UAV multispectral images and Sentinel-2 images and perform preprocessing;

[0027] A first determination module, which is used to combine ground survey data and visual interpretation to determine healthy and bacterial blight-affected rice sample points in the UAV multispectral image; Calculate various common vegetation indices based on the UAV image bands to determine the vegetation index with the highest classification accuracy for bacterial blight in rice and the corresponding threshold;

[0028] A construction module, which is used to spatially overlay the UAV multispectral image classification results with Sentinel-2 pixels, count the area proportion and vegetation index of the disease area in each Sentinel-2 pixel, and construct an index of the severity of bacterial blight on rice;

[0029] A second determination module, which is used to use the reflectance and vegetation index of the Sentinel-2 image as independent variables and the index of the severity of bacterial blight on rice as the dependent variable to determine the optimal model through multi-model comparison and cross-validation;

[0030] A drawing module, which is used to use the optimal model and the Sentinel-2 image to draw a spatial distribution map of the severity of bacterial blight in rice in the study area and conduct grade division.

[0031] In a third aspect, a computer storage medium is provided, in which a computer program is stored; when the computer program runs on a computer, the computer is enabled to execute the method according to any one of the first aspect.

[0032] In a fourth aspect, an electronic device is provided, including:

[0033] a memory for storing a computer program;

[0034] a processor for executing the computer program to implement the method according to any one of the first aspect.

[0035] The beneficial effects of the present invention are as follows:

[0036] 1. By integrating high-resolution images of unmanned aerial vehicles (UAVs) and Sentinel-2 satellite images, the present invention constructs an index for the severity of bacterial leaf blight (BLB) in rice, and through the comparative optimization of support vector machine (SVM), random forest (RF), and multiple linear regression (MLR) models, realizes the accurate inversion and spatial mapping of the disease severity. This method innovatively solves key technical problems such as cross-scale fusion of multi-source remote sensing data, quantitative extraction of disease characteristics, and optimization of large-scale inversion models, and has characteristics such as high monitoring accuracy and wide spatial coverage, providing a scientific basis and technical support for the accurate monitoring and prevention and control decision-making of rice diseases.

[0037] 2. The present invention realizes for the first time the large-scale and high-precision quantitative inversion and mapping of the severity of rice bacterial leaf blight based on multi-scale remote sensing data of space-air-ground; and comprehensively considers two factors, the damaged area and damaged intensity of BLB within a satellite pixel, and the proposed BLB_DSI index can accurately characterize the true damage degree of bacterial leaf blight at the Sentinel-2 pixel scale and can more effectively reflect the disease situation; in addition, the present invention also proposes a research framework for large-scale accurate inversion and mapping of the damage degree of rice bacterial leaf blight through multi-scale coordination of ground-UAV-satellite, and also has the potential to be extended to the monitoring of other crop diseases, providing new technical means and method ideas for the field of agricultural pest and disease monitoring. Description of the Drawings

[0038] Figure 1 is a flowchart of a method for large-area mapping of the severity of rice bacterial leaf blight based on the coordination of multi-source data of space-air-ground;

[0039] Figure 2 is a box plot of the spectral indices of UAVs for bacterial leaf blight and healthy rice samples calculated based on ground and visually interpreted samples;

[0040] Figure 3 is a schematic diagram of the classification results of bacterial leaf blight and healthy rice based on UAV multi-spectral;

[0041] Figure 4 Scatter density and corresponding accuracy graph between the BLB_DSI estimated by RF, SVM, and MLR and the reference BLB_DSI;

[0042] Figure 5 Mapping results of bacterial blight severity based on Sentinel-2 data. Detailed implementation manners

[0043] The present invention will be further described below in conjunction with embodiments. The descriptions of the following embodiments are only used to help understand the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several modifications can still be made to the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

[0044] Embodiment 1:

[0045] As an embodiment, a method for large-area mapping of bacterial blight severity of rice based on the collaboration of multi-source data of space, air, and ground provided by the present invention, as Figure 1 shown, includes the following steps:

[0046] S1. Obtain UAV multispectral images and Sentinel-2 images and perform preprocessing.

[0047] Specifically, obtaining UAV multispectral data includes: collecting data using 5 bands of the DJI P4 Multispectral camera, obtaining the spatial resolution of the acquired image of the image as 6 - 10.6 cm (the spatial resolution is different at different flight altitudes), and the central wavelengths of the bands are blue light (450 nm), green light (560 nm), red light (650 nm), red edge (730 nm), and near-infrared (840 nm). Use DJI Terra software (version 3.7.6) to perform preprocessing such as image mosaicking, radiometric correction, and geometric correction on the multispectral image. Then, through visual interpretation, mask the non-rice areas in the UAV multispectral image.

[0048] Obtaining Sentinel-2 satellite data includes: downloading Sentinel-2 image data (orbit number: T51RXV) that matches the acquisition time of the study area (October 15th and 17th, 2023) on the Google Earth Engine platform. Perform preprocessing such as cloud masking, band synthesis, and mosaicking on the downloaded Sentinel-2 Level-2A data.

[0049] S2. Combine ground survey data and visual interpretation to determine healthy and bacterial blight-affected rice sample points in the UAV multispectral image; calculate various common vegetation indices according to the UAV image bands, and determine the vegetation index and corresponding threshold with the highest classification accuracy for rice bacterial blight (such asFigure 2 as shown

[0050] Exemplarily, based on ground survey data and visual interpretation results, when NDVI is equal to 0.5, the classification accuracy of healthy rice and rice areas damaged by bacterial blight is the highest, and the classification accuracy reaches 0.989.

[0051] Specifically, S2 includes:

[0052] S201. Select sample areas of healthy rice and rice damaged by bacterial blight from the UAV multispectral images according to the ground survey data and visual interpretation results. For example, a total of 238,252 healthy rice samples and 289,286 rice samples damaged by bacterial blight are obtained.

[0053] S202. Use the spectral index threshold method and combine multiple common vegetation indices to classify healthy rice and rice damaged by bacterial blight in the UAV multispectral images;

[0054] As shown in Table 1, seven common spectral indices are used in the embodiments of the present application.

[0055] Table 1 Spectral index information for extracting rice damaged by bacterial blight in UAV images

[0056]

[0057] S203. Evaluate the classification effects of multiple common vegetation indices, and determine the vegetation index and corresponding threshold with the highest classification accuracy for rice bacterial blight.

[0058] Specifically, when classifying UAV images, 70% of the samples of healthy rice and rice damaged by bacterial blight are selected as the training set, and the remaining 30% are used as the validation set to evaluate the classification effect. The classification results are evaluated by overall accuracy (OA), precision, recall, and F1 score. The results show that the classification effect is the best using NDVI (threshold set to 0.5). The NDVI value of healthy rice is greater than 0.5, and the NDVI value of rice damaged by bacterial blight is less than 0.5, effectively distinguishing healthy rice and disease-affected areas (as Figure 3 shown), and the classification accuracy OA reaches 0.989.

[0059] S3. Superimpose the classification results of the UAV multispectral images with the Sentinel-2 pixel space, and count the area ratio and vegetation index of the disease area in each Sentinel-2 pixel to construct an index of the severity of rice damaged by bacterial blight.

[0060] S4. Using the reflectance and vegetation indices of Sentinel-2 images as independent variables and the severity index of rice affected by bacterial blight as the dependent variable, determine the optimal model through multi-model comparison and cross-validation.

[0061] S5. Using the optimal model and Sentinel-2 images, draw a spatial distribution map of the severity of rice bacterial blight in the study area and conduct grade classification.

[0062] Example 2:

[0063] Based on Example 1, Example 2 of this application provides a more specific large-area mapping method for the severity of rice bacterial blight based on the collaboration of multi-source data from space, air, and ground, including:

[0064] S1. Obtain and preprocess unmanned aerial vehicle (UAV) multispectral images and Sentinel-2 images.

[0065] S2. Combine ground survey data and visual interpretation to determine healthy and rice samples affected by bacterial blight in the UAV multispectral images; calculate various common vegetation indices based on the UAV image bands, and determine the vegetation index and corresponding threshold with the highest classification accuracy for rice bacterial blight.

[0066] S3. Spatially overlay the classification results of the UAV multispectral images with the Sentinel-2 pixels, and count the area ratio and vegetation indices of the disease areas within each Sentinel-2 pixel to construct the severity index of rice affected by bacterial blight.

[0067] Specifically, within each Sentinel-2 pixel, count the number of UAV pixels of healthy rice and rice affected by bacterial blight, and calculate their corresponding average vegetation index values (NDVI in this example of the application). Through the area ratio and damage intensity of the areas affected by bacterial blight within each satellite pixel, construct the severity index of bacterial blight (BLB_DSI) for each Sentinel-2 pixel. The specific process is as follows:

[0068] S301. Spatially overlay the UAV classification results with the pixels of the Sentinel-2 image to ensure that each Sentinel-2 pixel contains several UAV pixels.

[0069] This step is implemented through the fishnet tool in R language to ensure that each Sentinel-2 pixel contains several UAV pixels, thus facilitating the subsequent accurate extraction of rice disease information within each pixel.

[0070] S302. Count the number of UAV pixels affected by bacterial blight (i.e., the infected area) within each Sentinel-2 pixel, and compare it with the number of all rice pixels within this pixel, so as to calculate the proportion of the area damaged by bacterial blight. At the same time, calculate the average normalized difference vegetation index (NDVI) value of all UAV pixels within each Sentinel-2 pixel, and construct an index for the severity of rice damaged by bacterial blight.

[0071] In S302, to quantify the infection intensity, define the NDVI value of healthy rice as the theoretical maximum value (i.e., 1), compare the NDVI value of rice damaged by bacterial blight with that of healthy rice, and calculate the degree of deviation as an indicator of the infection intensity. Finally, by combining the damaged area and the damage intensity of rice damaged by bacterial blight, construct an index for the severity of rice damaged by bacterial blight (BLB_DSI), and its calculation formula is:

[0072]

[0073] where BLB_DSI is the severity of rice damaged by bacterial blight in the sentinel-2 pixel, N BIB represents the number of UAV pixels marked as rice damaged by bacterial blight within this pixel, and N total represents the number of all UAV pixels within this Sentinel-2 pixel. VI all represents the NDVI value of all UAV pixels within a Sentinel-2 pixel.

[0074] S4. Take the reflectance and vegetation index of Sentinel-2 images as independent variables, and the index for the severity of rice damaged by bacterial blight as the dependent variable, compare the accuracies of the support vector machine (SVM), random forest (RF) and multiple linear regression (MLR) models, and determine the optimal model through cross-validation and use it for mapping the severity of bacterial blight over a large area.

[0075] S4 includes:

[0076] S401. Extract the band reflectance and vegetation index of Sentinel-2 pixels (as shown in Table 1) as independent variables, and the index for the severity of rice damaged by bacterial blight as the dependent variable to construct a sample set.

[0077] In addition, perform Z-score standardization on all variables, and divide them into a training set (13,764) and a validation set (5,899) according to a 7:3 ratio.

[0078] S402. Use ten-fold cross-validation and grid search to determine the optimal model among the RF model, SVM model and MLR model.

[0079] Specifically, the R language caret package is used for model training and hyperparameter optimization. Ten-fold cross-validation and grid search are adopted to determine the optimal parameters for the RF model, SVM model, and MLR model. The optimal inversion model for the severity of rice BLB is determined through the coefficient of determination, mean absolute error, and root mean square error of validation. Among them, RF has the best validation accuracy, with R 2 being 0.91, RMSE being 8.58%, and MAE being 5.48%. Followed by SVM, while the MLR has the lowest validation accuracy, with R 2 being 0.816.

[0080] S5. Using the optimal model and Sentinel-2 images, draw the spatial distribution map of the severity of rice bacterial leaf blight in the study area and conduct grade classification.

[0081] Specifically, S5 includes:

[0082] S501. Apply the optimal model to the Sentinel-2 images of Ningbo City, input the band reflectance and vegetation index, and output the spatial distribution of the severity of rice BLB in Ningbo City, BLB_DSI.

[0083] S502. To more clearly display the severity of BLB over a large area, classify the disease severity (healthy, mild, moderate, and severe) and generate a spatial distribution map (as Figure 5 shown). Among them, according to the distribution of the predicted values of BLB_DSI, the BLB_DSI values are divided into five grades by the interval method (as shown in Table 2), and the thresholds are obtained as:

[0084] T a ≤BLB_DSI<T b

[0085] Table 2

[0086]

[0087] Among them, BLB_DSI is the severity index of rice infected by bacterial leaf blight, and T a and T b are the maximum and minimum values of BLB_DSI, respectively. In this embodiment, T1, T2, T3, T4, T5, and T6 are 0, 0.2, 0.4, 0.6, 0.8, and 1, respectively.

[0088] It should be noted that the same or similar parts in this embodiment and Embodiment 1 can be referred to each other and will not be elaborated in this application.

[0089] Embodiment 3:

[0090] Based on Embodiment 2, Embodiment 3 of the present application provides a multi-source data collaborative large-area mapping system for the severity of rice bacterial blight in space, air, and ground, including:

[0091] An acquisition module, configured to acquire and preprocess unmanned aerial vehicle (UAV) multispectral images and Sentinel-2 images;

[0092] A first determination module, configured to combine ground survey data and visual interpretation to determine healthy and rice bacterial blight-affected rice sample points in the UAV multispectral images; calculate multiple common vegetation indices based on the UAV image bands, and determine the vegetation index with the highest classification accuracy for rice bacterial blight and the corresponding threshold;

[0093] A construction module, configured to spatially overlay the classification results of the UAV multispectral images with Sentinel-2 pixels, count the area ratio of the disease area and the vegetation index within each Sentinel-2 pixel, and construct an index for the severity of rice affected by bacterial blight;

[0094] A second determination module, configured to use the reflectance and vegetation index of the Sentinel-2 images as independent variables and the index for the severity of rice affected by bacterial blight as the dependent variable, and determine the optimal model through multi-model comparison and cross-validation;

[0095] A mapping module, configured to use the optimal model and the Sentinel-2 images to draw a spatial distribution map of the severity of rice bacterial blight in the study area and perform grade division.

[0096] It should be noted that the system provided in this embodiment is the system corresponding to the method provided in Embodiment 2. Therefore, for the parts that are the same or similar in this embodiment and Embodiment 2, reference can be made to each other and will not be repeated in this application.

[0097] In summary, this application proposes a method for large-scale mapping of the severity of rice bacterial blight based on the collaboration of multi-source data from space, air, and ground. First, unmanned aerial vehicle (UAV) multispectral data and Sentinel-2 satellite images are acquired and preprocessed. The UAV data are processed through image mosaicking, radiometric correction, and geometric correction, while the Sentinel-2 images are processed through cloud masking and band synthesis to ensure data quality. Then, multiple vegetation indices (such as NDVI, EVI, etc.) are calculated, and the classification thresholds for healthy rice and rice with bacterial blight are determined in combination with ground survey data to distinguish healthy areas from affected areas. Then, the UAV classification results are spatially overlaid with Sentinel-2 pixels, and the severity index of rice affected by bacterial blight (BLB_DSI) is constructed by statistically calculating the proportion of diseased area and the degree of damage (represented by the value of the vegetation index) within each satellite pixel. Finally, using the random forest (RF), support vector machine (SVM), and multiple linear regression (MLR) models, with the reflectance and vegetation indices of Sentinel-2 as independent variables and BLB_DSI as the dependent variable, the optimal model is optimized through cross-validation, and a large-scale spatial distribution map of the severity of rice bacterial blight is generated. Through this method, the spatial distribution and severity of rice bacterial blight can be effectively identified, providing a scientific basis for agricultural disease monitoring and prevention and control decisions, and having important practical application value.

Claims

1. A method for large-area mapping of the severity of rice bacterial blight through multi-source data collaboration of space, sky and ground, characterized in that Including: S1. Obtain the multi - spectral images of unmanned aerial vehicles (UAVs) and Sentinel - 2 images and perform pre - processing; S2. Combine ground survey data and visual interpretation to determine the healthy and bacterial blight - affected rice sample points in the UAV multi - spectral images; Calculate multiple common vegetation indices based on the UAV image bands, and determine the vegetation index and corresponding threshold with the highest classification accuracy for rice bacterial blight; S3. Superimpose the classification results of the UAV multi - spectral images with the Sentinel - 2 pixel space, and statistically calculate the area ratio and vegetation index of the disease area within each Sentinel - 2 pixel to construct an index for the severity of rice affected by bacterial blight; S4. Use the reflectance and vegetation index of the Sentinel - 2 images as independent variables and the index for the severity of rice affected by bacterial blight as the dependent variable, and determine the optimal model through multi - model comparison and cross - validation; S5. Use the optimal model and the Sentinel - 2 images to draw a spatial distribution map of the severity of rice bacterial blight in the study area and conduct grade division.

2. The large-area mapping method for the severity of rice bacterial blight by synergistic use of space-air-ground multi-source data according to claim 1, wherein In S3, the calculation formula for the index of the severity of rice affected by bacterial blight is: Among them, BLB_DSI is the severity index of rice blast disease damage to rice, N BLB represents the number of UAV pixels of rice damaged by BLB within the Sentinel-2 pixel marked as Sentinel-2, N total represents the total number of UAV pixels within the Sentinel-2 pixel, represents the normalized difference vegetation index value of all UAV pixels within a Sentinel-2 pixel.

3. The method for large-area mapping of the severity of rice bacterial blight by integrating multi-source data from space, air, and ground according to claim 2, characterized in that S2 Including: S201. Select the sample areas of healthy rice and rice affected by bacterial blight from the UAV multi - spectral images according to the ground survey data and visual interpretation results; S202. Adopt the spectral index threshold method and combine multiple common vegetation indices to classify healthy rice and rice affected by bacterial blight in the UAV multi - spectral images; S203. Evaluate the classification effects of multiple common vegetation indices to determine the vegetation index and corresponding threshold with the highest classification accuracy for rice bacterial blight.

4. The method for large-area mapping of the severity of rice bacterial blight by synergistic use of space-air-ground multi-source data according to claim 3, characterized in that S3 Including: S301. Superimpose the UAV classification results with the pixels of the Sentinel - 2 images spatially to ensure that each Sentinel - 2 pixel contains several UAV pixels; S302. Statistically calculate the number of UAV pixels affected by bacterial blight within each Sentinel - 2 pixel and compare it with the number of all rice pixels within this pixel, so as to calculate the area ratio of the damage caused by bacterial blight; At the same time, calculate the average normalized vegetation index value of all UAV pixels within each Sentinel - 2 pixel to construct an index for the severity of rice affected by bacterial blight.

5. The large-area mapping method for the severity of rice bacterial blight by integrating multi-source data from space, air and ground according to claim 4, wherein S4 Including: S401. Extract the band reflectance and vegetation index of the Sentinel - 2 pixels as independent variables and the index for the severity of rice affected by bacterial blight as the dependent variable to construct a sample set; S402. Use ten - fold cross - validation and grid search to determine the optimal model among the RF model, SVM model, and MLR model.

6. A multi-source data collaborative mapping system for large-area mapping of the severity of rice bacterial blight in space, sky and ground, characterized in that, For implementing the method according to any one of claims 1 to 5, including: An acquisition module, configured to obtain the multi - spectral images of UAVs and Sentinel - 2 images and perform pre - processing; A first determination module, configured to combine ground survey data and visual interpretation to determine the healthy and bacterial blight - affected rice sample points in the UAV multi - spectral images; Calculate multiple common vegetation indices based on the UAV image bands, and determine the vegetation index and corresponding threshold with the highest classification accuracy for rice bacterial blight; A construction module for spatially overlaying the classification results of the drone multispectral images with Sentinel-2 pixels, calculating the area ratio of the disease-affected area and the vegetation index within each Sentinel-2 pixel, and constructing an index for the severity of bacterial blight damage to rice. A second determination module for determining the optimal model through multi-model comparison and cross-validation, with the reflectance and vegetation index of the Sentinel-2 image as independent variables and the index for the severity of bacterial blight damage to rice as the dependent variable. A plotting module for using the optimal model and the Sentinel-2 image to plot the spatial distribution map of the severity of bacterial blight in rice in the study area and conduct grade division.

7. A computer storage medium, characterized in that, The computer storage medium stores a computer program; when the computer program runs on a computer, it causes the computer to execute the method according to any one of claims 1 to 5.

8. An electronic device, characterized in that, It includes: A memory for storing the computer program; A processor for executing the computer program to implement the method according to any one of claims 1 to 5.

Images