A regional drought monitoring method and system based on surface spectral endmember space
Through the method based on the end element space of the surface spectral surface spectral, the random forest model is trained and the temperature vegetation drought index is constructed, which solves the problems of uncertain accuracy and insufficient spatial resolution, and realizes efficient and refined regional drought monitoring and provides high time and high spatial resolution monitoring results.
Patent Information
- Application Number
- CN202410006917.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-01-03
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-01-03
AI Technical Summary
In the existing regional drought monitoring methods, the NDVI index is affected by yellowed vegetation and fallen leaves residues, and the characterization accuracy is uncertain. The 1km spatial resolution of MODIS surface temperature data cannot meet the needs of high spatial resolution, and it is impossible to effectively monitor the arid areas. Vegetation information.
采用基于地表光谱端元空间的方法,通过获取高分辨率的反射率特征数据和地表温度数据,训练随机森林模型,映射反射率特征数据以获得高分辨率的地表温度数据,并结合端元丰度值构建温度植被干旱指数,提高监测模型的泛化能力和准确性。
The surface temperature data with high temporal resolution and high spatial resolution can be realized, which can more accurately characterize the degree of drought, support the refined management and monitoring of regional droughts, and provide temperature vegetation drought index with 5-day time resolution and 10m spatial resolution. Monitoring results.
Smart Images

Figure CN117969458B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of remote sensing technology, and in particular to a regional drought monitoring method and system based on surface spectral endmember space. Background Art
[0002] With the continuous intensification of global warming, problems such as land degradation, water source drying up and food production reduction caused by drought have seriously threatened human production and life. Therefore, it is very important to develop drought monitoring methods. Current regional drought monitoring methods include traditional monitoring methods and remote sensing-based monitoring methods. Traditional drought monitoring methods rely on ground meteorological stations with high accuracy, but it is difficult to achieve full regional coverage and accurate measurement. The remote sensing monitoring method can not only achieve accurate monitoring of the entire region, but also save time and effort, becoming the main method for regional drought monitoring. Among the existing regional drought monitoring methods, most of them are remote sensing drought monitoring models constructed based on the Normalized Difference Vegetation Index (NDVI) and surface temperature of MODIS remote sensing data.
[0003] In the existing technology, on the one hand, the NDVI index is affected by the withered vegetation and fallen leaf residues and the interference of different soil backgrounds, and its own characterization accuracy is uncertain; at the same time, a large number of existing studies have shown that the NDVI index is not capable of capturing vegetation information in areas with sparse vegetation, and the vegetation in arid areas is often sparse. Therefore, there are obvious deficiencies in using the NDVI index for regional drought monitoring models. On the other hand, at the regional scale, the need for refined land management such as farming often requires monitoring results with high spatial resolution. However, the MODIS surface temperature data commonly used in current regional drought remote sensing monitoring methods can only meet the needs of high temporal resolution, and its spatial resolution of 1 km obviously cannot meet the needs of high spatial resolution for regional management. Summary of the invention
[0004] The purpose of the present invention is to provide a regional drought monitoring method and system based on surface spectral endmember space, which can improve the generalization ability and accuracy of the temperature vegetation drought index model.
[0005] To achieve the above object, the present invention provides the following solutions:
[0006] A regional drought monitoring method based on surface spectral endmember space, the method comprising:
[0007] Acquire first reflectivity characteristic data and first surface temperature data of the study area, wherein the resolution of the first reflectivity characteristic data is a first resolution, and the resolution of the first surface temperature data is a second resolution, and the first resolution is greater than the second resolution; the first reflectivity characteristic data includes: end member abundance value, cover type and terrain variables;
[0008] Aggregating the first reflectivity characteristic data to obtain second reflectivity characteristic data; the resolution of the second reflectivity characteristic data is a second resolution;
[0009] A random forest model is trained based on the second reflectivity characteristic data and the first surface temperature data to obtain a trained random forest model as a mapping relationship model; the mapping relationship model is used to characterize the relationship between the reflectivity characteristic data and the surface temperature data;
[0010] The first reflectivity characteristic data is mapped using a mapping relationship model to obtain second surface temperature data; the resolution of the second surface temperature data is the first resolution;
[0011] The temperature vegetation drought index of the study area is determined based on the end member abundance values in the second surface temperature data and the first reflectivity characteristic data; the temperature vegetation drought index is used to characterize the degree of drought in the study area.
[0012] Optionally, the first resolution features of the study area are obtained, including:
[0013] Acquire surface reflectance data and elevation data; the resolutions of the surface reflectance data and the elevation data are both first resolutions;
[0014] Based on the surface reflectance data, the end-member abundance value is obtained through principal component analysis and end-member spectrum curve extraction;
[0015] The cover type was obtained through classification decision method based on the end-member abundance value;
[0016] The terrain variables are calculated based on the elevation data.
[0017] Optionally, before determining the temperature vegetation drought index of the study area based on the second surface temperature data and the end member abundance value in the first reflectivity characteristic data, the method further includes:
[0018] The second ground surface temperature data is corrected to obtain corrected second ground surface temperature data.
[0019] Optionally, correcting the second ground surface temperature data to obtain corrected second ground surface temperature data specifically includes:
[0020] aggregating the second land surface temperature data to obtain aggregated third land surface temperature data;
[0021] Performing residual correction on the second land surface temperature data based on the first land surface temperature data and the third land surface temperature data to obtain residual-corrected second land surface temperature data;
[0022] The second land surface temperature data after residual correction is input into the fitting relationship model to obtain the corrected second land surface temperature data.
[0023] Optionally, performing residual correction on the second land surface temperature data based on the first land surface temperature data and the third land surface temperature data to obtain the residual-corrected second land surface temperature data specifically includes:
[0024] Taking a residual of the first surface temperature data and the third surface temperature data to obtain a first residual; the resolution of the first residual is the second resolution;
[0025] Resampling the first residual to obtain a second residual; the resolution of the second residual is the first resolution;
[0026] The second residual is added to the second land surface temperature data to obtain the second land surface temperature data after residual correction.
[0027] Optionally, inputting the second surface temperature data after residual correction into the fitting relationship model to obtain the corrected second surface temperature data specifically includes:
[0028] constructing a linear correction model based on the first surface temperature data and the third surface temperature data;
[0029] The second surface temperature data after residual error correction is linearly corrected according to the linear correction model to obtain corrected second surface temperature data.
[0030] Optionally, determining the temperature vegetation drought index of the study area based on the second surface temperature data and the end member abundance value in the first reflectivity characteristic data specifically includes:
[0031] Based on the second surface temperature data and the end member abundance values in the first reflectivity characteristic data, the temperature vegetation drought index of the study area is calculated using the temperature vegetation drought index model;
[0032] The temperature vegetation drought index model is:
[0033]
[0034] Among them, TVDI is the temperature vegetation drought index of the study area, LST is the second surface temperature data, T max is the dry edge straight line temperature corresponding to the vegetation end member abundance value calculated based on the dry edge straight line equation, T min is the wet edge straight line temperature corresponding to the vegetation end member abundance value calculated based on the wet edge straight line equation.
[0035] Optionally, the construction process of the temperature vegetation drought index model is as follows:
[0036] Draw a scatter plot with the vegetation end-member abundance value in the end-member abundance value as the X-axis and the second surface temperature data as the Y-axis;
[0037] Select the top point set of the scatter plot to fit the dry edge straight line equation:
[0038] T max =a1*GV+b1;
[0039] In the formula, GV is the vegetation end-member abundance value in the end-member abundance value, a1, and b1 are parameter variables;
[0040] Select the point set at the bottom of the scatter plot to fit the wet edge straight line equation:
[0041] T min =a2*GV+b2;
[0042] In the formula, a2, and b2 are parameter variables;
[0043] A temperature vegetation drought index model was constructed based on the dry edge straight line equation and the wet edge straight line equation.
[0044] A regional drought monitoring system based on surface spectral endmember space, the system is applied to the above method, the system comprises:
[0045] An acquisition module is used to acquire first reflectivity characteristic data and first surface temperature data of a study area, wherein the resolution of the first reflectivity characteristic data is a first resolution, and the resolution of the first surface temperature data is a second resolution, and the first resolution is greater than the second resolution; the first reflectivity characteristic data includes: end member abundance value, cover type and terrain variable;
[0046] An aggregation module, used for aggregating the first reflectivity characteristic data to obtain second reflectivity characteristic data; the resolution of the second reflectivity characteristic data is a second resolution;
[0047] A training module, used for training a random forest model based on the second reflectivity characteristic data and the first surface temperature data, and obtaining a trained random forest model as a mapping relationship model; the mapping relationship model is used for characterizing the relationship between the reflectivity characteristic data and the surface temperature data;
[0048] A mapping module, used to map the first reflectivity characteristic data using a mapping relationship model to obtain second surface temperature data; the resolution of the second surface temperature data is the first resolution;
[0049] The calculation module is used to determine the temperature vegetation drought index of the study area based on the second surface temperature data and the end member abundance value in the first reflectivity characteristic data; the temperature vegetation drought index is used to characterize the degree of drought in the study area.
[0050] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method when executing the computer program.
[0051] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:
[0052] The present invention trains a random forest model based on second reflectivity characteristic data and first surface temperature data, maps the first reflectivity characteristic data after obtaining the mapping relationship, obtains the second surface temperature data, and determines the temperature vegetation drought index of the study area in combination with the end member abundance value in the first reflectivity characteristic data, thereby improving the generalization ability and accuracy of the temperature vegetation drought index model. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0054] Figure 1 A first flow chart of a method for regional drought monitoring based on surface spectral endmember space provided by an embodiment of the present invention;
[0055] Figure 2 A second flow chart of a method for regional drought monitoring based on surface spectral endmember space provided by an embodiment of the present invention;
[0056] Figure 3 A two-dimensional scatter plot of a temperature vegetation drought index model of a regional drought monitoring method based on surface spectral endmember space provided in an embodiment of the present invention;
[0057] Figure 4 A flow chart of endmember abundance value extraction of a regional drought monitoring method based on surface spectral endmember space provided by an embodiment of the present invention;
[0058] Figure 5 A scatter plot of pure pixels and mixed pixels of a regional drought monitoring method in a surface spectral endmember space provided by an embodiment of the present invention;
[0059] Figure 6 A flow chart of the classification of cover types provided by an embodiment of the present invention;
[0060] Figure 7 A structural diagram of a regional drought monitoring system based on surface spectral endmember space provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0062] The purpose of the present invention is to provide a regional drought monitoring method and system based on the surface spectral end-member space. A random forest model is trained based on second reflectivity characteristic data and first surface temperature data, and the first reflectivity characteristic data is mapped after obtaining a mapping relationship to obtain the second surface temperature data. The temperature vegetation drought index of the study area is determined in combination with the end-member abundance value in the first reflectivity characteristic data, thereby improving the generalization ability and accuracy of the temperature vegetation drought index model.
[0063] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0064] Embodiment 1
[0065] like Figure 1-Figure 2 As shown, the present invention provides a regional drought monitoring method based on surface spectral end-member space, the method comprising:
[0066] S1. Obtain the first reflectivity characteristic data and the first surface temperature data of the study area, the resolution of the first reflectivity characteristic data is the first resolution, the resolution of the first surface temperature data is the second resolution, and the first resolution is greater than the second resolution; the first reflectivity characteristic data includes: end member abundance value, cover type and terrain variables. The first resolution is used to represent high resolution, and the second resolution is used to represent low resolution.
[0067] Optionally, the first resolution features of the study area are obtained, including:
[0068] Acquire surface reflectance data (10m Sentinel-2 time series surface reflectance data) and elevation data (STRM 30m elevation data); the resolution of the surface reflectance data and elevation data are both first resolution.
[0069] The endmember abundance value was obtained based on the surface reflectance data through principal component analysis and endmember spectral curve extraction.
[0070] Using the acquired surface reflectance data, the endmember spectral curves of vegetation and soil habitats were extracted based on principal component analysis, and the endmember abundance value data of the time series in the study area was obtained using a linear mixed decomposition model, which included three steps: principal component analysis, endmember extraction based on geometric vertices, and calculation of the endmember abundance value of each pixel in the study area.
[0071] In actual application, Figure 4 As shown in the figure, the principal component analysis method is to perform principal component analysis on the surface reflectance data to obtain the matrix after principal component transformation. The type and number of end members are determined according to the existing research in the region. In arid areas, the monitoring is generally carried out on four types of end members: vegetation (GV), sand (SL), salt (SA) and dark matter (DA). According to a large number of research experiments, the original spectral information in most images can be expressed based on the first three principal components. Then, the corresponding end member type is determined according to the characteristics of the first three principal components in the matrix after principal component transformation of the image of each period, and the representative seasonal phase extracted by each end member is determined according to the contribution rate of the first three principal components of the image of each period.
[0072] like Figure 5 As shown in the figure, the extraction of endmember spectral curves mainly adopts the extraction method of two-dimensional scatter plot. According to the determined endmember type and representative seasonal phase, the corresponding principal component is used to construct a two-dimensional scatter plot. Based on the mathematical mechanism of the linear mixed model, the scatter plot of the principal component should be triangularly distributed. The three vertices of the triangle can be regarded as pure endmembers, and the interior of the triangle is a mixed pixel, which is a linear combination of pure endmembers. 200-400 pixels are selected as endmember pixels at the vertices of the principal component scatter plot, and the average reflectance of the endmember pixel is used as the endmember spectrum. After obtaining the endmember spectra in the study area, the endmember spectra and the image of the study area are placed in the linear spectral mixture decomposition model to obtain the endmember abundance values of all pixels in the study area.
[0073] The cover type was obtained through classification decision method based on the end member abundance value.
[0074] like Figure 6 As shown in the figure, the land cover type is obtained by classification decision-making based on the end-member abundance value, which specifically includes: first, according to the land cover classification system, about 1000 pixels are selected for each classification through field samples and high spatial resolution images of Google Earth, and the training samples (training set) and the validation samples are divided into 7:3 ratios. Then, based on the end-member abundance value, the segmentation variables and segmentation thresholds are determined using the end-member abundance values of the training samples (training set) of each land cover type, and then a decision tree is constructed based on the determined segmentation variables and segmentation thresholds to obtain the land cover type (classification result), and the classification result is evaluated for accuracy based on the test sample (test set), and the evaluation indicators include overall classification accuracy and Kappa coefficient.
[0075] The terrain variables are calculated based on the elevation data; the terrain variables include elevation, slope and terrain humidity index. The calculation formula of terrain humidity index is as follows:
[0076]
[0077] Among them, SCA represents the flow per unit area; slope represents the slope.
[0078] S2. Aggregate the first reflectivity characteristic data to obtain second reflectivity characteristic data; the resolution of the second reflectivity characteristic data is a second resolution.
[0079] S3. A random forest model is trained based on the second reflectivity characteristic data and the first surface temperature data (MODIS 900m surface temperature data) to obtain a trained random forest model as a mapping relationship model; the mapping relationship model is used to characterize the relationship between the reflectivity characteristic data and the surface temperature data.
[0080] S4. Map the first reflectivity characteristic data using a mapping relationship model to obtain second surface temperature data; the resolution of the second surface temperature data is the first resolution.
[0081] In the actual application process, different parameter settings of the random forest model will affect the fitting effect of the model. Therefore, the best parameters are determined before fitting. The parameters of the model mainly include the number of trees, the maximum depth and the maximum number of features. The parameter adjustment method uses a nested loop method to test all pre-set combinations of the three parameters in turn. The cross-validation method is used to evaluate each parameter combination and determine the optimal parameters based on R-square and RMSE. For example, the cross-validation method first divides the data set into 10 parts, and then uses each part as a validation set and the remaining 9 parts as training sets to calculate R-square and RMSE. Finally, the average value of 10 times is calculated as the final result, and the optimal parameters of the random forest are determined.
[0082] S5. Determine the temperature vegetation drought index of the study area based on the second surface temperature data and the end member abundance values in the first reflectivity characteristic data; the temperature vegetation drought index is used to characterize the degree of drought in the study area.
[0083] Optionally, determining the temperature vegetation drought index of the study area based on the second surface temperature data and the end member abundance value in the first reflectivity characteristic data specifically includes:
[0084] Based on the end-member abundance values in the second surface temperature data and the first reflectivity characteristic data, the temperature vegetation drought index of the study area is calculated using the temperature vegetation drought index model.
[0085] The temperature vegetation drought index model is:
[0086]
[0087] Among them, TVDI is the temperature vegetation drought index of the study area, LST is the second surface temperature data, T max is the dry edge straight line temperature corresponding to the vegetation end member abundance value calculated based on the dry edge straight line equation, T min is the wet edge straight line temperature corresponding to the vegetation end member abundance value calculated based on the wet edge straight line equation.
[0088] The construction process of the temperature vegetation drought index model is as follows:
[0089] like Figure 3 As shown, a scatter plot is drawn with the vegetation end-member abundance value in the end-member abundance value as the X-axis and the second surface temperature data as the Y-axis.
[0090] Select the top point set of the scatter plot to fit the dry edge straight line equation:
[0091] T max =a1*GV+b1.
[0092] Where GV is the vegetation end-member abundance value in the end-member abundance value, and a1, and b1 are parameter variables.
[0093] Select the point set at the bottom of the scatter plot to fit the wet edge straight line equation:
[0094] T min =a2*GV+b2.
[0095] Where a2, and b2 are parameter variables.
[0096] A temperature vegetation drought index model was constructed based on the dry edge straight line equation and the wet edge straight line equation.
[0097] Optionally, before determining the temperature vegetation drought index of the study area based on the second surface temperature data and the end member abundance values in the first reflectivity characteristic data, it also includes: correcting the second surface temperature data to obtain corrected second surface temperature data.
[0098] Optionally, after the corrected second surface temperature data, the method further includes: acquiring the surface temperature data at 10 m, and performing accuracy evaluation on the corrected second surface temperature data based on the surface temperature data at 10 m, and the accuracy evaluation indicators include: R-square, RMSE and MAE.
[0099] Correcting the second land surface temperature data to obtain corrected second land surface temperature data specifically includes:
[0100] The second land surface temperature data are aggregated to obtain aggregated third land surface temperature data.
[0101] The method further comprises: performing residual correction on the second surface temperature data based on the first surface temperature data, the third surface temperature data and the second surface temperature data to obtain the second surface temperature data after residual correction, specifically comprising: taking residuals of the first surface temperature data and the third surface temperature data to obtain a first residual; the resolution of the first residual is the second resolution; resampling the first residual to obtain a second residual; the resolution of the second residual is the first resolution; and adding the second residual to the second surface temperature data to obtain the second surface temperature data after residual correction.
[0102] The second surface temperature data after residual correction is input into the fitting relationship model to obtain the corrected second surface temperature data, specifically including: constructing a linear correction model based on the first surface temperature data and the third surface temperature data; performing linear correction on the second surface temperature data after residual correction according to the linear correction model to obtain the corrected second surface temperature data.
[0103] Embodiment 2
[0104] The present invention also discloses a regional drought monitoring system based on surface spectral end-member space, which is applied to the above method and includes:
[0105] The acquisition module 101 is used to obtain the first reflectivity characteristic data and the first surface temperature data of the study area, the resolution of the first reflectivity characteristic data is the first resolution, the resolution of the first surface temperature data is the second resolution, and the first resolution is greater than the second resolution; the first reflectivity characteristic data includes: end member abundance value, cover type and terrain variables.
[0106] The aggregation module 102 is used to aggregate the first reflectivity characteristic data to obtain second reflectivity characteristic data; the resolution of the second reflectivity characteristic data is a second resolution.
[0107] The training module 103 is used to train a random forest model based on the second reflectivity characteristic data and the first surface temperature data to obtain a trained random forest model as a mapping relationship model; the mapping relationship model is used to characterize the relationship between the reflectivity characteristic data and the surface temperature data.
[0108] The mapping module 104 is used to map the first reflectivity characteristic data using a mapping relationship model to obtain second surface temperature data; the resolution of the second surface temperature data is the first resolution.
[0109] The calculation module 105 is used to determine the temperature vegetation drought index of the study area based on the second surface temperature data and the end member abundance value in the first reflectivity characteristic data; the temperature vegetation drought index is used to characterize the drought degree of the study area.
[0110] Optionally, the system further includes: a correction module, used for correcting the second ground surface temperature data to obtain corrected second ground surface temperature data.
[0111] Embodiment 3
[0112] The present invention also discloses an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above method is implemented.
[0113] The present invention trains a random forest model based on second reflectivity characteristic data and first surface temperature data, maps the first reflectivity characteristic data after obtaining the mapping relationship, obtains the second surface temperature data, and determines the temperature vegetation drought index of the study area in combination with the end member abundance value in the first reflectivity characteristic data, thereby improving the generalization ability and accuracy of the temperature vegetation drought index model.
[0114] The present invention applies the end-member abundance value data to the prediction of surface temperature data, obtains a surface temperature prediction model with stronger generalization ability, and finally obtains surface temperature data with high temporal and spatial resolution. The purpose of the present invention is to develop a high temporal resolution drought refinement monitoring method based on the surface spectral end-member space to serve the regional drought refinement management. On the one hand, it effectively improves the shortcomings of the NDVI index in the current remote sensing drought monitoring model; on the other hand, through the fusion of high spatial resolution and high temporal resolution data, it effectively solves the problem of low spatial resolution in the current remote sensing drought monitoring model.
[0115] The present invention obtains end-member abundance values with clear physical meanings through spectral hybrid decomposition technology. Vegetation end-member abundance values can not only effectively characterize the characteristics of vegetation with different coverage, but also have temporal and spatial stability; at the same time, soil end-member abundance values can effectively express the differences in soil properties. Secondly, through end-member abundance values, cover types and terrain variables, based on the random forest algorithm, the fusion of high temporal resolution of MODIS surface temperature data and high spatial resolution of Sentinel surface reflectance data is realized. Based on the spectral end-member space, surface temperature data with both high temporal resolution and high spatial resolution can be obtained. Based on the vegetation end-member abundance values, the TVDI model is adopted to realize the refined regional drought monitoring of the entire surface space. Finally, the end-member abundance values and surface temperature data with high temporal and spatial resolution are coupled to construct a regional drought refinement detection method, which can quickly and efficiently provide monitoring results with high temporal and spatial resolution. The method of the present invention can quickly and low-costly realize regional refined drought monitoring, obtain the temperature vegetation drought index monitoring results with 5-day temporal resolution and 10m spatial resolution, and effectively support regional drought warning and management.
[0116] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0117] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of the present invention. At the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.
Claims
1. A regional drought monitoring method based on surface spectral endmember space, characterized in that: The method comprises: The first reflectivity characteristic data and the first surface temperature data of the study area are obtained, the resolution of the first reflectivity characteristic data is the first resolution, the resolution of the first surface temperature data is the second resolution, and the first resolution is greater than the second resolution; the first reflectivity characteristic data include: end member abundance value, cover type and terrain variable; the first resolution characteristics of the study area are obtained, specifically including: obtaining surface reflectivity data and elevation data; the resolutions of the surface reflectivity data and the elevation data are both the first resolution; based on the surface reflectivity data, the end member abundance value is obtained by principal component analysis and end member spectral curve extraction; based on The end-member abundance values are used to obtain the cover type through the classification decision method; the terrain variables are calculated based on the elevation data; among them, the end-member abundance values are obtained based on the surface reflectance data through the principal component analysis method and the end-member spectral curve extraction, which specifically includes: using the acquired surface reflectance data, extracting the end-member spectral curves of vegetation and soil habitats based on principal component analysis, and using the linear mixed decomposition model to obtain the end-member abundance value data of the time series in the study area, including three steps: principal component analysis, end-member extraction and calculation of the end-member abundance value of each pixel in the study area; principal component analysis is to perform principal component analysis on the acquired surface reflectance data respectively to obtain the principal component The type and number of endmembers are determined based on the matrix after the transformation. Four types of endmembers are monitored in arid areas: vegetation, sand, salt and dark matter. The corresponding endmember types are determined according to the characteristics of the first three principal components in the matrix after the principal component transformation of the images of each period, and the representative seasonal phase extracted from each endmember is determined according to the contribution rate of the first three principal components of the images of each period. The endmember spectral curve is extracted using the two-dimensional scatter plot extraction method. According to the determined endmember type and representative seasonal phase, a two-dimensional scatter plot is constructed using the corresponding principal components. Based on the mathematical mechanism of the linear mixed model, the scatter plot of the principal components is distributed in a triangle, and the three vertices of the triangle are The pixels at the ends of the triangles are regarded as pure endmembers, and the interior of the triangles are mixed pixels, which are linearly combined from pure endmembers. 200-400 pixels are selected as endmember pixels at the vertices of the main component scatter plot, and the average reflectance of the endmember pixels is taken as the endmember spectrum. The endmember abundance values of each pixel in the study area are calculated by putting the endmember spectrum and the image of the study area into the linear spectral mixture decomposition model after obtaining the endmember spectrum of each endmember in the study area, and obtaining the endmember abundance values of all pixels in the study area. The land cover type is obtained by the classification decision method based on the endmember abundance value, which specifically includes: firstly, according to the land cover classification system, pixels are selected for each classification through field samples and high spatial resolution images of Google Earth, and training samples and verification samples are divided into 7:3 ratios. Based on the endmember abundance value, the segmentation variables and segmentation thresholds are determined using the endmember abundance values of the training samples of each cover type, and then a decision tree is constructed based on the determined segmentation variables and segmentation thresholds to obtain the land cover type, and the classification results are evaluated for accuracy based on the test samples. The evaluation indicators include overall classification accuracy and Kappa coefficient. Aggregating the first reflectivity characteristic data to obtain second reflectivity characteristic data; the resolution of the second reflectivity characteristic data is a second resolution; A random forest model is trained based on the second reflectivity characteristic data and the first surface temperature data to obtain a trained random forest model as a mapping relationship model; the mapping relationship model is used to characterize the relationship between the reflectivity characteristic data and the surface temperature data; the optimal parameters are first determined before fitting the random forest model, and the parameter adjustment method is: a nested loop method is used to test all pre-set combinations of the three parameters in turn, and a cross-validation method is used to evaluate each parameter combination, and the optimal parameters are determined according to R-square and RMSE; the parameters of the random forest model mainly include three parameters: the number of trees, the maximum depth, and the maximum number of features; The first reflectivity characteristic data is mapped using a mapping relationship model to obtain second surface temperature data; the resolution of the second surface temperature data is the first resolution; The temperature vegetation drought index of the study area is determined based on the second surface temperature data and the end member abundance values in the first reflectivity characteristic data; the temperature vegetation drought index is used to characterize the degree of drought in the study area, specifically including: based on the second surface temperature data and the end member abundance values in the first reflectivity characteristic data, the temperature vegetation drought index of the study area is calculated using the temperature vegetation drought index model; the temperature vegetation drought index model is: Among them, TVDI is the temperature vegetation drought index of the study area, LST is the second surface temperature data, T max is the dry edge straight line temperature corresponding to the vegetation end member abundance value calculated based on the dry edge straight line equation, T min is the wet edge straight line temperature corresponding to the vegetation end member abundance value calculated based on the wet edge straight line equation; The construction process of the temperature vegetation drought index model is as follows: the vegetation end-member abundance value in the end-member abundance value is used as the X-axis, and the second surface temperature data is used as the Y-axis to draw a scatter plot; the top point set of the scatter plot is selected to fit the dry edge straight line equation: T max =a1*GV+b1; In the formula, GV is the vegetation end-member abundance value in the end-member abundance value, a1, and b1 are parameter variables; Select the point set at the bottom of the scatter plot to fit the wet edge straight line equation: <h2 style=";text-align:left;direction:ltr">T<h2 style=";text-align:left;direction:ltr"> min <h2 style=";text-align:left;direction:ltr"> (a2*GV+b2) In the formula, a2, and b2 are parameter variables; A temperature vegetation drought index model was constructed based on the dry edge straight line equation and the wet edge straight line equation.
2. The regional drought monitoring method based on surface spectral endmember space according to claim 1, characterized in that: Before determining the temperature vegetation drought index of the study area based on the second surface temperature data and the end member abundance value in the first reflectance characteristic data, it also includes: The second ground surface temperature data is corrected to obtain corrected second ground surface temperature data.
3. The regional drought monitoring method based on surface spectral endmember space according to claim 2 is characterized in that: Correcting the second land surface temperature data to obtain corrected second land surface temperature data specifically includes: aggregating the second land surface temperature data to obtain aggregated third land surface temperature data; Performing residual correction on the second land surface temperature data based on the first land surface temperature data and the third land surface temperature data to obtain residual-corrected second land surface temperature data; The second land surface temperature data after residual correction is input into the fitting relationship model to obtain the corrected second land surface temperature data.
4. The method for regional drought monitoring based on surface spectral endmember space according to claim 3, characterized in that: The method further comprises: performing residual correction on the second land surface temperature data based on the first land surface temperature data, the third land surface temperature data and the second land surface temperature data to obtain the residual-corrected second land surface temperature data. Taking a residual of the first surface temperature data and the third surface temperature data to obtain a first residual; the resolution of the first residual is the second resolution; Resampling the first residual to obtain a second residual; the resolution of the second residual is the first resolution; The second residual is added to the second land surface temperature data to obtain the second land surface temperature data after residual correction.
5. The method for regional drought monitoring based on surface spectral endmember space according to claim 3, characterized in that: Inputting the residual-corrected second surface temperature data into the fitting relationship model to obtain the corrected second surface temperature data specifically includes: constructing a linear correction model based on the first surface temperature data and the third surface temperature data; The second surface temperature data after residual error correction is linearly corrected according to the linear correction model to obtain corrected second surface temperature data.
6. A regional drought monitoring system based on surface spectral endmember space, characterized in that: The system is applied to the method according to any one of claims 1 to 5, and the system comprises: An acquisition module is used to acquire first reflectivity characteristic data and first surface temperature data of a study area, wherein the resolution of the first reflectivity characteristic data is a first resolution, and the resolution of the first surface temperature data is a second resolution, and the first resolution is greater than the second resolution; the first reflectivity characteristic data includes: end-member abundance values, cover types, and terrain variables; acquiring first resolution characteristics of the study area specifically includes: acquiring surface reflectivity data and elevation data; the resolutions of the surface reflectivity data and the elevation data are both the first resolution; obtaining end-member abundance values based on the surface reflectivity data through principal component analysis and end-member spectral curve extraction; obtaining cover types based on the end-member abundance values through classification decision-making methods; and obtaining terrain variables based on elevation data calculations. ; Among them, the endmember abundance value is obtained based on the surface reflectance data through principal component analysis and endmember spectral curve extraction, which specifically includes: using the acquired surface reflectance data, extracting the endmember spectral curves of vegetation and soil habitats based on principal component analysis, and using the linear mixed decomposition model to obtain the endmember abundance value data of the time series in the study area, including three steps: principal component analysis, endmember extraction and calculation of the endmember abundance value of each pixel in the study area; principal component analysis is to perform principal component analysis on the acquired surface reflectance data, obtain the matrix after principal component transformation, and determine the type and number of endmembers. In arid areas, monitoring is generally for four types of endmembers: vegetation, sand, salt and dark matter. The characteristics of the first three principal components in the matrix after principal component transformation of the images of each period are used to determine the endmember abundance value. The corresponding endmember type is determined, and the representative seasonal phase extracted from each endmember is determined according to the contribution rate of the first three principal components of the images of each period; the endmember spectral curve is extracted by using the two-dimensional scatter plot extraction method. According to the determined endmember type and representative seasonal phase, the corresponding principal component is used to construct a two-dimensional scatter plot. Based on the mathematical mechanism of the linear mixed model, the scatter plot of the principal component is triangularly distributed, and the three vertices of the triangle are regarded as pure endmembers. The interior of the triangle is a mixed pixel, which is composed of a linear combination of pure endmembers. 200-400 pixels are selected as endmember pixels at the vertices of the principal component scatter plot, and the average reflectance of the endmember pixels is used as the endmember spectrum; the endmember abundance value of each pixel in the study area is calculated after obtaining the spectrum of each endmember in the study area. The endmember spectra and the image of the study area were put into the linear spectral mixture decomposition model to obtain the endmember abundance values of all pixels in the study area; the land cover type was obtained through classification decision-making based on the endmember abundance values, specifically including: first, according to the land cover classification system, pixels were selected for each classification through field samples and high spatial resolution images of Google Earth, and training samples and validation samples were divided into 7:3 ratios; based on the endmember abundance values, the segmentation variables and segmentation thresholds were determined using the endmember abundance values of the training samples of each cover type; then a decision tree was constructed based on the determined segmentation variables and segmentation thresholds to obtain the land cover type; and the classification results were evaluated for accuracy based on the test samples, and the evaluation indicators included overall classification accuracy and Kappa coefficient; An aggregation module, used for aggregating the first reflectivity characteristic data to obtain second reflectivity characteristic data; the resolution of the second reflectivity characteristic data is a second resolution; A training module is used to train a random forest model based on the second reflectivity feature data and the first surface temperature data to obtain a trained random forest model as a mapping relationship model; the mapping relationship model is used to characterize the relationship between the reflectivity feature data and the surface temperature data; the optimal parameters are first determined before fitting the random forest model, and the parameter adjustment method is: a nested loop method is used to test all pre-set combinations of three parameters in turn, and a cross-validation method is used to evaluate each parameter combination, and the optimal parameters are determined according to R square and RMSE; the parameters of the random forest model mainly include three parameters: the number of trees, the maximum depth, and the maximum number of features; A mapping module, used to map the first reflectivity characteristic data using a mapping relationship model to obtain second surface temperature data; the resolution of the second surface temperature data is the first resolution; A calculation module is used to determine the temperature vegetation drought index of the study area based on the second surface temperature data and the end member abundance value in the first reflectivity characteristic data; the temperature vegetation drought index is used to characterize the degree of drought in the study area, specifically including: based on the second surface temperature data and the end member abundance value in the first reflectivity characteristic data, using the temperature vegetation drought index model to calculate the temperature vegetation drought index of the study area; the temperature vegetation drought index model is: Among them, TVDI is the temperature vegetation drought index of the study area, LST is the second surface temperature data, T max is the dry edge straight line temperature corresponding to the vegetation end member abundance value calculated based on the dry edge straight line equation, T min is the wet edge straight line temperature corresponding to the vegetation end member abundance value calculated based on the wet edge straight line equation; The construction process of the temperature vegetation drought index model is as follows: the vegetation end-member abundance value in the end-member abundance value is used as the X-axis, and the second surface temperature data is used as the Y-axis to draw a scatter plot; the top point set of the scatter plot is selected to fit the dry edge straight line equation: T max =a1*GV+b1; In the formula, GV is the vegetation end-member abundance value in the end-member abundance value, a1, and b1 are parameter variables; Select the point set at the bottom of the scatter plot to fit the wet edge straight line equation: <h2 style=";text-align:left;direction:ltr">T<h2 style=";text-align:left;direction:ltr"> min <h2 style=";text-align:left;direction:ltr"> (a2*GV+b2) In the formula, a2, and b2 are parameter variables; A temperature vegetation drought index model was constructed based on the dry edge straight line equation and the wet edge straight line equation.
7. An electronic device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 5 when executing the computer program.
Citation Information
Patent Citations
Agricultural drought remote sensing monitoring method
CN105929406A
Random forest-based multifactor remote sensing surface temperature space downscaling method
CN107748736A
A method for reducing that scale of surface temperature space
CN109635309A
Land system state evaluation method and device and storage medium
CN113469586A
Vegetation coverage inversion method and device for hyperspectral remote sensing image data
CN116824410A
