Surface water extraction method based on satellite hyperspectral remote sensing data

Abstract

The present invention discloses a method for extracting surface water based on satellite hyperspectral remote sensing data, which includes the following main steps: step 1, preprocessing the acquired multiple satellite hyperspectral remote sensing data; step 2, selecting characteristic pixels of different objects on the image, and constructing the spectral characteristic curve of the object; step 3, analyzing the spectral characteristic curves of different types of objects, constructing a new water extraction model, and processing the image using the new model; step 4, constructing an algorithm to separate the water body from other objects in the processed image. The present invention proposes a multiple water index construction algorithm that is suitable for a variety of different spectral data environments, based on which water body information can be accurately extracted.

Description

Technical Field

[0001] The invention relates to the technical field of remote sensing spectrum data processing, in particular to a surface water body extraction method based on satellite hyperspectral remote sensing data. Background Art

[0002] Water resources are the basic conditions for the survival and development of humans and other organisms. Water extraction plays an important role in water resource investigation, flood disaster assessment and prediction, and environmental monitoring. Remote sensing technology, with its advantages of high temporal and spatial resolution and low cost, can quickly extract and monitor a wide range of water information, becoming one of the important technical means for water extraction and water quality monitoring.

[0003] At present, the research methods of water body extraction based on remote sensing images mainly include single-band threshold method, inter-band relationship method and water body index method. The single-band threshold method is relatively simple. It distinguishes water bodies from other land objects by setting thresholds based on the reflectivity difference between water bodies and the environment. The inter-band relationship law separates water bodies from the background by constructing a logical relationship between bands and using the spectral characteristics of land objects. The water body index law constructs parameters related to water bodies based on the higher reflectivity of water bodies in the blue-green band and the lower reflectivity of the near-infrared band using some specific bands. In general, the single-band threshold method and the inter-band relationship method use the information of a single or a small number of bands, and cannot completely distinguish water bodies from backgrounds, resulting in unsatisfactory extraction effects. The water body index method is a more effective method because of its simplicity and the use of the spectral characteristics of remote sensing images.

[0004] The existing water body index is mainly designed for multispectral remote sensing data such as Landsat and high-resolution series, while there is little research on water body index for hyperspectral data. Hyperspectral remote sensing has the advantage of unified image and spectrum, and can obtain reflectance information of ground objects in dozens or even hundreds of bands, so that the spectral characteristic curve of the ground objects can be used for identification more accurately, effectively improving the application ability of remote sensing. At present, there is still a lot of room for development in the use of hyperspectral remote sensing data for water body extraction. Summary of the invention

[0005] The technical problem to be solved by the present invention is to overcome various deficiencies of the prior art and provide a method for extracting surface water based on satellite hyperspectral remote sensing data.

[0006] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows.

[0007] A surface water extraction method based on satellite hyperspectral remote sensing data, the method includes the following main steps:

[0008] Step 1: preprocess the acquired multiple satellite hyperspectral remote sensing data;

[0009] Step 2: Select characteristic pixels of different objects on the image and construct the spectral characteristic curve of the objects;

[0010] Step 3: Analyze the spectral characteristic curves of different types of objects, build a new water body extraction model, and use the new model to process the image;

[0011] Step 4: Construct an algorithm to separate water bodies and other land objects from the processed images.

[0012] As a preferred technical solution of the present invention, in step 1, preprocessing the hyperspectral remote sensing image data specifically includes:

[0013] Radiation correction: The DN value in the hyperspectral satellite image is converted into a radiation brightness value using the sensor response function, thereby eliminating the error of the sensor itself and determining the accurate radiation value at the sensor entrance;

[0014] Atmospheric correction: The radiation-corrected image is used as input, and the 6S model is used to perform atmospheric correction on the result of the radiation correction to eliminate the influence of solid particles and water vapor in the atmosphere on the image;

[0015] Geometric correction: Use the reference image to perform geometric correction on the image resulting from atmospheric correction to obtain hyperspectral radiance data with geographic coordinates.

[0016] As a preferred technical solution of the present invention, in step 2, the feature pixel is constructed as a pure pixel of a certain type of land object, and the pure pixel is constructed as the current pixel is located in the non-edge area of ​​a certain land object category, and the pixel only contains one type of land object information; wherein, selecting feature pixels of different land objects on the hyperspectral remote sensing image specifically includes: determining the number of land object categories on the image, and uniformly selecting a certain number of pure pixels for each category of land objects as feature pixels.

[0017] As a preferred technical solution of the present invention, in step 2, the step of constructing the ground feature spectral characteristic curve includes:

[0018] Step 2.1, calculate the JM distance between different types of objects to characterize the separability between different types of objects;

[0019] Step 2.2, the reflectance of all characteristic pixels of each type of ground object is counted, and the average value is taken to represent the spectral characteristic curve of the ground object;

[0020] Step 2.3: Perform the above processing on all types of ground objects to obtain the spectral characteristic curves of all types of ground objects.

[0021] As a preferred technical solution of the present invention, in step 3, the method for analyzing the spectral characteristic curves of different categories of ground objects is: summarizing the spectral characteristic curves of all categories of ground objects, analyzing the characteristics of the spectral curves of different categories of ground objects, and finding the bands that distinguish water from other ground objects.

[0022] As a preferred technical solution of the present invention, in step 3, a new water body extraction model is constructed and the new model is used to process the image. Specifically, for the pre-processed satellite hyperspectral remote sensing data, the spectral reflectance function of the interval segment is definitely integrated according to the spectral order. The spectral order is distinguished according to the physical parameters of the spectrum. Generally, when the wavelength λ is used, the endpoints of the integral interval segment are characterized by the specified wavelength value; then, based on the difference of the definite integral obtained in the constant interval, the first-order driving derivative and / or the higher-order driving derivative of the spectrum are characterized, and the characterization parameter of any order driving derivative or its combination is constructed as a surface water body index, and the surface water body in the satellite hyperspectral remote sensing data is separated and extracted after hierarchical clustering post-processing of the data; wherein, when the interval length of the difference is constant, the difference characterizes the average value of the continuous derivatives in the interval; when the object of investigation is the mean of the interval derivative, this first-order derivative that drives the data can be characterized by the difference of two groups of definite integrals under the constant interval length; further, for the difference of The first-order mean derivative represented by interval difference, when the surface driving factor of this spectral data cannot well distinguish the spectral object, a high-order driving factor with more refined data connotation, or a combination of high-order and low-order driving factors should be used; for this purpose, a multi-level driving derivative parameterization model oriented to the separation target and combined with the separation effect is constructed; in the multi-level driving derivative parameterization model, the high-order driving derivative is obtained based on the first-order driving derivative through the following iterative method: for the second-order driving derivative, its difference in a constant interval length is obtained by subtracting two adjacent first-order driving derivatives; in this way, the second-order driving derivative contains the connotation information of the two first-order driving derivatives, and is constructed by the difference of adjacent differences, which corresponds to the interactive information between three groups of definite integrals under the constant interval length; for the third-order driving derivative, its difference in a constant interval length is obtained by subtracting two adjacent second-order driving derivatives; and it is iterated in this way; for any order of driving derivative, since a constant interval length is used, the difference of the driving derivative of the previous order directly represents the driving derivative of the next order.

[0023] As a preferred technical solution of the present invention, in step 4, the construction algorithm divides the input image into two categories: water bodies and other landforms. First, the input data is divided into n groups, and then n objects are randomly selected as the initial sub-data centers. Then, the distance between each object and each sub-data center is calculated, and each object is assigned to the data center closest to it; the data centers and the objects assigned to them are constructed into a sub-data group and labeled; each time a sample object is assigned, the data center returns to recalculate all existing objects containing sub-data group information; this data process is iterated until all data are traversed.

[0024] As a preferred technical solution of the present invention, the present invention also includes step 5: using a confusion matrix or other algorithms to evaluate the accuracy of water body extraction results; in step 5, each column of the confusion matrix represents a predicted category, and the total number of each column represents the number of data predicted to be in that category; each row of the confusion matrix represents the true category of the data, and the total number of data in each row represents the number of data instances in that category; the overall accuracy and kappa coefficient are calculated based on the confusion matrix to evaluate the classification accuracy.

[0025] The beneficial effects of adopting the above technical scheme are as follows: the present invention proposes a multiple water body index construction algorithm that is adapted to a variety of different spectral data environments, based on which water body information can be accurately extracted; at the same time, compared with the traditional method of setting a single threshold to separate water bodies and non-water bodies, the present invention further introduces a data grouping algorithm based on the data mean center to separate water bodies and non-water bodies, thereby improving the accuracy of remote sensing water body extraction; and achieving a water body extraction accuracy of more than 90% on remote sensing data with a resolution of ten meters, greatly improving the accuracy and efficiency of remote sensing identification of water bodies, which has important scientific research and practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 The following are the spectral characteristic curves of five typical ground objects.

[0027] Figure 2 It is a comparison diagram of the actual water body distribution and the details of the water body extracted by the method used in the present invention. DETAILED DESCRIPTION

[0028] The following examples illustrate the invention in detail.

[0029] As used in the specification of the present application and the appended claims, the term "if" can be interpreted as "when..." or "once" or "in response to determination" or "in response to detection" according to the context. Similarly, the phrase "if it is determined" or "if [the described condition or event] is detected" can be interpreted as meaning "once it is determined" or "in response to determination" or "once [the described condition or event] is detected" or "in response to detection [the described condition or event]" according to the context. In addition, in the description of the specification of the present application and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance. The reference "one embodiment" or "some embodiments" described in the specification of the present application means that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Thus, the sentences "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in the specification are not necessarily all referring to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "include", "comprising", "having" and variations thereof all mean "including but not limited to", unless specifically emphasized otherwise.

[0030] Example 1

[0031] The surface water extraction method based on satellite hyperspectral remote sensing data of the present invention includes the following main step framework:

[0032] Step 1: preprocess the acquired multiple satellite hyperspectral remote sensing data;

[0033] Step 2: Select characteristic pixels of different objects on the image and construct the spectral characteristic curve of the objects;

[0034] Step 3: Analyze the spectral characteristic curves of different types of objects, build a new water body extraction model, and use the new model to process the image;

[0035] Step 4: Construct an algorithm to separate water bodies and other land objects from the processed images.

[0036] Step 5: Use confusion matrix or other algorithms to evaluate the accuracy of water body extraction results.

[0037] Example 2

[0038] In step 1, the multiple satellite hyperspectral remote sensing data include 2 or more hyperspectral satellite images. Specifically, in step 1, the preprocessing of the hyperspectral remote sensing image data includes radiation correction, atmospheric correction and geometric correction. The specific steps are as follows:

[0039] Radiation correction: The hyperspectral satellite image is input into the PIE-Hyp software, and the sensor response function is used to convert the DN value in the remote sensing image into a radiation brightness value, thereby eliminating the error of the sensor itself and determining the accurate radiation value at the sensor entrance.

[0040] Atmospheric correction: The radiation-corrected image is used as input, and the 6S model of the PIE-Hyp software is used to perform atmospheric correction on the result of the radiation correction to eliminate the influence of solid particles, water vapor, etc. in the atmosphere on the image.

[0041] Geometric correction: The reference image of PIE-Hyp software is used to perform geometric correction on the result image of atmospheric correction to obtain hyperspectral radiance data with geographic coordinates.

[0042] For example, we selected two Zhuhai No. 1 hyperspectral data in the Chengdu Plain area of ​​Sichuan Province. One of the images is located in a flat area, with the Tuojiang River, a tributary of the Yangtze River, in the center of the image, and many small rivers, ponds and reservoirs distributed around it; the other image is located in a mountainous area, with the Qingyi River in the center of the image, alluvial plains on both sides, and mainly mountains around it, and some small rivers and ponds are also distributed. The effectiveness and applicability of the present invention are tested using images distributed in different terrains. The two hyperspectral satellite images are respectively input into the PIE-Hyp software, and the images are sequentially corrected for radiation, atmosphere and geometry using the tools provided by the software, and the resulting data are saved for subsequent processing.

[0043] Example 3

[0044] In step 2, the characteristic pixel refers to a pure pixel of a certain type of object, and the pure pixel refers to the current pixel being located in the non-edge area of ​​a certain object category, and the pixel only contains one type of object information. Specifically, in step 2, selecting characteristic pixels of different objects on the image specifically includes: determining the number of object categories on the image, and uniformly selecting a certain number of pure pixels for each category of objects as characteristic pixels.

[0045] Constructing the spectral characteristic curve of the ground object includes: step 2.1, calculating the direct JM distance of different ground object categories. If it is close to 2, it means that the spectra of different categories of ground objects are better. Specifically, in step 2.1, JM distance refers to Jeffreys-Matusita distance, which refers to a common indicator used to measure the spectral separability between two groups in remote sensing problems, and is used to characterize the separability between different categories of ground objects. Step 2.2, the reflectance of all characteristic pixels of each type of ground object is counted, and the average value is taken to characterize the spectral characteristic curve of this type of ground object; step 2.3, the above processing is performed on all categories of ground objects to obtain the spectral characteristic curves of all categories of ground objects.

[0046] In this embodiment, a certain number of pure pixels are selected from the preprocessed image for the five types of objects, namely water, buildings, roads, vegetation and shadows. The pure pixels must be located in the non-edge area of ​​the object category. The number of pixels for each type of object is determined according to the area occupied by the object category, and generally follows the principle of uniform distribution.

[0047] To test the separability of the selected pixels, the ENVI software was used to calculate the JM distance between different land types. If the value is close to 2, it means that the pixels of the selected land type are well separable, otherwise the characteristic pixels need to be reselected. After selecting pure pixels, the reflectance of all characteristic pixels of each type of land object is statistically analyzed and averaged to obtain the spectral characteristic curves of each type of land object.

[0048] Example 4

[0049] In step 3, analyzing the spectral characteristic curves of different types of objects specifically includes summarizing the spectral characteristic curves of all types of objects, analyzing the characteristics of the spectral curves of different types of objects, and finding the bands that distinguish water from other objects. Specifically, for the pre-processed satellite hyperspectral remote sensing data, the spectral reflectance function of the interval segment is definitely integrated according to the spectral order. The spectral order is distinguished according to the physical parameters of the spectrum. Generally, when the wavelength λ is used, the endpoints of the integral interval segment are characterized by the specified wavelength values, such as 550-700nm, 700-850nm, etc.; then, based on the difference of the definite integral obtained in the constant interval, the first-order driving derivative and / or the higher-order driving derivative of the spectrum are characterized, and the characterization parameter of any order driving derivative or its combination is constructed as a surface water body index, and the surface water body in the satellite hyperspectral remote sensing data is separated and extracted after hierarchical clustering post-processing of the data.

[0050] Obviously, when the length of the interval for constructing the difference is constant, the difference represents the average value of the continuous derivatives within the interval; when the object of investigation is the mean of the interval derivatives, this first-order derivative that drives the data can be represented by the difference between two sets of definite integrals under a constant interval length.

[0051] Furthermore, we realize that when the first-order mean derivative (represented by interval difference), the surface driving factor of spectral data, cannot well distinguish spectral objects, a higher-order driving factor with more refined data connotation, or a combination of high-order and low-order driving factors, should be used. To this end, we propose a multi-level driving derivative parameterization model that is oriented to the separation target and combines the separation effect. Among them, the higher-order driving derivative is obtained based on the first-order driving derivative through the following iterative method: for the second-order driving derivative, its difference in a constant interval length is obtained by subtracting the two adjacent first-order driving derivatives; in this way, the second-order driving derivative contains the connotation information of the two first-order driving derivatives, which is constructed by the difference of the adjacent differences, corresponding to the interactive information between the three groups of definite integrals under the constant interval length; for the third-order driving derivative, its difference in a constant interval length is obtained by subtracting the two adjacent second-order driving derivatives; and so on. For any order of driving derivative, since a constant interval length is used, the difference of the driving derivative of the previous order directly represents the driving derivative of the next order.

[0052] Both theory and practice have shown that when constructing a surface water separation algorithm in an application, the order of the driving derivative is derived with the accuracy value that can achieve the separation effect as the standard termination symbol of the data process; generally, the first or second order can meet conventional requirements.

[0053] For example, we constructed characteristic spectral curves using characteristic pixels of five types of objects, including water bodies, buildings, roads, vegetation, and shadows, selected from two Zhuhai No. 1 hyperspectral remote sensing images. Figure 1 As shown. Analyzing the reflectivity characteristics of different types of land objects in different bands, the reflectivity curves of water bodies in different bands are relatively flat, the reflectivity of other land objects is low before 700nm, and the reflectivity increases to varying degrees after 700nm. According to the reflectivity spectral curve characteristics of various land objects, the reflectivity spectral curve of water bodies is relatively flat, which means that its first-order derivative is close to 0, while the first-order derivative of the spectral curves of other land objects should be positive. The bands (550nm~700nm and 700nm~850nm) that are conducive to distinguishing water bodies from other land objects are selected to construct a water body extraction model. Specifically, the constructed new hyperspectral water index (Hyperspectral Water Index, HWI) is as follows:

[0054]

[0055] R(λ) represents the reflectance of the wavelength λ on the hyperspectral image, and the characteristic band is selected from the band between (b06~b16) and (b16~b23). The construction of the HWI model is mainly based on the idea of ​​integration and difference, using the difference approximation to replace the first-order derivative to simplify the calculation. The new model is used to process the image, including the calculation of the HWI index for the preprocessed hyperspectral remote sensing image. The preprocessed hyperspectral image data is calculated using the HWI model to obtain a single-band result image for subsequent processing.

[0056] This is the first-order case; for higher-order cases, the above method can be used for iteration. When applied, a comprehensive analysis method with different order parameters is generally used, and a higher weight is assigned to the lower order. Generally, only one or two orders can meet conventional requirements.

[0057] Example 5

[0058] In step 4, the algorithm is used to separate water bodies and land objects in the image processed by the new model. Specifically, the results calculated using the HWI model are used as input, and the algorithm is used to divide the input image into two categories: water bodies and other land objects. First, the input data is divided into n groups, and n objects are randomly selected as the initial sub-data centers. Then, the distance between each object and each sub-data center is calculated, and each object is assigned to the data center closest to it; the data center and the objects assigned to them are constructed into a sub-data group and labeled; each time a sample object is assigned, the data center returns to recalculate all existing objects containing sub-data group information; this data process is iterated until all data are traversed. Therefore, we input the single-band image processed by the HWI model into the ENVI software for data center clustering, and divide the image into two categories: water bodies and non-water bodies. The local details of the classification results of the two images are shown as follows. Figure 2 shown.

[0059] Example 6

[0060] In step 5, the confusion matrix algorithm is used, which is a standard format for expressing accuracy evaluation, and is expressed in the form of a matrix with n rows and n columns. Each column of the confusion matrix represents the predicted category, and the total number of each column represents the number of data predicted to be in that category; each row represents the true category of the data, and the total number of data in each row represents the number of data instances in that category. According to the confusion matrix, the overall accuracy and kappa coefficient can be calculated to evaluate the classification accuracy. Specifically, in step 5, the accuracy evaluation indicators are the overall accuracy and kappa coefficient, which can be calculated according to the confusion matrix. The overall accuracy is the sum of correctly classified pixels divided by the total number of pixels; the kappa coefficient is obtained by multiplying the total number of pixels in all surface true classifications by the sum of the diagonal of the confusion matrix, and then subtracting the sum of all categories by the product of the total number of surface true pixels of a certain class and the total number of classified pixels in that class, and then dividing it by the square of the total number of pixels minus the sum of all categories by the product of the total number of surface true pixels of a certain class and the total number of classified pixels in that class. In this embodiment, sample points are randomly selected on the two images to test the water body classification accuracy, and a total of 666 sample points are selected, including 165 water body sample points and 501 non-water body sample points. The overall accuracy of the two regions calculated is 92.11% and 93.50%, respectively, and the kappa coefficients are 0.842 and 0.850, respectively.

[0061] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0062] In each embodiment, the hardware implementation of the technology can directly adopt existing intelligent devices, including but not limited to industrial computers, PCs, smart phones, handheld stand-alone machines, floor-standing stand-alone machines, etc. Its input device preferably adopts a screen keyboard, its data storage and calculation module adopts existing memory, calculator, controller, its internal communication module adopts existing communication ports and protocols, and its remote communication adopts existing GPRS networks, World Wide Internet, etc. It can be clearly understood by technicians in the relevant field that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of each functional unit and module are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the above-mentioned method embodiment, which will not be repeated here. In the embodiments provided by the present invention, it should be understood that the disclosed device / terminal equipment and method can be implemented in other ways. For example, the device / terminal equipment embodiment described above is only schematic, for example, the division of modules or units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms. The unit described as a separate component may or may not be physically separated, and the component displayed as a unit may or may not be a physical unit, that is, it may be located in one place, or it may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0063] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A surface water extraction method based on satellite hyperspectral remote sensing data, characterized in that: The method includes the following main steps: Step 1: preprocess the acquired multiple satellite hyperspectral remote sensing data; Step 2: Select characteristic pixels of different objects on the image and construct the spectral characteristic curve of the objects; Step 3: Analyze the spectral characteristic curves of different types of objects, build a new water body extraction model, and use the new model to process the image; Step 4: construct an algorithm to separate water bodies from other land objects in the processed images; In step 3, a new water body extraction model is constructed and the new model is used to process the image. Specifically, the pre-processed satellite hyperspectral remote sensing data is subjected to the definite integral of the spectral reflectance function of the interval segment according to the spectral order. The spectral order is distinguished according to the physical parameters of the spectrum, generally using wavelength. When the endpoints of the integral interval are characterized by the specified wavelength values; then, the first-order driving derivative and / or high-order driving derivative of the spectrum are characterized based on the difference of the definite integral obtained in the constant interval, and the characterization parameter of any order driving derivative or its combination is constructed as a surface water body index, and the surface water body in the satellite hyperspectral remote sensing data is separated and extracted after hierarchical clustering post-processing of the data; wherein, when the interval length of the constructed difference is constant, the difference characterizes the average value of the continuous derivatives in the interval; when the object of investigation is the mean of the interval derivative, the first-order derivative that drives the data can be characterized by the difference of two groups of definite integrals under the constant interval length; further, for the first-order mean derivative represented by the interval difference, when the surface driving factor of the spectral data cannot well distinguish the spectral object, a high-order driving factor with more refined data connotation, or a combination of high-order and low-order driving factors should be used; To this end, a multi-level driven derivative parameterization model is constructed for separation goals and combined with separation effects; in the multi-level driven derivative parameterization model, high-order driven derivatives are obtained based on first-order driven derivatives through the following iterative method: for second-order driven derivatives, their difference in a constant interval length is obtained by subtracting two adjacent first-order driven derivatives; in this way, the second-order driven derivative contains the connotation information of the two first-order driven derivatives, and is constructed by the difference of adjacent differences, which corresponds to the interactive information between three groups of definite integrals under a constant interval length; for third-order driven derivatives, their difference in a constant interval length is obtained by subtracting two adjacent second-order driven derivatives; and so on and so forth, it is iterated; for any order of driven derivatives, since a constant interval length is used, the difference of the previous order driven derivative directly represents the next order driven derivative.

2. The surface water extraction method based on satellite hyperspectral remote sensing data according to claim 1 is characterized in that: In step 1, the preprocessing of the hyperspectral remote sensing image data specifically includes: Radiation correction: The DN value in the hyperspectral satellite image is converted into a radiation brightness value using the sensor response function, thereby eliminating the error of the sensor itself and determining the accurate radiation value at the sensor entrance; Atmospheric correction: The radiation-corrected image is used as input, and the 6S model is used to perform atmospheric correction on the result of the radiation correction to eliminate the influence of solid particles and water vapor in the atmosphere on the image; Geometric correction: Use the reference image to perform geometric correction on the image resulting from atmospheric correction to obtain hyperspectral radiance data with geographic coordinates.

3. The surface water extraction method based on satellite hyperspectral remote sensing data according to claim 1 is characterized in that: In step 2, the characteristic pixel is constructed as a pure pixel of a certain type of land object, and the pure pixel is constructed as the current pixel is located in the non-edge area of ​​a certain land object category, and the pixel only contains one type of land object information; wherein, selecting characteristic pixels of different land objects on the hyperspectral remote sensing image specifically includes: determining the number of land object categories on the image, and uniformly selecting a certain number of pure pixels for each category of land objects as characteristic pixels.

4. The surface water extraction method based on satellite hyperspectral remote sensing data according to claim 1 is characterized in that: In step 2, the step of constructing the ground feature spectral characteristic curve includes: Step 2.1, calculate the JM distance between different types of objects to characterize the separability between different types of objects; Step 2.2, the reflectance of all characteristic pixels of each type of ground object is counted, and the average value is taken to represent the spectral characteristic curve of the ground object; Step 2.3: Perform the above processing on all types of ground objects to obtain the spectral characteristic curves of all types of ground objects.

5. The method for extracting surface water based on satellite hyperspectral remote sensing data according to claim 1, characterized in that: In step 3, the method for analyzing the spectral characteristic curves of different types of ground objects is: summarizing the spectral characteristic curves of all types of ground objects, analyzing the characteristics of the spectral curves of different types of ground objects, and finding the bands that distinguish water from other ground objects.

6. The surface water extraction method based on satellite hyperspectral remote sensing data according to claim 1 is characterized in that: In step 4, the algorithm is constructed to divide the input image into two categories: water bodies and other objects. First, the input data is divided into n groups, and n objects are randomly selected as the initial sub-data centers. Then, the distance between each object and each sub-data center is calculated, and each object is assigned to the data center closest to it. The data centers and the objects assigned to them are constructed into a sub-data group and labeled. After each sample object is assigned, the data center returns to recalculate all existing objects containing sub-data group information. This data process is iterated until all the data are traversed.

7. The method for extracting surface water based on satellite hyperspectral remote sensing data according to claim 1, characterized in that: The present invention also includes step 5: using a confusion matrix or other algorithms to evaluate the accuracy of water body extraction results; in step 5, each column of the confusion matrix represents a predicted category, and the total number of each column represents the number of data predicted to be in this category; each row of the confusion matrix represents the true category of the data, and the total number of data in each row represents the number of data instances in this category; the overall accuracy and kappa coefficient are calculated based on the confusion matrix to evaluate the classification accuracy.

Images