Automatic spectral endmember extraction method and device

Abstract

The present invention relates to a method and device for automatically extracting spectral endmembers, which obtain satellite remote sensing images and calculate vegetation index images of the satellite remote sensing images; use the corresponding pixel values in each vegetation index image as the coordinate points of a two-dimensional coordinate system, construct triangular feature space scatter points pixel by pixel, and determine the effective scatter point range; finally, automatically search for spectral endmembers based on the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space to complete the extraction. Based on this, it solves the problems of time-consuming, laborious and low automation degree such as manual intervention and expert prior knowledge required in the selection of spectral endmembers, and improves the efficiency and accuracy of spectral endmember extraction.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing technology, and particularly to a method and device for automatically extracting spectral endmembers. Background Art

[0002] Remote sensing has become an important means of earth observation due to its ability to rapidly acquire information on the attributes of ground objects. In recent years, new types of satellite multi / hyperspectral remote sensing data, such as high-resolution, medium-high-resolution, and high-temporal low-resolution data, have emerged continuously. These images not only have rich spectral and spatial information, but also the data acquisition with a revisit period and global coverage range has brought new opportunities for large-scale and long-term remote sensing applications.

[0003] Photosynthetic Vegetation (PV), Bare Soil (BS), and Non-Photosynthetic Vegetation (NPV) are important components of the soil-vegetation ecosystem, including agriculture, forests, and grasslands. Due to the spectral characteristics of vegetation with a sharp increase in near-infrared reflectance, vegetation indices can be used to distinguish PV from BS and NPV well. However, NPV and BS often have similar spectral reflectances in the range of 400 - 2500 nm, making it difficult to distinguish them, which also poses a challenge to the integrated inversion of NPV-BS-PV coverage information. Researchers have constructed spectral indices that can characterize the features of PV and NPV by combining the spectral reflectance spectra and their various transformation forms of PV, NPV, and BS in the visible-near-infrared range (400 - 1000 nm) to distinguish the three of PV, BS, and NPV. For example, based on the Linear spectral unmixing technology and the Spectral mixture analysis (SMA) model, a triangular space technology has been developed by constructing a two-dimensional scatter plot through two spectral indices that characterize PV and NPV. According to the position of the scatter points in the constructed triangular space, the pixel three-component coverage information of NPV-BS-PV can be inverted.

[0004] However, currently, based on the triangular space technology, it is possible to achieve integrated inversion of NPV-BS-PV coverage information. There are many problems in inverting NPV-BS-PV coverage information on large spatial and long temporal scales, specifically including the following points: First, the inversion process of NPV-BS-PV coverage information is based on linear spectral unmixing and depends on the determination of spectral endmembers. In the triangular space technology, the determination of spectral endmembers highly relies on expert prior knowledge and manual intervention. Second, the degree of automation for manual extraction of spectral endmembers is low. Inverting PV-NPV-BS coverage information on large spatial and long temporal scales is time-consuming and laborious, and it is difficult to meet the requirements for high-precision integrated remote sensing inversion and mapping of PV-NPV-BS coverage. Summary of the Invention

[0005] Based on this, it is necessary to provide a method and device for automatic extraction of spectral endmembers in view of the deficiencies existing in the traditional method for extracting spectral endmembers.

[0006] A method for automatic extraction of spectral endmembers includes the steps of:

[0007] Obtain satellite remote sensing images;

[0008] Calculate the vegetation index images of the satellite remote sensing images;

[0009] Take the corresponding pixel values in each vegetation index image as the coordinate points of a two-dimensional coordinate system, and construct triangular feature space scatter points pixel by pixel, and determine the effective scatter point range;

[0010] According to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space, automatically search for spectral endmembers to complete the extraction.

[0011] The above method for automatic extraction of spectral endmembers obtains satellite remote sensing images and calculates the vegetation index images of the satellite remote sensing images; takes the corresponding pixel values in each vegetation index image as the coordinate points of a two-dimensional coordinate system, and constructs triangular feature space scatter points pixel by pixel, and determines the effective scatter point range; finally, according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space, automatically search for spectral endmembers to complete the extraction. Based on this, it solves the problems of time-consuming, laborious and low automation degree such as manual intervention and expert prior knowledge in the selection of spectral endmembers, and improves the efficiency and accuracy of spectral endmember extraction.

[0012] In one embodiment, the process of obtaining satellite remote sensing images includes the steps of:

[0013] Obtain satellite remote sensing images and remove the water body part in the satellite remote sensing images.

[0014] In one embodiment, the vegetation index images include NDVI vegetation index images and NSSI vegetation index images.

[0015] In one embodiment, the process of taking the corresponding pixel values in each vegetation index image as the coordinate points of a two-dimensional coordinate system, constructing a triangular feature space scatter plot pixel by pixel, and determining the effective scatter point range includes the steps:

[0016] Take the corresponding pixel values of the NDVI vegetation index image as the x-axis coordinate points of the two-dimensional coordinate system, and take the corresponding pixel values of the NSSI vegetation index image as the y-axis coordinate points of the two-dimensional coordinate system;

[0017] According to the distribution law of the triangular feature space scatter plot, set a threshold as the pixel frequency for removing pixel noise points, and respectively remove the noise points from the x-axis data and y-axis data according to the set statistical results of the pixel frequency, obtaining the frequency ranges of the pixel values of the x-axis data and y-axis data after removal;

[0018] According to the set statistical results and frequency ranges, respectively determine the minimum and maximum values of the NDVI vegetation index and the NSSI vegetation index as the extreme value parameters;

[0019] Construct the boundary range in the triangular feature space according to the extreme value parameters as the effective scatter point range.

[0020] In one embodiment, the process of automatically searching for spectral endmembers to complete the extraction according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space includes the steps:

[0021] Take the y-axis direction in the rectangular boundary where the maximum value of the x-axis is located as the search direction for spectral endmembers, determine the search range on the x-axis with the maximum value of the x-axis as the axis by setting the buffer size, and determine the possible range of the x-axis of the spectral endmembers;

[0022] Statistically calculate the different pixel values in all y-axis directions within the possible range of the x-axis and calculate the frequency, and determine the search interval on the y-axis by setting the frequency threshold;

[0023] Perform cyclic search and statistical calculation with a fixed step size on the search interval of the y-axis to reduce the selection of spectral endmember candidate points;

[0024] In the cyclic search, statistically calculate the number of pixels in the rectangular frame with the maximum value of the x-axis and the center point determined by the candidate point as the axis through a preset search distance, and save the statistical results in the cycle;

[0025] Select the center point position corresponding to the rectangular frame with the largest number of pixels in the statistical results of the number of pixels obtained in the cycle as the spectral endmember of PV.

[0026] In one embodiment, the process of automatically searching for spectral endmembers to complete the extraction according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space includes the steps:

[0027] Take the rectangular boundary where the maximum value of the y-axis lies as the search range of the spectral endmember. Determine the search range on the y-axis by setting the buffer size with the maximum value of the y-axis as the axis, and determine the possible range of the y-axis of the spectral endmember.

[0028] Statistically calculate the different pixel values in all y-axis directions within the possible range of the y-axis and calculate the frequency. Determine the search interval on the x-axis by setting the frequency threshold.

[0029] Perform cyclic search and statistical calculation with a fixed step size on the search interval of the x-axis to reduce the selection of spectral endmember candidate points.

[0030] During the cyclic search, statistically calculate the number of pixels in the rectangular box with the center point determined by the maximum value of the y-axis and the candidate point according to the preset search distance, and save the statistical results in the cycle.

[0031] Select the center point position corresponding to the rectangular box with the largest number of pixels in the statistical results of the number of pixels obtained in the cycle as the spectral endmember of NPV.

[0032] In one embodiment, according to the effective scatter point range and the distribution law of the spectral endmember in the triangular feature space, automatically search for the spectral endmember to complete the extraction process, including the steps:

[0033] According to the distribution law of the triangular feature space, use the minimum value as the BS spectral endmember value.

[0034] In one embodiment, it further includes the steps:

[0035] Use the spectral endmember and the vegetation index image as the input parameters of the pixel three-component model to perform integrated remote sensing inversion of PV-NPV-BS coverage.

[0036] In one embodiment, the process of using the spectral endmember and the vegetation index image as the input parameters of the pixel three-component model to perform integrated remote sensing inversion of PV-NPV-BS coverage is as follows:

[0037] f PV +f NPV +f BS =1

[0038] NDVI M =f PV ·NDVI PV +f NPV ·NDVI NPV +f BS ·NDVI BS

[0039] NSSI M =fPV ·NSSI PV +f NPV ·NSSI NPV +f BS ·NSSI BS

[0040] Among them, f PV 、f NPV and f BS are the proportions of PV, NPV, and BS in the pixel respectively, NDVI PV is the NDVI value of the PV spectral endmember, NDVI NPV is the NDVI value of the NPV spectral endmember, NDVI BS is the NDVI value of the BS spectral endmember, NSSI PV is the NSSI value of the PV spectral endmember, NSSI NPV is the NSSI value of the NPV spectral endmember, NSSI BS is the NSSI value of the BS spectral endmember, NDVI M and NSSI M are the NDVI and NSSI pixel values of any pixel respectively;

[0041] Among them, based on the obtained f NPV 、f PV and f BS , dimension stacking is performed and the PV-NPV-BS coverage result is output by the pixel trisection model.

[0042] A spectral endmember automatic extraction device, comprising:

[0043] An image acquisition module for acquiring satellite remote sensing images;

[0044] An image calculation module for calculating the vegetation index images of the satellite remote sensing images;

[0045] A range determination module for taking the corresponding pixel values in each vegetation index image as the coordinate points of a two-dimensional coordinate system, constructing a triangular feature space scatter point pixel by pixel, and determining the effective scatter point range;

[0046] An endmember extraction module for automatically searching for spectral endmembers to complete the extraction according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space.

[0047] The above spectral endmember automatic extraction device acquires satellite remote sensing images and calculates the vegetation index images of the satellite remote sensing images; takes the corresponding pixel values in each vegetation index image as the coordinate points of a two-dimensional coordinate system, constructs triangular feature space scatter points pixel by pixel, and determines the effective scatter point range; finally, automatically searches for spectral endmembers to complete the extraction according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space. Based on this, it solves the problems of time-consuming, laborious and low automation degree such as manual intervention and expert prior knowledge in the selection of spectral endmembers, and improves the efficiency and accuracy of spectral endmember extraction.

[0048] A computer storage medium stores computer instructions thereon. When the computer instructions are executed by a processor, the spectral endmember automatic extraction method of any of the above embodiments is implemented.

[0049] The above computer storage medium acquires satellite remote sensing images and calculates the vegetation index images of the satellite remote sensing images; takes the corresponding pixel values in each vegetation index image as the coordinate points of a two-dimensional coordinate system, constructs triangular feature space scatter points pixel by pixel, and determines the effective scatter point range; finally, automatically searches for spectral endmembers to complete the extraction according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space. Based on this, it solves the problems of time-consuming, laborious and low automation degree such as manual intervention and expert prior knowledge in the selection of spectral endmembers, and improves the efficiency and accuracy of spectral endmember extraction.

[0050] A computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the spectral endmember automatic extraction method of any of the above embodiments is implemented.

[0051] The above computer device acquires satellite remote sensing images and calculates the vegetation index images of the satellite remote sensing images; takes the corresponding pixel values in each vegetation index image as the coordinate points of a two-dimensional coordinate system, constructs triangular feature space scatter points pixel by pixel, and determines the effective scatter point range; finally, automatically searches for spectral endmembers to complete the extraction according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space. Based on this, it solves the problems of time-consuming, laborious and low automation degree such as manual intervention and expert prior knowledge in the selection of spectral endmembers, and improves the efficiency and accuracy of spectral endmember extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flowchart of the spectral endmember automatic extraction method in an embodiment;

[0053] Figure 2 It is a flowchart of the spectral endmember automatic extraction method in another embodiment;

[0054] Figure 3 It is a schematic diagram of the maximum and minimum parameters;

[0055] Figure 4 Flow chart of the method for automatically extracting spectral endmembers in yet another embodiment;

[0056] Figure 5 Structural diagram of the module of the device for automatically extracting spectral endmembers in one embodiment;

[0057] Figure 6 Schematic diagram of the internal structure of a computer in one embodiment. Detailed implementation manners

[0058] In order to better understand the purpose, technical solution and technical effect of the present invention, the present invention will be further described and explained below in conjunction with the accompanying drawings and embodiments. At the same time, it is declared that the embodiments described below are only used to explain the present invention and are not used to limit the present invention.

[0059] The embodiment of the present invention provides a method for automatically extracting spectral endmembers.

[0060] Figure 1 Flow chart of the method for automatically extracting spectral endmembers in one embodiment, as Figure 1 shown, the method for automatically extracting spectral endmembers in one embodiment includes steps S100 to S103:

[0061] S100, obtaining satellite remote sensing images;

[0062] S101, calculating vegetation index images of the satellite remote sensing images;

[0063] S102, taking the corresponding pixel values in each vegetation index image as the coordinate points of a two-dimensional coordinate system, constructing a triangular feature space scatter point pixel by pixel, and determining the effective scatter point range;

[0064] S103, automatically searching for spectral endmembers according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space to complete the extraction.

[0065] Among them, the satellite remote sensing images include satellite images of various resolutions. Such images are obtained as the extraction basis.

[0066] In one embodiment, Figure 2 Flow chart of the method for automatically extracting spectral endmembers in another embodiment, as Figure 2 shown, the process of obtaining satellite remote sensing images in step S100 includes the steps:

[0067] S200, obtaining satellite remote sensing images and removing the water body part in the satellite remote sensing images.

[0068] Exclude the water body part and retain the non-water body part to reduce the subsequent data processing volume and improve the extraction accuracy. Among them, the non-water body part in the image can be extracted by calculating the effective data range of the Normalized Difference Water Index (NDWI), and the reflectance remote sensing image after excluding the water body is obtained as the satellite remote sensing image after the exclusion process.

[0069] Further, calculate the vegetation index images of the satellite remote sensing image to calculate the spectral endmembers. Among them, the spectral endmembers include Photosynthetic Vegetation (PV), BareSoil (BS), and / or Non-Photosynthetic Vegetation (NPV). For the convenience of writing, hereinafter referred to as PV spectral endmember, NPV spectral endmember, and BS spectral endmember.

[0070] In one embodiment, calculate two or more types of vegetation index images to support the determination of two-dimensional coordinates in the subsequent two-dimensional coordinate system. As a preferred implementation manner, the vegetation index images include the Normalized Difference Vegetation Index (NDVI) vegetation index image and the NPV-Soil Separation Index (NSSI) remote sensing index image.

[0071] Based on the pixel values of multiple types of vegetation index images, construct a two-dimensional coordinate system to provide a basis for the determination of the triangular feature space.

[0072] In one embodiment, as Figure 2 shown, the process of taking the corresponding pixel values in each vegetation index image as the coordinate points of the two-dimensional coordinate system, constructing the scatter points of the triangular feature space pixel by pixel, and determining the effective scatter point range in step S102 includes steps S201 to S204:

[0073] S201, take the corresponding pixel value of the NDVI vegetation index image as the x-axis coordinate point of the two-dimensional coordinate system, and take the corresponding pixel value of the NSSI vegetation index image as the y-axis coordinate point of the two-dimensional coordinate system.

[0074] Taking the NDVI vegetation index image and the NSSI vegetation index image as examples, take the corresponding pixel values of the two types of vegetation index images as the coordinate points of the two-dimensional coordinate system to complete the deployment of the scatter points.

[0075] S202. According to the scatter distribution law of the triangular feature space, by setting a threshold as the pixel frequency for removing pixel noise points, and based on the set statistical results of the pixel frequency, the noise points are removed from the x-axis data and y-axis data respectively, obtaining the frequency ranges of the pixel values of the x-axis data and y-axis data after removal;

[0076] Among them, according to the scatter distribution law of the triangular feature space, by setting a threshold "threshold", which is used as the frequency for removing pixel noise points, based on the pixel frequency statistical results, the noise points are removed from the x (NDVI) -axis and y (NSSI) -axis data respectively, and the frequency ranges of the pixel values of the x (NDVI) -axis and y (NSSI) -axis data after removal are obtained as f threshold —f 1-threshold .

[0077] S203. According to the set statistical results and frequency ranges, determine the minimum and maximum values of the NDVI vegetation index and the NSSI vegetation index respectively as the extreme value parameters;

[0078] Figure 3 For the schematic diagram of the extreme value parameters, as Figure 3 shown, according to the above pixel frequency statistical results and the frequency range f threshold —f 1-threshold of the valid pixel values, the minimum value MIN threshold and the maximum value MAX 1-threshold of each of the two vegetation indices are determined respectively, and a total of 4 extreme value parameters are obtained;

[0079] S204. Construct the boundary range in the triangular feature space according to the extreme value parameters as the valid scatter range.

[0080] In one embodiment, as Figure 3 shown, based on the positions of the 4 determined extreme value parameters, they are used as 4 corner points in the triangular feature space and a rectangular boundary range of the valid scatter pixel values is constructed. It should be noted that the size of the extreme value parameters can be adjusted by setting the threshold "threshold", and the shape and size of the boundary range can be changed to adapt to different satellite remote sensing images.

[0081] Among them, the pixel scatter points where different spectral endmembers are located have the largest interval in the triangular feature space and are distributed in different data boundary ranges. According to the valid scatter range determined by the above steps and the distribution law of the spectral endmembers in the triangular space, the search range of the spectral endmembers can be continuously narrowed through an automatic search algorithm and the automatic search of the spectral endmembers can be completed, effectively overcoming the manual intervention in the process of spectral endmember extraction and realizing the automatic extraction of spectral endmembers.

[0082] In one embodiment, asFigure 2 As shown, in step S103, according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space, the process of automatically searching for spectral endmembers to complete the extraction includes steps S205 to S209:

[0083] S205, taking the y-axis direction in the rectangular boundary where the maximum value of the x-axis is located as the search direction for spectral endmembers, determining the search range on the x-axis by setting the buffer size with the maximum value of the x-axis as the axis, and determining the possible range of the x-axis of the spectral endmember;

[0084] Among them, taking the y(NSSI)-axis direction in the rectangular boundary where the maximum value of the x(NDVI)-axis MAX 1-threshold is located as the search direction for the PV spectral endmember, and determining the search range on the x-axis by setting the buffer size buffer_distance (as Figure 3 shown) with MAX 1-threshold as the axis, and determining the possible range of the x-axis of the PV spectral endmember PV x_range =[MAX 1-threshold -buffer_distance, MAX 1-threshold +buffer_distance].

[0085] S206, counting all different pixel values in the y-axis direction within the possible range of the x-axis and performing frequency calculation, and determining the search interval on the y-axis by setting a frequency threshold;

[0086] Count all different pixel values in the y-axis direction within the possible range PV x_range of the x-axis and perform frequency calculation, and determine the search interval on the y-axis by setting a frequency threshold.

[0087] S207, performing cyclic search and statistical calculation with a fixed step size on the search interval of the y-axis to reduce the selection of spectral endmember candidate points;

[0088] S208, in the cyclic search, counting the number of pixels in the rectangular frame with the maximum value of the x-axis and the center point determined by the candidate point through a preset search distance and saving the statistical results in the cycle;

[0089] In the cyclic search, counting the number of pixels in the rectangular frame with (MAX 1-threshold , y i ) as the center point through a preset search distance search_distance and saving the statistical results in the cycle, where y i represents the candidate point and is the pixel value of the candidate spectral endmember in the cyclic traversal.

[0090] S209. Select the center point position corresponding to the rectangle with the largest number of pixels in the pixel count results obtained in the loop as the spectral endmember of PV.

[0091] Among them, select the center point position corresponding to the rectangle with the largest number of pixels in the pixel count results obtained in the loop as the spectral endmember of PV (MAX 1-threshold , y PV ).

[0092] In one embodiment, Figure 4 is a flowchart of the spectral endmember automatic extraction method for another embodiment. As shown in Figure 4 , in step S103, according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space, automatically search for spectral endmembers to complete the extraction process, including steps S300 to S304:

[0093] S300. Use the rectangular boundary where the maximum value of the y-axis is located as the search range of the spectral endmember. Determine the search range on the y-axis by setting the buffer size with the maximum value of the y-axis as the axis, and determine the possible range of the y-axis of the spectral endmember;

[0094] Among them, use the rectangular boundary where the maximum value MAX 1-threshold of the y-axis is located as the search range of the NPV spectral endmember. Determine the search range on the y-axis by setting the buffer size buffer_distance with MAX 1-threshold as the axis, and determine the possible range of the y-axis of the NPV spectral endmember:

[0095] NPV y_range = [MAX 1-threshold - buffer_distance, MAX 1-threshold + buffer_distance];

[0096] S301. Count all different pixel values in the y-axis direction within the possible range of the y-axis and perform frequency calculation. Determine the search interval on the x-axis by setting a frequency threshold;

[0097] Among them, count all different pixel values in the x-axis direction within NPV y_range and perform frequency calculation. Determine the search interval on the x-axis by setting the frequency threshold.

[0098] S302. Perform cyclic search and statistical calculation with a fixed step size on the search interval of the x-axis to reduce the selection of spectral endmember candidate points;

[0099] S303. In the cyclic search, perform pixel count in the rectangle with the center point determined by the maximum value of the y-axis and the candidate point through the preset search distance, and save the statistical results in the loop;

[0100] Among them, in the loop search, the number of pixels in the rectangular box is counted with (x i , MAX 1-threshold ) as the center point according to the preset search distance search_distance, and the statistical results in the loop are saved, where x i represents the alternative point, which is the pixel value of the alternative spectral endmember in the loop traversal;

[0101] S304. Select the center point position corresponding to the rectangular box with the largest number of pixels in the statistical results of the number of pixels obtained in the loop as the spectral endmember of NPV.

[0102] Among them, select the center point position corresponding to the rectangular box with the largest number of pixels in the statistical results of the number of pixels obtained in the loop and use it as the spectral endmember of NPV (x NPV , MAX 1-threshold ).

[0103] In one embodiment, as Figure 4 shown, in step S103, according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space, the process of automatically searching for spectral endmembers to complete the extraction further includes step S305:

[0104] S305. According to the distribution law of the triangular feature space, use the minimum value as the BS spectral endmember value.

[0105] Among them, according to the distribution law of the triangular feature space of BS-PV-NPV, the MIN threshold value of two vegetation indices (x-axis and y-axis) is selected as the spectral endmember value during the search process of the BS spectral endmember position.

[0106] In one embodiment, as Figure 2 shown, the automatic extraction method of spectral endmembers in another implementation manner further includes step S210:

[0107] S210. Use the spectral endmember and the vegetation index image as the input parameters of the pixel three-component model to perform integrated remote sensing inversion of PV-NPV-BS coverage.

[0108] Based on the pixel coverage types in the soil-vegetation ecosystem and the triangular feature space, two preset premises for constructing the pixel three-component model are: 1) The pixel values of mixed pixels are only contributed by PV, NPV, and BS in different proportions, and 2) The spectral endmembers of PV, NPV, and BS exist at the three vertices in the triangular space; Use the obtained spectral endmember parameters and two vegetation index images as the input parameters of the pixel three-component model to perform integrated remote sensing inversion of PV-NPV-BS coverage.

[0109] In one of the embodiments, according to two premise settings in the pixel three-component model, f is calculated according to the following algorithm PV , f NPV and f BS ;

[0110] f PV + f NPV + f BS = 1

[0111] NDVI M = f PV ·NDVI PV + f NPV ·NDVI NPV + f BS ·NDVI BS

[0112] NSSI M = f PV ·NSSI PV + f NPV ·NSSI NPV + f BS ·NSSI BS

[0113] where f PV , f NPV and f BS are the proportions of PV, NPV, and BS in the pixel respectively, NDVI PV is the NDVI value of the PV spectral endmember, NDVI NPV is the NDVI value of the NPV spectral endmember, NDVI BS is the NDVI value of the BS spectral endmember, NSSI PV is the NSSI value of the PV spectral endmember, NSSI NPV is the NSSI value of the NPV spectral endmember, NSSI BS is the NSSI value of the BS spectral endmember, NDVI M and NSSI M are the NDVI and NSSI pixel values of any pixel respectively;

[0114] According to the above algorithm, the results of f NPV , f PV and f BS are obtained, and their dimensions are stacked and the PV-NPV-BS coverage result is output.

[0115] The spectral endmember automatic extraction method of any of the above embodiments obtains a satellite remote sensing image and calculates the vegetation index images of the satellite remote sensing image; takes the corresponding pixel values in each vegetation index image as the coordinate points of a two-dimensional coordinate system, constructs a triangular feature space scatter plot pixel by pixel, and determines the effective scatter point range; finally, automatically searches for spectral endmembers to complete the extraction according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space. Based on this, it solves the problems of time-consuming, laborious and low automation degree such as manual intervention and expert prior knowledge in the selection of spectral endmembers, and improves the efficiency and accuracy of spectral endmember extraction.

[0116] An embodiment of the present invention also provides a spectral endmember automatic extraction device.

[0117] Figure 5 It is a module structure diagram of a spectral endmember automatic extraction device in an embodiment, as Figure 5 shown. A spectral endmember automatic extraction device in an embodiment includes:

[0118] An image acquisition module 100, configured to acquire a satellite remote sensing image;

[0119] An image calculation module 101, configured to calculate the vegetation index images of the satellite remote sensing image;

[0120] A range determination module 102, configured to take the corresponding pixel values in each vegetation index image as the coordinate points of a two-dimensional coordinate system, construct a triangular feature space scatter plot pixel by pixel, and determine the effective scatter point range;

[0121] An endmember extraction module 103, configured to automatically search for spectral endmembers to complete the extraction according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space.

[0122] The above spectral endmember automatic extraction device obtains a satellite remote sensing image and calculates the vegetation index images of the satellite remote sensing image; takes the corresponding pixel values in each vegetation index image as the coordinate points of a two-dimensional coordinate system, constructs a triangular feature space scatter plot pixel by pixel, and determines the effective scatter point range; finally, automatically searches for spectral endmembers to complete the extraction according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space. Based on this, it solves the problems of time-consuming, laborious and low automation degree such as manual intervention and expert prior knowledge in the selection of spectral endmembers, and improves the efficiency and accuracy of spectral endmember extraction.

[0123] An embodiment of the present invention also provides a computer storage medium, on which computer instructions are stored, and when the instructions are executed by a processor, the spectral endmember automatic extraction method of any of the above embodiments is implemented.

[0124] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0125] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of the present invention, in essence, or the part that contributes to the related technology can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a terminal, or a network device, etc.) to execute all or part of the methods of the embodiments of the present invention. The aforementioned storage medium includes: various media such as a removable storage device, RAM, ROM, a magnetic disk, or an optical disc that can store program codes.

[0126] Corresponding to the above computer storage medium, in one embodiment, a computer device is also provided. The computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements any one of the spectral endmember automatic extraction methods in the above embodiments.

[0127] The computer device can be a terminal, and its internal structure diagram can be as Figure 6As shown in the figure. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a method for automatically extracting spectral endmembers. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, trackball, or touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0128] For the above computer device, obtain satellite remote sensing images and calculate the vegetation index images of the satellite remote sensing images; use the corresponding pixel values in each vegetation index image as the coordinate points of a two-dimensional coordinate system, and construct a triangular feature space scatter point pixel by pixel, and determine the effective scatter point range; finally, according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space, automatically search for spectral endmembers to complete the extraction. Based on this, it solves the problems of time-consuming, laborious, and low automation degree such as the need for manual intervention and expert prior knowledge in the selection of spectral endmembers, and improves the efficiency and accuracy of spectral endmember extraction.

[0129] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0130] The above embodiments only represent several implementation manners of the present invention. The description is relatively specific and detailed, but it cannot be understood as a limitation to the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several deformations and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the invention patent should be subject to the appended claims.

Claims

1. A method for automatically extracting spectral endmembers, characterized in that, Including the steps: Obtain satellite remote sensing images; Calculate each vegetation index image of the satellite remote sensing image; the vegetation index images include NDVI vegetation index images and NSSI vegetation index images; Take the corresponding pixel values in each of the vegetation index images as the coordinate points of a two-dimensional coordinate system, and construct a triangular feature space scatter point pixel by pixel, and determine the effective scatter point range; According to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space, automatically search for spectral endmembers to complete the extraction; The process of taking the corresponding pixel values in each of the vegetation index images as the coordinate points of a two-dimensional coordinate system, constructing a triangular feature space scatter point pixel by pixel, and determining the effective scatter point range includes the steps: Take the corresponding pixel value of the NDVI vegetation index image as the x-axis coordinate point of the two-dimensional coordinate system, and take the corresponding pixel value of the NSSI vegetation index image as the y-axis coordinate point of the two-dimensional coordinate system; According to the distribution law of the triangular feature space scatter points, set a threshold as the pixel frequency for removing pixel noise points, and respectively remove noise points from the x-axis data and y-axis data according to the set statistical results of the pixel frequency, and obtain the frequency ranges of the pixel values of the x-axis data and the y-axis data after removal; According to the set statistical results and the frequency ranges, respectively determine the minimum and maximum values of the NDVI vegetation index and the NSSI vegetation index as the extreme value parameters; Construct the boundary range in the triangular feature space according to the extreme value parameters as the effective scatter point range; The process of automatically searching for spectral endmembers to complete the extraction according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space includes the steps: Take the y-axis direction in the rectangular boundary where the maximum value of the x-axis is located as the search direction of the spectral endmember, determine the search range on the x-axis with the maximum value of the x-axis as the axis by setting the buffer size, and determine the possible range of the x-axis of the spectral endmember; Statistically calculate the different pixel values in all y-axis directions within the possible range of the x-axis and calculate the frequency, and determine the search interval on the y-axis by setting the frequency threshold; Perform cyclic search and statistical calculation with a fixed step size on the search interval of the y-axis to reduce the selection of spectral endmember candidate points; In the cyclic search, statistically calculate the number of pixels in the rectangular frame with the maximum value of the x-axis and the center point determined by the candidate point as the center, and save the statistical results in the cycle; Select the center point position corresponding to the rectangular frame with the largest number of pixels in the statistical results of the number of pixels obtained in the cycle as the spectral endmember of PV; Take the rectangular boundary where the maximum value of the y-axis is located as the search range of the spectral endmember, determine the search range on the y-axis with the maximum value of the y-axis as the axis by setting the buffer size, and determine the possible range of the y-axis of the spectral endmember; Statistically calculate the different pixel values in all y-axis directions within the possible range of the y-axis and calculate the frequency, and determine the search interval on the x-axis by setting the frequency threshold; Perform cyclic search and statistical calculation with a fixed step size on the search interval of the x-axis to reduce the selection of spectral endmember candidate points; In the loop search, the number of pixels in the rectangular box is counted with the center point determined by the maximum value of the y-axis and the alternative point at a preset search distance, and the statistical results in the loop are saved; Among the statistical results of the number of pixels obtained in the loop, the center point position corresponding to the rectangular box with the largest number of pixels is selected as the spectral endmember of the NPV; According to the distribution law of the triangular feature space, the minimum value is used as the BS spectral endmember value.

2. The method for automatically extracting spectral endmembers according to claim 1, wherein The process of obtaining the satellite remote sensing image includes the steps of: Obtain the satellite remote sensing image and remove the water body part in the satellite remote sensing image.

3. The spectral endmember automatic extraction method according to claim 1, wherein It also includes the steps of: Using the spectral endmember and the vegetation index image as input parameters of the pixel three-component model for integrated remote sensing inversion of PV-NPV-BS coverage.

4. The spectral endmember automatic extraction method according to claim 3, wherein The process of using the spectral endmember and the vegetation index image as input parameters of the pixel three-component model for integrated remote sensing inversion of PV-NPV-BS coverage is as follows: f PV +f NPV +f BS =1 NDVI M = f PV · NDVI PV + f NPV · NDVI NPV + f BS · NDVI BS NSSI M = f PV ·NSSI PV + f NPV ·NSSI NPV + f BS ·NSSI BS Among them, f PV , f NPV and f BS are the proportions of PV, NPV, and BS in the pixel, respectively. NDVI PV is the NDVI value of the PV spectral endmember, NDVI NPV is the NDVI value of the NPV spectral endmember, NDVI BS is the NDVI value of the BS spectral endmember, NSSI PV is the NSSI value of the PV spectral endmember, NSSI NPV is the NSSI value of the NPV spectral endmember, NSSI BS is the NSSI value of the BS spectral endmember, NDVI M and NSSI M are the NDVI and NSSI pixel values of any pixel, respectively; Among them, based on the obtained f above NPV , f PV and f BS , dimension stacking is performed and the PV-NPV-BS coverage result is output by the pixel trisection model.

5. A spectral endmember automatic extraction device, characterized in that, It includes: An image acquisition module for acquiring satellite remote sensing images; An image calculation module for calculating the vegetation index images of the satellite remote sensing image; the vegetation index images include NDVI vegetation index images and NSSI vegetation index images; A range determination module for taking the corresponding pixel values in each of the vegetation index images as the coordinate points of the two-dimensional coordinate system, constructing triangular feature space scatter points pixel by pixel, and determining the effective scatter point range; An endmember extraction module for automatically searching for spectral endmembers to complete the extraction according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space; The process of taking the corresponding pixel values in each of the vegetation index images as the coordinate points of the two-dimensional coordinate system, constructing triangular feature space scatter points pixel by pixel, and determining the effective scatter point range includes the steps of: Taking the corresponding pixel value of the NDVI vegetation index image as the x-axis coordinate point of the two-dimensional coordinate system, and taking the corresponding pixel value of the NSSI vegetation index image as the y-axis coordinate point of the two-dimensional coordinate system; According to the distribution law of the triangular feature space scatter points, by setting a threshold as the pixel frequency for removing pixel noise points, and respectively removing noise points from the x-axis data and the y-axis data according to the set statistical results of the pixel frequency, obtaining the frequency range of the pixel values of the x-axis data after removal and the frequency range of the pixel values of the y-axis data after removal; According to the set statistical results and the frequency range, respectively determine the minimum and maximum values of the NDVI vegetation index and the NSSI vegetation index as the extreme value parameters; Construct the boundary range in the triangular feature space according to the extreme value parameters as the effective scatter point range; The process of automatically searching for spectral endmembers to complete the extraction according to the effective scatter point range and the distribution law of spectral endmembers in the triangular feature space includes the steps of: Taking the y-axis direction in the rectangular boundary where the maximum value of the x-axis is located as the search direction of the spectral endmember, determining the search range on the x-axis with the maximum value of the x-axis as the axis by setting the buffer size, and determining the possible range of the x-axis of the spectral endmember; Statistically calculate the different pixel values in all y-axis directions within the possible range of the x-axis and perform frequency calculations. Determine the search interval on the y-axis by setting a frequency threshold; Perform cyclic searches and statistical calculations with a fixed step size on the search interval of the y-axis to reduce the selection of spectral endmember candidate points; During the cyclic search, statistically calculate the number of pixels in the rectangular box with the maximum value of the x-axis and the center point determined by the candidate point according to the preset search distance, and save the statistical results during the cycle; Select the center point position corresponding to the rectangular box with the largest number of pixels in the statistical results of the number of pixels obtained during the cycle as the spectral endmember of PV; Use the rectangular boundary where the maximum value of the y-axis is located as the search range of the spectral endmember. Determine the search range on the y-axis with the maximum value of the y-axis as the axis by setting the buffer size, and determine the possible range of the y-axis of the spectral endmember; Statistically calculate the different pixel values in all y-axis directions within the possible range of the y-axis and perform frequency calculations. Determine the search interval on the x-axis by setting a frequency threshold; Perform cyclic searches and statistical calculations with a fixed step size on the search interval of the x-axis to reduce the selection of spectral endmember candidate points; During the cyclic search, statistically calculate the number of pixels in the rectangular box with the maximum value of the y-axis and the center point determined by the candidate point according to the preset search distance, and save the statistical results during the cycle; Select the center point position corresponding to the rectangular box with the largest number of pixels in the statistical results of the number of pixels obtained during the cycle as the spectral endmember of NPV; According to the distribution law of the triangular feature space, use the minimum value as the spectral endmember value of BS.

Images