Method, system, device, and medium for detecting forest disturbances based on fully polarized SAR images

Abstract

The present invention discloses a method for detecting forest disturbances based on a full-polarization SAR image, which includes obtaining a full-polarization SAR image within a target area, performing image classification for the reference time phase and the monitoring time phase, performing pre-processing including introduction of the SAR image, polarization matrix transformation of the SAR image, Refined Lee filtering, and false-color synthesis on the classified full-polarization SAR image to obtain a false-color synthesized SAR image, inputting the false-color synthesized SAR image into an optimized surface change detection model to extract surface change patches, combining with a radar vegetation index RVI image to obtain initial forest disturbance patches, performing an overlay check on the initial forest disturbance patches to obtain forest disturbance patches within the target area, and generating a visualization result. It has efficient, accurate, and reliable detection capabilities, can observe the ground all day long, all-weather, and penetrate clouds and fog, can be widely applied in fields such as forest resource monitoring and environmental protection, and can better serve ecological protection restoration supervision and management.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing technology, and more specifically, to a method, system, device and medium for detecting forest land disturbances based on full polarization SAR images.

Background Art

[0002] In recent years, our country has highly emphasized ecological environment protection and ecological civilization construction, implemented a series of effective ecological environment protection measures, and significantly improved the quality of the ecological environment.

[0003] Forests have functions such as soil and water conservation, water source nourishment, wind prevention and sand fixation, and biodiversity protection, and are an important component of the global ecosystem and ecological balance. Especially for areas with important ecological functions and fragile ecological environments, overdevelopment of forests is an important factor leading to a decline in the regional ecological service function. In recent years, illegal destruction incidents of forest resources have frequently occurred despite being prohibited, seriously threatening national ecological security. Therefore, strengthening the monitoring of forest land development is urgent and extremely important for enhancing the diversity, stability and sustainability of the ecosystem.

[0004] Remote sensing technology is widely applied in the field of monitoring surface cover changes. Currently, surface change detection technology based on high-resolution optical remote sensing images has become the main means for operations such as forest land changes. However, optical images are greatly affected by weather, and there are often problems such as high forest cover rates, difficulty in obtaining optical data in cloudy and rainy areas, or poor quality of the obtained images, which to a certain extent affect the accuracy and timeliness of the remote sensing monitoring data results.

[0005] Compared with optical satellites, the number of SAR satellites is relatively small, and the data sources somewhat limit their extensive applications. At the same time, due to the influence of elements such as speckle noise, geometric distortion, layover, and shadow in SAR images, it is relatively difficult to monitor human activities, and there are relatively few related applied research. In recent years, SAR satellite technology has developed rapidly. Multiple satellites with different high-resolution bands have already been launched, and the polarization mode has also evolved from single polarization to dual polarization, multi-polarization, and full polarization, further expanding the application fields and directions of SAR satellite remote sensing.

[0006] Therefore, how to design a method for detecting forest land disturbances based on full-polarization SAR images to compensate for the lack of coverage of high-resolution optical images in cloudy and rainy areas and optimize the monitoring of forest land development activities is an urgent problem that those skilled in the art should solve.

Summary of the Invention

Problems to be Solved by the Invention

[0007] In view of this, the present invention provides a method for detecting forest land disturbances based on full-polarization SAR images, applies it to the monitoring of forest land disturbances in important ecological spaces, discovers forest land disturbance activities in a timely manner, serves ecological protection and restoration supervision and management, and can further improve the diversity, stability, and sustainability of the ecosystem.

Means for Solving the Problems

[0008] To achieve the above object, the present invention adopts the following technical solutions:

[0009] According to a first aspect, the present invention provides a method for detecting forest land disturbances based on full-polarization SAR images, and the method includes: S1, obtaining a full-polarization SAR image within a target area, and performing image classification on the full-polarization SAR image for the reference time phase and the monitoring time phase; S2, performing preprocessing including polarization matrix transformation, Refined Lee filtering, and false color composition on the classified full-polarization SAR image to obtain a false color composite SAR image; S3. Input the false - color composite SAR image into an optimized surface change detection model, extract surface change patches, combine them with the radar vegetation index (RVI) image to obtain initial forest disturbance patches; S4. Perform an overlay check on the initial forest disturbance patches to obtain forest disturbance patches within the target area and generate a visualization result. The method includes the above steps.

[0010] Here, as for S2, performing pre - processing on the segmented full - polarization SAR image to obtain a false - color composite SAR image specifically includes the following:

[0011] S21. Introduction of the SAR image: Convert the full - polarization SAR images of the reference time - phase and the monitoring time - phase into a standard data format and perform radiometric correction processing to obtain full - polarization single - look complex data X: X = {X i : i = 1, 2, …, M} Here, X i = {xpq i = Ipq i + jQpq i : p, q ∈ {H, V}} M is the total number of pixels, p is the antenna transmission mode, q is the antenna reception mode, i is the pixel index, j is the imaginary unit, xpq i is the complex scattering coefficient of the i - th pixel in the pq polarization mode, Ipq i is the real part of xpq i Qpq i is the imaginary part of xpq i H represents horizontal polarization and V represents vertical polarization.

[0012] S22. Polarization matrix transformation of the SAR image: Based on the full - polarization single - look complex data X, construct a polarization scattering matrix S, S = {S i : i = 1, 2, …, M} Define a target scattering vector set K using the polarization scattering matrix S, K = {K i : i = 1, 2, …, M} Convert the target scattering vector set K into a polarization coherent matrix T, T = {T i : i = 1, 2, …, M} where * is the conjugate operation and < > indicates multi-view processing.

[0013] S23, Refined Lee filtering: Performing Refined Lee filtering on the polarization coherent matrix T to obtain the filtered polarization coherent matrix T1 includes the following: where N is the number of views and m indicates the pixel index of the image after multi-view processing. If the resolution in the azimuth direction is changed to 1 / N of the original and the resolution in the range direction is not changed, can be obtained. where indicates the average value of the same boundary alignment window within the local window where the m-th pixel is located, and w is the filtering weight value.

[0014] S24, false color synthesis: Perform full-polarization false color synthesis on the filtered polarization conversion matrix T1 using the Pauli synthesis method to obtain the false color synthesis SAR image S1 and the false color synthesis SAR image S2.

[0015] Furthermore, in the above S3, the optimization of the surface change detection model is S31, creating a tag image dataset based on the surface disturbance interpretation flag; S32, training a surface change detection model using the tag image dataset; S33, classifying the false color synthesis SAR image using the trained surface change detection model to obtain a classification result. Step S34: Perform vector transformation on the classification result, calculate the cross-entropy loss value using the cross-entropy loss function, and perform model evaluation; Step S35: If the cross-entropy loss value is smaller than the cross-entropy loss threshold, output the classification result; otherwise, return to S31 and supplement the tag image dataset using the classification result.

[0016] Furthermore, in step S31, the construction of the surface disturbance interpretation flag includes: Step S311: Obtain two high-resolution optical remote sensing images in the previous and subsequent periods that match the full-polarization SAR image phase; Step S312: Input the two high-resolution optical remote sensing images in the previous and subsequent periods into the pre-trained neural network model to extract newly added surface disturbance patches; Step S313: Combine with the false-color composite SAR image to construct the surface disturbance interpretation flag.

[0017] Furthermore, in step S3, the Freeman method is adopted to calculate the radar vegetation index RVI, RVI = Fv / (Fv + Fd + Fs) where Fv, Fd, and Fs respectively represent the odd scattering component, even scattering component, and volume scattering component obtained by decomposing the scattering matrix after the polarization matrix transformation of the SAR image.

[0018] Furthermore, in step S4, performing the overlay check on the initial forest land disturbance patch includes: Step S411: Perform an overlay check on the initial forest land disturbance patch and the forest land data to obtain the forest land disturbance patch; Step S412: Perform a spatial overlay check on the forest land disturbance patch and the gradient data to obtain the forest land disturbance patch within the preset gradient range.

[0019] Furthermore, in step S4, generating the visualization result includes: S421. A step of deriving forest land disturbance patches within the target area from the vector data layer; S422. A step of combining with existing attribute information and performing marking on the vector data layer; S423. A step of generating a plurality of types of visualization results including a forest land property classification map, a forest land disturbance heat map, and a forest land disturbance trend graph based on the marked vector data layer.

[0020] According to a second aspect, the present invention provides a forest land disturbance detection system based on a full-polarization SAR image, and the system includes: A data acquisition module for acquiring a full-polarization SAR image within a target area and performing image classification of a reference time phase and a monitoring time phase on the full-polarization SAR image; A data preprocessing module for performing preprocessing including introduction of the SAR image, polarization matrix conversion of the SAR image, Refined Lee filtering, and false color synthesis on the classified full-polarization SAR image to obtain a false color synthesized SAR image; A feature extraction module for inputting the false color synthesized SAR image into an optimized surface change detection model, extracting surface change patches, and combining them with a radar vegetation index RVI image to obtain initial forest land disturbance patches; An overlay check module for performing an overlay check on the initial forest land disturbance patches to obtain forest land disturbance patches within the target area and generate visualization results.

[0021] According to a third aspect, the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the above forest land disturbance detection method is realized.

[0022] According to a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program that realizes the above forest land disturbance detection method when executed by a processor.

[0023] For the description of the second to fourth aspects of the present invention, reference may be made to the detailed description of the first aspect. Also, for the beneficial effects of the description of the second to fourth aspects, they can be analyzed by referring to the beneficial effects of the first aspect, so they are omitted here.

Advantages of the Invention

[0024] As can be seen from the above technical solutions, the method for detecting forest disturbances based on fully polarized SAR images according to the present invention has the following beneficial effects compared with the prior art:

[0025] 1. This method uses fully polarized SAR images, combines image classification of the reference time phase and the monitoring time phase, and improves the accuracy of detecting and analyzing forest disturbances in the target area. The fully polarized SAR image has weather independence and remote sensing penetrability, and can obtain high-quality image data under various weather conditions.

[0026] 2. By performing preprocessing on the fully polarized SAR image, including introducing the SAR image, performing polarization matrix transformation of the SAR image, Refined Lee filtering, and false color composition, a high-quality false color composite SAR image is generated. It can provide more intuitive and clear ground surface information, which is beneficial for subsequent detection of surface changes and patch extraction.

[0027] 3. Based on the optimized surface change detection model, surface change patches can be extracted more accurately. Also, by combining with the radar vegetation index RVI image, the disturbance situation of the forest can be better distinguished, and the accuracy and reliability of the detection results can be improved.

[0028] 4. Finally, by performing overlay checks on the initial forest disturbance patches, the detection accuracy of forest disturbances in the target area can be further improved. The false detection rate and detection omission rate are reduced, and a more reliable and accurate visualization result is generated, providing important reference information for the protection and management of forests.

Brief Description of the Drawings

[0029] To more clearly explain the technical solutions in the embodiments of the present invention or the prior art, the following briefly introduces the attached drawings that need to be used in the description of the embodiments or the prior art. Obviously, the attached drawings in the following description are only the embodiments of the present invention. Based on the provided attached drawings, other attached drawings can also be obtained by those skilled in the art without creative efforts.

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Embodiments for Carrying Out the Invention

[0030] The following clearly and completely describes the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0031] The detection method according to the embodiment of the present application can be applied to a detection server. The above detection server may be hardware or software. When the detection server is hardware, it may be realized as a distributed server cluster providing detection services, or may be realized as a single server. When the detection server is software, it may be installed on the above-listed servers. This may be realized as a plurality of software or software modules, or may be realized as a single software or software module, and is not particularly limited herein.

[0032] Example 1: As shown in FIG. 1, this embodiment provides a method for detecting forest land disturbance based on a full polarization SAR image, with a region within a certain range of a certain city as a target region. The method includes the following:

[0033] S1. Obtain a full polarization SAR image within the target region, and perform image classification on the full polarization SAR image for the reference time phase and the monitoring time phase.

[0034] The domestic Gaofen-3 satellite has various polarization modes such as single polarization, dual polarization, and full polarization. It adopts Gaofen-3 full polarization data that covers a certain area of a city, with a spatial resolution of 8 meters. To improve the identification accuracy, an interpretation flag library is constructed, and Gaofen-1 optical data is used as auxiliary data, with a spatial resolution of 2 meters.

[0035] S2. For the segmented full polarization SAR image, perform preprocessing including polarization matrix transformation, Refined Lee filtering, and false color composition of the SAR image to obtain a false color composite SAR image.

[0036] S3. Input the false color composite SAR image into an optimized surface change detection model, extract surface change patches, and combine them with the radar vegetation index RVI image to obtain initial forest land disturbance patches.

[0037] S4. Perform an overlay check on the initial forest land disturbance patches to obtain forest land disturbance patches within the target area and generate a visualization result.

[0038] This method has efficient, accurate, and reliable detection capabilities, can observe the ground all day, all-weather, and penetrate clouds and fog, and can be widely applied in fields such as forest land resource monitoring and environmental protection. In particular, it is applicable to the remote sensing monitoring of forest land development activities in cloudy and rainy areas where the effective coverage of optical images is insufficient, can better serve the ecological protection and restoration supervision and management, and can enhance the biodiversity, stability, and sustainability of the ecosystem.

[0039] The following further elaborates on the above steps:

[0040] As shown in Figure 2, in S2 of this embodiment, performing preprocessing on the segmented full polarization SAR image to obtain a false color composite SAR image specifically includes the following:

[0041] S21. Introduction of SAR Images: Convert the full-polarization SAR images of the reference time phase and the monitoring time phase into the standard data format, and perform radiometric correction processing to obtain the full-polarization single-look composite data X: X = {X i : i = 1, 2, …, M} Here, X i = {xpq i = Ipq i + jQpq i : p, q ∈ {H, V}} M is the total number of pixels, p is the antenna transmission mode, q is the antenna reception mode, i is the pixel index, j is the imaginary unit, and xpq i is the complex scattering coefficient of the i-th pixel in the pq polarization mode, Ipq i is the real part of xpq i , Qpq i is the imaginary part of xpq i , H represents horizontal polarization, and V represents vertical polarization.

[0042] As shown in Figure 3, the single-look composite data that has not undergone subsequent processing has a lot of noise and low image readability, and cannot be directly used for human activity patch interpretation.

[0043] S22. Polarization Matrix Conversion of SAR Images: Construct the polarization scattering matrix S based on the full-polarization single-look composite data X: S = {S i : i = 1, 2, …, M} Define the target scattering vector set K using the polarization scattering matrix S: K = {K i : i = 1, 2, …, M} Convert the target scattering vector set K to the polarization coherent matrix T: T = {T i : i = 1, 2, …, M} Here, * represents the conjugate operation, and < > represents the multi-view processing.

[0044] As shown in FIG. 4, the converted polarization coherent matrix includes nine images.

[0045] S23, Refined Lee filtering: Performing Refined Lee filtering on the polarization coherent matrix T, adopting a 7*7 Refined Lee filtering with a window size, and obtaining a filtered polarization coherent matrix T1 includes the following: TIFF0007717301000011.tif20166 Here, N represents the number of views, m represents the pixel index of the image after multi-view processing, Changing the resolution in the azimuth direction to 1 / N of the original and keeping the resolution in the range direction unchanged can be achieved as follows: TIFF0007717301000012.tif10166 Here, represents the average value of the same boundary alignment window within the 7*7 local window where the Mth pixel is located, and w is the filtering weighting value.

[0046] As shown in FIG. 5, the image after filtering using a 7*7 Refined Lee with a window size has appropriate noise and maximally retains the detailed information of the image.

[0047] S24, false color synthesis: Performing full polarization false color synthesis on the filtered polarization conversion matrix T1 using the Pauli synthesis method to obtain a false color synthesis SAR image S1 and a false color synthesis SAR image S2.

[0048] As shown in FIG. 6, the SAR image after false color synthesis has rich color information, significantly improved display of different ground object images, and is beneficial for the interpretation of human activities.

[0049] As shown in FIG. 7, in this embodiment S3, the optimization of the surface change detection model includes the following: S31, creating a tagged image dataset based on the surface disturbance interpretation flag, S32, training a surface change detection model using the tagged image dataset, S33. Classify the false-color composite SAR image using the trained surface change detection model to obtain a classification result. S34. Perform vector conversion on the classification result, calculate the cross-entropy loss value using the cross-entropy loss function, and conduct model evaluation. S35. If the cross-entropy loss value is smaller than the cross-entropy loss threshold, output the classification result; otherwise, return to S31 and supplement the tag image dataset using the classification result.

[0050] As shown in Figure 8, the construction of the surface disturbance interpretation flag includes the following: S311. Obtain two high-resolution optical remote sensing images in the previous and later periods that match the full-polarization SAR image phase. S312. As shown in Figure 9, input the two high-resolution optical remote sensing images in the previous and later periods into the pre-trained neural network model to extract newly added surface disturbance patches. S313. As shown in Figure 10, combine with the false-color composite SAR image to construct the surface disturbance interpretation flag.

[0051] As shown in Figure 11, adopt the Freeman method to calculate the radar vegetation index RVI: RVI = Fv / (Fv + Fd + Fs) Here, Fv, Fd, and Fs respectively represent the odd scattering component, even scattering component, and volume scattering component obtained by decomposing the scattering matrix after the polarization matrix transformation of the SAR image.

[0052] As shown in Figure 12, combine with the radar vegetation index RVI to obtain the initial forest land disturbance patches.

[0053] As shown in Figure 13, in S4 of this embodiment, performing an overlay check on the initial forest land disturbance patches includes the following: S411. Perform an overlay check on the initial forest land disturbance patches and forest land data to obtain forest land disturbance patches. S412. Perform a spatial overlay check on the forest land disturbance patches and slope data to obtain forest land disturbance patches within a preset slope range.

[0054] In the following, the overlay check performed on the initial forest land disturbance patches will be described in detail:

[0055] Performing an overlay check on the initial forest land disturbance patches means using GIS (Geographic Information System) software to perform an overlay check on the initial forest land disturbance patches, especially a spatial overlay check between the forest land data and the slope data:

[0056] Data preparation: Ensure a vector data layer that includes forest land data and contains the boundaries and related information of the forest land. Slope data: A vector or raster data layer containing slope information is also required. This is slope data calculated based on a DEM (Digital Elevation Model). Initial forest land disturbance patches: This is the layer that needs to be checked and may be a vector layer containing the locations and extents of forest land disturbances.

[0057] Data import: Open ArcGIS software and create a new map project. Use the "Add Data" function to import the forest land data, slope data, and initial forest land disturbance patches into the map project.

[0058] Spatial overlay analysis: Overlay with forest land data: In the ArcGIS toolbox, find the "Overlay Analysis" tool. Select the "Intersect" or "Identity" tool according to the desired results. If you only want to find the disturbance patches that overlap with the forest land, you can use the "Intersect" tool. If you want to retain other information of the initial forest land disturbance patches, you can use the "Identity" tool. In the pop-up dialog box, set the input element (initial forest land disturbance patches) and the overlay element (forest land data).

[0059] Set the position and name of the output layer, and click "OK" to execute the analysis. As a result, a new layer containing forest land disturbance patches overlapping with forestry land will be obtained. Overlay with gradient data (assuming the gradient data is raster data): If the gradient data is raster data, in order to facilitate the overlay analysis, it is necessary to convert it into vector data using the "Raster to Vector" tool. After the conversion is completed, use the "Spatial Selection" tool or the "Location-based Selection" tool to select the vector gradient data within the preset gradient range. Next, use the same overlay analysis tool as the forestry land data (such as "Intersect" or "Identify") to overlay the selected vector gradient data and the forestry land disturbance patches. When the parameters are set and the analysis is executed, as a result, a new layer containing forestry land disturbance patches within the preset gradient range will be obtained.

[0060] Processing of results: Perform necessary processing and cleanup on the overlay analysis results, such as deleting duplicate items and merging adjacent patches. If necessary, the style and symbols of the layer can be modified to better display the results. The results can be exported to the required formats such as Shapefile, GeoJSON, KMZ, etc. using the export function of ArcGIS so that they can be shared with other software and platforms.

[0061] Through the above steps, an overlay check of forestry land data and gradient data is performed on the initial forest land disturbance patches based on the ArcGIS software, and forestry land disturbance patches within the required preset gradient range are generated.

[0062] As shown in Figure 14, generating visualization results includes the following: S421, Export the forest land disturbance patches within the target area from the vector data layer, S422, Combine with the existing attribute information and mark the vector data layer, Based on the marked vector data layer, generate multiple types of visualization results including a forest land property classification map, a forest land disturbance heat map, and a forest land disturbance trend graph.

[0063] Forest land property classification map: A visualization map that classifies forest land disturbance patches according to their properties and is represented by different colors or symbols. Through this map, the situations of different types of forest land disturbances within the target area, such as forest logging, fires, natural disasters, etc., can be clearly understood.

[0064] Forest land disturbance heat map: Can display the forest land disturbance heat areas within the target area. The heat areas are usually represented as high-density points, thermodynamic diagrams, or density diagrams and are used to indicate the concentration degree and distribution situation of forest land disturbances. It helps to determine the areas where focused actions are needed.

[0065] Forest land disturbance trend graph: A forest land disturbance trend graph can be generated based on time-series data. Such a visualization result can show the change trend of forest land disturbances over time and helps analysts understand the evolution of forest land disturbances and influencing factors.

[0066] The technical solution in this embodiment realizes accurate detection and visualization presentation of forest land disturbances by combining methods such as full-polarization SAR images, radar vegetation index RVI, optimization of surface change detection models, and construction of surface disturbance interpretation flags. It plays an important role in aspects such as forest resource management, environmental monitoring, and natural disaster assessment, and provides more effective data support and decision-making basis for related fields.

[0067] Example 2: As shown in FIG. 15, this embodiment provides a forest land disturbance detection system based on full-polarization SAR images, and the system includes A data acquisition module for acquiring full-polarization SAR images within the target area and performing image classification for the reference time phase and the monitoring time phase on the full-polarization SAR images, For all polarized SAR images, perform preprocessing including the introduction of SAR images, SAR image polarization matrix conversion, Refined Lee filtering, and false color synthesis to obtain a false color synthesized SAR image, and a data preprocessing module for Input the false color synthesized SAR image into an optimized surface change detection model, extract surface change patches, and combine them with a radar vegetation index RVI image to obtain an initial forest disturbance patch, and a feature extraction module for Perform an overlay check on the initial forest disturbance patch to obtain a forest disturbance patch within the target area and generate a visualization result, and an overlay check module for

[0068] Example 3: As shown in Figure 16, this example provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the disturbance detection method in the above example is realized.

[0069] Example 4: This example provides a computer-readable storage medium storing a computer program that realizes the forest disturbance detection method in the above example when executed by a processor.

[0070] In the technical solution of the above example, all polarized SAR images can collect more surface detail information according to different polarization states. Compared with conventional non-polarized SAR images, all polarized SAR images can provide additional information about surface structures and materials, which is particularly important for forest disturbance detection. Since interference in forest areas often causes changes in surface structures, all polarized SAR images can help identify and monitor forest disturbance events by capturing these fine structural changes.

[0071] The preprocessing of SAR images includes steps such as image introduction, polarization matrix conversion, and Refined Lee filtering. These steps can remove or reduce the influence of noise and improve image quality. False color synthesis can highlight important features while suppressing irrelevant background information, making the detection of forest land disturbances more accurate.

[0072] After being optimized, the surface change detection model can process the preprocessed SAR images more efficiently and accurately extract surface change patches. The overlay check process can determine the specific scope and degree of forest land disturbances by comparing SAR images in different time periods, reducing the error in a single measurement and ensuring the accuracy of the final disturbance patches. By converting complex SAR image data into easy-to-understand visualization results, it can provide decision-makers and relevant parties with rapid and accurate information, helping them respond in a timely manner and take appropriate protection measures.

[0073] It is achieved by the effective utilization of fully polarized SAR images, the application of preprocessing and false color synthesis, the optimization of the surface change detection model, the overlay check, and the generation of visualization results. This method has relatively high accuracy and reliability in actual applications and can effectively support forest land management and protection work.

[0074] In the above embodiments according to the present application, it should be understood that the disclosed methods, systems, devices, and media may be implemented in other ways. The embodiments of the methods, systems, devices, and media described above are merely exemplary. For example, the division of modules or units is merely a logical function division, and there may be other division methods when actually implemented. Each functional unit may be integrated into one processing unit, each unit may physically exist independently, or two or more units may be integrated into one unit.

[0075] Combining the various example units and algorithm steps described in the embodiments disclosed in this specification can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware manner or a software manner depends on the specific application of the technical solution and design constraints. Those skilled in the art can use different methods for each specific application to implement the described functions, but such implementation should not be considered as exceeding the scope of this application.

[0076] Here, the computer program may include computer program code that may be in the form of source code, object code, an executable file, or some intermediate form, etc. The computer-readable medium may include any entity or device that can carry the computer program code, a recording medium, a USB disk, a removable hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, an electrical carrier signal, an electrical communication signal, and a software distribution medium, etc.

[0077] The above embodiments are only used to explain the technical solutions of this application and do not limit it. Although this application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in each of the above embodiments can be modified, or some or all of their technical features can be equivalently replaced. These modifications or replacements do not deviate from the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of each embodiment of this application, and all should be included in the protection scope of this application.

Claims

1. A method for detecting forest disturbances based on a full-polarization SAR image, comprising: S1, obtaining a full-polarization SAR image within a target area and performing image classification of a reference time phase and a monitoring time phase on the full-polarization SAR image; S2, performing preprocessing including polarization matrix transformation, Refined Lee filtering, and false-color composition on the classified full-polarization SAR image to obtain a false-color composite SAR image; S3, inputting the false-color composite SAR image into an optimized surface change detection model, extracting surface change patches, and combining them with a radar vegetation index (RVI) image to obtain initial forest disturbance patches; S4, performing an overlay check on the initial forest disturbance patches to obtain forest disturbance patches within the target area and generating a visualization result, wherein, in S4, performing the overlay check on the initial forest disturbance patches includes S411, performing an overlay check on the initial forest disturbance patches and forest land data to obtain forest land disturbance patches; and S412, performing a spatial overlay check on the forest land disturbance patches and gradient data to obtain forest land disturbance patches within a preset gradient range. A method for detecting forest disturbances based on a full-polarization SAR image, characterized by the above.

2. As S2, performing preprocessing on the classified full-polarization SAR image to obtain a false-color composite SAR image specifically includes the following: S21, introduction of the SAR image: converting the full-polarization SAR images of the reference time phase and the monitoring time phase into a standard data format and performing radiometric correction processing to obtain full-polarization single-look complex data X; X = {X ｉ : i = 1, 2, …, M} Here, X ｉ = {xpq ｉ = Ipq ｉ + jQpq ｉ : p, q ∈ {H, V}} M is the total number of pixels, p is the antenna transmission method, q is the antenna reception method, i is the pixel index, j is the imaginary unit, and xpq ｉ is the complex scattering coefficient of the i-th pixel in the pq polarization mode, and Ip ｉ is xpq ｉ 's real part, Qpq ｉ is xpq ｉ 's imaginary part, H represents horizontal polarization, and V represents vertical polarization. S22, polarization matrix transformation of the SAR image: constructing a polarization scattering matrix S based on the full-polarization single-look complex data X; S = {S ｉ : i = 1, 2,..., M} defining a target scattering vector set K using the polarization scattering matrix S; K = {K ｉ : i = 1, 2, …, M} converting the target scattering vector set K into a polarization coherent matrix T, T = {T ｉ : i = 1, 2,..., M} where * represents a conjugate operation and < > represents a multi-view process; S23, Refined Lee filtering: performing Refined Lee filtering processing on the polarization coherent matrix T to obtain a filtered polarization coherent matrix T1, which specifically includes the following: where N represents the number of views and m represents the pixel index of the image after multi-view processing. If the resolution in the azimuth direction is changed to 1 / N of the original and the resolution in the range direction is not changed, it can be obtained, where, represents the average value of the same boundary alignment window within the local window where the m-th pixel is located, and w is the filtering weighting value. S24, false color synthesis: Perform full polarization false color synthesis on the filtered polarization conversion matrix T1 using the Pauli synthesis method to obtain a false color synthesized SAR image S1 and a false color synthesized SAR image S2. A method for detecting forest land disturbances based on a full polarization SAR image according to claim 1, characterized in that.

3. In the above S3, the optimization of the surface change detection model is S31, based on the surface disturbance interpretation flag, creating a tag image dataset; S32, using the tag image dataset to train a surface change detection model; S33, classifying the false color synthesized SAR image by the trained surface change detection model to obtain a classification result; S34, performing vector conversion on the classification result, calculating a cross-entropy loss value using a cross-entropy loss function, and performing model evaluation; S35, when the cross-entropy loss value is smaller than the cross-entropy loss threshold, outputting the classification result; otherwise, returning to S31 and supplementing the tag image dataset using the classification result. A method for detecting forest land disturbances based on a full polarization SAR image according to claim 1, characterized in that it includes the above steps.

4. In the above S31, the construction of the surface disturbance interpretation flag is S311, obtaining two high-resolution optical remote sensing images in the previous and later periods that match the full polarization SAR image time phase; S312, inputting the two high-resolution optical remote sensing images in the previous and later periods into a pre-trained neural network model to extract new additional surface disturbance patches; S313, combining with the false color synthesized SAR image to construct a surface disturbance interpretation flag. A method for detecting forest land disturbances based on a full polarization SAR image according to claim 3, characterized in that it includes the above steps.

5. In the above S3, the Freeman method is adopted to calculate the radar vegetation index RVI, RVI = Fv / (Fv + Fd + Fs) Here, Fv, Fd, and Fs are obtained by decomposing the scattering matrix after polarization matrix transformation of the SAR image, and represent the odd scattering component, even scattering component, and volume scattering component respectively. The method for detecting forest land disturbance based on the full polarization SAR image according to claim 1 is characterized by this.

6. In S4, generating the visualization result includes: S421, a step of deriving forest land disturbance patches within the target area from the vector data layer; S422, a step of combining with existing attribute information and marking the vector data layer; S423, a step of generating a plurality of types of visualization results including a forest land property classification map, a forest land disturbance heat map, and a forest land disturbance trend graph based on the marked vector data layer. The method for detecting forest land disturbance based on the full polarization SAR image according to claim 1 is characterized by this.

7. A forest land disturbance detection system based on a full polarization SAR image, comprising: A data acquisition module for acquiring a full polarization SAR image within a target area and performing image classification of the reference time phase and the monitoring time phase on the full polarization SAR image; [[ID=⑨]]A data preprocessing module for performing preprocessing including introduction of the SAR image, polarization matrix transformation of the SAR image, Refined Lee filtering, and false color synthesis on the classified full polarization SAR image to obtain a false color synthesis SAR image; A feature extraction module for inputting the false color synthesis SAR image into an optimized surface change detection model, extracting surface change patches, combining with a radar vegetation index RVI image, and obtaining initial forest land disturbance patches; A superposition check module for performing a superposition check on the initial forest land disturbance patches to obtain forest land disturbance patches within the target area and generate a visualization result, Performing a superposition check on the initial forest land disturbance patches includes: Performing a superposition check on the initial forest land disturbance patches and forest land data to obtain forest land disturbance patches outside the forest land; Performing a spatial superposition check on the forest land disturbance patches outside the forest land and gradient data to obtain forest land disturbance patches within a preset gradient range. The forest land disturbance detection system based on the full polarization SAR image is characterized by this.

8. An electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for detecting forest land disturbance according to any one of claims 1 to 6 is realized. An electronic device characterized by this.

9. A computer-readable storage medium, characterized in that when executed by a processor, it stores a computer program that realizes the method for detecting forest land disturbance according to any one of claims 1 to 6.

Images