Multi-temporal remote sensing image ecological surface change pattern spot extraction method and device

Abstract

The invention discloses a multi-temporal remote sensing image ecological surface change pattern spot extraction method and device. The method comprises the following steps: performing multi-dimensional feature extraction from a multi-temporal remote sensing image of a preset target area to obtain a multi-temporal high-dimensional feature image; classifying the multi-temporal high-dimensional feature image according to a pre-trained surface feature classification model to obtain a multi-temporal classification result image, and determining a classification change annotation image according to a classification result of each pixel in the multi-temporal classification result image; generating a change intensity image according to the multi-temporal high-dimensional feature image; according to the classification change labeling image and the change intensity image, generating final classification change labeling vector fishing net data; and performing pattern spot identification on the final classification change labeling vector fishing net data to obtain an earth surface change pattern spot of the preset target area.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and more specifically, to a method and apparatus for extracting ecological surface change patches from multi-temporal remote sensing images. Background Technology

[0002] In current remote sensing monitoring and land cover change research, change detection and patch extraction techniques based on multi-temporal remote sensing imagery have become key supporting tools in fields such as resource and environmental surveys, ecological protection monitoring, and urban expansion monitoring. Traditional remote sensing change extraction methods mainly rely on pixel-level classification differences or change index analysis, which often face the following limitations:

[0003] 1. Unstable accuracy of change detection: Simple band difference or threshold segmentation methods are easily affected by image noise, terrain shadows and seasonal changes, resulting in a high rate of false positives and false negatives;

[0004] 2. Lack of spatial structure representation ability: The pixel-level analysis results are severely fragmented, making it difficult to accurately represent the spatial boundaries and connectivity of actual ground features;

[0005] 3. High subjectivity in change judgment: Traditional change recognition relies on human experience to set thresholds and lacks a stable adaptive judgment mechanism;

[0006] 4. Lack of standardization in patch extraction: Clustering or surface generation methods often produce "jagged" edges or disconnected patches during boundary construction, which limits the application of subsequent mapping and statistical analysis. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a method and apparatus for extracting ecological surface change patches from multi-temporal remote sensing images.

[0008] According to one aspect of the present invention, a method for extracting ecological surface change patches from multi-temporal remote sensing images is provided, comprising:

[0009] Multi-dimensional feature extraction is performed on multi-temporal remote sensing images of a preset target area to obtain multi-temporal high-dimensional feature images;

[0010] The multi-temporal high-dimensional feature images are classified according to the pre-trained land cover classification model to obtain multi-temporal classification result images. Based on the classification results of each pixel in the multi-temporal classification result images, the classification change label images are determined.

[0011] Generate intensity change images based on multi-temporal high-dimensional feature images;

[0012] Based on the classification change labeled image and the change intensity image, the final classification change labeled vector fishing net data is generated;

[0013] The final classified change-annotated vector fishing net data is used to identify landform change patches in the preset target area.

[0014] Optionally, the multi-temporal remote sensing images include: optical images and SAR images of the first temporal phase; and optical images and SAR images of the second temporal phase, wherein the first temporal phase and the second temporal phase are images from two different times within the same quarter;

[0015] The features of multi-temporal high-dimensional feature images include optical image features and SAR image features, among which,

[0016] Optical image features include: BLUE band, GREEN band, RED band, NIR band, SWIR1 band, SWIR2 band, Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Soil-Regulated Vegetation Index (SAVI), Normalized Difference Building Index (NDBI), Normalized Difference Water Index (NDWI), and Normalized Difference in Snow Index (NDSI).

[0017] SAR image features include: vertical polarization (VV), vertical-horizontal polarization (VH), vertical polarization entropy (VVEn), vertical polarization variance (VVVa), vertical polarization contrast (VVCon), vertical polarization correlation (VVCor), and vertical polarization angle (VVASM), vertical-horizontal polarization entropy (VHEn), vertical-horizontal polarization variance (VHVa), vertical-horizontal polarization contrast (VHCon), vertical-horizontal polarization correlation (VHCor), and vertical-horizontal polarization angle (VHASM).

[0018] Optionally, based on the classification results of each pixel in the multi-temporal classification result image, a classification change annotation image is determined, including:

[0019] Based on the classification results of each pixel in the multi-temporal classification result image, the pixel value of each pixel in the classification change annotation image is determined, and the classification change annotation image is generated.

[0020] Optionally, the method for determining the pixel values ​​of the classified change labeled image is as follows:

[0021] If pixel p is classified the same as c in the multi-temporal classification result image... 1,p =c 2,p The expression for calculating the pixel value is:

[0022]

[0023] If pixel p is classified differently in the multi-temporal classification result image, c 1,p ≠c 2,p The expression for calculating the pixel value is:

[0024]

[0025] In the formula, The image pixel value represents the classification change of pixel p; c 1,p c is the pixel value of pixel p in the first phase classification result image of the multi-temporal classification result image; 2,p is the pixel value of pixel p in the second phase classification result image of the multi-temporal classification result image; N is the number of land cover categories.

[0026] Optionally, the expression for calculating the pixel value of the intensity change image is:

[0027]

[0028] In the formula, ||ΔX p || represents the intensity value of the change in pixel p; The pixel value of pixel p in the first phase of the high-dimensional feature image in the multi-temporal high-dimensional feature image; denoted as pixel value of pixel p in the second phase of the high-dimensional feature image in the multi-temporal high-dimensional feature image.

[0029] Optionally, based on the classification change-annotated image and the change intensity image, final classification change-annotated vector fishing net data is generated, including:

[0030] Based on the categorical change-annotated images and change intensity images, high-precision categorical change-annotated vector data and uncertain change-annotated vector data are obtained;

[0031] Multi-temporal fused images are obtained by fusing optical image features and SAR image features in multi-temporal high-dimensional feature images;

[0032] The pseudo-change results of the uncertain change labeled vector data are updated by using multi-temporal fusion imagery to obtain updated uncertain change labeled vector data.

[0033] The updated uncertain change annotation vector data is synchronized to the high-precision classification change annotation vector data to obtain the final classification change annotation vector fishing net data.

[0034] Optionally, based on the categorical change-annotated image and the change intensity image, high-precision categorical change-annotated vector data and uncertain change-annotated vector data are obtained, including:

[0035] Estimate the set of change thresholds V based on the categorical change-labeled images and change intensity images:

[0036]

[0037] In the formula, ΔT *The image is labeled for classification changes; ||ΔX|| is the image for intensity changes, P represents the set of all pixel locations in the image, and set V contains the values ​​of all pixels that meet the conditions in the image for intensity changes;

[0038] Sort the set of change thresholds V from smallest to largest, and take the change intensity value corresponding to the first quartile as the optimal change threshold λ for the change intensity image.

[0039] Based on the optimal change threshold λ, the change intensity image is calculated to obtain a high-precision classified change label image and an uncertain change label image;

[0040] Based on the high-precision classified change-annotated image and the uncertain change-annotated image, high-precision classified change-annotated vector data and uncertain change-annotated vector data are generated respectively.

[0041] Optionally, the expression for calculating the pixel value of a high-precision classified change-annotated image is:

[0042]

[0043] The expression for calculating the pixel value of an image with uncertain changes is as follows:

[0044]

[0045] In the formula, p represents a pixel; c 1,p This represents the pixel value of pixel p in the first phase of the multi-phase classification result image.

[0046] According to another aspect of the present invention, a device for extracting ecological surface change patches from multi-temporal remote sensing images is provided, comprising:

[0047] The extraction module is used to extract multi-dimensional features from multi-temporal remote sensing images of a preset target area to obtain multi-temporal high-dimensional feature images;

[0048] The classification module is used to classify multi-temporal high-dimensional feature images according to a pre-trained land cover classification model, obtain multi-temporal classification result images, and determine the classification change label images based on the classification results of each pixel in the multi-temporal classification result images.

[0049] The first generation module is used to generate intensity change images based on multi-temporal high-dimensional feature images;

[0050] The second generation module is used to generate the final classification change labeled vector fishing net data based on the classification change labeled image and the change intensity image.

[0051] The identification module is used to identify patches in the final classified change-annotated vector fishing net data to obtain surface change patches in the preset target area.

[0052] According to another aspect of the present invention, a computer-readable storage medium is provided, the storage medium storing a computer program for performing the methods described in any of the above aspects of the present invention.

[0053] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: a processor; a memory for storing executable instructions of the processor; the processor being configured to read the executable instructions from the memory and execute the instructions to implement the method described in any of the preceding aspects of the present invention.

[0054] Therefore, the overall design of this invention embodies a fully integrated approach from remote sensing spectral texture features, data fusion, machine learning classification, change intensity modeling to spatial structure construction. It has strong practicality, scalability and technological innovation, and is suitable for the task of extracting and mapping ecological surface change patches from large-area, multi-temporal remote sensing images. Attached Figure Description

[0055] Exemplary embodiments of the present invention can be more fully understood by referring to the following figures:

[0056] Figure 1 This is a schematic flowchart of a method for extracting ecological surface change patches from multi-temporal remote sensing images provided by an exemplary embodiment of the present invention;

[0057] Figure 2 This is a schematic diagram of a patch extraction example provided by an exemplary embodiment of the present invention;

[0058] Figure 3 This is a schematic diagram of the structure of a multi-temporal remote sensing image ecological surface change patch extraction device provided in an exemplary embodiment of the present invention;

[0059] Figure 4 This is the structure of an electronic device provided in an exemplary embodiment of the present invention. Detailed Implementation

[0060] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0061] It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention.

[0062] Those skilled in the art will understand that the terms "first," "second," etc., in the embodiments of the present invention are only used to distinguish different steps, devices, or modules, and do not represent any specific technical meaning, nor do they indicate a necessary logical order between them.

[0063] It should also be understood that in the embodiments of the present invention, "multiple" can refer to two or more, and "at least one" can refer to one, two or more.

[0064] It should also be understood that any component, data or structure mentioned in the embodiments of the present invention can generally be understood as one or more unless explicitly defined or given contrary instructions in the context.

[0065] Furthermore, the term "and / or" in this invention is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this invention generally indicates that the preceding and following related objects have an "or" relationship.

[0066] It should also be understood that the description of the various embodiments in this invention emphasizes the differences between the various embodiments, and the similarities or similarities can be referred to each other. For the sake of brevity, they will not be described in detail.

[0067] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0068] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use.

[0069] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, they should be considered part of the specification.

[0070] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0071] The embodiments of this invention can be applied to electronic devices such as terminal devices, computer systems, and servers, and can operate together with a wide range of other general-purpose or special-purpose computing system environments or configurations. Well-known examples of terminal devices, computing systems, environments, and / or configurations suitable for use with electronic devices such as terminal devices, computer systems, and servers include, but are not limited to: personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputer systems, mainframe computer systems, and distributed cloud computing environments including any of the above systems, etc.

[0072] Electronic devices such as terminal devices, computer systems, and servers can be described in the general context of computer system executable instructions (such as program modules) executed by a computer system. Typically, program modules can include routines, programs, object programs, components, logic, data structures, etc., which perform specific tasks or implement specific abstract data types. Computer systems / servers can be implemented in distributed cloud computing environments, where tasks are executed by remote processing devices linked through communication networks. In distributed cloud computing environments, program modules can reside on local or remote computing system storage media, including storage devices.

[0073] Exemplary methods

[0074] Figure 1 This is a schematic flowchart of a method for extracting ecological surface change patches from multi-temporal remote sensing images according to an exemplary embodiment of the present invention. This embodiment can be applied to electronic devices, such as... Figure 1 As shown, the method 100 for extracting ecological surface change patches from multi-temporal remote sensing images includes the following steps:

[0075] Step 101: Extract multi-dimensional features from multi-temporal remote sensing images of the preset target area to obtain multi-temporal high-dimensional feature images;

[0076] Step 102: Classify the multi-temporal high-dimensional feature images according to the pre-trained land cover classification model to obtain multi-temporal classification result images, and determine the classification change annotation images based on the classification results of each pixel in the multi-temporal classification result images.

[0077] Step 103: Generate intensity change images based on multi-temporal high-dimensional feature images;

[0078] Step 104: Generate the final classified change-annotated vector fishing net data based on the classified change-annotated image and the change intensity image;

[0079] Step 105: Perform patch recognition on the final classified change labeled vector fishing net data to obtain surface change patches in the preset target area.

[0080] Specifically, addressing the technical problems existing in the background art, this invention proposes a method for extracting ecological surface change patches based on multi-temporal remote sensing image feature construction, classification change annotation (CCA), change vector analysis (CVA), and regular fishing net patch construction, the specific implementation of which is as follows:

[0081] 1. Multi-temporal remote sensing imagery (This invention takes Landsat 9OLI optical imagery as an example, but is not limited to the optical imagery data of this satellite; it also takes C-band synthetic aperture radar (SAR) data carried by Sentinel-1A as an example, but is not limited to the SAR imagery data of this satellite).

[0082] The processing of optical images is as follows;

[0083] The T1 and T2 images are determined based on optical images. T1 and T2 are images taken at two different times during the same season and in the same area (e.g., early summer and late summer). The features include 12 features such as band and spectral characteristics.

[0084] Band selection: {BLUE,GREEN,RED,NIR,SWIR1,SWIR2}, as shown in Table 1.

[0085] Table 1. Bands used in Landsat 9OLI imagery

[0086] Band Name Band number Center wavelength (μm) BLUE B2 0.45–0.51 GREEN B3 0.53–0.59 RED B4 0.64–0.67 NIR B5 0.85–0.88 SWIR1 B6 1.57–1.65 SWIR2 B7 2.11–2.29

[0087] Spectral feature selection: NDVI (Normalized Difference Vegetation Index), EVI (Enhanced Vegetation Index), SAVI (Soil-Regulated Vegetation Index), NDBI (Normalized Building Index), NDWI (Normalized Water Index), NDSI (Normalized Snow Difference Index), {NDVI, EVI, SAVI, NDBI, NDWI, NDSI}.

[0088]

[0089] Where a and b are aerosol resistance parameters, with default values ​​of 6 and 7.5 respectively; d is the gain factor d = 2.5; c is the adjustment factor used to minimize the impact of soil noise c = 1; L is usually taken as 0.5.

[0090] The processing of SAR images is as follows:

[0091] Sentinel-1IW GRD format data at two time points in the same time region as the T1 and T2 images were identified. After radiometric calibration, spatial filtering, and topographic correction, the data were resampled to the same resolution (30 meters) as Landsat9OLI. The VV and VH backscattering coefficients were obtained, and five commonly used texture statistics with low redundancy were calculated for VV and VH. The five texture features are: entropy, variance, contrast, correlation, and angular second moment. The texture calculation window size was set to 5*5. The gray-level co-occurrence matrix included four directions: 0, 45, 90, and 135. To eliminate the influence of angle, the mean of the statistics in the four directions was used to calculate 5 texture feature values ​​for VV and VH. 5 texture features for VV and VH were calculated (the entropy, variance, contrast, correlation, and second moment of angle for VV and VH are represented as VVEn, VVVa, VVCon, VVCor, VVASM, VHEn, VHVa, VHCon, VHCor, and VHASM, respectively), for a total of 12 feature quantities.

[0092] 2. Classification Change Labeling (CCA): Using 6 bands, 6 spectral features, 2 backscattering coefficients, 5 texture features (VV), and 5 texture features (VH), with {BLUE,GREEN,RED,NIR,SWIR1,SWIR2,NDVI,EVI,SAVI,NDBI,NDWI,NDSI,VV,VH,VVEn,VVVa,VVCon,VVCor,VVASM,VHEn,VHVa,VHCon,VHCor,VHASM} as input factors, and utilizing existing land use data or manually labeled samples, a random forest algorithm is used to train a model for land cover classification. The T1 and T2 images are classified separately to obtain multi-temporal classification result images c1 and c2. Assuming N classification categories, the classification labels of pixel p at times T1 and T2 are denoted as follows:

[0093] c 1,p ∈{1,2,…,N},c 2,p ∈{1,2,…,N}

[0094] Among them, c 1,p It is the classification label of pixel p at time T1, c 2,p This is the classification label of pixel p at time T2. Assume the image labeled with the classification change determined by c1 and c2 is ΔT. * Case where pixel p remains unchanged: If pixel p has the same classification in both T1 and T2 images. 1,p =c 2,p If , it means there is no change, then:

[0095]

[0096] Pixel p changes: If pixel p is classified differently in T1 and T2 images, c 1,p ≠c 2,p , indicating a change, then:

[0097]

[0098] In summary, the categorical change-labeled image ΔT is obtained. * All cell values.

[0099] Specifically, refer to the "Guidelines for Land and Sea Use Classification in Territorial Spatial Survey, Planning, and Land Use Control (Trial Implementation)" and select the first-level category. Assume N=4 (i.e., 4 categories; taking the first-level classification as an example, Category 1: Forest Land; Category 2: Grassland; Category 3: Cultivated Land; Category 4: Residential Land; N can be set according to the actual situation). The numerical values ​​assigned to each pixel are marked as follows:

[0100] Table 2 Classification Change Labels

[0101]

[0102]

[0103] 3. Change Vector Analysis (CVA): Based on the change vector analysis of the T1 and T2 time-image data, 24 features were extracted from remote sensing image pixel p in the T1 and T2 time-images. The feature vectors are defined as follows:

[0104]

[0105] in These represent the values ​​of the BLUE feature in pixel p at time phases T1 and T2, respectively.

[0106] Change vector (difference vector): The change vector ΔX represents the feature difference between two time phases, and the change vector ΔX of pixel p. p :

[0107]

[0108] Change intensity (vector magnitude): Change intensity is the Euclidean norm of the change vector, ‖ΔX‖. The larger this value, the more significant the change in that pixel. Pixel p change intensity ‖ΔX p || is defined as:

[0109]

[0110] The intensity change image ‖ΔX‖ is determined based on the intensity change of all pixels.

[0111] 4. Estimate the optimal change threshold based on the Classification Change Annotation (CCA) from step 2 and the Change Vector Analysis (CVA) from step 3: In actual change detection, directly applying automatic segmentation methods to the full-map change vector analysis may be affected by interference from non-changed areas, leading to threshold shift. Therefore, utilizing existing labeled classifying change pixels can more accurately estimate the optimal change threshold V.

[0112]

[0113] The image with categorical change annotation is ΔT * (Step 2), Change in intensity image ||ΔX|| (Step 3), P represents the set of all pixel locations in the image, and set V contains the values ​​of all pixels that meet the condition (classification change label is greater than 0) in the change in intensity image.

[0114] Sort the set V in ascending order, and take the change intensity value corresponding to the first quartile as the CVA threshold λ. Based on the threshold λ, calculate the higher-precision classified change annotation image ΔT for pixel p:

[0115]

[0116] Based on the high-precision classification change annotation image ΔT, a corresponding vector fishing net is generated, and each raster value is assigned to the corresponding vector surface attribute table to obtain the high-precision classification change annotation vector data ΔT. shp .

[0117] For uncertain annotations, the following is represented:

[0118]

[0119] Generate the corresponding vector fishing net ΔT based on the raster image ΔT' rows and columns. shp′ Each raster value is assigned to the corresponding vector surface attribute table to create a scale-adaptive raster image of the fishing net unit.

[0120] 5. Data Fusion: Principal Component Analysis (PCA) was employed to extract spectral feature information from six bands and six spectral features {BLUE, GREEN, RED, NIR, SWIR1, SWIR2, NDVI, EVI, SAVI, NDBI, NDWI, NDSI}, selecting the first two principal components as the spectral feature data of the image. The average of five texture features (VV, VVVa, VVCon, VVCor, VVASM, VHEn, VHVa, VHCon, VHCor, VHASM) was used as the texture feature information of the image. Image fusion at the feature layer was performed based on the Gram-Schmidt (GS) method, a technique already in use. The GS fusion method can achieve optimal matching of the bands involved in the fusion, is not limited by band, and effectively preserves spatial texture and spectral feature information. Images G1 and G2 were fused from the two time points respectively.

[0121] 6. Manual interpretation: Basic vector fishing net ΔT shp′ For ΔT in the attribute table shp′ The T1 and G1 images and T2 and G2 images at positions ≠ 0 are used to determine the values ​​of ΔT and remove them. shp' The pseudo-change results are updated, the erroneous result labels are updated, and the vector fishing net ΔT is updated synchronously. shp Above (i.e., T1 and G1 correspond to the same pixel category, and will be adjusted to a negative number).

[0122] Based on the results of the pseudo-change determination in Table 2, T1 and G1 correspond to the same pixel category. ΔT shp′ The corresponding pixel is adjusted to Otherwise, adjust to

[0123] 7. Patch Extraction: In remote sensing change monitoring and patch identification, traditional pixel-based classification often lacks spatial connectivity and shape regularity, leading to fragmented patch boundaries and unstable morphology. To improve the spatial representation quality of changed areas, this invention employs a patch extraction method based on a vector fishing net. Using regular grid surfaces as analysis units, it identifies continuous changed areas in space through topological relationships and attribute consistency, achieving spatial patch identification and reconstruction. Using a vector fishing net ΔT... shp Based on this, identify continuous (4-point adjacency, diagonal adjacency is excluded) mesh groups, that is, only when two meshes share a complete boundary (shared edge) are they considered adjacent. Combine the meshes into polygons and assign them the same value, such as... Figure 2 The image shown is an example of a patch extraction.

[0124] Therefore, the present invention has the following beneficial technical effects:

[0125] 1. Integrating multi-source features to enhance change sensitivity: Based on the introduction of six original bands of optical images, six typical remote sensing indices (NDVI, EVI, SAVI, NDBI, NDWI, and NDSI), and the backscattering coefficient of Sar images, this invention further introduces texture features to achieve multi-dimensional feature coupling analysis of ground feature changes. It fully explores the spectral and structural change information of targets such as vegetation, water bodies, and buildings, improves the response capability to changes in different ground features, and enhances the distinguishability of change identification.

[0126] 2. Improve change detection robustness by constructing change priors based on Classification of Change (CCA): Land cover classification is performed on T1 and T2 temporal images using a random forest classification model, and a classified change-labeled image ΔT is constructed using change labels. * This provides supervision information for subsequent CVA change vectors and effectively suppresses the impact of noise disturbances on threshold determination.

[0127] 3. Introduce change vector magnitude and adaptive threshold estimation to enhance change judgment accuracy: The intensity of multidimensional feature changes is measured by Euclidean norm, and the change intensity distribution is constructed by combining existing change sample statistics. The change threshold is adaptively extracted in the change magnitude image, which significantly improves the separability of changed and non-changed regions.

[0128] 4. Constructing Regular Vector Fishing Net Units to Ensure Map Patch Structure Stability: Unlike traditional pixel-level connected region extraction methods, this invention uses regular vector fishing net units as the basic spatial processing unit. The fishing net unit scale is adaptive, and the grid size is set based on different data sources: the side length of the fishing net unit should not be less than the spatial resolution of the source data. For example, when using 30-meter remote sensing data, the lower limit of the fishing net size is 30 meters × 30 meters to avoid amplifying statistical errors. Combining topological relationships and attribute consistency to identify changing maps ensures that the map patch boundaries are regular and the structure is clear, meeting the needs of mapping and statistical analysis of change results.

[0129] 5. Image fusion of feature layers using principal component analysis and Gram-Schmidt effectively preserves spatial texture and spectral features. Based on this, a mechanism for simultaneous manual verification and vector annotation updates is integrated to enhance the credibility of the results: a manual verification mechanism is provided for change identification results to identify pseudo-changes or erroneous changes and update attribute fields, achieving two-way correction between classification results and patch annotations, ensuring the accuracy and standardization of the final change results.

[0130] The overall design of this invention embodies a fully integrated approach, encompassing remote sensing spectral texture features, data fusion, machine learning classification, change intensity modeling, and spatial structure construction. It possesses strong practicality, scalability, and technological innovation, and is suitable for the extraction and mapping of ecological surface change patches from large-area, multi-temporal remote sensing images.

[0131] Exemplary device

[0132] Figure 3 This is a schematic diagram of the structure of a multi-temporal remote sensing image ecological surface change patch extraction device provided in an exemplary embodiment of the present invention. Figure 3 As shown, the device 300 includes:

[0133] Extraction module 310 is used to extract multi-dimensional features from multi-temporal remote sensing images of a preset target area to obtain multi-temporal high-dimensional feature images;

[0134] The classification module 320 is used to classify multi-temporal high-dimensional feature images according to a pre-trained land cover classification model, obtain multi-temporal classification result images, and determine classification change annotation images based on the classification results of each pixel in the multi-temporal classification result images.

[0135] The first generation module 330 is used to generate intensity change images based on multi-temporal high-dimensional feature images;

[0136] The second generation module 340 is used to generate the final classification change labeled vector fishing net data based on the classification change labeled image and the change intensity image.

[0137] The identification module 350 is used to identify the surface change patches in the final classified change-annotated vector fishing net data to obtain the surface change patches in the preset target area.

[0138] Exemplary electronic devices

[0139] Figure 4 This is the structure of an electronic device provided in an exemplary embodiment of the present invention. For example... Figure 4 As shown, the electronic device 40 includes one or more processors 41 and a memory 42.

[0140] The processor 41 may be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions.

[0141] The memory 42 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 41 may execute the program instructions to implement the methods of the software programs of the various embodiments of the present invention described above, and / or other desired functions. In one example, the electronic device may also include an input device 43 and an output device 44, these components being interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0142] In addition, the input device 43 may also include, for example, a keyboard, a mouse, etc.

[0143] The output device 44 can output various information to the outside. The output device 44 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0144] Of course, for the sake of simplicity, Figure 4 Only some of the components of this electronic device relevant to the present invention are shown, omitting components such as buses, input / output interfaces, etc. In addition, the electronic device may include any other suitable components depending on the specific application.

[0145] Exemplary computer program products and computer-readable storage media

[0146] In addition to the methods and apparatus described above, embodiments of the present invention may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the methods according to various embodiments of the present invention described in the "Exemplary Methods" section above.

[0147] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of the present invention. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0148] Furthermore, embodiments of the present invention may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps of the methods according to various embodiments of the present invention described in the "Exemplary Methods" section above.

[0149] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0150] The basic principles of the present invention have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in the present invention are merely examples and not limitations, and should not be considered as essential features of each embodiment of the present invention. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the present invention to the necessity of employing the aforementioned specific details.

[0151] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For system embodiments, since they largely correspond to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0152] The block diagrams of devices, systems, devices, and systems involved in this invention are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, systems, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0153] The methods and systems of the present invention may be implemented in many ways. For example, they may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above-described order of steps for the methods is for illustrative purposes only, and the steps of the methods of the present invention are not limited to the order specifically described above unless otherwise specifically stated. Furthermore, in some embodiments, the present invention may also be implemented as a program recorded on a recording medium, the program comprising machine-readable instructions for implementing the methods according to the present invention. Thus, the present invention also covers recording media storing programs for performing the methods according to the present invention.

[0154] It should also be noted that in the systems, apparatus, and methods of the present invention, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered equivalents of the present invention. The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the invention. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of the invention. Therefore, the invention is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0155] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of the invention to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for extracting ecological surface change patches from multi-temporal remote sensing images, characterized in that, include: Multi-dimensional feature extraction is performed on multi-temporal remote sensing images of a preset target area to obtain multi-temporal high-dimensional feature images; The multi-temporal high-dimensional feature images are classified according to the pre-trained land cover classification model to obtain multi-temporal classification result images. Based on the classification results of each pixel in the multi-temporal classification result images, the classification change label images are determined. Generate intensity change images based on multi-temporal high-dimensional feature images; Based on the classified change-annotated image and the change intensity image, the final classified change-annotated vector fishing net data is generated; The final classified change-annotated vector fishing net data is used to identify landform change patches in the preset target area.

2. The method according to claim 1, characterized in that, The multi-temporal remote sensing imagery includes: optical imagery and SAR imagery of a first temporal phase; and optical imagery and SAR imagery of a second temporal phase, wherein the first temporal phase and the second temporal phase are images taken at two different times within the same quarter; The features of the multi-temporal high-dimensional feature image include optical image features and SAR image features, wherein... The optical image features include: BLUE band, GREEN band, RED band, NIR band, SWIR1 band, SWIR2 band, Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Soil-Regulated Vegetation Index (SAVI), Normalized Difference Building Index (NDBI), Normalized Difference Water Index (NDWI), and Normalized Difference in Snow Index (NDSI). The SAR image features include: vertical polarization VV, vertical-horizontal polarization VH, vertical polarization entropy VVEn, vertical polarization variance VVVa, vertical polarization contrast VVCon, vertical polarization correlation VVCor, vertical polarization angle VVASM, vertical-horizontal polarization entropy VHEn, vertical-horizontal polarization variance VHVa, vertical-horizontal polarization contrast VHCon, vertical-horizontal polarization correlation VHCor, and vertical-horizontal polarization angle VHASM.

3. The method according to claim 1, characterized in that, Based on the classification results of each pixel in the multi-temporal classification image, the classification change annotation image is determined, including: Based on the classification results of each pixel in the multi-temporal classification result image, the pixel value of each pixel in the classification change annotation image is determined, and the classification change annotation image is generated.

4. The method according to claim 3, characterized in that, The method for determining the pixel values ​​of the classified change-annotated image is as follows: If pixel p is classified the same as c in the multi-temporal classification result image... 1,p =c 2,p The expression for calculating the pixel value is: If pixel p is classified differently in the multi-temporal classification result image, c 1,p ≠c 2,p The expression for calculating the pixel value is: In the formula, The image pixel value represents the classification change of pixel p; c 1,p c is the pixel value of pixel p in the first phase classification result image of the multi-temporal classification result image; 2,p is the pixel value of pixel p in the second phase classification result image of the multi-temporal classification result image; N is the number of land cover categories.

5. The method according to claim 1, characterized in that, The expression for calculating the pixel value of the intensity-changing image is as follows: In the formula, ||ΔX p || represents the intensity value of the change in pixel p; The pixel value of pixel p in the first phase of the high-dimensional feature image in the multi-temporal high-dimensional feature image; denoted as pixel value of pixel p in the second phase of the high-dimensional feature image in the multi-temporal high-dimensional feature image.

6. The method according to claim 1, characterized in that, Based on the categorized change-annotated image and the change intensity image, final categorized change-annotated vector fishing net data is generated, including: Based on the classified change-annotated image and the change intensity image, high-precision classified change-annotated vector data and uncertain change-annotated vector data are obtained; A multi-temporal fused image is obtained by fusing optical image features and SAR image features in the multi-temporal high-dimensional feature image. The pseudo-change results of the uncertain change labeled vector data are updated using the multi-temporal fusion image to obtain updated uncertain change labeled vector data. The updated uncertain change annotation vector data is synchronized to the high-precision classification change annotation vector data to obtain the final classification change annotation vector fishing net data.

7. The method according to claim 6, characterized in that, Based on the classified change-annotated image and the change intensity image, high-precision classified change-annotated vector data and uncertain change-annotated vector data are obtained, including: Based on the classified change-annotated images and change intensity images, estimate the change threshold set V: In the formula, ΔT * The image is labeled for classification changes; ||ΔX|| is the image for intensity changes, P represents the set of all pixel locations in the image, and set V contains the values ​​of all pixels that meet the conditions in the image for intensity changes; Sort the set of change thresholds V from smallest to largest, and take the change intensity value corresponding to the first quartile as the optimal change threshold λ for the change intensity image. Based on the optimal change threshold λ, the change intensity image is calculated to obtain a high-precision classified change labeling image and an uncertain change labeling image; Based on the high-precision classified change-annotated image and the uncertain change-annotated image, the high-precision classified change-annotated vector data and the uncertain change-annotated vector data are generated respectively.

8. The method according to claim 7, characterized in that, The pixel value calculation expression for the high-precision classified change-annotated image is as follows: The expression for calculating the pixel value of the uncertain change-annotated image is as follows: In the formula, p represents a pixel; c 1,p This represents the pixel value of pixel p in the first phase of the multi-phase classification result image.

9. A device for extracting ecological surface change patches from multi-temporal remote sensing images, characterized in that, include: The extraction module is used to extract multi-dimensional features from multi-temporal remote sensing images of a preset target area to obtain multi-temporal high-dimensional feature images; The classification module is used to classify multi-temporal high-dimensional feature images according to a pre-trained land cover classification model, obtain multi-temporal classification result images, and determine the classification change label images based on the classification results of each pixel in the multi-temporal classification result images. The first generation module is used to generate intensity change images based on multi-temporal high-dimensional feature images; The second generation module is used to generate final classification change labeled vector fishing net data based on the classification change labeled image and the change intensity image. The identification module is used to identify the surface change patches in the final classified change-annotated vector fishing net data to obtain the surface change patches in the preset target area.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program for performing the method described in any one of claims 1-8.

Images