SAR change detection method based on global neighborhood information enhanced difference map

Abstract

A SAR change detection method based on global neighborhood information-enhanced difference maps, relating to the field of image change detection technology. This invention aims to address the problems of low detection efficiency and accuracy in existing SAR change detection methods. The invention includes: generating an initial difference image using dual-temporal SAR image pairs; then obtaining a segmentation map containing structural information and difference features using the initial difference image; and finally obtaining the weight AX of SAR image X using the segmentation map containing structural information and difference features. i,j The weights AY of the SAR image Y i,j Then, using the weight AX of the SAR image X i,j The weights AY of the SAR image Y i,j Update SAR images X and Y to obtain SAR images PGX and PGY; use SAR images PGX and PGY to obtain a final difference map; perform automatic thresholding segmentation on the final difference map to obtain the SAR image change regions. This invention is used to detect SAR image change regions.

Description

Technical Field

[0001] This invention relates to the field of image change detection technology, and in particular to a SAR change detection method based on a global neighborhood information-enhanced difference map. Background Technology

[0002] With the advancement of remote sensing and computer vision technologies, remote sensing image change detection has been widely used in environmental monitoring, terrain change assessment, and natural disaster monitoring. SAR image difference maps, due to their ability to penetrate clouds and their all-weather imaging capabilities, are widely used in the field of change detection. Therefore, SAR change detection based on difference maps has become a research focus in this field.

[0003] Current SAR image change detection methods largely rely on local neighborhood information for difference map generation, failing to fully capture large-scale image contextual relationships and resulting in inaccurate identification of changed regions. While graph theory-based image modeling methods can capture global relationships between pixels or superpixels by constructing graph structures, existing graph theory models are computationally intensive and still inadequate in capturing subtle changes. This leads to low efficiency in difference map generation and difficulty in fully utilizing contextual information in high-resolution images with complex backgrounds or noise interference, ultimately resulting in low accuracy and efficiency in SAR image change detection. Summary of the Invention

[0004] The purpose of this invention is to address the problems of low detection efficiency and accuracy in existing SAR change detection methods, and to propose a SAR change detection method based on a global neighborhood information-enhanced difference map.

[0005] The SAR change detection method based on global neighborhood information-enhanced difference maps includes the following steps:

[0006] Step 1: Generate an initial difference image using dual-temporal SAR image pairs, and then use the initial difference image to obtain a segmentation map containing structural information and difference features;

[0007] The dual-temporal SAR image pair includes: SAR image X and SAR image Y;

[0008] Step 2: Obtain the weights AX of SAR image X using the segmentation map containing structural information and differential features. i,j The weights AY of the SAR image Y i,j Then, using the weight AX of the SAR image X i,j The weights AY of the SAR image Y i,j Update SAR images X and Y to obtain images PGX and PGY;

[0009] Step 3: Use the PGX and PGY images obtained in Step 2 to obtain the final difference map;

[0010] Step 4: Perform automatic thresholding on the final difference map to obtain the change areas in the SAR image.

[0011] Further, in step one, an initial difference image is generated using dual-temporal SAR image pairs, and then a segmentation map containing structural information and difference features is obtained using the initial difference image. Specifically:

[0012] Step 1: Generate the initial difference image (DI) using dual-temporal SAR image pairs;

[0013] The pixel values ​​at (m, n) in the initial difference image are as follows:

[0014]

[0015] Where X and Y are a pair of dual-temporal SAR images, DI(m,n) is the pixel value at (m,n) in the initial difference image, X(m,n) is the pixel value at (m,n) in SAR image X, and Y(m,n) is the pixel value at (m,n) in SAR image Y.

[0016] Steps 1 and 2: Generate a pseudo-color image using dual-temporal SAR image pairs and the initial difference image DI;

[0017] Step 13: Perform superpixel segmentation on the pseudocolor image to obtain a segmentation map containing structural information and differential features.

[0018] Furthermore, the generation of pseudo-color images using dual-temporal SAR image pairs and the initial difference image DI in steps one and two specifically involves:

[0019] The pixels X(m,n) and Y(m,n) in the dual-temporal image pair and the pixels DI(m,n) in the initial difference image are stitched together along the channel dimension. After all pixels are stitched together, a pseudo-color image is obtained.

[0020] Furthermore, the superpixel segmentation of the pseudocolor image in steps one and three to obtain a segmentation map containing structural information and differential features specifically involves using the SLIC algorithm to perform superpixel segmentation on the pseudocolor image.

[0021] Furthermore, in step two, the weight AX of the SAR image X is obtained using a segmentation map containing structural information and differential features. i,j The weights AY of the SAR image Y i,j Then, using the weight AX of the SAR image X i,j The weights AY of the SAR image Y i,j Update SAR images X and Y to obtain images PGX and PGY, specifically:

[0022] Step 2: In the SAR image X, obtain the pixel values ​​corresponding to the superpixel positions in the segmentation map containing structural information and differential features, and then obtain the weight AX of the SAR image X. i,j :

[0023]

[0024] Among them, X i X′ is the average pixel value in the i-th superpixel region of the SAR image X. j is the average pixel value in the j-th superpixel region of SAR image X, and ∈ is the smoothing parameter;

[0025] Step 22: Obtain the pixel values ​​corresponding to the superpixel positions in the segmentation map containing structural information and differential features in the SAR image Y, and then obtain the weights AY of the SAR image Y. i,j ;

[0026] Steps 2 and 3: Treat the region corresponding to each superpixel in SAR images X and Y as a node, and obtain the self-center graph selfGX of node i in SAR images X and Y. i selfGY i ;

[0027] Step 24: Utilize the self-centered graph selfGX i selfGY i Update the SAR image to obtain the updated SAR images PGX and PGY.

[0028] Further, in step two, the pixel values ​​corresponding to the superpixel positions in the segmentation map containing structural information and differential features in the SAR image Y are obtained, and then the weights AY of the SAR image Y are obtained. i,j Specifically:

[0029]

[0030] Among them, Y i ′ is the average pixel value in the i-th superpixel region of the SAR image Y, where Y is the average pixel value. j ′ is the average pixel value in the j-th superpixel region of the SAR image Y.

[0031] Furthermore, in steps two and three, the region corresponding to each superpixel in SAR images X and Y is treated as a node, and the self-center graph selfGX of node i in SAR images X and Y is obtained. i selfGY i Specifically:

[0032] selfGX i =(XV i XEi )

[0033] selfGY i =(YV i ,YE i )

[0034] Among them, selfGX i XV is the self-center map of the i-th node in the SAR image X. i XE is the set consisting of the i-th node and its neighboring nodes in the SAR image X. i It is the set of connecting edges between the i-th node and its neighboring nodes in the SAR image X, selfGY i YV is the self-center map of the i-th node in the SAR image Y. i YE is the set consisting of the i-th node and its neighboring nodes in the SAR image Y. i It is the set of connecting edges between the i-th node and its neighboring nodes in the SAR image Y.

[0035] Furthermore, in step two or four, the self-centered graph selfGX is utilized. i selfGY i Update the SAR image to obtain the updated SAR images PGX and PGY, specifically:

[0036] X in SAR image X i The updated version is:

[0037]

[0038] Y in SAR image Y i The updated version is:

[0039]

[0040] Among them, PGX i It is the updated X i PGY i It is the updated Y i X j It is X i The neighboring nodes of Y j It is Y i The neighboring node of X, j is X i The label of the neighboring node, N i It is X i The set of neighboring nodes.

[0041] Furthermore, in step three, the final difference map is obtained using the images PGX and PGY obtained in step two. The value of the i-th pixel in the final difference map is as follows:

[0042]

[0043] Among them, DIfinal i This represents the value of the i-th pixel in the final difference map.

[0044] Furthermore, in step four, the final difference map is automatically thresholded and segmented using K-means clustering, maximum entropy, or iterative thresholding.

[0045] The beneficial effects of this invention are as follows:

[0046] This invention designs a graph structure construction and feature aggregation propagation architecture from the perspective of global image information propagation. It fully utilizes the spatial context information and temporal difference feature information of SAR images to enhance the context information of SAR images in complex noise environments, improving the accuracy and robustness of difference map generation, and thus enhancing the accuracy of SAR change detection. Simultaneously, this invention reduces computational load and improves the detection efficiency of SAR change regions. The invention comprises three parts: initial difference image generation and pseudo-color image construction, global structure map construction, and feature information propagation and difference information optimization. The initial difference image generation and pseudo-color image construction part is responsible for generating an initial difference map by performing a logarithmic ratio operator on dual-temporal SAR images, and fusing the dual-temporal images and difference images to generate a pseudo-color map for subsequent superpixel segmentation. The global structure map construction part provides a superpixel map containing structural information generated through superpixel segmentation of the pseudo-color image, constructing a global robust structure map. The feature information propagation and difference information optimization part, based on a global neighborhood information propagation mechanism, aggregates and propagates superpixel features to generate a reconstructed SAR image, and optimizes the difference information based on the reconstructed image to generate the final difference map, ultimately obtaining the SAR change region. This invention obtains global contextual relationships in images through a global information propagation mechanism, effectively suppressing high noise interference while enhancing image features. Therefore, this invention can effectively enhance the context of dual-temporal SAR images in application scenarios such as high noise and complex terrain, improving the accuracy and stability of change area detection, meeting practical needs, and being easy to implement. Attached Figure Description

[0047] Figure 1 This is a flowchart of the present invention;

[0048] Figure 2 Flowchart for difference image generation and pseudo-color image construction;

[0049] Figure 3 A schematic diagram illustrating superpixel partitioning and global robust structure graph construction;

[0050] Figure 4This is a schematic diagram of graph structure feature aggregation for the global neighborhood information propagation mechanism. Detailed Implementation

[0051] Specific implementation method one: as follows Figure 1 As shown, the specific process of the SAR change detection method based on the global neighborhood information-enhanced difference map in this embodiment is as follows:

[0052] Step 1: Generate an initial difference image using dual-temporal SAR image pairs, and then use the initial difference image to obtain a segmentation map containing structural information and difference features. Specifically:

[0053] Step 11, as follows Figure 2 As shown, the logarithmic ratio operator is used to generate the initial difference image DI from dual-temporal SAR image pairs:

[0054] The pixel values ​​at (m, n) in the initial difference image are as follows:

[0055]

[0056] Where X and Y are a pair of dual-temporal SAR images, DI(m,n) is the pixel value at (m,n) in the initial difference image, X(m,n) is the pixel value at (m,n) in SAR image X, Y(m,n) is the pixel value at (m,n) in SAR image Y, and || is the absolute value symbol;

[0057] Steps 1 and 2: Generate a pseudo-color image using dual-temporal SAR image pairs and the initial difference image (DI):

[0058] The pixels X(m,n) and Y(m,n) in the dual-temporal image pair and the pixels DI(m,n) in the initial difference image are merged by multi-channel cascade integration (stitching along the channel dimension) to obtain a pseudo-color image.

[0059] Step 13: Perform superpixel segmentation on the pseudocolor image to obtain a segmentation map containing structural information and differential features:

[0060] The superpixel segmentation is achieved through the Simple Linear Iterative Clustering (SLIC) algorithm to reduce the number of nodes and computational complexity;

[0061] The superpixel count is 5000, the superpixel compactness factor is 1, and the superpixel color space is Lab color space.

[0062] Step 2: Obtain the weights AX of SAR image X using the segmentation map containing structural information and differential features. i,j The weights AY of the SAR image Y i,j Then, using the weight AX of the SAR image X i,j The weights AY of the SAR image Yi,j Update SAR images X and Y to obtain SAR images PGX and PGY:

[0063] Step 2: 1. Obtain the pixel values ​​corresponding to the superpixel positions in the segmentation map of SAR image X, and then obtain the weights AX of SAR image X. i,j Specifically:

[0064]

[0065] Among them, X i X′ is the average pixel value in the i-th superpixel region of the SAR image X. j is the average pixel value in the j-th superpixel region of SAR image X, ∈ is the smoothing parameter, and is a non-zero constant; the superpixel X in SAR image X i and weight AX i,j A global robust structure diagram that makes up SAR image X;

[0066] Step 22, as follows Figure 3 As shown, the pixel values ​​corresponding to the superpixel positions in the segmentation map of the SAR image Y are obtained, and then the weights AY of the SAR image Y are obtained. i,j Specifically:

[0067]

[0068] Among them, Y i ′ is the average pixel value in the i-th superpixel region of the SAR image Y, where Y is the average pixel value. j ′ is the average pixel value in the j-th superpixel region of the SAR image Y, ∈ is the smoothing parameter, and is a constant that is not equal to 0;

[0069] Superpixel Y in SAR image Y i And weight AY i,j A global robust structure map that makes up the SAR image Y is constructed. This global robust structure map uses each superpixel region in the segmentation image as a node and utilizes the feature similarity between nodes to build neighborhood connections.

[0070] Steps two and three, such as Figure 4 As shown, each superpixel region in SAR images X and Y is treated as a node, and the self-center graph selfGX of node i in SAR images X and Y is obtained. i selfGY i :

[0071] selfGX i =(XV i XE i )

[0072] selfGYi =(YV i ,YE i )

[0073] Among them, selfGX i XV is the self-center map of the i-th node in the SAR image X. i XE is the set consisting of the i-th node and its neighboring nodes in the SAR image X. i It is the set of edges connecting the i-th node in the SAR image X and its neighboring nodes, selfGY i YV is the self-center map of the i-th node in the SAR image Y. i YE is the set consisting of the i-th node and its neighboring nodes in the SAR image Y. i It is the set of connecting edges between the i-th node and its neighboring nodes in the SAR image Y;

[0074] Step 2.4: Global Neighborhood Information Propagation Mechanism Utilizing Self-Centered Graph (selfGX) i selfGY i The superpixels are updated by information propagation, thereby updating the SAR image and obtaining the updated SAR image PGX and PGY:

[0075] SAR image X in X i The updated version is:

[0076]

[0077] Y in SAR image i The updated version is:

[0078]

[0079] Among them, PGX i It is the updated X i PGY i It is the updated Y i X j It is X i The neighboring nodes of Y j It is Y i The neighboring node of X, j is X i The label of the neighboring node, N i It is X i The set of neighboring nodes.

[0080] Step 3: Based on the global neighborhood information propagation mechanism, feature aggregation and propagation are performed in the graph structure to reconstruct the SAR image with enhanced contextual information. The final difference map DIfinal is obtained using the SAR images PGX and PGY obtained in Step 2.

[0081] The final pixel value in the difference image is as follows:

[0082]

[0083] Among them, DIfinal i This represents the value of the i-th pixel in the final difference map.

[0084] The construction of image PGX and PGY involves assigning the updated representation homogeneous values ​​to each superpixel PGX. i and PGY i The encompassed set of pixels maintains the consistency and scalability of features within the region, thereby achieving stable transmission of global neighborhood features in the local space. This processing mechanism not only ensures the robustness of representations in each region but also improves the fidelity of structural details in the overall image within the framework of global information propagation.

[0085] The process of constructing an image PGX essentially involves updating the PGX representation of each superpixel region through assignment operations. i The feature is uniformly propagated to all pixels in the corresponding region to achieve intrinsic consistency within that region. This process implicitly assumes that local features within each superpixel maintain spatial homogeneity, thus achieving this by filling the corresponding update value PGX within the range of each superpixel. i This process gradually completes the construction of the overall image PGX. Correspondingly, the generation mechanism of PGY is similar to that of PFX, which involves updating the PGY feature of each superpixel. i This is extended to the entire pixel group under its jurisdiction to achieve a balanced distribution of features in the updated region. This operation not only consolidates the feature stability of each region but also preserves structural details, achieving feature representation that combines spatial consistency and robustness within the framework of global neighborhood information propagation.

[0086] Step 4: Perform automatic thresholding on the final difference map to obtain the change areas in the SAR image;

[0087] The automatic thresholding methods include, but are not limited to, the Otsu method, K-means clustering, maximum entropy method, iterative thresholding method, and machine learning-based thresholding methods.

[0088] In this embodiment, the dual-temporal SAR image can be a SAR image that has undergone preprocessing operations such as registration and radiometric correction. The dual-temporal SAR image consists of two SAR images of the same region, the same size, but different temporal phases. After being input into a dual-temporal SAR difference map generation architecture based on global neighborhood information propagation, the final difference map is obtained at the output layer. The dual-temporal SAR difference map generation architecture based on global neighborhood information propagation includes two logarithmic ratio operator operations, one superpixel segmentation operation, and one feature aggregation and propagation operation. The first logarithmic ratio operator operation generates an initial difference map for the dual-temporal SAR image and constructs a pseudo-color image based on the initial difference map. The superpixel segmentation operation processes the pseudo-color image and constructs a global robust structure map based on feature similarity. The feature aggregation and propagation operation processes the global robust structure map to obtain the reconstructed dual-temporal SAR image. The second logarithmic ratio operator operation processes the reconstructed dual-temporal SAR image to generate the final difference map, ultimately obtaining the SAR change region.

[0089] This invention innovatively proposes a global neighborhood information propagation model for generating dual-temporal SAR difference maps. The input to this model is a pair of dual-temporal SAR images, and the output is a difference map. The model includes a global neighborhood information propagation module and a difference feature extraction module, providing assurance for high-precision image analysis such as change detection, anomaly detection, and target recognition, and exhibiting certain methodological advantages. This invention utilizes deep learning and image processing techniques to better meet the needs of change detection and monitoring in different scenarios. The global neighborhood information propagation model maintains a certain degree of stability in feature extraction, edge preservation, and noise suppression of the input image.

Claims

1. A SAR change detection method based on a global neighborhood information-enhanced difference map, characterized in that: The specific process of the method is as follows: Step 1: Generate an initial difference image using dual-temporal SAR image pairs, and then use the initial difference image to obtain a segmentation map containing structural information and difference features; The dual-temporal SAR image pair includes: SAR image X and SAR image Y; Step 2: Obtain the weights AX of SAR image X using the segmentation map containing structural information and differential features. i,j The weights AY of the SAR image Y i,j Then, using the weight AX of the SAR image X i,j The weights AY of the SAR image Y i,j Update SAR images X and Y to obtain images PGX and PGY, specifically as follows: Step 2: In the SAR image X, obtain the pixel values ​​corresponding to the superpixel positions in the segmentation map containing structural information and differential features, and then obtain the weight AX of the SAR image X. i,j : Where, X′ i X′ is the average pixel value in the i-th superpixel region of the SAR image X. j is the average pixel value in the j-th superpixel region of SAR image X, and ∈ is the smoothing parameter; Step 22: Obtain the pixel values ​​corresponding to the superpixel positions in the segmentation map containing structural information and differential features in the SAR image Y, and then obtain the weights AY of the SAR image Y. i,j Specifically: Among them, Y i Y′ is the average pixel value in the i-th superpixel region of the SAR image Y. j It is the average pixel value in the j-th superpixel region of the SAR image Y; Steps 2 and 3: Treat the region corresponding to each superpixel in SAR images X and Y as a node, and obtain the self-center graph selfGX of node i in SAR images X and Y. i selfGY i Specifically: selfGX i =(XV i ,XE i ) selfGY i =(YV i ,YE i ) Among them, selfGX i XV is the self-center map of the i-th node in the SAR image X. i XE is the set consisting of the i-th node and its neighboring nodes in the SAR image X. i It is the set of connecting edges between the i-th node and its neighboring nodes in the SAR image X, selfGY i YV is the self-center map of the i-th node in the SAR image Y. i YE is the set consisting of the i-th node and its neighboring nodes in the SAR image Y. i It is the set of connecting edges between the i-th node and its neighboring nodes in the SAR image Y; Step 24: Utilize the self-centered graph selfGX i selfGY i Update the SAR image to obtain the updated SAR images PGX and PGY, specifically: X in SAR image X i The updated version is: Y in SAR image Y i The updated version is: Among them, PGX i It's the updated X i PGY i It is the updated Y i X j It is X i The neighboring nodes of Y j It is Y i The neighboring node of X, j is X i The label of the neighboring node, N i It is X i The set of neighboring nodes; Step 3: Use the PGX and PGY images obtained in Step 2 to obtain the final difference map; Step 4: Perform automatic thresholding on the final difference map to obtain the change areas in the SAR image.

2. The SAR change detection method based on a global neighborhood information-enhanced difference map according to claim 1, characterized in that: In step one, an initial difference image is generated using dual-temporal SAR image pairs, and then a segmentation map containing structural information and difference features is obtained using the initial difference image. Specifically: Step 1: Generate the initial difference image (DI) using dual-temporal SAR image pairs; The pixel values ​​at (m, n) in the initial difference image are as follows: Where X and Y are a pair of dual-temporal SAR images, DI(m,n) is the pixel value at (m,n) in the initial difference image, X(m,n) is the pixel value at (m,n) in SAR image X, and Y(m,n) is the pixel value at (m,n) in SAR image Y. Steps 1 and 2: Generate a pseudo-color image using dual-temporal SAR image pairs and the initial difference image DI; Step 13: Perform superpixel segmentation on the pseudocolor image to obtain a segmentation map containing structural information and differential features.

3. The SAR change detection method based on a global neighborhood information-enhanced difference map according to claim 2, characterized in that: The generation of pseudo-color images using dual-temporal SAR image pairs and initial difference images (DI) in steps one and two specifically involves: The pixels X(m,n) and Y(m,n) in the dual-temporal image pair and the pixels DI(m,n) in the initial difference image are stitched together along the channel dimension. After all pixels are stitched together, a pseudo-color image is obtained.

4. The SAR change detection method based on a global neighborhood information-enhanced difference map according to claim 3, characterized in that: The step one and three involves performing superpixel segmentation on the pseudocolor image to obtain a segmentation map containing structural information and differential features. Specifically, the SLIC algorithm is used to perform superpixel segmentation on the pseudocolor image.

5. The SAR change detection method based on a global neighborhood information-enhanced difference map according to claim 4, characterized in that: In step three, the final difference map is obtained using the images PGX and PGY obtained in step two. The value of the i-th pixel in the final difference map is as follows: Among them, DIfinal i This represents the value of the i-th pixel in the final difference map.

6. The SAR change detection method based on a global neighborhood information-enhanced difference map according to claim 5, characterized in that: The automatic thresholding of the final difference map in step four is achieved using K-means clustering, maximum entropy, or iterative thresholding.

Images