A dynamic context aggregation method for building remote sensing image change detection

Abstract

The present invention discloses a method for detecting changes in building remote sensing images using dynamic context aggregation, belonging to the field of image detection. The method comprises: constructing a dataset and performing sample partitioning; constructing a DCA-NET network model; adjusting parameters of the constructed network model; inputting the partitioned dataset into the model for detection to obtain change results; and quantitatively evaluating the change results. By introducing a dynamic context attention mechanism, the present invention enables the model to better integrate local and global information, thereby improving the ability to detect target change areas. Compared to traditional technologies that fail to fully consider the interrelationships between neighborhood features and that limit standard convolution operations to fixed local receptive fields, the present invention can flexibly adjust the shape and size of the convolution kernel, improving the model's ability to capture both large-scale global structures and small-scale detailed features.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a method for detecting changes in building remote sensing images based on dynamic context aggregation. Background Art

[0002] Remote sensing-based change detection aims to detect and identify differences in ground objects using two or more images taken at the same geographic location at different times. Remote sensing change detection provides an important technical tool for gaining a deeper understanding of various surface changes caused by natural and human activities. It has been widely applied in disaster monitoring, forest resource surveys, urban planning, agricultural production, and other fields. Over the past few decades, change detection technology has been extensively researched, and a variety of change detection algorithms have been developed. Traditional change detection methods, such as image differencing, background subtraction, and thresholding, are often based on handcrafted spatiotemporal features, identifying changes in categories through clustering or thresholding. However, with the emergence of massive, multimodal remote sensing data, the limitations of these methods in complex scenarios have become increasingly apparent. In particular, handcrafted features struggle to describe the essential differences between land classes, becoming a bottleneck in improving change detection accuracy. Deep learning methods, which adaptively learn from low-level visual cues to high-level semantic features, highlight the essential differences between land classes. In recent years, they have been widely introduced into remote sensing image change detection research, significantly improving the accuracy of remote sensing change detection. Existing change detection methods are primarily based on deep learning networks. These include early network models such as FC-EF, FC-Siam-conc, and FC-Siam-diff based on the U-Net architecture, as well as improved models such as DSIFN, which introduces an attention mechanism, and BIT, based on the Transformer architecture. In terms of feature extraction, existing technologies typically employ multi-scale feature fusion. For example, SNUNet maintains high-resolution information by establishing dense skip connections between the encoder and decoder, while UNet++ employs full-scale skip connections to learn low-level positional information and high-level semantic information. Regarding feature interaction, TFIM proposes a temporal feature interaction module to enhance the perception of changing areas, while A2Net improves the ability to capture detailed changes through gradual feature aggregation.

[0003] However, the current technology has the following deficiencies: (1) Insufficient multi-scale feature interaction. Existing methods based on multi-scale contextual interaction fail to fully consider the relationship between neighborhood features and lack an effective dynamic adjustment mechanism during feature learning. (2) Convolution operations are limited to local receptive fields. Traditional convolution operations use fixed-size convolution kernels for feature extraction, which makes it difficult to capture both large-scale global structural features and small-scale local detail features at the same time. When performing feature extraction, the shape and size of the convolution kernel in existing methods are often fixed, and cannot be adaptively adjusted according to target features of different scales, which affects the feature expression ability of the model. Summary of the Invention

[0004] To solve the above problems, the present invention provides a building remote sensing image change detection method based on dynamic context aggregation.

[0005] To implement the above technology, the steps are as follows:

[0006] S1. Build a data set and divide samples;

[0007] The dataset was obtained by augmenting the open source dataset;

[0008] The data augmentation used for open source datasets includes: geometric transformation and image enhancement;

[0009] Geometric changes include: random horizontal flip, random vertical flip, and random rotation;

[0010] Image enhancement includes: random adjustment of brightness, random adjustment of contrast, random adjustment of color and random addition of Gaussian noise;

[0011] The sample was divided in a ratio of 8:1:1;

[0012] S2, build DCA-NET network model;

[0013] DCA-NET consists of a twin network structure, including an encoder, a decoder, a fusion module, and a dynamic context attention (DCA) module. The DCA module includes an upper GCP layer and a lower DCK layer. The encoder includes four patch embedding layers and an encoder block. The encoder block includes a DCA module and a convolutional multilayer perceptron (MLP) unit.

[0014] The build steps are as follows:

[0015] S2.1. Input the image into the encoder for encoding. The steps are as follows:

[0016] S2.1.1. Input one image (H×W×3) from a pair of images in the dataset into the patch embedding layer in the encoder to split it into smaller patches;

[0017] S2.1.2. The image after the patch embedding layer enters the GCP layer in the DCA module for context fusion operation;

[0018] Specifically, the input image passes through the upper GCP layer in the DCA module, and the key (Key), query (Query) and value (Value) in the input two-dimensional feature map are defined as K = XW K ,Q=XW Q ,V=XW vUnlike the traditional self-attention mechanism that encodes each key independently through 1×1 convolution, the DCA module performs group convolution on all neighboring keys within a grid, thereby achieving contextual key representation. The learned contextual keys reflect the static contextual information between local neighboring keys and are regarded as the static contextual representation of the input.

[0019] The present invention concatenates the context key with the query and performs two consecutive 1×1 convolutions (W θ Using ReLU activation, W δ Without activation function) calculate the attention matrix, the expression is as follows:

[0020] A=[K 1 ,Q]W θ W δ

[0021] Where W θ is the weight matrix of the first 1×1 convolution, W δ is the weight matrix of the second 1×1 convolution; K 1 Represents a context key;

[0022] For each attention head, the local attention at each spatial position in the attention matrix A is learned jointly based on the query features and the key features of the context, rather than relying solely on isolated query and key pairs; in this way, the additional guidance of static contextual information enhances the learning ability of self-attention; then, using the contextual attention matrix A, all values ​​are aggregated to calculate the weighted feature map K 2 :

[0023] K 2 =V*A

[0024] Where * represents the weighted sum operation;

[0025] S2.1.3. The image after the patch embedding layer enters the GCK layer in the DCA module for variable convolution operation;

[0026] The sampling grid of regular convolution is centered at (0,0), while irregular convolution kernels do not have a fixed center at various sizes. Therefore, to adapt to different convolution kernel sizes, the algorithm of this invention defines the upper left corner (0,0) as the sampling origin. Based on this, the initial sampling coordinates of irregular convolution are determined, and the corresponding convolution operation at position F0 is defined as:

[0027] Conv(F0)=∑w×(F0+F n )

[0028] Where w represents the convolution parameter; F n Indicates offset at position F0;

[0029] The sampling position of standard convolution is fixed, which can only extract information within a local window and cannot capture contextual features in a wider range. Deformable convolution, on the other hand, compensates for this limitation to a certain extent by learning offsets and dynamically adjusting the sampling grid. Specifically, the sampling grids of standard and deformable convolutions are still regular and cannot adapt to convolution kernels of arbitrary shapes or numbers of parameters. To solve this problem, DCA introduces a more flexible variable convolution operation DCK layer. DCA's deformable convolution first generates an offset through the convolution layer, whose dimension is (B, 2N, H, W), where N represents the size of the convolution kernel.

[0030] S2.1.4. The DCP output and DCK output are concatenated in the channel dimension as the output of the DCA module, which reduces the number of feature channels and promotes the further fusion of information from different modules.

[0031] Specifically, by fusing static and dynamic context representations, the DCA module can form more discriminative feature representations, thereby improving the accuracy of change detection. The outputs of these two components are then merged in the channel dimension to achieve complementary advantages. Subsequently, the merged features are integrated through a 1x1 convolutional layer. This step not only reduces the number of feature channels but also promotes the further fusion of information from different modules.

[0032] S2.1.5. Downsample the output of the DCA module;

[0033] Specifically, the output of the DCA module is downsampled by 1 / 4 to obtain a Image The output of the DCA module is downsampled by 1 / 8 to obtain a Image The output of the DCA module is downsampled by 1 / 16 to obtain a Image The output of the DCA module is downsampled by 1 / 32 to obtain a Image

[0034] At the same time, another image is downsampled after the operations S2.1.1 to S2.1.4, including: performing a 1 / 4 downsampling operation to obtain a Image The output of the DCA module is downsampled by 1 / 8 to obtain a Image The output of the DCA module is downsampled by 1 / 16 to obtain a Image The output of the DCA module is downsampled by 1 / 32 to obtain a Image

[0035] S2.2, performing a fusion operation on the encoded image;

[0036] Specifically, images and images image and images image and images and images and images Perform image feature fusion to obtain multi-scale images and

[0037] S2.3. Input the fused image into the decoder for decoding;

[0038] Specifically, the multi-scale image and After connecting along the channel dimension, the network is input into a 1×1 convolutional layer for dimensionality reduction. The spatial resolution of the feature map is gradually improved through the transposed convolution operation. The result of the transposed convolution operation passes through two 3×3 convolutional layers and then undergoes the first element-by-element addition operation with the result of the transposed convolution operation. The result of the first element-by-element addition operation passes through the transposed convolution operation, passes through two 3×3 convolutional layers, and then undergoes the second element-by-element addition operation with the result of the transposed convolution operation. The result of the second element-by-element addition operation passes through a 3×3 convolutional layer to obtain the final output result.

[0039] S3, adjusting the parameters of the constructed network model;

[0040] S4, input the divided data set into the model for detection to obtain the change results;

[0041] The training set is input into the model for model training. After the trained model is input into the test set for testing, the validation set is input into the model for model verification. The verified model is used for change detection.

[0042] S5. Quantitatively evaluate the change results;

[0043] Quantitative evaluation methods include: F1-score (F1), Intersection over Union (IoU) and Overall Accuracy (OA);

[0044] Beneficial effects of the present invention:

[0045] To address the problems in existing technologies where the interrelationships between neighborhood features are not fully considered and standard convolution operations are limited to fixed local receptive fields, the present invention uses dual-time images (captured by the same type of sensor and with similar features) as input and adopts a weighted twin network structure to effectively extract image features. The model uses rich contextual information to optimize the learning of the dynamic attention matrix, further enhancing the ability to recognize changing areas. At the same time, a novel coordinate generation algorithm is used to define the initial positions of convolution kernels of different sizes, allowing the shape and size of the convolution kernel to be flexibly adjusted, thereby improving the model's ability to capture both large-scale global structures and small-scale detail features.

[0046] By introducing a dynamic contextual attention mechanism, the model can better integrate local and global information, thereby improving the ability to detect target changing areas; secondly, the proposed coordinate generation algorithm enables the initial position of the convolution kernel to be flexibly adjusted, enhancing the adaptability of the model to targets of different scales and shapes. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a flow chart of the steps of the present invention;

[0048] Figure 2 It is the overall technical roadmap of the embodiment of the present invention;

[0049] Figure 3 is a change detection network of the present invention;

[0050] Figure 4 It is a schematic diagram of the DCA structure of the present invention;

[0051] Figure 5 is a schematic diagram of an encoder block of the present invention;

[0052] Figure 6 It is the dynamic context attention module proposed by the present invention;

[0053] Figure 7 is a decoder used in the present invention;

[0054] Figure 8 Comparison of change detection results of different methods on the CDD-CD dataset;

[0055] Figure 9 Comparison of change detection results of different methods on the LEVIR-CD dataset. DETAILED DESCRIPTION

[0056] In order to enable those skilled in the art to better understand the present invention, the technical solution is now further described in detail in conjunction with the embodiments of the present invention and the accompanying drawings. It should be pointed out that the embodiments are only examples of part of the present invention, not all of them. All other embodiments that can be implemented by those of ordinary skill in the art based on the embodiments of the present invention without the need for creative work should be deemed to fall within the scope of protection of the present invention.

[0057] like Figure 1 and Figure 2 As shown, a dynamic context aggregation building remote sensing image change detection method includes the following steps:

[0058] S1. Build a data set and divide samples;

[0059] The dataset was obtained by augmenting the open source dataset;

[0060] The open-source LEVIR-CD dataset contains 637 pairs of dual-temporal remote sensing images acquired from Google Earth, with a spatial resolution of 0.5 m / pixel. The open-source CDD-CD dataset contains 11 pairs of remote sensing images with seasonal variations, of which 7 pairs have a resolution of 4725 × 2700 pixels and 4 pairs have a resolution of 1900 × 1000 pixels. To facilitate deep learning model training, the images in the open-source dataset were cut into 256 × 256 pixel blocks.

[0061] The data augmentation used for open source datasets includes: geometric transformation and image enhancement;

[0062] Geometric changes include: random horizontal flip, random vertical flip, and random rotation;

[0063] Image enhancement includes: random adjustment of brightness, random adjustment of contrast, random adjustment of color and random addition of Gaussian noise;

[0064] Data augmentation methods can effectively expand the training samples and improve the model's adaptability to various scenarios. During the training process, the above transformations are randomly applied to the input images, enabling the model to learn more robust feature representations.

[0065] The sample was divided in a ratio of 8:1:1;

[0066] The final results are shown in Table 1:

[0067] Table 1: Dataset division

[0068]

[0069] S2, such as Figure 3 As shown, the DCA-NET network model is constructed;

[0070] Change detection is a multi-scale, multi-layer feature learning problem. In remote sensing imagery, different types of changing objects (such as buildings, vegetation, and water bodies) often have diverse appearances and morphologies. To address the limitations of existing methods in capturing multi-scale and complex background changes, this paper proposes a Dynamic Contextual Attention Network (DCA-NET), which uses a dynamic context aggregation strategy to enhance the model's change detection capabilities.

[0071] like Figure 3 As shown in Figure 2, DCA-NET consists of a twin network structure, including: encoder, decoder, fusion module and dynamic context attention (DCA) module; the DCA module includes: upper GCP layer and lower DCK layer, as shown in Figure 2. Figure 4 As shown in Figure 2, the encoder includes four patch embedding layers and an encoder block. The encoder block includes a DCA module and a convolutional multilayer perceptron (MLP) unit, as shown in Figure 2. Figure 5 As shown;

[0072] This architecture design enhances the model's feature expression and detection accuracy in complex scenarios by combining static and dynamic context information and multimodal feature fusion;

[0073] The build steps are as follows:

[0074] S2.1. Input the image into the encoder for encoding. The steps are as follows:

[0075] S2.1.1. Input one image (H×W×3) from a pair of images in the dataset into the patch embedding layer of the encoder to split it into smaller patches for multi-scale processing.

[0076] S2.1.2, such as Figure 6 As shown in Figure 2, the image after the patch embedding layer enters the GCP layer in the DCA module for context fusion operation;

[0077] Specifically, the input image passes through the upper GCP layer in the DCA module, and the key (Key), query (Query) and value (Value) in the input two-dimensional feature map are defined as K = XW K ,Q=XW Q ,V=XW vUnlike the traditional self-attention mechanism that encodes each key independently through 1×1 convolution, the DCA module performs convolution on all neighboring keys in the grid through group convolution, thereby realizing contextual key representation. The learned context keys reflect the static context information between local neighboring keys and are regarded as the static context representation of the input. The present invention concatenates the context keys with the query and performs two consecutive 1×1 convolutions (W θ Using ReLU activation, W δ Without activation function) calculate the attention matrix, the expression is as follows:

[0078] A=[K 1 ,Q]W θ W δ

[0079] Where W θ is the weight matrix of the first 1×1 convolution, W δ is the weight matrix of the second 1×1 convolution; K 1 Represents a context key;

[0080] For each attention head, the local attention at each spatial position in the attention matrix A is learned jointly based on the query features and the key features of the context, rather than relying solely on isolated query and key pairs; in this way, the additional guidance of static contextual information enhances the learning ability of self-attention; then, using the contextual attention matrix A, all values ​​are aggregated to calculate the weighted feature map K 2 :

[0081] K 2 =V*A

[0082] Where * represents the weighted sum operation;

[0083] The weighted feature map captures the dynamic feature interactions between the input data, which we refer to as the dynamic context representation of the input data. Therefore, the output of the GCP layer is the result of fusing the static context and dynamic context through the attention mechanism, thereby enhancing the model's ability to express complex feature relationships. This mechanism enables the local attention matrix to be learned jointly based on the context key and query features, rather than relying solely on isolated query and key pairs, significantly enhancing the ability of the self-attention mechanism to handle feature interactions.

[0084] S2.1.3, such as Figure 6 As shown in Figure 2, the image after the patch embedding layer enters the GCK layer in the DCA module for variable convolution operation;

[0085] Convolutional neural networks (CNNs) are based on convolution operations, mapping features to corresponding locations using a regular sampling grid. Typically, the sampling grid for convolution is regular, but to adapt irregularly shaped convolution kernels to the sampling grid, DCA introduces an algorithm that supports convolution kernels of any size to generate the initial sampling coordinates of the convolution kernel. The algorithm first generates a regular sampling grid, then creates an irregular grid for the remaining sampling points, and finally concatenates the two to generate a complete sampling grid.

[0086] The sampling grid of regular convolution is centered at (0,0), while irregular convolution kernels do not have a fixed center at various sizes. Therefore, to adapt to different convolution kernel sizes, the algorithm of this invention defines the upper left corner (0,0) as the sampling origin. Based on this, the initial sampling coordinates of irregular convolution are determined, and the corresponding convolution operation at position F0 is defined as:

[0087] Conv(F0)=∑w×(F0+F n )

[0088] Where w represents the convolution parameter; F n Indicates offset at position F0;

[0089] The sampling position of standard convolution is fixed, and it can only extract information within a local window, but cannot capture contextual features in a larger range. Deformable convolution compensates for this limitation to a certain extent by learning offsets and dynamically adjusting the sampling grid. Specifically, the sampling grids of standard convolution and deformable convolution are still regular and cannot adapt to convolution kernels of arbitrary shapes or numbers of parameters. To solve this problem, DCA introduces a more flexible variable convolution operation DCK layer. DCA's deformable convolution first generates an offset through the convolution layer, whose dimension is (B, 2N, H, W), where N represents the size of the convolution kernel.

[0090] In this embodiment, N=3, the generated offset is added to the original sampling point coordinates to obtain a new sampling position; finally, the feature information at the position is obtained by interpolation and resampling;

[0091] This method effectively enhances the expressive power of the convolution kernel by flexibly adjusting the sampling position. It not only effectively enhances the flexibility of the convolution kernel, but also improves the feature extraction efficiency of the model at different scales and in complex backgrounds.

[0092] S2.1.4. The DCP output and DCK output are concatenated in the channel dimension as the output of the DCA module, which reduces the number of feature channels and promotes the further fusion of information from different modules.

[0093] Specifically, by fusing static and dynamic context representations, the DCA module can form more discriminative feature representations, thereby improving the accuracy of change detection. The outputs of these two components are then merged in the channel dimension to achieve complementary advantages. Subsequently, the merged features are integrated through a 1x1 convolutional layer. This step not only reduces the number of feature channels but also promotes the further fusion of information from different modules.

[0094] S2.1.5. Downsample the output of the DCA module;

[0095] Specifically, the output of the DCA module is downsampled by 1 / 4 to obtain a Image The output of the DCA module is downsampled by 1 / 8 to obtain a Image The output of the DCA module is downsampled by 1 / 16 to obtain a Image The output of the DCA module is downsampled by 1 / 32 to obtain a Image

[0096] At the same time, another image is downsampled after the operations S2.1.1 to S2.1.4, including: performing a 1 / 4 downsampling operation to obtain a Image The output of the DCA module is downsampled by 1 / 8 to obtain a Image The output of the DCA module is downsampled by 1 / 16 to obtain a Image The output of the DCA module is downsampled by 1 / 32 to obtain a Image

[0097] S2.2, performing a fusion operation on the encoded image;

[0098] Specifically, images and images image and images image and images and images and images Perform image feature fusion to obtain multi-scale images and

[0099] S2.3. Input the fused image into the decoder for decoding;

[0100] like Figure 7 As shown in Figure 2, the decoder of DCA-NET consists of multiple layers of convolution and transposed convolution operations to gradually restore high-resolution change detection maps.

[0101] The decoder input is the multi-scale features output by the encoder. These features are first concatenated along the channel dimension and reduced in dimension through a 1×1 convolutional layer to reduce computational complexity and prevent overfitting. The decoder uses transposed convolution operations to gradually increase the spatial resolution of the feature map and further optimizes the feature representation through a series of residual blocks to maintain consistency between local details and global structure. The residual blocks enhance the expressiveness of the feature map through skip connections without increasing the computational burden. Finally, a convolutional layer is applied to generate a prediction score consisting of two channels, where one channel predicts the unchanged area and the other predicts the changed area. The binary change map is obtained by using the Argmax operation along the channel dimension.

[0102] Specifically, the multi-scale image and After connecting along the channel dimension, the network is input into a 1×1 convolutional layer for dimensionality reduction. The spatial resolution of the feature map is gradually improved through the transposed convolution operation. The result of the transposed convolution operation passes through two 3×3 convolutional layers and then undergoes the first element-by-element addition operation with the result of the transposed convolution operation. The result of the first element-by-element addition operation passes through the transposed convolution operation, passes through two 3×3 convolutional layers, and then undergoes the second element-by-element addition operation with the result of the transposed convolution operation. The result of the second element-by-element addition operation passes through a 3×3 convolutional layer to obtain the final output result.

[0103] S3, adjusting the parameters of the constructed network model;

[0104] In this example, all experiments were implemented based on the PyTorch deep learning framework, and the experimental hardware platform used a computing environment equipped with a single NVIDIA Tesla T4 GPU. To improve the robustness and generalization ability of the model, the present invention adopted a variety of data augmentation strategies, including random flipping, proportional cropping, color jittering, and Gaussian blurring.

[0105] During training, the batch size is set to 6 and the initial learning rate is set to 3.5e -4 , and the linear decay strategy is used to dynamically adjust the learning rate until the end of training; the model optimization uses Pixel-wise Cross Entropy as the loss function, and the parameters are updated through the back propagation algorithm. The number of training iterations is 300;

[0106] S4, input the divided data set into the model for detection to obtain the change results;

[0107] The training set is input into the model for model training. After the trained model is input into the test set for testing, the validation set is input into the model for model verification. The verified model is used for change detection.

[0108] Figure 8 and Figure 9 The following is a comparison of the change detection results of different methods on the CDD-CD and LEVIR-CD datasets; from the visualization results, it can be observed that although each method can detect the approximate change area of ​​the building, there are significant differences in the details; among the compared methods, the FC-Siam-Conc and FC-Siam-Diff methods are greatly affected by pseudo-changes, resulting in a high missed detection rate; DTCDSCN has difficulty in effectively distinguishing important features in a complex noisy environment, and the detection area is incomplete; the BIT model has the problem of loss of mid- and bottom-level detail information during the upsampling process, and the detection area is incomplete. ChangeFormer still needs to be improved in terms of target integrity and boundary clarity, especially the detection effect on large-scale buildings is poor. In contrast, the DCA-Net proposed in this invention can effectively extract small-scale and large-scale spatial features through the dynamic context attention mechanism, making the building boundary outline more complete; at the same time, with the help of multi-scale feature fusion strategy, it significantly improves the false detection and missed detection phenomena;

[0109] S5. Quantitatively evaluate the change results;

[0110] Quantitative evaluation methods include: F1-score (F1), Intersection over Union (IoU) and Overall Accuracy (OA); the expressions are as follows:

[0111]

[0112] Where TP represents the number of samples correctly predicted as positive; TN represents the number of samples correctly predicted as negative; FP represents the number of samples incorrectly predicted as positive; FN represents the number of samples incorrectly predicted as negative; positive refers to pixels in the changed class; negative refers to pixels in the non-changed class.

[0113] The comparison of the evaluation results with other mainstream methods is shown in Table 2

[0114] Table 2: Performance comparison of different models on the LEVIR-CD dataset and CDD-CD dataset

[0115]

[0116] The present invention is compared and analyzed with other methods on the LEVIR-CD and CDD-CD datasets. FC-Siam-Di, FC-EF, FC-Siam-Conc, DTCDSCN, BIT, ChangeFormer, and ScratchFormer are all currently widely used methods. Table 2 shows the comparison results of each method on the datasets. On the LEVIR-CD dataset, the proposed method achieves an F1-score of 91.69%, an improvement of 1.29 percentage points over the currently superior ChangeFormer; IoU reaches 85.89%, an improvement of 3.41 percentage points; and OA reaches 99.11%, an improvement of 0.07 percentage points. On the CDD-CD dataset, the performance advantage is even more significant, with an F1-score of 95.5%, an improvement of 5.67 percentage points over ChangeFormer; IoU reaches 91.77%, an improvement of 10.24 percentage points; and OA reaches 99.12%, an improvement of 1.44 percentage points. Experimental results show that the proposed DCA model can effectively improve the accuracy of change detection.

[0117] In summary, the DCANet proposed in this patent effectively extracts both small- and large-scale spatial features through a dynamic contextual attention mechanism, making building boundary outlines more complete. Furthermore, a multi-scale feature fusion strategy significantly improves false positives and missed detections, demonstrating stable detection performance for buildings of varying sizes and densities. In a qualitative comparison of the two datasets, DCANet's detection results were closest to the true value labels, effectively addressing issues such as fragmented and incomplete detection results and fully demonstrating the superiority of the proposed method.

Claims

1. A method for detecting building remote sensing image changes using dynamic context aggregation, characterized by: The following steps are involved: S1. Build a data set and divide samples; The dataset was obtained by augmenting the open source dataset; The data augmentation used for open source datasets includes: geometric transformation and image enhancement; Geometric changes include: random horizontal flip, random vertical flip, and random rotation; Image enhancement includes: random adjustment of brightness, random adjustment of contrast, random adjustment of color and random addition of Gaussian noise; The sample was divided in a ratio of 8:1:1; S2, build DCA-NET network model; The DCA-NET comprises an encoder, a decoder, a fusion module, and a dynamic context attention module; wherein the dynamic context attention module comprises an upper GCP layer and a lower DCK layer; the encoder comprises four patch embedding layers and an encoder block; the encoder block comprises a DCA module and a convolutional multilayer perceptron unit; The steps for constructing the DCA-NET network model are as follows: S2.

1. Input the image into the encoder for encoding; S2.2, performing a fusion operation on the encoded image; S2.

3. Input the fused image into the decoder for decoding; The steps of the encoding operation are as follows: S2.1.

1. Input one image from a pair of images in the dataset into the patch embedding layer of the encoder to split it into small patches; S2.1.

2. The image after the patch embedding layer enters the GCP layer in the DCA module for context fusion operation; S2.1.

3. The image after the patch embedding layer enters the GCK layer in the DCA module for variable convolution operation; S2.1.

4. Concatenate the DCP output and the DCK output in the channel dimension as the output of the DCA module to reduce the number of feature channels. S2.1.

5. Downsample the output of the DCA module; The downsampling operations include: 1 / 4 downsampling, 1 / 8 downsampling, 1 / 16 downsampling and 1 / 32 downsampling; S3, adjusting the parameters of the constructed network model; S4, input the divided data set into the model for detection to obtain the change results; S5. Quantitatively evaluate the change results; Quantitative evaluation methods include: F1-score, intersection over union, and overall accuracy.

Images