Double-branch multi-scale fusion SAR flood intelligent detection method

Abstract

The invention provides a double-branch multi-scale fusion SAR flood intelligent detection method, and belongs to the technical field of remote sensing image processing. According to the technical scheme, the double-branch multi-scale fusion SAR flood intelligent detection method comprises the following steps that S1, an SAR image is processed, and an SAR flood detection data set is constructed; s2, constructing a dual-branch multi-scale fusion network comprising a twin encoder and a dual-branch decoder; s3, training and verifying the network by using the SAR flood detection data set to obtain a trained flood detection model; and S4, inputting a to-be-detected dual-time-phase SAR image pair into the flood detection model, and outputting a flood detection result map with the same size as the input image. The method has the beneficial effects that the double-branch multi-scale fusion network provided by the invention adaptively fuses global and local features and cooperates with multi-scale change information to realize accurate identification of a large-range submerging region and a small-scale fine region, and the flood detection precision is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image processing, and in particular to a dual-branch multi-scale fusion SAR flood intelligent detection method. Background Art

[0002] Floods are one of the most common natural disasters worldwide, significantly impacting human society and ecosystems. In flood monitoring, remote sensing technology, by comparing pre- and post-disaster imagery, can detect changes and quickly identify flooded areas and high-risk zones, providing critical support for accurately delineating the affected area. Synthetic aperture radar, with its cloud-penetrating and all-weather observation capabilities, is widely used in disaster assessment, emergency response, and risk prevention and control.

[0003] Traditional change detection methods can be categorized into three types: those based on transformation, algebraic operations, and classification. The first category enhances change signatures through techniques like change vector analysis and principal component analysis; the second category employs arithmetic strategies like ratio calculation, regression analysis, and difference normalization to identify changes; and the third category utilizes traditional machine learning models like support vector machines and K-means clustering for pixel-level classification. However, these methods rely heavily on artificial feature design and generally suffer from limited generalization capabilities.

[0004] With the continuous development of technologies such as deep learning and computer vision, change detection is becoming a research hotspot at the intersection of remote sensing and computer vision. Unlike traditional methods, deep learning-based approaches can integrate feature extraction and change detection into a single network, directly producing detection results in an end-to-end manner. Among deep learning change detection models, convolutional neural networks (CNNs) have become the mainstream architecture due to their powerful image processing capabilities. Frameworks such as FC-EF, FC-Siam-conc, and FC-Siam-diff, and their variants, have been widely used in change detection, achieving promising results. However, most current methods focus on multi-scale feature extraction. In fusing multi-scale information, they simply reuse skip connections to combine low-level features in the encoder with high-level features in the decoder, ignoring the interrelationships between information at each scale. This leads to insufficient utilization of multi-scale features and difficulty in identifying flooded areas of varying sizes. Furthermore, these research frameworks often process semantic and difference information independently, ignoring the correlation between the two.

[0005] Floods typically occur over large areas. CNN-based methods are limited by the size of the convolution kernel, making them ineffective at extracting large-scale floods. Transformer-based methods leverage self-attention to capture global information and are widely used in change detection tasks. Methods such as BIT, ChangeFormer, and DAM-Net have all achieved good results. However, the application of Transformers requires large-scale training samples and substantial computing resources and memory. Furthermore, compared to CNNs, pixel-level feature extraction struggles to focus on local information, making it difficult to detect small areas of change and poorly detecting change boundaries.

[0006] To address the above challenges, the present invention aims to provide a dual-branch multi-scale fusion SAR flood intelligent detection method to provide important insights for rescue planning and post-disaster assessment, thereby supporting disaster relief and post-disaster recovery efforts. Summary of the Invention

[0007] The purpose of the present invention is to overcome the problems in the background technology and provide a dual-branch multi-scale fusion SAR flood intelligent detection method.

[0008] The technical concept of this invention is as follows: The SAR image flood detection model consists of a twin encoder and a dual-branch decoder. The encoder utilizes the first three stages of ResNet18 as the network backbone and adjusts the stride of the first 7×7 convolutional layer from the default value of 2 to 1. This extracts multi-level features while retaining more spatial detail and reducing computational complexity, thus better suiting the requirements of change detection tasks. The Global Local Adaptation Module (GLAM) then dynamically fuses global and local information using adaptive parameters α and β. This adaptive fusion mechanism allows the model to prioritize change-related information while effectively suppressing irrelevant details. In the decoder, the semantic branch utilizes a 3D-Attention module to sequentially interact with bi-temporal features and enhance temporal, channel, and spatial features. The difference branch extracts difference information using element-wise subtraction. A ResidualBlock is then applied to process difference and semantic information. Subsequently, the Scale Information Aggregation Module (SIAM) is used to fuse the multi-level features and identify flood areas of different shapes in both the semantic and difference branches. Finally, the dual-branch fusion produces the final flood detection result.

[0009] In order to achieve the above-mentioned object of the invention, the present invention adopts a technical solution specifically as follows: a dual-branch multi-scale fusion SAR flood intelligent detection method, comprising the following steps:

[0010] S1, processes SAR images and constructs SAR flood detection dataset;

[0011] S2, construct a dual-branch multi-scale fusion network consisting of a twin encoder and a dual-branch decoder;

[0012] S3, using the SAR flood detection dataset to train and verify the network to obtain a trained flood detection model;

[0013] S4, inputting the dual-phase SAR image pair to be detected into the flood detection model, and outputting a flood detection result image with the same size as the input image.

[0014] Furthermore, in step S1, constructing a SAR flood detection dataset includes:

[0015] S1.1, select SAR image pairs with uniform resolution and complete spatial registration;

[0016] S1.2, perform pixel-level annotation on the registered SAR image pairs, with annotation types including flood-changed areas and non-changed areas;

[0017] S1.3, perform synchronous cropping on the SAR image pair to obtain sub-image pairs with consistent sizes and discard images with insufficient size;

[0018] S1.4, screen samples with a change area ratio ≥ 3.5% and construct a standard dataset consisting of a training set and a validation set.

[0019] Furthermore, the dual-branch multi-scale fusion network includes:

[0020] S2.1, construct a dual-branch multi-scale fusion network, where:

[0021] The GLAM module in the twin encoder dynamically fuses global and local information through adaptive parameters α and β;

[0022] The dual-branch decoder contains:

[0023] The 3D-Attention module in the dual-branch decoder is used to enhance the spatiotemporal channel interaction of bi-phase features;

[0024] The SIAM module in the dual-branch decoder is used to fuse multi-level features and identify flood areas in the semantic branch and the difference branch;

[0025] S2.2: Based on step S2.1, set the parameter values ​​required for network training according to the SAR flood detection training dataset, including the initial learning rate, optimizer, and loss function.

[0026] Furthermore, in step S3, the training and verification include:

[0027] S3.1, set the maximum number of training iterations for the dual-branch multi-scale fusion network;

[0028] S3.1, input the dataset into the dual-branch multi-scale fusion network constructed in step 2 for model training and verification, and optimize the network parameters during the training process;

[0029] S3.1, after network training is completed, save the network parameters with the best performance on the validation set.

[0030] Furthermore, in step S4, the processing of the SAR image to be detected includes:

[0031] S4.1, obtain a pair of SAR images to be tested with the same resolution and number of polarization channels as the training dataset;

[0032] S4.2, perform synchronous cropping and edge padding according to the training data size to construct the test dataset;

[0033] S4.3, input the dual-temporal image pairs in the SAR flood detection test dataset into the trained dual-branch multi-scale fusion network to obtain pixel-level detection results;

[0034] S4.4, the detection results of the SAR flood detection test dataset are stitched according to the cropping method in step S4.2, and the padded parts of the image pairs with insufficient size in step S4.2 are deleted, and finally a complete SAR flood detection result is obtained.

[0035] Compared with existing technologies, the present invention offers the following advantages: It proposes a dual-branch, multi-scale fusion method for intelligent SAR flood detection—a dual-branch, multi-scale fusion network. The GLAM in the network dynamically coordinates the self-attention mechanism and dilated convolution design by combining adaptive factors. The self-attention mechanism enables the model to focus on large-scale flood-affected areas, while the dilated convolution focuses on smaller flooded areas, thereby achieving more effective and accurate integration of global and local contextual information. The 3D-Attention module in the present invention utilizes the attention mechanism and interactively adjusts multidimensional features in the temporal, channel, and spatial dimensions, enabling the model to gradually acquire more refined multidimensional features. The SIAM in the present invention utilizes grouped and multi-scale convolution designs, which not only reduces feature redundancy but also leverages multi-level variation information to effectively detect irregular flood areas. The dual-branch, multi-scale fusion network not only adaptively fuses global and local information but also effectively combines multi-scale variation features derived from difference and semantic information, achieving accurate detection of both large-scale flooded areas and small, fragmented areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0037] Figure 1 is a flow chart of a method provided by an embodiment of the present invention;

[0038] Figure 2 This is a diagram of a dual-branch multi-scale fusion network structure provided by an embodiment of the present invention;

[0039] Figure 3 This is a diagram of the GLAM structure provided by an embodiment of the present invention;

[0040] Figure 4 This is a structural diagram of a 3D-Attention module provided by an embodiment of the present invention;

[0041] Figure 5 1 is a SAIM structure diagram provided by an embodiment of the present invention;

[0042] Figure 6 This is a visualization of flood detection results for a portion of the test set provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. Of course, the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0044] Example: Figure 1 As shown, a dual-branch multi-scale fusion SAR flood intelligent detection method includes the following steps:

[0045] Step 1: Create a SAR flood detection dataset;

[0046] Specifically, step 1 includes the following sub-steps:

[0047] Step 1.1: Select SAR image pairs with uniform resolution and quality that meet the requirements and complete the spatial registration operation;

[0048] Step 1.2: Perform pixel-level annotation on the registered SAR image pair to clearly divide the pixels in the SAR image pair into changed and non-changed areas;

[0049] Step 1.3: Perform synchronous cropping on the SAR image pairs to obtain several pairs of sub-SAR images with the same length and width. Images with insufficient length or width are discarded.

[0050] Step 1.4: Arrange the SAR image pair files and the corresponding change label files, select valid samples with a change area ratio of ≥ 3.5%, and construct a SAR flood detection standard dataset consisting of a training set and a validation set;

[0051] Step 2: Design a dual-branch multi-scale fusion network based on the deep learning framework;

[0052] Specifically, step 2 includes the following sub-steps:

[0053] Step 2.1: The dual-branch multi-scale fusion network designed by the present invention adopts an encoder-decoder structure, such as Figure 2 As shown. The encoder part uses the first three stages of ResNet18 as the backbone, and adjusts the stride of the first 7×7 convolutional layer from 2 to 1 to extract multi-scale features. GLAM uses adaptive parameters α and β to dynamically fuse global and local information. The decoder part adopts a dual-branch decoder. The semantic branch interacts with the dual-phase features through the 3D-Attention module to enhance the temporal, channel and spatial features; the difference branch extracts difference information through element subtraction. Subsequently, ResidualBlock processes the difference and semantic information, and SIAM fuses multi-level features and identifies flood areas of different shapes. Finally, the dual-branch fusion obtains the final flood detection result. The dual-phase SAR image pair is input at the input of the network, and the flood detection result of the same size as the input SAR image is obtained at the output. Before training, it is necessary to set the input of the dual-branch multi-scale fusion network, and set the input size and output size of the dual-branch multi-scale fusion network according to the length, width and number of polarization channels of the images in the SAR image pair dataset;

[0054] In order to effectively perceive large-scale and small-scale flood areas, the present invention designs GLAM using the self-attention mechanism and dilated convolution with different dilation rates, such as Figure 3 As shown in the figure. This module can better adaptively couple the dual-phase features by sharing parameters in the twin network, thereby achieving effective fusion of global and local information to reduce the interference of redundant information on change detection. The GLAM operation formula is as follows:

[0055] Q=F in W Q ,K=F in W K ,V=F in W V (1)

[0056] F G =Softmax(Q·K T )·V

[0057]

[0058]

[0059] Where, F in is the input feature, W Q 、W K and W V is a learnable weight matrix, Softmax(·) is the activation function, represents a convolution operation with a kernel size of 3×3 and a dilation rate of r. Concat(·) is a concatenation operation. F out is the output feature, α and β are learnable weight parameters used to balance the global feature F G and local features F L contribution.

[0060] In order to enhance the model's ability to represent water body characteristics, the present invention designs a 3D-Attention module in the form of an attention mechanism, such as Figure 4 As shown in Figure 2. This module connects the temporal, channel, and spatial attention mechanisms in series, enabling the model to interact with information across different dimensions, thereby extracting richer multidimensional feature representations. The 3D-Attention module operation formula is as follows:

[0061]

[0062]

[0063]

[0064] Where, and is the input bi-temporal feature, R(·) represents the reshaping operation, MLP(·) is the multi-layer perceptron, AvgPool(·) and MaxPool(·) represent the average pooling and maximum pooling operations respectively, F T and F C are the features after time information enhancement and channel information enhancement, F 3D It is the output feature after information interaction between different dimensions.

[0065] In order to effectively capture the irregular change characteristics of the broken flood area, the present invention uses group convolution and multi-scale convolution to design SIAM, such as Figure 5 As shown in Figure 2. This module not only reduces feature redundancy but also utilizes multi-scale change information to effectively perceive irregular flood areas. The SIAM operation formula is as follows:

[0066]

[0067]

[0068] In the formula, Chunk(·) represents grouping, Up(·) represents bilinear interpolation upsampling, and f k,i represents the kth group of features at the i-th scale, f k Represents the kth group of features after grouping, i∈[1, 2, 3], k∈[1, 2, 3, 4], Denotes a double convolution operation, each of which includes a convolution operation with a kernel size of 3×3, a dilation rate of r, normalization, and a ReLU activation function. agg It is the result of multi-scale change information fusion.

[0069] Step 2.2: Based on step 2.1, set the parameter values ​​required for network training according to the SAR flood detection training dataset, including the initial learning rate, optimizer, and loss function;

[0070] Step 3: Use the SAR flood detection dataset to train and optimize the dual-branch multi-scale fusion network;

[0071] Specifically, step 3 includes the following sub-steps:

[0072] Step 3.1: Set the maximum number of training iterations for the dual-branch multi-scale fusion network;

[0073] Step 3.2: Input the SAR flood detection dataset into the dual-branch multi-scale fusion network constructed in step 2 for model training and verification. During the training process, the dual-branch multi-scale fusion network will continuously optimize the network parameters to achieve the best learning effect;

[0074] Step 3.3: After network training is completed, save the network parameters that perform best in the validation set for model testing;

[0075] Step 4: Use the trained dual-branch multi-scale fusion network for flood detection.

[0076] Specifically, step 4 includes the following sub-steps:

[0077] Step 4.1: Obtain the SAR image pair to be detected. The resolution and number of polarization channels of the SAR image pair must be the same as those of the SAR flood detection training dataset.

[0078] Step 4.2: Synchronously crop the SAR image pairs according to the length and width of the images in the SAR image training dataset and sort them to obtain several pairs of sub-SAR images that meet the input requirements of the dual-branch multi-scale fusion network. Fill in the SAR images whose length and width are insufficient to meet the input requirements, and finally form the SAR flood detection test dataset;

[0079] Step 4.3: Input the dual-temporal image pairs in the SAR flood detection test dataset into the trained dual-branch multi-scale fusion network;

[0080] Step 4.4: Obtain the SAR flood detection results, i.e., the pixel-level detection results of the SAR flood detection test dataset;

[0081] Step 4.5: Perform image stitching on the detection results of the SAR flood detection test dataset according to the cropping method in step 4.2, and delete the padded parts of the image pairs with insufficient size in step 4.2 to finally obtain the complete SAR flood detection results.

[0082] The effects of the embodiments of the present invention can be further verified by the following experiments:

[0083] 1) Experimental environment

[0084] The proposed model is implemented on PyTorch and trained using a single NVIDIA RTX4090 GPU.

[0085] 2) Experimental content

[0086] The method of the present invention, a dual-branch multi-scale fusion SAR flood intelligent detection method, was used to perform flood detection on a SAR flood detection dataset. To verify the advantages of this method in flood detection, we compared it with different change detection methods.

[0087] 3) Accuracy evaluation

[0088] In order to quantitatively evaluate the effect of the present invention, the present invention selects overall accuracy (OA), F1 score (F1), kappa coefficient (KC) and intersection over union (IoU) as evaluation indicators.

[0089] Table 1 Accuracy evaluation of flood detection results of the method of the present invention and other methods

[0090]

[0091] 4) Experimental results analysis

[0092] Figure 6 The experimental results in Table 1 show that compared with other methods, the dual-branch multi-scale fusion network proposed in this paper achieves flood detection results that are closer to the ground truth map of flooded areas, effectively distinguishes flooded from non-flooded areas, significantly reduces false alarms and missed detections, and can effectively detect small flooded areas. Its quantitative evaluation indicators also outperform by a significant margin. Therefore, the proposed method significantly improves flood detection performance in multiple aspects.

[0093] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A dual-branch multi-scale fusion SAR flood intelligent detection method, characterized by: The following steps are involved: S1, processes SAR images and constructs SAR flood detection dataset; S2, construct a dual-branch multi-scale fusion network consisting of a twin encoder and a dual-branch decoder; S3, using the SAR flood detection dataset to train and verify the network to obtain a trained flood detection model; S4, inputting the dual-phase SAR image pair to be detected into the flood detection model, and outputting a flood detection result image with the same size as the input image.

2. The dual-branch multi-scale fusion SAR flood intelligent detection method according to claim 1 is characterized in that: In step S1, constructing a SAR flood detection dataset includes: S1.1, select SAR image pairs with uniform resolution and complete spatial registration; S1.2, perform pixel-level annotation on the registered SAR image pairs, with annotation types including flood-changed areas and non-changed areas; S1.3, perform synchronous cropping on the SAR image pair to obtain sub-image pairs with consistent sizes and discard images with insufficient size; S1.4, screen samples with a change area ratio ≥ 3.5% and construct a standard dataset consisting of a training set and a validation set.

3. The dual-branch multi-scale fusion SAR flood intelligent detection method according to claim 1 is characterized in that: In step S2, the dual-branch multi-scale fusion network includes: S2.1, construct a dual-branch multi-scale fusion network, where: The GLAM module in the twin encoder dynamically fuses global and local information through adaptive parameters α and β; The dual-branch decoder contains: The 3D-Attention module in the dual-branch decoder is used to enhance the spatiotemporal channel interaction of bi-phase features; The SIAM module in the dual-branch decoder is used to fuse multi-level features and identify flood areas in the semantic branch and the difference branch; S2.2: Based on step S2.1, set the parameter values ​​required for network training according to the SAR flood detection training dataset, including the initial learning rate, optimizer, and loss function.

4. The dual-branch multi-scale fusion SAR flood intelligent detection method according to claim 1 is characterized in that: In step S3, the training and verification includes: S3.1, set the maximum number of training iterations for the dual-branch multi-scale fusion network; S3.1, input the dataset into the dual-branch multi-scale fusion network constructed in step 2 for model training and verification, and optimize the network parameters during the training process; S3.1, after network training is completed, save the network parameters with the best performance on the validation set.

5. The dual-branch multi-scale fusion SAR flood intelligent detection method according to claim 1 is characterized in that: In step S4, the processing of the SAR image to be detected includes: S4.1, obtain a pair of SAR images to be tested with the same resolution and number of polarization channels as the training dataset; S4.2, perform synchronous cropping and edge padding according to the training data size to construct the test dataset; S4.3, input the dual-temporal image pairs in the SAR flood detection test dataset into the trained dual-branch multi-scale fusion network to obtain pixel-level detection results; S4.4, the detection results of the SAR flood detection test dataset are stitched according to the cropping method in step S4.2, and the padded parts of the image pairs with insufficient size in step S4.2 are deleted, and finally a complete SAR flood detection result is obtained.

Images