Collaborative optimization method for speckle noise suppression and structure preservation in SAR images

Abstract

The present invention discloses a collaborative optimization system and method for SAR image speckle noise suppression and structure preservation. The system consists of a structure extraction module based on an edge detection network, a feature fusion module based on convolution, and a speckle noise suppression module based on a bilateral residual block network. The system is stored in a 1B-level signal processor. The method first converts the original SAR image with speckle noise into a i The image structure information XS extracted by inputting into the Stu module i ; Then X i and XS i The feature fusion is performed through convolution and attention mechanism to obtain the feature fusion information UX i ;Finally UX i After despeckling using the Dep module, the resulting SAR image exhibits a collaborative optimization of speckle noise suppression and structure preservation. This method leverages the coupling between the speckle noise suppression module and the structure extraction module. Guided by structural information, this method not only removes speckle but also better preserves the edges and texture details of the SAR image.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image processing, and is a collaborative optimization method for suppressing speckle noise and preserving structure of Synthetic Aperture Radar (SAR) images based on deep learning. Background Art

[0002] As an advanced imaging technology, SAR systems are capable of acquiring high-resolution images of the Earth's surface in all weather conditions, making them widely used in military surveillance, environmental monitoring, geological exploration, and other fields. However, during the SAR imaging process, factors such as the Doppler effect, beam interference, and surface scattering properties often produce a unique noise phenomenon called speckle noise. This noise, caused by the coherence of the radar signal, manifests as randomly distributed high-frequency speckle noise in the image, severely impacting the visual quality of SAR images and subsequent image analysis.

[0003] According to the book Radar Imaging Technology edited by Liu Yongtan, first edition, published in March 2014 by Harbin Institute of Technology Press, pages 109, 113-114, Synthetic Aperture Radar (SAR) is installed on a moving platform and transmits and receives pulses at a certain repetition frequency to form echo signals. The structural block diagram of the SAR system is shown in the figure below. Figure 1 As shown, the SAR system consists of an onboard radar system, a satellite platform and data downlink system, and a ground system. Synthetic aperture radar imaging processing is performed in the ground system. The ground system receives echo signals from the satellite platform and data downlink system via a ground receiving station. These echo signals are processed by a SAR signal processor to produce SAR images, which are then stored in a backup file operation system.

[0004] The presence of speckle noise poses a significant challenge to SAR image processing. Traditional denoising methods, such as mean filtering, median filtering, and wavelet transforms, while able to suppress speckle noise to a certain extent, often result in the loss of detailed image information, particularly around edges and textures. Because structural information in SAR images is crucial for tasks such as object recognition and target detection, effectively removing speckle noise while preserving image edges and detail has become a pressing issue in SAR image processing.

[0005] With the rapid development of deep learning technology in recent years, researchers have proposed a series of deep learning-based despeckling methods for SAR images. These methods leverage the powerful feature extraction capabilities of neural networks to automatically learn complex patterns in images, thereby removing noise while better preserving the image's structural information. However, existing methods still face significant challenges when dealing with heterogeneous areas (such as strong scatterers like buildings and vehicles), which can easily lead to blurred edges or loss of detail. Summary of the Invention

[0006] Speckle noise is a significant problem in SAR image processing because it destroys texture details, thereby limiting the accuracy of its high-precision applications. The impact of speckle noise on SAR images is related to their structural characteristics. Images are usually divided into homogeneous regions, inhomogeneous regions, and extremely inhomogeneous regions. In the latter two regions, structures such as edges and strong point targets will produce more complex speckle distributions than in homogeneous regions. The despeckling filter based on the multiplicative model performs well in homogeneous regions, but performs poorly in the other two regions. In order to solve the problem of loss of image structural information when suppressing speckle noise, the present invention proposes a structure-preserving SAR image despeckling collaborative optimization method, which utilizes the coupling between the despeckling module and the structure extraction module. Guided by structural information, the proposed method can better preserve edge and texture details while removing speckles.

[0007] The present invention provides a collaborative optimization system for SAR image speckle noise suppression and structure preservation, which consists of a structure extraction module based on an edge detection network, a feature fusion module based on convolution, and a speckle noise suppression module based on a bilateral residual block network;

[0008] The Stu module first extracts image features of different scales through the encoder block to generate a multi-scale structure map. Second, it uses the upsampling block to improve the feature map resolution and generate a refined structure map. Third, it uses the structure fusion layer to fuse the intermediate structure maps generated by all upsampling blocks and generate the final structure extraction feature map through the learned filter.

[0009] The Fus module processes SAR images through convolution operations to generate preliminary SAR feature maps. It also processes SAR images through convolution operations to generate attention matrices. It also performs weighted fusion of preliminary SAR feature maps and attention matrices. Finally, it generates higher-quality fused features through convolution fusion.

[0010] The Dep module first extracts image features preliminarily through an input layer consisting of a convolutional layer, a batch normalization layer, and a RELU activation function; secondly, it suppresses speckle noise through a dual residual block consisting of two convolution operations with different receptive fields, while achieving global feature extraction and local detail preservation; thirdly, it maps the denoising result back to the image through an output layer consisting of a convolutional layer and a hyperbolic tangent activation function, and generates the final speckle noise suppression result.

[0011] The collaborative optimization system for SAR image speckle noise suppression and structure preservation of the present invention is applied to prevent the loss of image structure information caused by suppressing speckle noise. The collaborative optimization comprises the following steps:

[0012] Step 1: Obtain the SAR image to be processed;

[0013] The Stu-Fus-Dep collaborative model obtains any single-view amplitude image from the 1B-level signal processor, that is, the SAR image with speckle noise, which is recorded as the image to be processed X i ;

[0014] Step 2: Apply Stu network to extract feature information of digital binary image;

[0015] Step 21, processing the SAR image in the Stu module;

[0016] The Stu network includes an encoder block, an upsampling block, and a structure fusion layer based on the DexiNed architecture;

[0017] The encoder block is used to extract X at different scales i The characteristics of i ;

[0018] The upsampling block receives AX i , use convolution and transposed convolution to generate an intermediate structure map, denoted as BX i ;

[0019] The structure fusion layer maps the intermediate structure BX generated by all upsampling blocks i Fusion is performed to generate the final structure extraction feature map, denoted as CX i ;

[0020] Step 22, obtaining digitized binary image feature information;

[0021] According to the trained Stu model F Stu CX i Processing is performed to obtain the binary structure label image XS i ;

[0022] Step 3: Apply the Fus module to complete feature fusion based on convolution and attention;

[0023] Step 31, feature extraction;

[0024] X i Through the convolution operation of the 3×3 convolution layer, the preliminary feature map of the image X is generated. i ';

[0025] Step 32, generate an attention matrix;

[0026] XS i Generate the spatial attention matrix XS through a 1×1 convolution operation i ';

[0027] Step 33, feature weighting;

[0028] X i ' and XS i 'Multiply element by element to form a weighted feature map XXS i ';

[0029] Step 34, second weighted fusion;

[0030] XXS i ′Fuse again through the convolution layer to obtain the weighted fusion output feature information, recorded as UX i ;

[0031] Step 4: Speckle noise suppression is completed through Dep;

[0032] Step 41, processing the feature fusion information in the Dep module;

[0033] The input layer initially extracts UX i Features, denoted as UX i ';

[0034] Use the first 6 double residual blocks connected to UX i 'Processing, get the denoising result RX i ;

[0035] The output layer will RX i Mapped back to the image space, denoted as OX i ;

[0036] Step 42, obtaining speckle noise suppression results;

[0037] According to F Dep OX i Processing is performed to obtain the speckle noise suppression result X o ;

[0038] The F Dep It comes from the output of the trained Dep module.

[0039] Technical effects of the present invention:

[0040] ① Efficient removal of speckle noise: The present invention uses a bilateral residual block network to suppress speckle noise, which significantly reduces noise interference, improves the clarity and readability of SAR images, and makes subsequent image analysis more accurate and reliable.

[0041] ② Structural Information Preservation Enhancement: By introducing a structure extraction network, the present invention can preserve edges and details in the image while denoising. This structure-preserving function is particularly important for tasks such as object recognition and target detection in SAR images, effectively improving the practicality of the images.

[0042] ③ Reduce edge blur: The collaborative optimization framework effectively avoids the edge blurring common in traditional denoising methods by combining structure extraction with denoising modules. Experiments show that edge regions in denoised images remain sharp, ensuring image resolution in areas with strong scatterers and fine details.

[0043] ④ Construction of a highly adaptable dataset: Using echo simulation and SAR imaging processing technology, a training dataset based on SAR images was constructed, enabling the algorithm to cope with complex ground scattering environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is the block diagram of the SAR platform system structure.

[0045] Figure 2 This is a flow chart of the collaborative optimization method for SAR image speckle noise suppression and structure preservation proposed in the present invention.

[0046] Figure 3 This is the loss function design of the collaborative optimization network for SAR image speckle noise suppression and structure preservation proposed in the present invention.

[0047] Figure 4 This is an example of a SAR image from the Stu network dataset.

[0048] Figure 5 It is from Figure 4 The structural information extracted is an example of a labeled image from the Stu network dataset.

[0049] Figure 6 This is an example of a labeled image from the Dep network dataset.

[0050] Figure 7 is Figure 6 The simulated image generated by the echo simulation and imaging processing technology of the SAR system is an example of the speckle noise suppression network dataset image.

[0051] Figure 8 This is the SAR image effect before speckle noise suppression.

[0052] Figure 9 yes Figure 8 SAR image results after speckle noise suppression. DETAILED DESCRIPTION

[0053] The present invention will be further described in detail below with reference to the accompanying drawings and examples. The parameters listed are merely exemplary embodiments of the present invention. It should be noted that those skilled in the art may make various improvements and modifications without departing from the principles of the present invention, and such improvements and modifications are also within the scope of protection of the present invention.

[0054] The present invention relates to a collaborative optimization method for SAR image speckle noise suppression and structure preservation, and the method aims to solve the problem in the prior art that image structural information is lost when suppressing speckle noise. The collaborative optimization of SAR image speckle noise suppression and structure preservation of the present invention consists of a structure extraction module based on an edge detection network, a feature fusion module based on convolution, and a speckle noise suppression module based on a bilateral residual block network, referred to as the Stu-Fus-Dep collaborative model, and the Stu-Fus-Dep collaborative model is stored in a Class 1B signal processor. The SAR image speckle noise suppression based on the collaborative optimization of the present method combines the Dep model and the Stu module in deep learning. Through the collaborative optimization of the two, the edges and texture details in the SAR image are effectively preserved while removing the speckle noise.

[0055] The speckle noise suppression module based on the bilateral residual block network (i.e., Dep module) refers to X.Liu, M.Suganuma, Z.Sun, and T.Okatani, “Dual Residual Networks Leveraging the Potential of Paired Operations for Image Restoration,” in Proc.IEEE CVPR, 2019, pp.7000-7009.

[0056] The edge detection network-based structure extraction module (i.e., Stu module) is based on X. Soria, E. Riba, and A. Sappa, "Dense Extreme Inception Network: Towards a Robust CNN Model for Edge Deception," in Proc. IEEE WACV, Snowmass, CO, USA, 2020, pp. 1912–1921. It is referred to as the DexiNed architecture.

[0057] Structure extraction module based on edge detection network (Stu module)

[0058] In the present invention, the Stu module first extracts image features of different scales through the encoder block to generate a multi-scale structure map; secondly, it improves the feature map resolution through the upsampling block to generate a refined structure map; thirdly, it fuses the intermediate structure maps generated by all upsampling blocks through the structure fusion layer, and generates the final structure extraction feature map through the learned filter.

[0059] Convolution-based feature fusion module (Fus module)

[0060] In the present invention, the Fus module first processes the SAR image through convolution operation to generate a preliminary SAR feature map; secondly, it processes the SAR image through convolution operation to generate an attention matrix; thirdly, the preliminary SAR feature map and the attention matrix are weightedly fused; fourthly, higher quality fusion features are generated through convolution fusion.

[0061] Speckle noise suppression module (Dep module) based on bilateral residual block network

[0062] In the present invention, the Dep module firstly extracts image features through an input layer composed of a convolution layer, a batch normalization layer and a RELU activation function; secondly, it suppresses speckle noise through a dual residual block composed of convolution operations with two different receptive fields, while achieving global feature extraction and local detail preservation; thirdly, it maps the denoising result back to the image through an output layer composed of a convolution layer and a hyperbolic tangent activation function, and generates the final speckle noise suppression result.

[0063] In the present invention, the SAR image with speckle noise is recorded as the original SAR image.

[0064] The processing flow of the present invention is as follows Figure 2 First, the original SAR image X with speckle noise is i Input into the structure extraction module based on edge detection network, the extracted image structure information is recorded as XS i ; Then, the image to be processed X i and XS i Feature fusion is performed through convolution and attention mechanism, and the information after feature fusion is recorded as UX i Finally, UX i After despeckled by the speckle noise suppression module based on the bilateral residual block network, the obtained SAR image with the coordinated optimization of speckle noise suppression and structure preservation is denoted as X o .

[0065] Construction of structure extraction module based on edge detection network, namely Stu model

[0066] Construction step A, obtain training images;

[0067] A number of SAR single-view complex images with spots are selected from the 1B signal processor as training images, which are recorded as the training image set TRIM = {image1, image2, ..., image a ,…,image A}.

[0068] image1 represents the first training image.

[0069] image2 represents the second training image.

[0070] image a Represents the a-th training image.

[0071] image A Represents the last training image.

[0072] For the sake of convenience, image a Also called any training image. a The pixel matrix is ​​represented as

[0073] N a is the total number of pixels in the azimuth direction.

[0074] N r is the total number of distance pixels.

[0075] χ=1,2,3,…,N a , represents the xth pixel point in the azimuth direction.

[0076] γ=1,2,3,…,N r , represents the γth pixel point in the distance direction.

[0077] trx χ,γ Represents the pixel at the xth row and γth column of the training image.

[0078] trx 1,γ Indicates the Nth row of the training image r Column of pixels.

[0079] trx χ,1 Represents the Nth training image a The pixel in row and column 1.

[0080] Represents the Nth training image aRow N r The pixel in the column is also the last pixel in the training image.

[0081] Construction step B, distinction between structure area and background area;

[0082] With the help of auxiliary information of optical remote sensing images in online maps, according to the characteristics of SAR images, TRIM={image1,image2,…,image a ,…,image A}, the pixels in the structure area to be retained are marked as 1, and the pixels in the background area are marked as 0, and the training binary label image set is obtained, which is recorded as TRI, and TRI={ri1,ri2,…,ri a ,…,ri A}.

[0083] ri1 represents the training binary label image after the training image image1 is distinguished between the structure area and the background area.

[0084] ri2 represents the training binary label image after the training image image2 is distinguished between the structure area and the background area.

[0085] ri a Represents the training image image a Training binary label image after distinguishing structure area and background area.

[0086] ri A Represents the training image image A Training binary label image after distinguishing structure area and background area.

[0087] Construction step C, build the Stu training network based on the DexiNed architecture;

[0088] In the present invention, the DexiNed architecture includes multiple sets of filters that receive an image as input and then predict a structure map with the same resolution.

[0089] In the present invention, the Stu training network is provided with an encoder block, an upsampling block and a structure fusion layer according to the DexiNed architecture.

[0090] The encoder block contains an encoder consisting of 6 convolutional blocks, which are used to extract image features of different scales and generate multi-scale structure maps. Each encoder consists of two convolutional layers with 3×3 convolution kernels, and is equipped with batch normalization and ReLU activation function.

[0091] The upsampling block receives the feature maps from the encoder block and generates an intermediate structure map using convolution and transposed convolution.

[0092] The structure fusion layer fuses the intermediate structure maps generated by all upsampling blocks and generates the final structure extraction feature map.

[0093] Construct step D and set the loss function of Stu training network;

[0094] Set the Stu network loss function during training

[0095] In the present invention, the Stu network loss function adopts the BCDN loss function of the network output and the structure label, and the relationship between the constructed Stu module loss function is:

[0096] Represents the loss function required by Stu to train the network.

[0097] Represents the BCDN loss output by the Stu training network.

[0098] Represents the output digitized binary image feature information.

[0099] In the present invention, the loss function is used in the Stu training network. For TRIM={image1,image2,…,image a ,…,image A} and TRI={ri1,ri2,…,ri a ,…,ri A}After model training, the trained Stu module F is obtained Stu .

[0100] Construction of the speckle noise suppression module of the bilateral residual block network, namely the Dep model

[0101] Build step 1, get the training image;

[0102] A number of SAR single-view complex images with spots are selected from the 1B signal processor as training images, which are recorded as the training image set TRIM = {image1, image2, ..., image a ,…,image A}.

[0103] The image a The pixel matrix is ​​represented as

[0104]

[0105] Build step 2 to obtain a noise-free label dataset;

[0106] TRIM={image1,image2,…,image a ,…,image A} as a noise-free reference tag, and then the echo simulation and imaging processing technology of the SAR system is used to obtain the imaging result of the simulated data through the range-Doppler algorithm, which is recorded as SM_TRIM={sm1,sm2,…,sm a ,…,sm A}.

[0107] sm1 represents the imaging result of the training image image1 after echo simulation and imaging processing by the SAR system.

[0108] sm2 represents the imaging result of the training image image2 after echo simulation and imaging processing by the SAR system.

[0109] sm a Represents the training image image a Imaging results after echo simulation and imaging processing by the SAR system.

[0110] sm A Represents the training image image A Imaging results after echo simulation and imaging processing by the SAR system.

[0111] Specifically, image a The amplitude is used as the scattering coefficient in the echo simulation, and the imaging result of the simulated data is obtained through the range-Doppler algorithm to generate image a The corresponding simulation data sm a .

[0112] In order to accurately capture the complexity and diversity of ground work scenarios, sm a The echo simulation process uses an image with a size of 1024×1024 a The amplitude is used as the scattering coefficient in the echo simulation. Considering the complexity of the ground working scene, 4×4 scattering points are placed in each resolution unit during the simulation to serve the working conditions. The simulated image sm after imaging a The size is 256 × 256. Finally, in order to ensure the consistency of the noisy input of the Stu network with the noise-free reference label, the image with a size of 1024 × 1024 is a Downsample to 256×256.

[0113] Construction step 3: build the Dep training network based on the DuRB architecture;

[0114] In this paper, the Dep training network based on the DuRB architecture utilizes dual residual modules, combined with large and small receptive field convolution operations, to efficiently extract global features and local details. Through outer residual connections and modular stacking, the Dep training network can remove noise while preserving image texture and details.

[0115] In the present invention, the Dep training network consists of an input layer, 6 dual residual blocks and an output layer.

[0116] The input layer consists of a convolutional layer, a batch normalization layer, and a RELU activation function, which is used to preliminarily extract image features.

[0117] The dual residual block employs two convolution operations with different receptive fields. The first convolution uses a large kernel to capture information at a wider range. The second convolution uses a small kernel to preserve fine details in the image. This paired operation design ensures that the network balances global feature extraction with the preservation of local details.

[0118] The output layer contains a convolutional layer and a hyperbolic tangent activation function to map the denoising result back to the image space.

[0119] Build step 4 and set the loss function of the Dep training network;

[0120] Set the Dep network loss function L during training 训练 .

[0121] In the present invention, the Dep network loss function adopts the mean square error, and the constructed Dep model loss function relationship is:

[0122] Indicates that the pixel carries information loss. Calculated sm a Output after Dep network and the noise-free reference label image a The mean square error between

[0123] Indicates the loss of pixel structure information. Quantified Output after passing through Stu network and the noise-free reference label image a Output after passing through Stu network The mean square error between

[0124] λ is and In the present invention, the value of λ is 0.5.

[0125] At this point, the Dep model training is completed.

[0126] In the present invention, see Figure 3 As shown in the Dep training network based on the loss function L 训练 For SM_TRIM={sm1,sm2,…,sm a ,…,sm A} and TRIM={image1,image2,…,image a ,…,image A}After model training, the trained Dep module F is obtained Dep .

[0127] See also Figure 2 As shown in FIG, a collaborative optimization method for SAR image speckle noise suppression and structure preservation of the present invention is denoted as the Stu-Fus-Dep collaborative model, and the specific steps are as follows:

[0128] Step 1: Obtain the SAR image to be processed;

[0129] In the present invention, the Stu-Fus-Dep collaborative model obtains any single-view amplitude image, i.e., a SAR image with speckle noise, from a level 1B signal processor, and records it as the image to be processed X. i The X i The available matrix can be expressed as

[0130] N a is the total number of pixels in the azimuth direction.

[0131] N r is the total number of distance pixels.

[0132] q=1,2,3,…,N a , represents the qth pixel point in the azimuth direction.

[0133] p=1,2,3,…,N r , represents the p-th pixel point in the distance direction.

[0134] x q,p Represents the pixel at the qth row and pth column of the single-view amplitude image.

[0135] Indicates the Nth row of the single-view amplitude image r Column of pixels.

[0136] Represents the Ntha The pixel in row and column 1.

[0137] Represents the Nth a Row N r The pixel point of the column is also the last pixel point of the single-view amplitude image.

[0138] Step 2: Apply Stu network to extract feature information of digital binary image;

[0139] Step 21, processing the SAR image in the Stu module;

[0140] In the present invention, the Stu network includes an encoder block, an upsampling block and a structure fusion layer based on the DexiNed architecture.

[0141] The encoder block is used to extract X at different scales i The characteristics of i .

[0142] The upsampling block receives AX i , use convolution and transposed convolution to generate an intermediate structure map, denoted as BX i .

[0143] The structure fusion layer maps the intermediate structure BX generated by all upsampling blocks i Fusion is performed to generate the final structure extraction feature map, denoted as CX i .

[0144] Step 22, obtaining digitized binary image feature information;

[0145] According to the trained Stu model F Stu CX i Processing is performed to obtain the binary structure label image XS i .

[0146] Step 3: Apply the Fus module to complete feature fusion based on convolution and attention;

[0147] Step 31, feature extraction;

[0148] X i Through the convolution operation of the 3×3 convolution layer, the preliminary feature map of the image X is generated. i This operation can not only effectively capture the local spatial information of the image, but also enhance the expressiveness of the features while keeping the feature size of the input and output consistent.

[0149] Step 32, generate an attention matrix;

[0150] XS iGenerate the spatial attention matrix XS through a 1×1 convolution operation i ′. This operation can reduce the feature dimension and realize the spatial weight distribution of features.

[0151] Step 33, feature weighting;

[0152] X i ' and XS i 'Multiply element by element to form a weighted feature map XXS i This weighted operation can adaptively enhance the feature expression of key areas in the image while weakening the influence of unimportant areas.

[0153] Step 34, second weighted fusion;

[0154] XXS i ′Fuse again through the convolution layer to obtain the weighted fusion output feature information, recorded as UX i This operation enhances the ability to capture global information by increasing the receptive field and re-aggregating features, while generating higher quality fusion features.

[0155] Step 4: Speckle noise suppression is completed through Dep;

[0156] Step 41, processing the feature fusion information in the Dep module;

[0157] The input layer initially extracts UX i Features, denoted as UX i ′.

[0158] Use the first 6 double residual blocks connected to UX i 'Processing, get the denoising result RX i This operation is used to balance global feature extraction and local detail preservation.

[0159] The output layer will RX i Mapped back to the image space, denoted as OX i .

[0160] Step 42, obtaining speckle noise suppression results;

[0161] According to F Dep OX i Processing is performed to obtain the speckle noise suppression result X o .

[0162] The F Dep It comes from the output of the trained Dep module.

[0163] The main contributions and innovations of the method proposed in this invention include three parts: network structure, network loss function and data set and construction.

[0164] In terms of network structure, the present invention designs a collaboratively optimized network structure that combines a speckle noise suppression module and a structure extraction module. Guided by structural information, the proposed method suppresses speckle noise while better preserving edge and texture details.

[0165] In terms of loss function, this paper defines a new loss function that takes into account both spatial details and structural information.

[0166] Regarding dataset construction, this paper proposes a novel method for constructing SAR datasets, leveraging echo simulation and SAR system imaging processing techniques. This method uses the amplitude of real SAR images as the scattering coefficient in echo simulation and uses a range-Doppler algorithm to obtain imaging results from the simulated data, thereby generating a dataset based on real SAR images.

[0167] Example 1

[0168] This embodiment uses TerraSAR-X satellite data for training and testing. The image resolution ranges from 0.8 meters to 1.4 meters, including both single-view and multi-view images.

[0169] When the Stu network is trained, it contains 800 pairs of samples with a size of 256×256. One of the pairs is the training set sample of the Stu network. Figure 4 and Figure 5 As shown, Figure 4 is a SAR image (the scale of the image is pixel coordinates), Figure 5 for Figure 4 During training, the parameters of the 3×3 convolution kernel were initialized using a Xavier normal distribution, the parameters of the 1×1 convolution kernel were initialized using a normal distribution with mean 0, and the bias was initialized to 0. The learning rate was set to a constant 0.0001, and the number of image samples input to the network was 32 at a time. To prevent overfitting, a weight decay coefficient of 0.0001 was set. The network was trained for a total of 20 epochs.

[0170] The dataset used in the Dep network uses 4096 pairs of image slices with a size of 256×256. One pair is the training set sample of the Dep network. Figure 6 and Figure 7 shown. Figure 6 is the original SAR image, Figure 7 Based on Figure 6 The generated echo simulation and SAR system simulation results. During the training process, the Adam optimization algorithm is used to optimize the parameters. The optimizer's β1 and β2 parameters are set to 0.9 and 0.999 respectively, and the smoothing term ε is set to 1×10 -8The training process uses an exponentially decaying learning rate, with an initial value of 0.001 and a decay rate of half every 10 training cycles. When initializing the Dep network parameters, the convolution kernel weights are randomly initialized using a Gaussian distribution with a mean of 0 and a standard deviation of 0.02. The bias parameters are initialized to 0. The number of image samples fed into the network at a time during training is 4, and a total of 100 Dep networks are trained.

[0171] The test image is a slice of TerraSAR-X satellite data with a resolution of 1m. Figure 8 The original SAR image is first applied to the original image using the Stu module to extract the digital binary image feature information. Then, the original image and its corresponding digital binary image feature information are put into the Fus module for feature fusion. The result of feature fusion is then put into the Dep module for speckle noise suppression, and the final speckle noise suppression result is output. Figure 9 The following are the test results of the SAR image speckle noise suppression method proposed in this invention. By comparing the two images, it can be found that the method proposed in this invention can significantly suppress speckle noise, and the image details, especially in the high-feature structure area, are effectively preserved. Table 1 lists the quantitative indicators of the results. ENL stands for equivalent visual number, which is an extension of SAR resolution and indicates the size of the equivalent ground area represented by one pixel. TCR stands for target clutter ratio, and the change in target clutter ratio ΔTCR can reflect the structure preservation ability of the algorithm. EPI stands for edge preservation index, which is an indicator used to evaluate the edge preservation ability of image processing algorithms.

[0172] Table 1 Quantitative evaluation results of SAR image speckle noise suppression effect

[0173]

[0174] According to the results in Table 1, after speckle noise suppression, the equivalent view count of the image is significantly improved, while the target clutter ratio changes little and the edge preservation index is large, indicating that the algorithm can well suppress speckle noise while retaining the structural information of the image.

Claims

1. A collaborative optimization system for SAR image speckle noise suppression and structure preservation, characterized by: The system consists of a structure extraction module based on edge detection network, namely the Stu module, a feature fusion module based on convolution, namely the Fus module, and a speckle noise suppression module based on bilateral residual block network, namely the Dep module. The Stu module first extracts image features of different scales through the encoder block to generate a multi-scale structure map. Second, it uses the upsampling block to improve the feature map resolution and generate a refined structure map. Third, it uses the structure fusion layer to fuse the intermediate structure maps generated by all upsampling blocks and generate the final structure extraction feature map through the learned filter. The Fus module processes SAR images through convolution operations to generate preliminary SAR feature maps. It also processes SAR images through convolution operations to generate attention matrices. It also performs weighted fusion of preliminary SAR feature maps and attention matrices. Finally, it generates higher-quality fused features through convolution fusion. The first aspect of the Dep module is to preliminarily extract image features through an input layer consisting of a convolutional layer, a batch normalization layer, and a RELU activation function; Secondly, speckle noise is suppressed through a dual residual block consisting of two convolution operations with different receptive fields, simultaneously extracting global features and preserving local details. Thirdly, an output layer consisting of a convolution layer and a hyperbolic tangent activation function is used to map the denoising result back to the image and generate the final speckle noise suppression result. The method of constructing the speckle noise suppression module of the bilateral residual block network, namely the Dep module, has the following steps: Build step 1, get the training image; Select multiple SAR single-view complex images with spots from the 1B signal processor as training images, which are recorded as the training image set ; Represents the first training image; Represents the second training image; Indicates the training images; represents the last training image; described The pixel matrix is ​​represented as ; is the total number of pixels in the azimuth direction; is the total number of distance pixels; , indicating the direction of pixels; , represents the distance direction pixels; Represents the first Rank Column pixels; Represents the first row of the training image Column pixels; Represents the first The pixel in the first column of the row; Represents the first Rank The pixel point of the column is also the last pixel point of the training image; Build step 2 to obtain a noise-free label dataset; Will As a noise-free reference tag, the echo simulation and imaging processing technology of the SAR system is then used to obtain the imaging result of the simulated data through the range-Doppler algorithm, which is recorded as ; Represents training images Imaging results after echo simulation and imaging processing by the SAR system; Represents training images Imaging results after echo simulation and imaging processing by the SAR system; Represents training images Imaging results after echo simulation and imaging processing by the SAR system; Represents training images Imaging results after echo simulation and imaging processing by the SAR system; described The amplitude is used as the scattering coefficient in the echo simulation, and the imaging result of the simulated data is obtained through the range-Doppler algorithm, thereby generating Corresponding simulation data ; In order to accurately capture the complexity and diversity of ground work scenes, The echo simulation process uses a 1024×1024 size The amplitude is used as the scattering coefficient in the echo simulation; considering the complexity of the ground working scene, 4×4 scattering points are placed in each resolution unit during the simulation to serve the working conditions; the simulated image after imaging The size is 256×256; finally, in order to ensure the consistency of the noisy input of the Stu network with the noise-free reference label, the size of Downsample to 256×256; Construction step 3: build the Dep training network based on the DuRB architecture; The Dep training network consists of an input layer, 6 dual residual blocks, and an output layer; The input layer consists of a convolutional layer, a batch normalization layer, and a RELU activation function, which is used to preliminarily extract image features; The dual residual block uses two convolution operations with different receptive fields. The first convolution uses a large-scale convolution kernel to capture a wider range of information. The second convolution uses a small-sized convolution kernel to preserve fine details in the image. This paired operation design ensures that the network can balance global feature extraction and local detail preservation. The output layer contains a convolutional layer and a hyperbolic tangent activation function to map the denoising result back to the image space; Build step 4 and set the loss function of the Dep training network; Set the Dep module loss function during training ; The Dep module loss function uses mean square error, so the constructed Dep module loss function relationship is: ; Indicates that the pixel carries information loss; Calculated Output after Dep network With noiseless reference label The mean square error between ; Indicates the loss of pixel structure information; Quantified Output after passing through Stu network and noise-free reference labels Output after passing through Stu network The mean square error between ; yes and The linear combination coefficient of ; In the present invention, The value is 0.5; At this point, the Dep module training is completed; In Dep training network, based on loss function right and After model training, the trained Dep module is obtained .

2. The system for collaborative optimization of SAR image speckle noise suppression and structure preservation according to claim 1, characterized in that: The system is stored in the Class 1B signal processor.

3. The system for collaborative optimization of SAR image speckle noise suppression and structure preservation according to claim 1, characterized in that: The system is used to solve the problem in the prior art that image structure information is lost when suppressing speckle noise.

4. The system for collaborative optimization of SAR image speckle noise suppression and structure preservation according to claim 1, characterized in that: The method of constructing the structure extraction module based on the edge detection network, namely the Stu module, has the following steps: Construction step A, obtain training images; Select multiple SAR single-view complex images with spots from the 1B signal processor as training images, which are recorded as the training image set ; described The pixel matrix is ​​represented as ; Construction step B, distinction between structure area and background area; Will The pixels in the structure area that need to be retained are marked as 1, and the pixels in the background area are marked as 0, and the training binary label image set is obtained, which is recorded as ,and ; Represents training images The training binary label image after distinguishing the structure area and background area; Represents training images The training binary label image after distinguishing the structure area and background area; Represents training images The training binary label image after distinguishing the structure area and background area; Represents training images The training binary label image after distinguishing the structure area and background area; Construction step C, build the Stu training network based on the DexiNed architecture; The Stu training network is equipped with encoder blocks, upsampling blocks, and structure fusion layers based on the DexiNed architecture; The encoder block contains an encoder consisting of 6 convolutional blocks, which are used to extract image features of different scales and generate multi-scale structure maps. Each encoder consists of two convolutional layers with 3×3 convolution kernels, and is equipped with batch normalization and ReLU activation function. The upsampling block receives the feature map from the encoder and generates an intermediate structure map using convolution and transposed convolution; The structure fusion layer fuses the intermediate structure maps generated by all upsampling blocks and generates the final structure extraction feature map; Construct step D and set the loss function of Stu training network; Set the Stu module loss function during training ; The constructed Stu module loss function relationship is ; Represents the loss function required by Stu to train the network; represents the BCDN loss output by the Stu training network; Represents the output digital binary image feature information; In Stu training network, based on loss function right and After model training, the trained Stu module is obtained .

5. The collaborative optimization method for a system for collaborative optimization of SAR image speckle noise suppression and structure preservation according to claim 1, 2, 3 or 4, characterized in that The steps are as follows: Step 1: Obtain the SAR image to be processed; The Stu-Fus-Dep collaborative model obtains any single-view amplitude image from the 1B-level signal processor, that is, the SAR image with speckle noise, and records it as the image to be processed. ; The available matrix can be expressed as ； , indicating the direction of pixels; , represents the distance direction pixels; Represents the first Rank Column pixels; Represents the first row of the single-view amplitude image Column pixels; Represents the first The pixel in the first column of the row; Represents the first Rank The pixel point of the column is also the last pixel point of the single-view amplitude image; Step 2: Apply Stu network to extract feature information of digital binary image; Step 21, processing the SAR image in the Stu module; The Stu module includes an encoder block, an upsampling block, and a structure fusion layer based on the DexiNed architecture; The encoder block is used to extract different scales The characteristics of ; Upsampling block receives , using convolution and transposed convolution to generate intermediate structure mapping, denoted as ; The structure fusion layer maps the intermediate structures generated by all upsampling blocks Fusion is performed and the final structure extraction feature map is generated, which is recorded as ; Step 22, obtaining digitized binary image feature information; Based on the trained Stu module right Processing is performed to obtain a binary structure label image ; Step 3: Apply the Fus module to complete feature fusion based on convolution and attention; Step 31, feature extraction; Will Generate the preliminary feature map of the image through the convolution operation of the 3×3 convolution layer ; Step 32, generate an attention matrix; Will Generate a spatial attention matrix through a 1×1 convolution operation ; Step 33, feature weighting; Will and Multiply element by element to form a weighted feature map ; Step 34, second weighted fusion; Will Through the convolution layer, the weighted fusion output feature information is obtained again, which is recorded as ; Step 4: Speckle noise suppression is completed through the Dep module; Step 41, processing the feature fusion information in the Dep module; Initial extraction of input layer The characteristics of ; Use the first 6 double residual block pairs connected Process and get the denoising result ; The output layer will Mapped back to the image space, denoted as ; Step 42, obtaining speckle noise suppression results; in accordance with right Processing is performed to obtain the speckle noise suppression result ; described It comes from the output of the trained Dep module.

Images