SAR (Synthetic Aperture Radar) image target classification system based on complex domain feature map network

Abstract

The invention discloses an SAR image target classification system based on a complex domain feature map network, and the system comprises the following steps: dividing an SAR data set into a training set, a verification set and a test set according to a division proportion, and carrying out the data augmentation of the training set; dividing the SAR images into a support set and a query set; respectively processing the support set and the query set through a complex domain transformation feature extraction module to obtain complex domain features of the SAR; updating by using a graph network in the system to obtain a support set image which is most similar to the query set image; and printing a final classification accuracy result in a display window. The invention relates to the field of image processing, and discloses an SAR image target classification system based on a complex domain feature map network, which solves the problem that the potential advantages of a complex domain in SAR image classification are neglected while most features are focused on a spatial domain or a time-frequency domain in the prior art, and improves the accuracy of SAR image classification.

Description

Technical Field

[0001] The invention relates to the field of image processing, and in particular to a SAR image target classification system based on a complex domain feature graph network. Background Art

[0002] Synthetic Aperture Radar (SAR) is a system that performs ground imaging based on radar technology. Compared with traditional optical imaging technology, SAR has many advantages that are difficult to quantify. It has the unique ability to observe the earth around the clock and in all weather conditions. It is widely used in military, ocean, communication, geological exploration, disaster monitoring and other fields. Studying the complex domain characteristics of SAR images helps to better deal with interference, enrich feature dimensions, accurately capture changes and grasp the details of changes.

[0003] Traditional image classification methods mostly focus on feature extraction and analysis based on the spatial domain or time-frequency domain, but often destroy the intrinsic connection between the real and imaginary parts of SAR images, ignoring the potential advantages of the complex domain in SAR image classification. SAR images are complex-valued in nature, which means that each pixel can be represented by a complex number. The complex number consists of two different parts. The real part and the imaginary part provide different information about the target object in different ways. The complex domain method can improve the classification accuracy and robustness by comprehensively considering the intensity and phase information, and provide a more refined feature representation than the spatial domain or time-frequency domain. Therefore, studying the complex domain of SAR images is of vital help to the analysis of SAR images. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] In view of the shortcomings of the prior art, the present invention provides a SAR image target classification system based on a complex domain feature map network, which solves the limitation that the current research on SAR images only targets the real domain and improves the classification accuracy of SAR images.

[0006] (II) Technical solution

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: A SAR image target classification system based on a complex domain feature map network comprises the following steps:

[0008] S1: Divide the SAR dataset into training set, validation set, and test set according to the division ratio, and perform data augmentation on the training set;

[0009] S2: Divide the SAR images into support set and query set;

[0010] S3: The support set and the query set are respectively subjected to the complex domain transformation feature extraction module to obtain the complex domain features of SAR;

[0011] S4: Use the graph network in the system to train the network model to obtain the support set image that is most similar to the query set image;

[0012] S5: Print the final classification accuracy result in the display window;

[0013] The optimal model training described in step S3 includes the following steps:

[0014] The first step is complex domain transformation: the support set and query set are transformed by Hilbert transform respectively. The goal of Hilbert transform is to find the adjoint function y(t) for the real function x(t), so that the complex function z(t) = x(t) + iy(t) can be analyzed on the real axis t∈R and extended to the upper half of the complex plane. By convolving the signal, a complex-valued signal can be generated from the real-valued signal I(x). Therefore, the complex domain images of the support set and query set can be obtained by Hilbert transform.

[0015] The second step is feature extraction: extract complex domain features from complex domain images through the ResNet12 backbone network. It is divided into four stages, of which the first stage is the processing of input. The input is It has been successively obtained through the convolution layer, BN layer, LeakyReLU activation function, and maximum pooling layer. Output features of the last stage That is, it is the output feature of the backbone network, where B is the size of each batch.

[0016] As an optimization, the specific steps of training the graph network training network model described in step S4 are as follows:

[0017] (1) Constructing the complex domain node initialization and update module: Extracting complex domain node features through the backbone network, 5 nodes are composed of 4 support set samples and 1 query set sample, and the nodes are initialized. Let v i Yes V i The node characteristics, V i = {v i |i=1,2,...,N×K+T}, when the true label of the support set is consistent with the predicted label of the query set, it is 1, otherwise it is 0, Given the node at layer l-1 and edge The node features are updated through the process of domain aggregation. The features of the l-th layer nodes are updated by aggregating the features of other nodes in a way that is proportional to the edge features of the nodes. The aggregated features are then transformed to update the features of the l-th layer nodes. Node features at layer l-1 Update the node features of layer l Edge features at layer l-1 Used to quantify the contribution of the corresponding neighboring nodes

[0018] (2) Constructing the hybrid edge initialization and update module: Nodes effectively capture the similarity information between each other through similarity measurement edges, thereby generating preliminary edge features. The similarity measurement edges and relative measurement edges are spliced ​​to form the hybrid edges of the nodes, thereby forming the hybrid edges of the graph. The similarity measurement edges can only reflect the similarity of node features, while the relative measurement edges encode the relative positions or topological relationships between nodes. Nodes aggregate features with their adjacent nodes. This dynamic feature fusion process enables the hybrid edges to more comprehensively reflect the multi-dimensional relationships between nodes and effectively distinguish nodes with similar features but different categories. The relative measurement edge is denoted as e ij1 Represents the first dimension edge feature between node i and node j, and the similarity measurement edge is recorded as e ij2 , the dissimilarity measure edge is recorded as e ij3 , initialize the relative metric edge Combined with the similar and dissimilar structures between nodes, the edges of the graph are initialized After the node features are updated, new node features will be obtained, thereby realizing the update of mixed edge features.

[0019] As an optimization, the update of the mixed edge features uses similarity metric edges and relative metric edges to construct the mixed edges of the graph. The relative metric edges are measured and encoded by calculating the distance between nodes after the misalignment, while the similarity edges are encoded by directly measuring the distance between nodes. After the mixed edge update is completed, the nodes will be classified according to the characteristics of the points.

[0020] As an optimization, in step S1, the SAR dataset includes four categories: ships, airplanes, oil tanks, and bridges. The proportion of the number of target images varies greatly. Therefore, data enhancement is performed on the training set to balance the number of samples of different categories in the dataset.

[0021] As an optimization, the data augmentation methods in step S1 include rotating the image, flipping the image horizontally, flipping the image vertically, and random cropping.

[0022] As an optimization, the angle of the rotated image is randomly chosen between -45 and 45 degrees.

[0023] As an optimization, the classification accuracy result in step S5 is obtained by predicting the label result of the query set with the true label of the support set and the mixed edge.

[0024] (III) Beneficial effects

[0025] The present invention provides a SAR image target classification system based on a complex domain feature graph network, which has the following beneficial effects:

[0026] The present invention can extract more refined feature representations than those in the spatial domain or time-frequency domain by studying the complex domain features of SAR images, and use deep learning methods to make SAR image recognition work intelligent. Good recognition results can be obtained through low-cost complex domain transformation feature extraction and mixed edge distance similarity calculation, which solves the problem that traditional recognition methods are based on manually designed features and are difficult to fully mine complex target information in SAR images. Some subtle and representative features may not be effectively extracted, resulting in limited recognition accuracy. The invention can detect and identify military targets and monitor battlefield environments to a certain extent, making certain contributions to the military field. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is a flow chart of the system of the present invention;

[0028] Figure 2 A framework diagram for model training of the present invention;

[0029] Figure 3 It is a framework diagram of the complex domain feature extraction module and graph network module of the present invention. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] See also Figure 1-3 , the present invention provides a technical solution:

[0032] like Figure 2 As shown, model training and retrieval are completed in the following steps:

[0033] The first step is data set preprocessing: SAR images are divided into training set, validation set, and test set according to the proportion, and data augmentation is performed on the training set. The main methods include rotating images, horizontally flipping images, vertically flipping images, and random cropping. The angle of rotation of images is randomly selected between -45 and 45 degrees. Each data set is divided into a support set and a query set.

[0034] The second step is model training: construct a complex domain transformation feature extraction module, use the graph network to train the network model, and finally achieve classification.

[0035] The complex domain transformation feature extraction module includes the following steps:

[0036] (1) Complex domain transformation: The support set and query set are transformed through the Hilbert transform. The goal of the Hilbert transform is to find the adjoint function y(t) for the real function x(t), so that the complex function z(t) = x(t) + iy(t) can be analyzed on the real axis t∈R and extended to the upper half of the complex plane. By convolving the signal, the complex signal can be generated from the real-valued signal I(x).

[0037] (2) Feature extraction: The support set and query set after Hilbert transformation are extracted through the ResNet12 backbone network. It is divided into 4 stages, of which the first 3 stages are preprocessing of the input. It has been successively obtained through the convolution layer, BN layer, LeakyReLU activation function, and maximum pooling layer. Output features of the last stage That is, the output feature of the backbone network, where B is the number of samples in each batch, C is the number of channels, and H and W represent the height and width of the feature respectively;

[0038] The specific steps for training the graph network model are as follows:

[0039] (1) Complex domain node initialization and update module: The complex domain node features are extracted through the backbone network. Five nodes are composed of four support set samples and one query set sample. The nodes are initialized. Let v i Yes V i The node characteristics, V i = {v i |i=1,2,...,N×K+T}, when the true label of the support set is consistent with the predicted label of the query set, it is 1, otherwise it is 0, Given the node at layer l-1 and edge The node features are updated through the process of domain aggregation. The features of the l-th layer nodes are updated by aggregating the features of other nodes in a way that is proportional to the edge features of the nodes. The aggregated features are then transformed to update the features of the l-th layer nodes. Node features at layer l-1 Update the node features of layer l Edge features at layer l-1 Used to quantify the contribution of the corresponding neighboring nodes

[0040] (2) Constructing the hybrid edge initialization and update module: Nodes effectively capture the similarity information between each other through similarity measurement edges, thereby generating preliminary edge features. The similarity measurement edges and relative measurement edges are spliced ​​to form the hybrid edges of the nodes, thereby forming the hybrid edges of the graph. The similarity measurement edges can only reflect the similarity of node features, while the relative measurement edges can effectively distinguish nodes with similar features but different categories by encoding the relative positions or topological relationships between nodes. The relative measurement edges are denoted as e ij1 Represents the first dimension edge feature between node i and node j, and the similarity measurement edge is recorded as e ij2 , the dissimilarity measure edge is recorded as e ij3 , initialize the relative metric edge Combined with the similar and dissimilar structures between nodes, the edges of the graph are initialized After the node features are updated, new node features will be obtained, thereby realizing the update of mixed edge features.

[0041] The third step is image classification function: after the first round of node and mixed edge feature updates, the final node features are used to generate the prediction results of the mixed edge labels, that is, where y ij ∈[0,1] is regarded as two nodes v i and v j The probability of being from the same class. Therefore, the classification result can jointly predict the label result of the query set by having the true label of the support set and the mixed edge

[0042] Based on the above steps, Figure 1 As shown in FIG. 1 , a SAR image target classification system based on a complex domain feature map network is shown in FIG. 1 . The specific operation steps are as follows:

[0043] 1. A SAR image target classification system based on a complex domain feature map network, characterized in that it comprises the following steps:

[0044] S1: Divide the SAR dataset into training set, validation set, and test set according to the division ratio, and perform data augmentation on the training set;

[0045] S2: Divide the SAR images into support set and query set;

[0046] S3: The support set and the query set are respectively subjected to the complex domain transformation feature extraction module to obtain the complex domain features of SAR;

[0047] S4: Use the graph network in the system to update the support set image that is most similar to the query set image;

[0048] S5: Print the final classification accuracy result in the display window.

[0049] In the optimization process of the entire network of the present invention, the network gradually learns the differences between nodes within a class and nodes between classes, and the different features between similarity metric edges and relative metric edges. This process can be regarded as a process in which an expert in a related field learns corresponding knowledge. When the model finds the best network model through continuous optimization, the entire network is equivalent to an expert with rich relevant knowledge. At this point, the model has the ability to quickly identify SAR images.

[0050] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A SAR image target classification system based on complex domain feature graph network, characterized by: The following steps are involved: S1: Divide the SAR dataset into training set, validation set, and test set according to the division ratio, and perform data augmentation on the training set; S2: Divide the SAR images into support set and query set; S3: The support set and the query set are respectively subjected to the complex domain transformation feature extraction module to obtain the complex domain features of SAR; S4: Use the graph network in the system to train the network model to obtain the support set image that is most similar to the query set image; S5: Print the final classification accuracy result in the display window; The complex domain transformation feature extraction module described in step S3 includes the following steps: The first step is complex domain transformation: the support set and query set images are respectively transformed by Hilbert transform. The goal of Hilbert transform is to find the adjoint function y(t) for the real function x(t), so that the complex function z(t) = x(t) + iy(t) can be analyzed on the real axis t∈R and extended to the upper half of the complex plane by convolving the signal. The second step is feature extraction: the complex domain image is extracted through the ResNet12 backbone network. It is divided into four stages, of which the first stage is the processing of the input. The input is It has been successively obtained through the convolution layer, BN layer, LeakyReLU activation function, and maximum pooling layer. Output features of the last stage That is, it is the output feature of the backbone network, where B is the size of each batch.

2. The SAR image target classification system based on complex domain feature graph network according to claim 1, characterized in that: The specific steps of training the graph network training network model described in step S4 are as follows: (1) Complex domain node initialization and update module: The complex domain node features are extracted through the backbone network. Five nodes are composed of four support set samples and one query set sample. The nodes are initialized. Let v i Yes V i The node characteristics, V i = {v i |i=1,2,...,N×K+T}, when the true label of the support set is consistent with the predicted label of the query set, it is 1, otherwise it is 0, Given the node in layer l-1 and edge The node features are updated through the process of domain aggregation. The features of the l-th layer nodes are updated by aggregating the features of other nodes in a way that is proportional to the edge features of the nodes. The aggregated features are then transformed to update the features of the l-th layer nodes. Node features at layer l-1 Update the node features of the lth layer Edge features at layer l-1 Used to quantify the contribution of the corresponding neighboring nodes (2) Constructing the hybrid edge initialization and update module: Nodes effectively capture the similarity information between each other through similarity measurement edges, thereby generating preliminary edge features. The similarity measurement edges and relative measurement edges are spliced ​​to form the hybrid edges of the nodes, thereby forming the hybrid edges of the graph. The relative measurement edge is recorded as e ij1 , represents the edge feature of the first dimension between node i and node j, and the similarity measurement edge is recorded as e ij2 , the dissimilarity measure edge is recorded as e ij3 , initialize the relative metric edge Combined with the similar and dissimilar structures between nodes, the edges of the graph are initialized After the node features are updated, new node features will be obtained, thereby realizing the update of mixed edge features.

3. The SAR image target classification system based on complex domain feature graph network according to claim 2, characterized in that: The update of mixed edge features uses similarity metric edges and relative metric edges to construct the mixed edges of the graph. After the mixed edge update is completed, the nodes will be classified according to the characteristics of the points.

4. The SAR image target classification system based on complex domain feature graph network according to claim 1, characterized in that: In step S1, the SAR dataset includes four categories: ships, aircraft, oil tanks, and bridges. Data enhancement is performed on the training set to balance the number of samples in different categories in the dataset.

5. The SAR image target classification system based on complex domain feature graph network according to claim 1, characterized in that: The data augmentation method in step S1 includes rotating the image, horizontally flipping the image, vertically flipping the image, and randomly cropping.

6. The SAR image target classification system based on complex domain feature graph network according to claim 5, characterized in that: The angle of the rotated image is randomly chosen between -45 and 45 degrees.

7. The SAR image target classification system based on complex domain feature graph network according to claim 1, characterized in that: In step S5, the classification accuracy result is obtained by predicting the label result of the query set with the true label of the support set and the mixed edge.

Images