Radar target recognition method and device based on high-resolution one-dimensional range profile visual map

Abstract

The application discloses a radar target recognition method and device based on high-resolution one-dimensional range image visual graph, which comprises the following steps: extracting the target time sequence in the preprocessed radar high-resolution one-dimensional range image by using a noise threshold; obtaining a node set according to the amplitude value in the target time sequence; traversing the slope of node i and node j in the node set to obtain the edge set corresponding to the node pair in the target time sequence; wherein the range of node i is 0 to L-1, the range of node j is i+1 to L-1, and L is the number of the middle node of the node set; updating the node set multiple times according to the edge set to obtain multiple node sets with different update times; performing graph classification according to the multiple node sets with different update times to obtain a classification result, so as to realize the target recognition of the radar high-resolution one-dimensional range image. The application has higher recognition accuracy and improves the performance of the high-resolution radar system target recognition.

Description

Technical Field

[0001] This invention belongs to the field of radar target recognition, specifically relating to a radar target recognition method and apparatus based on a high-resolution one-dimensional range image visualization. Background Technology

[0002] With the continuous development of radar technology, the ability to accurately identify targets is of paramount importance in numerous fields, including military and civilian applications. High-resolution one-dimensional range profiles from radar reflect the distribution of scattering points of a target along the radar's line-of-sight, containing rich information about the target's structure, and have become an important research direction in target recognition. In today's complex electromagnetic environment and diverse target scenarios, traditional low-resolution radar struggles to meet the demands of high-precision target identification. High-resolution radar can acquire more detailed target features, and high-resolution one-dimensional range profiles, due to their relatively simple data acquisition and high processing efficiency, have attracted significant attention in target recognition applications.

[0003] However, traditional methods that use convolutional networks or temporal networks to directly extract the structural features of a target's one-dimensional range image can only obtain the target's spatial structure from local or sequential relationships. When using convolutional networks to process one-dimensional range images, the convolutional window determines that it can only capture a portion of the information in the input data. As the number of network layers increases, local features may be gradually transferred, but the integration of global features may become inadequate. When using temporal networks to process high-resolution one-dimensional range images, the target's structure may not be a simple temporal sequence relationship, but rather a nonlinear and spatial structure involving multiple feature dimensions.

[0004] From the perspective of extracting as much structural information as possible from the target distance image, traditional methods of directly extracting features from time series have fixed relative positions in the time series. When extracting features from convolutional networks, the window scanning range during extraction becomes limited. The sequential scanning of time series networks makes it difficult to capture the nonlinear relationships between different parts of the target. For example, there is still room for improvement in extracting features from various structural locations of an aircraft together with adjacent locations. Summary of the Invention

[0005] To address the aforementioned problems in the prior art, this invention provides a radar target identification method and apparatus based on a high-resolution one-dimensional range image view.

[0006] The technical problem to be solved by this invention is achieved through the following technical solution:

[0007] In a first aspect, the present invention provides a radar target identification method based on a high-resolution one-dimensional range profile, the method comprising:

[0008] Target time series in preprocessed high-resolution one-dimensional range profiles of radar are extracted using a noise threshold.

[0009] Obtain a node set based on the amplitude values in the target time series;

[0010] Traverse the slopes of nodes i and j in the node set to obtain an edge set corresponding to the nodes in the target time series; wherein, the range of node i is from 0 to L - 1, the range of node j is from i + 1 to L - 1, and L is the number of nodes in the node set;

[0011] Update the node set multiple times according to the edge set to obtain the node sets with multiple different update times;

[0012] Perform graph classification according to the node sets with multiple different update times to obtain a classification result, so as to achieve target recognition of the radar high-resolution one-dimensional range profile.

[0013] Optionally, the method for using a noise threshold to extract the target time series in the preprocessed radar high-resolution one-dimensional range profile includes:

[0014] Perform amplitude normalization on the radar high-resolution one-dimensional range profile to obtain the preprocessed radar high-resolution one-dimensional range profile;

[0015] Obtain the noise threshold according to the preprocessed radar high-resolution one-dimensional range profile;

[0016] Use the noise threshold to search for the first range cell greater than the noise threshold from front to back in the preprocessed radar high-resolution one-dimensional range profile as the starting cell;

[0017] Use the noise threshold to search for the first range cell greater than the noise threshold from back to front in the preprocessed radar high-resolution one-dimensional range profile as the ending cell;

[0018] Extract the target time series according to the starting cell and the ending cell.

[0019] Optionally, the noise threshold is expressed as follows:

[0020]

[0021] where threshold represents the noise threshold, K is a preset coefficient, x norm =[x norm,1 , x norm,2 ,..., x norm,n ,..., x norm,N represents the preprocessed radar high-resolution one-dimensional range profile, a, b ∈ [0, N], a < b, and N is the number of range cells in the preprocessed radar high-resolution one-dimensional range profile.

[0022] Optionally, the step of traversing the slopes of nodes i and j in the node set to obtain the edge set corresponding to the nodes in the target time series includes:

[0023] When traversing, if the current slope value of node i and node j in the node set corresponding to the current time is greater than the slope threshold, an undirected edge is added between node i and node j at the current time, the slope threshold is updated to the current slope value, and the traversal continues.

[0024] After the traversal is completed, the set of edges is obtained, consisting of the undirected edges between nodes i and j at each time step.

[0025] Optionally, updating the node set multiple times based on the edge set to obtain multiple updated node sets includes:

[0026] The node set is subjected to word embedding processing to obtain the node set after word embedding;

[0027] Obtain the neighbor nodes v of any node v in the node set after word embedding. neighbor The importance of node v; wherein, the neighboring nodes v of node v neighbor It is obtained based on the node v and the set of edges;

[0028] The normalized attention coefficient corresponding to node v is obtained based on the node importance and the softmax function.

[0029] The node v is updated according to the attention coefficient to obtain the updated node v;

[0030] The updated node v is updated again until a preset number of updates is reached, resulting in multiple nodes v with different number of updates, and thus a set of nodes with multiple number of updates.

[0031] Optionally, the neighbor node v neighbor The importance of node v is represented as follows:

[0032] e neighbor =LeakyReLU(a T [Wv||Wv neighbor ]);

[0033] Among them, e neighbor This indicates that the neighbor node v neighbor For the node importance of node v, LeakyReLU is the activation function, a T Let W be the first learnable parameter and W be the second learnable parameter.

[0034] Optionally, the normalized attention coefficient corresponding to node v is represented as follows:

[0035]

[0036] Where, α neighbor This represents the normalized attention coefficient corresponding to node v, where exp is the exponential function, and M... neighbor Let v represent the set of neighboring nodes of node v. neighbor 'Refers to the other neighboring nodes of node v.

[0037] Optionally, the graph classification based on the set of nodes with multiple different update counts includes:

[0038] The node sets with different update counts are input into the graph pooling layer to obtain multiple pooled node sets with different update counts.

[0039] The set of nodes with different number of updates after multiple pooling is input into the fully connected layer to obtain the classification result.

[0040] Secondly, the present invention provides a radar target identification device based on a high-resolution one-dimensional range profile, the device comprising:

[0041] The sequence extraction module is used to extract the target time sequence from the preprocessed radar high-resolution one-dimensional range image using a noise threshold.

[0042] The first set acquisition module is used to obtain a set of nodes based on the amplitude value in the target time series;

[0043] The second set acquisition module is used to traverse the slopes of nodes i and j in the node set to obtain the edge set corresponding to the nodes in the target time series; wherein, the range of node i is 0 to L-1, the range of node j is i+1 to L-1, and L is the number of nodes in the node set;

[0044] An update module is used to update the node set multiple times based on the edge set to obtain multiple node sets with different update counts;

[0045] The identification module is used to perform graph classification based on the set of nodes with multiple different update counts to obtain classification results, so as to realize target identification of the radar high-resolution one-dimensional range image.

[0046] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0047] In the above technical solution, by constructing a set of nodes and edges based on the target time series, the resulting visual structure is classified into graphs. This allows for deeper feature extraction, resulting in a higher recognition accuracy compared to traditional time series methods that directly extract features through convolutional networks or temporal networks. This improves the target recognition performance of high-resolution radar systems.

[0048] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0049] Figure 1 This is a flowchart of a radar target recognition method based on a high-resolution one-dimensional range profile provided by an embodiment of the present invention;

[0050] Figure 2a This is a schematic diagram of an embodiment of the present invention for extracting a high-resolution one-dimensional distance image;

[0051] Figure 2b This is a schematic diagram of an extracted high-resolution one-dimensional distance image provided in an embodiment of the present invention;

[0052] Figure 3 This is a flowchart of constructing an edge set proposed in an embodiment of the present invention;

[0053] Figure 4 This is a flowchart of a graph classification method proposed in an embodiment of the present invention;

[0054] Figure 5 This is a comparison chart of the classification accuracy of different models proposed in an embodiment of the present invention;

[0055] Figure 6 This is a comparison chart of the classification accuracy of different graph structures proposed in an embodiment of the present invention;

[0056] Figure 7 This is a block diagram of a radar target identification device based on a high-resolution one-dimensional range image, as proposed in an embodiment of the present invention. Detailed Implementation

[0057] To facilitate understanding of the present invention, a brief description of the relevant prior art and the inventive concept of the present invention will be given first.

[0058] Traditional methods for directly extracting structural features from high-resolution one-dimensional range images of targets using convolutional networks or temporal networks can only obtain the spatial structure of the target from local or sequential relationships. When using convolutional networks to process high-resolution one-dimensional range images, the convolutional window determines that it can only capture a portion of the information in the input data. As the number of network layers increases, local features may be gradually propagated, but the integration of global features may become insufficient. When using temporal networks to process high-resolution one-dimensional range images, the structure of the target may not be a simple temporal sequence relationship, but rather a nonlinear and spatial structure involving multiple feature dimensions. Constructing the high-resolution one-dimensional range image as a visual structure allows each scattering point of the target to be regarded as a node attribute in the graph, and the connecting edge between each pair of nodes can represent their physical or spatial relationship. The scattering points of some targets may exhibit certain regular structures in the range image, such as specific shapes or symmetries. These local regularities can be captured and propagated to the whole graph through the connection relationship of the graph structure, thus providing rich target structural information for target recognition. Furthermore, through the graph structure, graph neural networks can integrate global information and propagate information through the connections between nodes, thereby capturing the global scattering characteristics of the target. The connections between different scattering points can reflect the relationships between various parts of the target structure. In this way, graph neural networks can more comprehensively understand the global features of the target, such as its structure and shape, thereby improving recognition accuracy. Therefore, based on the above inventive concept, this invention proposes a radar target recognition method based on a high-resolution one-dimensional range profile.

[0059] Figure 1 This is a flowchart of a radar target recognition method based on a high-resolution one-dimensional range profile provided by an embodiment of the present invention, as follows: Figure 1 As shown, the method may include the following steps:

[0060] S101. Extract the target time series from the preprocessed radar high-resolution one-dimensional range image using a noise threshold.

[0061] Optionally, S101 may include:

[0062] Amplitude normalization is performed on the high-resolution one-dimensional range image of the radar to obtain the preprocessed high-resolution one-dimensional range image of the radar.

[0063] The noise threshold is obtained from the preprocessed high-resolution one-dimensional range image of the radar.

[0064] Using a noise threshold, the first range cell larger than the noise threshold is searched from front to back in the preprocessed high-resolution one-dimensional range image of the radar and used as the starting cell.

[0065] Using a noise threshold, the first range cell in the preprocessed high-resolution one-dimensional range image of the radar is searched from back to front to be the end cell.

[0066] The target time series is extracted according to the starting unit and the ending unit.

[0067] It can be understood that before target recognition of the radar high-resolution one-dimensional range profile, due to the amplitude sensitivity, translation sensitivity, and azimuth sensitivity of the one-dimensional range profile, preprocessing operations need to be performed first. Amplitude sensitivity refers to the degree of dependence of the one-dimensional range profile on the amplitude change of the target reflection signal. Even for the same radar and the same target, there may be certain differences in the echo amplitude. For a high-resolution one-dimensional range profile sample x s =[x1, x2,..., x N , the two-norm is used for amplitude normalization to eliminate this difference, and the preprocessed high-resolution one-dimensional range profile is obtained:

[0068]

[0069] In addition, Figure 2a is a schematic diagram of the pre-extraction high-resolution one-dimensional range profile provided by an embodiment of the present invention, Figure 2b is a schematic diagram of the post-extraction high-resolution one-dimensional range profile provided by an embodiment of the present invention. As shown in Figure 2a and Figure 2b , translation sensitivity refers to the influence of the target moving along the range direction in physical space on the one-dimensional range profile. When the target displacement is greater than the range resolution cell, the target peak will jump from one cell to another, resulting in the misalignment of the range profiles of the same target. And considering the influence of the noise region, if traditional methods such as centroid alignment are used to align the one-dimensional range profile without noise removal, the noise will also be constructed as nodes in the construction of the graph structure, which will bring interference of useless signals and huge computational amounts. The noise threshold refers to setting a suitable threshold through the amplitude at the noise. When the signal amplitude of the range cell is greater than this threshold, it is determined that the cell is the target range profile, so as to extract the cell where the target is located, eliminate the translation sensitivity, and remove the noise region.

[0070] Optionally, the noise threshold is expressed as follows:

[0071]

[0072] where threshold represents the noise threshold, K is a preset coefficient, x norm =[x norm,1 , x norm,2 ,..., x norm,n ,..., x norm,N represents the preprocessed radar high-resolution one-dimensional range profile, a, b ∈ [0, N], a < b, and N is the number of range cells in the preprocessed radar high-resolution one-dimensional range profile.

[0073] It is worth mentioning that azimuth sensitivity refers to the sensitivity of a high-resolution one-dimensional range image to changes in the relative azimuth angle between the target and the radar. High-resolution one-dimensional range images obtained from observing a target from multiple angles exhibit significant differences. Therefore, this invention uses data that covers as many azimuth angles as possible for target observation for training, enabling the model to learn target azimuth information from different angles and thus enhancing the model's azimuth generalization ability.

[0074] S102. Obtain the node set based on the amplitude value in the target time series.

[0075] It is understandable that a target time series corresponds to a graph sample, and a graph sample consists of a set of nodes and a set of edges. This section aims to transform the target time series into a graph sample, where the node set y = [y1, y2, ..., y...]. L [ represents the amplitude values ​​of all data points in the target time series.]

[0076] S103. Traverse the slopes of nodes i and j in the node set to obtain the edge set corresponding to the nodes in the target time series; where the range of node i is 0 to L-1, the range of node j is i+1 to L-1, and L is the number of nodes in the node set.

[0077] Optionally, S103 may include:

[0078] When traversing, if the current slope value of node i and node j in the node set corresponding to the current time is greater than the slope threshold, add an undirected edge between node i and node j at the current time, update the slope threshold to the current slope value, and continue traversing.

[0079] After the traversal is complete, we obtain a set of edges consisting of the undirected edges between nodes i and j at each time step.

[0080] Understandable, Figure 3 This is a flowchart of constructing an edge set proposed in an embodiment of the present invention, such as... Figure 3As shown, num represents the target time series, node i and node j are two pointers, len(num) is the length of the target time series, slope is the slope threshold, and nowSlope is the current slope value. Traditional methods first calculate a straight line between two data points in the time series, then iterate through the midpoints between the two points. If a midpoint exists with a slope greater than the straight line, the two points are considered invisible. This invention iterates through and calculates the slopes of nodes i and j, storing the current slope value as the maximum slope value (slope). As node j continues to traverse, it may encounter a larger nowSlope. When nowSlope is greater than slope, nodes i and j are considered visible, and the slope is updated. An undirected edge is connected between nodes i and j in the graph's edge set; otherwise, nodes i and j are considered invisible.

[0081] S104. Update the node set multiple times based on the edge set to obtain multiple node sets with different update counts.

[0082] Optionally, S104 may include:

[0083] Perform word embedding processing on the node set to obtain the node set after word embedding;

[0084] Get the neighbor nodes v of any node v in the set of nodes after word embedding. neighbor The importance of node v; where node v's neighboring nodes v neighbor It is obtained from node v and the set of edges;

[0085] The normalized attention coefficient corresponding to node v is obtained based on node importance and the softmax function.

[0086] Update node v based on the attention coefficient to obtain the updated node v;

[0087] The updated node v is updated again until the preset number of updates is reached, resulting in multiple nodes v with different number of updates, and thus multiple sets of nodes with different number of updates.

[0088] It is understandable that, at this point, the amplitude values ​​of all data points in a target time series are considered as a set of nodes. This can be used to construct an edge set. With the node set and edge set, a high-resolution one-dimensional distance image is constructed as a graph sample, which can then be used for feature extraction and classification. The obtained graph sample is fed into a graph neural network for message propagation and aggregation. Finally, the extracted features are fed into a classifier for graph classification. This invention uses a graph attention network to implement the graph classification task. The node set of the target time series is y = [y1, y2, ..., y...]. LAfter word embedding, the magnitude value of each point is mapped to a high-dimensional vector, resulting in the word-embedded node set V = [v0, v1, v2, ..., v]. M ] T Where M is the number of nodes, and the vector of each node's magnitude value after word embedding is... m = 1, 2, ..., M, and their dimensions are the size F of the word embedding dimension.

[0089] Attention mechanisms are unsupervised, self-learning weight coefficient calculation mechanisms. The example of a single amplitude node in a target time series can be used to explain message propagation and aggregation. For a node v in a target time series and its neighboring nodes, node v is first linearly transformed into a higher-level feature. Then, attention mechanisms a are used to calculate attention coefficients, and the neighboring nodes v... neighbor The importance of node v is represented as follows:

[0090] e neighbor =LeakyReLU(a T [Wv||Wv neighbor ]);

[0091] Among them, e neighbor Represents neighbor node v neighbor For the node importance of node v, LeakyReLU is the activation function, a T Let W be the first learnable parameter and W be the second learnable parameter.

[0092] Optionally, the normalized attention coefficients corresponding to node v are represented as follows:

[0093]

[0094] Where, α neighbor This represents the normalized attention coefficient corresponding to node v, where exp is the exponential function, and M... neighbor Let v represent the set of neighboring nodes of node v. neighbor 'Represents the other neighboring nodes of node v.

[0095] The updated node v can be represented as:

[0096]

[0097] It is worth mentioning that v' represents the updated node v. At this point, node v has completed a feature extraction and updated the vector value. This node v' contains the information of its neighboring nodes and becomes the new feature value. This operation should be performed on all nodes in the graph sample.

[0098] S105. Perform graph classification based on multiple sets of nodes with different update counts to obtain classification results, thereby achieving target recognition of high-resolution one-dimensional range images of radar.

[0099] Optionally, S105 may include:

[0100] Input multiple sets of nodes with different update counts into the graph pooling layer to obtain multiple pooled sets of nodes with different update counts.

[0101] The set of nodes with different number of updates after pooling is input into the fully connected layer to obtain the classification result.

[0102] Understandable, Figure 4 This is a flowchart of a graph classification method proposed in this invention. An attention mechanism is used to focus attention on information useful for extracting target structural features from each distance unit, allowing the model to converge in the correct direction. After message propagation and aggregation, the nodes in the graph sample store feature values, which need to be extracted and integrated into a global feature. Since this invention uses a graph classification task, pooling and activation function layers are added to the graph samples after message propagation and aggregation. The pooling layer downsamples the nodes of a graph sample, reducing the number of nodes while preserving the approximate original structure of the edge connections between nodes. This invention introduces a TopKPooling layer as a pooling layer to reduce the graph size and extract global information, such as... Figure 4 As shown, after each convolution operation, the TopKPooling layer selects nodes with the highest ratio and their corresponding edge information based on the importance scores of node features, thus achieving a downsampling operation on the graph data. This not only helps reduce computation but also preserves key information in the graph to a certain extent, allowing the model to focus on more representative nodes and structures for subsequent learning and feature extraction.

[0103] After multiple convolutional, pooling, and readout operations, the model integrates features extracted from the graph data at different levels using global average pooling, resulting in multiple fixed-dimensional feature vectors. These vectors are then summed to fuse multi-scale information. Subsequently, these fused features undergo further feature transformation and mapping through two fully connected layers. Between the fully connected layers, the ReLU activation function introduces non-linearity, enhancing the model's expressive power and enabling it to learn more complex data patterns. The final layer of the model is a fully connected layer, which maps the input features to the final output dimension and normalizes them using a softmax function to obtain the predicted probability distribution for each category. This design allows the model to perform multi-class classification predictions based on the input graph data, outputting the log probability value of each sample belonging to different categories. This allows for optimization and performance evaluation using the cross-entropy loss function during subsequent training and evaluation.

[0104] In one implementation, Figure 5 This is a comparison chart of the classification accuracy of different models proposed in this embodiment of the invention. This example uses high-resolution one-dimensional distance image data of three types of aircraft to compare with traditional deep learning network models. Under the same experimental data, the accuracy of CNN, RNN, LSTM, and the method proposed in this invention are as follows: Figure 5 As shown; Figure 6 This is a comparison chart of classification accuracy for different graph structures proposed in this embodiment of the invention. This example uses the same graph neural network model but different graph structures to construct graphs for three types of aircraft, obtaining a comparison of classification performance. Under the same measured data, the comparison results of classification performance between different graph structures are as follows: Figure 6 As shown.

[0105] Figure 7 This is a block diagram of a radar target identification device based on a high-resolution one-dimensional range image, as proposed in an embodiment of the present invention. Figure 7 As shown, the device 700 may include:

[0106] The sequence extraction module 701 is used to extract the target time sequence in the preprocessed radar high-resolution one-dimensional range image using a noise threshold.

[0107] The first set acquisition module 702 is used to obtain a set of nodes based on the magnitude value in the target time series;

[0108] The second set acquisition module 703 is used to traverse the slopes of nodes i and j in the node set to obtain the edge set corresponding to the nodes in the target time series; where the range of node i is 0 to L-1, the range of node j is i+1 to L-1, and L is the number of nodes in the node set;

[0109] Update module 704 is used to update the node set multiple times based on the edge set to obtain multiple node sets with different update counts;

[0110] The identification module 705 is used to perform graph classification based on multiple sets of nodes with different update counts to obtain classification results, so as to realize target identification of radar high-resolution one-dimensional range image.

[0111] It should be noted that the terms "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention.

[0112] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0113] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings and the disclosure in carrying out the claimed invention. In the description of the invention, the word "comprising" does not exclude other components or steps, "a" or "an" does not exclude a plurality, and "a plurality" means two or more, unless otherwise explicitly specified. Furthermore, while different embodiments may describe certain measures, this does not mean that these measures cannot be combined to produce good results.

[0114] For the embodiments of the device / electronic device / storage medium, since they are basically similar to the method embodiments, the description is relatively simple, and relevant parts can be referred to in the description of the method embodiments.

[0115] It should be noted that the apparatus of the present invention is an apparatus that applies the radar target recognition method based on the above-mentioned high-resolution one-dimensional range image view. Therefore, all embodiments of the radar target recognition method based on the above-mentioned high-resolution one-dimensional range image view are applicable to this apparatus and can achieve the same or similar beneficial effects.

[0116] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A radar target identification method based on a high-resolution one-dimensional range profile, characterized in that, The method includes: Target time series in preprocessed high-resolution one-dimensional range profiles of radar are extracted using a noise threshold. The node set is obtained based on the amplitude value in the target time series; By traversing the slopes of nodes i and j in the node set, the edge set corresponding to the nodes in the target time series is obtained; wherein, the range of node i is 0 to L-1, the range of node j is i+1 to L-1, and L is the number of nodes in the node set; The node set is updated multiple times based on the edge set to obtain multiple node sets with different update counts; Graph classification is performed on the set of nodes with multiple different update counts to obtain classification results, thereby enabling target identification of the radar high-resolution one-dimensional range image.

2. The radar target recognition method based on a high-resolution one-dimensional range profile as described in claim 1, characterized in that, The extraction of target time series from preprocessed high-resolution one-dimensional range profiles using a noise threshold includes: The high-resolution one-dimensional range image of the radar is normalized to obtain the preprocessed high-resolution one-dimensional range image of the radar. The noise threshold is obtained based on the preprocessed high-resolution one-dimensional range profile of the radar. Using the noise threshold, the preprocessed radar high-resolution one-dimensional range image is searched from front to back for the first range cell that is greater than the noise threshold, which is then used as the starting cell; Using the noise threshold, the preprocessed radar high-resolution one-dimensional range image is searched from back to front for the first range cell that is greater than the noise threshold, which is then used as the ending cell; The target time series is extracted based on the starting unit and the ending unit.

3. The radar target recognition method based on a high-resolution one-dimensional range profile as described in claim 2, characterized in that, The noise threshold is expressed as follows: where threshold represents the noise threshold, K is a preset coefficient, and x norm = [x norm,1 , x norm,2 ,..., x norm,n ,..., x norm,N represents the preprocessed one-dimensional high-resolution range profile of the radar, a, b ∈ [0, N], a < b, and N is the number of range cells in the preprocessed one-dimensional high-resolution range profile of the radar.

4. The radar target recognition method based on a high-resolution one-dimensional range profile as described in claim 1, characterized in that, The slope of node i and node j in the node set is traversed to obtain the edge set corresponding to the node in the target time series, including: When traversing, if the current slope value of node i and node j in the node set corresponding to the current time is greater than the slope threshold, an undirected edge is added between node i and node j at the current time, the slope threshold is updated to the current slope value, and the traversal continues. After the traversal is completed, the set of edges is obtained, consisting of the undirected edges between nodes i and j at each time step.

5. The radar target recognition method based on a high-resolution one-dimensional range profile as described in claim 1, characterized in that, The step of updating the node set multiple times based on the edge set to obtain multiple updated node sets includes: The node set is subjected to word embedding processing to obtain the node set after word embedding; Obtain the neighbor nodes v of any node v in the node set after word embedding. neighbor The importance of node v; wherein, the neighboring nodes v of node v neighbor It is obtained based on the node v and the set of edges; The normalized attention coefficient corresponding to node v is obtained based on the node importance and the softmax function. The node v is updated according to the attention coefficient to obtain the updated node v; The updated node v is updated again until a preset number of updates is reached, resulting in multiple nodes v with different number of updates, and thus a set of nodes with multiple number of updates.

6. The radar target recognition method based on a high-resolution one-dimensional range profile as described in claim 5, characterized in that, The neighbor node v neighbor The importance of node v is represented as follows: e neighbor =LeakyReLU(a T [Wv||Wv neighbor ]); Among them, e neighbor This indicates that the neighbor node v neighbor For the node importance of node v, LeakyReLU is the activation function, a T Let W be the first learnable parameter and W be the second learnable parameter.

7. The radar target recognition method based on a high-resolution one-dimensional range profile as described in claim 6, characterized in that, The normalized attention coefficient corresponding to node v is represented as follows: Where, α neighbor This represents the normalized attention coefficient corresponding to node v, where exp is the exponential function, and M... neighbor Let v represent the set of neighboring nodes of node v. neighbor 'Refers to the other neighboring nodes of node v.

8. The radar target recognition method based on a high-resolution one-dimensional range profile as described in claim 6, characterized in that, The graph classification based on the set of nodes with multiple different update counts includes: The node sets with different update counts are input into the graph pooling layer to obtain multiple pooled node sets with different update counts. The set of nodes with different number of updates after multiple pooling is input into the fully connected layer to obtain the classification result.

9. A radar target identification device based on a high-resolution one-dimensional range image, characterized in that, The device includes: The sequence extraction module is used to extract the target time sequence from the preprocessed radar high-resolution one-dimensional range image using a noise threshold. The first set acquisition module is used to obtain a set of nodes based on the amplitude value in the target time series; The second set acquisition module is used to traverse the slopes of nodes i and j in the node set to obtain the edge set corresponding to the nodes in the target time series; wherein, the range of node i is 0 to L-1, the range of node j is i+1 to L-1, and L is the number of nodes in the node set; An update module is used to update the node set multiple times based on the edge set to obtain multiple node sets with different update counts; The identification module is used to perform graph classification based on the set of nodes with multiple different update counts to obtain classification results, so as to realize target identification of the radar high-resolution one-dimensional range image.

Images