Radar target clustering recognition method and device, electronic equipment and storage medium

Abstract

The application provides a radar target clustering recognition method and device, electronic equipment and a storage medium, wherein the method comprises: acquiring a plurality of sample pairs, each sample pair comprising a target high resolution range profile (HRRP) and a corresponding deep dimension reduction feature; inputting the deep dimension reduction feature in each sample pair into a manifold mapping to obtain a manifold mapping feature, and clustering the manifold mapping feature to obtain a target clustering result; acquiring a target deep dimension reduction feature of a radar target HRRP to be classified, and inputting the target deep dimension reduction feature into the manifold mapping to obtain a target manifold mapping feature; and clustering the target manifold mapping feature according to the target clustering result to determine the clustering label of the radar target HRRP. The scheme can improve the accuracy of the radar target clustering recognition result.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of data processing technology, and in particular to a radar target clustering and identification method, device, electronic device and storage medium. Background Technology

[0002] Currently, target classification and recognition based on artificial intelligence generally rely on deep learning-based classification and recognition of optical images of targets. However, due to environmental and other factors, image-based target clustering and recognition are often limited in many situations, resulting in lower clustering and recognition accuracy. Summary of the Invention

[0003] This invention provides a radar target clustering identification method, apparatus, electronic device, and storage medium, which can improve the accuracy of radar target clustering identification results.

[0004] In a first aspect, embodiments of the present invention provide a radar target clustering and identification method, including:

[0005] Acquire multiple sample pairs, each of which includes the target high-resolution one-dimensional distance image (HRRP) and the corresponding deep dimensionality reduction features;

[0006] The deep dimensionality reduction features of each sample pair are input into the manifold mapping to obtain the manifold mapping features, and the manifold mapping features are clustered to obtain the target clustering result;

[0007] The deep dimensionality reduction features of the radar target to be classified (HRRP) are obtained, and the deep dimensionality reduction features are input into the manifold mapping to obtain the target manifold mapping features.

[0008] Based on the target clustering results, the target manifold mapping features are clustered to determine the clustering label of the radar target HRRP.

[0009] Secondly, embodiments of the present invention also provide a radar target clustering and identification device, comprising:

[0010] The first acquisition unit is used to acquire multiple sample pairs, each sample pair including the target high-resolution one-dimensional distance image HRRP and the corresponding deep dimensionality reduction features;

[0011] The mapping clustering unit is used to input the deep dimensionality reduction features of each sample pair into the manifold mapping to obtain the manifold mapping features, and to cluster the manifold mapping features to obtain the target clustering result;

[0012] The second acquisition unit is used to acquire the target deep dimensionality reduction features of the radar target to be classified (HRRP), and input the target deep dimensionality reduction features into the manifold mapping to obtain the target manifold mapping features;

[0013] The clustering identification unit is used to cluster the target manifold mapping features according to the target clustering results, so as to determine the clustering label of the radar target HRRP.

[0014] Thirdly, embodiments of the present invention also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the method described in any embodiment of this specification.

[0015] Fourthly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the methods described in any embodiment of this specification.

[0016] This invention provides a radar target clustering identification method, device, electronic device, and storage medium. Since manifold mapping can adapt to the stretching of spatial distance, it maps deep dimensionality reduction features to the same three-dimensional space. In this three-dimensional space, the distance and position between each point can be visualized. Thus, clustering the manifold mapping features obtained after manifold mapping can achieve high-accuracy clustering of target HRRP data, which is suitable for target clustering problems and can improve the accuracy of radar target clustering identification results. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a radar target clustering and identification method provided in an embodiment of the present invention;

[0019] Figure 2 This is a flowchart of a testing method provided in an embodiment of the present invention;

[0020] Figures 3-6 This is a traversal diagram of all HRRP data for four typical UAV targets provided in an embodiment of the present invention;

[0021] Figure 7 This is a visualization result of the manifold mapping feature Y provided in an embodiment of the present invention;

[0022] Figure 8 This is a clustering visualization result of manifold mapping feature Y provided in an embodiment of the present invention;

[0023] Figure 9 This is a hardware architecture diagram of an electronic device provided in an embodiment of the present invention;

[0024] Figure 10 This is a structural diagram of a radar target clustering and identification device provided in an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0026] Please refer to Figure 1 This invention provides a radar target clustering and identification method, which includes:

[0027] Step 100: Obtain multiple sample pairs, each sample pair including the target high-resolution one-dimensional distance image HRRP and the corresponding deep dimensionality reduction features;

[0028] Step 102: Input the deep dimensionality reduction features of each sample pair into the manifold mapping to obtain the manifold mapping features, and cluster the manifold mapping features to obtain the target clustering result;

[0029] Step 104: Obtain the target deep dimensionality reduction features of the radar target to be classified by HRRP, and input the target deep dimensionality reduction features into the manifold mapping to obtain the target manifold mapping features;

[0030] Step 106: Cluster the target manifold mapping features according to the target clustering results to determine the clustering label of the radar target HRRP.

[0031] In this embodiment of the invention, since manifold mapping can adapt to the stretching of spatial distance, it maps deep dimensionality reduction features to the same three-dimensional space. In this three-dimensional space, the distance and position between each point can be visualized. Thus, clustering the manifold mapping features obtained after manifold mapping can achieve high-accuracy clustering of target HRRP data, which is suitable for target clustering problems and can improve the accuracy of radar target clustering and identification results.

[0032] The following description Figure 1 The execution method for each step is shown.

[0033] First, for step 100, multiple sample pairs are obtained, each sample pair including the target high-resolution one-dimensional distance image HRRP and the corresponding deep dimensionality reduction features.

[0034] In this embodiment of the invention, the high-resolution range profile (HRRP) is the vector sum of the projections of the complex echoes of the target scattering points onto the radar beam obtained by broadband radar signals. It provides information on the distribution of the target scattering points along the range direction. Its characteristic is that by emitting a high-frequency signal of a certain wavelength, the high-resolution one-dimensional range profile is obtained by reflecting the imaging time and position, thus possessing important structural features of the target.

[0035] To identify the category to which a target belongs, it is necessary to obtain the target HRRP for targets of known categories, and then obtain the corresponding deep dimensionality reduction features for each target HRRP. The target HRRP and the corresponding deep dimensionality reduction features are treated as a sample pair.

[0036] The target can be a drone, a ship, or the like.

[0037] In this embodiment of the invention, the method for obtaining the corresponding deep dimensionality reduction features for the target HRRP can be achieved through traditional dimensionality reduction feature extraction methods, such as principal component analysis, random forest, and forward feature selection.

[0038] Then, for step 102, the deep dimensionality reduction features in each sample pair are input into the manifold mapping to obtain the manifold mapping features, and the manifold mapping features are clustered to obtain the target clustering result.

[0039] Manifold mapping is a visualization algorithm that allows for more complex mappings and typically provides better visualizations. Manifold mapping is suitable for stretching spatial distances and can map deep dimensionality-reduced features to the same three-dimensional space, in which the distances and positions between points can be visualized.

[0040] In one embodiment of the present invention, the manifold mapping can be a t-SNE manifold mapping. A t-SNE manifold mapping finds a two-dimensional representation of the data while preserving the distances between data points as much as possible. The t-SNE manifold mapping provides a random two-dimensional representation of each data point and then attempts to bring closer points in the original feature space closer together and farther points further apart. The t-SNE manifold mapping focuses on closer points rather than preserving the distances between farther points. In other words, it attempts to preserve information that indicates which points are closer together. Therefore, using a t-SNE manifold mapping to perform manifold mapping on deep dimensionality reduction features allows the resulting manifold mapping features to visually represent the clustering results of the features.

[0041] After obtaining the manifold mapping features of multiple samples, clustering can be performed on these features to obtain the target clustering results. Since the category of the target corresponding to each sample is known, the categories of each cluster obtained after clustering the manifold mapping features are also known. Therefore, the target clustering results can be used as a category library to determine the category of subsequent radar targets (HRRP) to be classified.

[0042] Next, for step 104, the deep dimensionality reduction features of the radar target to be classified (HRRP) are obtained, and the deep dimensionality reduction features are input into the manifold mapping to obtain the target manifold mapping features.

[0043] In this embodiment of the invention, after obtaining the HRRP of the radar target to be classified, its deep dimensionality reduction features can be determined by traditional dimensionality reduction feature extraction methods such as principal component analysis. In addition, to improve the speed and accuracy of obtaining deep dimensionality reduction features, in one embodiment of the invention, the following method can also be used to obtain the deep dimensionality reduction features of the HRRP of the radar target to be classified:

[0044] Specifically, after acquiring multiple sample pairs and before acquiring the target deep dimensionality reduction features of the radar target HRRP to be classified, the method may further include: using the target HRRP in the sample pair as input and the deep dimensionality reduction features in the sample pair as output to train the deep autoencoder network to obtain a trained deep network architecture.

[0045] Therefore, obtaining the target deep dimensionality reduction features of the radar target HRRP to be classified may include: inputting the radar target HRRP to be classified into the deep network architecture, and obtaining the target deep dimensionality reduction features based on the output of the deep network architecture.

[0046] Among them, deep autoencoders can perform high-fidelity dimensionality reduction on one-dimensional images. The deep autoencoder network outputs deep dimensionality reduction features relative to an equivalent function. Multiplying the network architecture with the network architecture yields the input target HRRP. By using the deep network architecture trained on the deep autoencoder network, the output deep dimensionality reduction features can be made more accurate.

[0047] In a preferred embodiment, the deep autoencoder network has four layers. The hyperparameters of the deep autoencoder network include the number of neurons in each layer, the ideal activity level of hidden layer neurons, neuron weight parameters, weight descent parameters, and sparsity penalty factor weights. After obtaining the trained deep network architecture, the corresponding hyperparameters are saved. The hyperparameters of the trained deep network architecture include the number of neurons in each layer and the neuron weight parameters. This allows the trained deep network architecture to output accurate deep dimensionality reduction features for the target HRRP.

[0048] Furthermore, to determine whether the deep autoencoder network is fully trained, its accuracy can be tested using a test sample set. For details, please refer to [link to relevant documentation]. Figure 2 The testing method may include:

[0049] Step 200: Input the test sample set into the trained deep autoencoder network to obtain deep dimensionality reduction features X; input the deep dimensionality reduction features X into the manifold mapping to obtain the manifold mapping features Y; the test samples in the test sample set are target HRRPs, and each target HRRP is labeled with its class label;

[0050] By inputting the test sample set into the trained deep autoencoder network, the deep dimensionality reduction feature X corresponding to the test sample set can be output through the saved hyperparameters. Specifically, the dimension of the deep dimensionality reduction feature in the last layer of the four-layer network can be set to 30. For a test sample set consisting of N test samples, the deep dimensionality reduction feature X is an N-row, 30-column matrix vector.

[0051] The dimension of the manifold mapping feature can be set according to actual needs or the required accuracy. For example, the dimension of the manifold mapping feature can be set to 3-dimensional, which facilitates visualization and cluster analysis. In this case, the manifold mapping feature Y is an N-row, 3-column matrix vector.

[0052] Step 202: Cluster the manifold mapping feature Y according to the target clustering result to output the clustering label of each test sample and the clustering accuracy of all test samples;

[0053] Since the target clustering result is obtained by clustering multiple sample pairs using a trained deep autoencoder network, the category of each cluster obtained after clustering is known. Therefore, after clustering the manifold map feature Y, the cluster to which each test sample falls is determined to output the cluster label of each test sample. This cluster label is used to characterize the category of the cluster. The correctness of the clustering is determined based on the known category of the test sample and the corresponding cluster label. Then, the clustering accuracy is determined based on the number of test samples.

[0054] Step 204: Determine that the deep autoencoder network is well trained based on the clustering accuracy.

[0055] One way to determine whether a deep autoencoder network is well trained is to check whether the clustering accuracy exceeds a set threshold. If it does, it means that the network is well trained; otherwise, it means that the network is not well trained and needs to be trained with more sample pairs.

[0056] Finally, for step 106, the target manifold mapping features are clustered according to the target clustering results to determine the clustering label of the radar target HRRP.

[0057] Based on the clustering labels, the category to which the cluster to which the target manifold mapping features fall can be determined, thus realizing the clustering and identification results of radar targets.

[0058] It should be noted that the clustering methods in the embodiments of the present invention can all be implemented using k-means clustering, OPTICS, Divisive, quantum clustering, etc.

[0059] The following describes an embodiment of the invention using clustering of HRRP data from four typical UAV targets (UAV 1, UAV 2, UAV 3, and UAV 4). Please refer to... Figure 3-6 The graphs show the traversal of all HRRP data for four typical UAV targets. Each vertical line in the graph represents an HRRP data sample from a single observation view, and the horizontal line represents the corresponding observation view.

[0060] After dimensionality reduction of the HRRP of the four types of UAVs through a deep network architecture and t-SNE manifold mapping, the visualization results of the manifold mapping feature Y are as follows: Figure 7 As shown, according to Figure 7 It can be seen that most drones of the same type are in the same cluster, with a very few drones located in clusters of other types. The K-means algorithm is then used to cluster the manifold mapping feature Y, and the visualization results are as follows: Figure 8 As shown.

[0061] The final clustering accuracy was 99.5%, compared to... Figure 7 Dimensionality reduction results and Figure 8 The clustering results also show that the four types of drone targets have high clustering accuracy.

[0062] The foregoing has described specific embodiments of the invention. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims may be performed in a different order than those in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0063] like Figure 9 , Figure 10 As shown, this embodiment of the invention provides a radar target clustering and identification device. The device embodiment can be implemented through software, hardware, or a combination of both. From a hardware perspective, as... Figure 9The diagram shown is a hardware architecture diagram of an electronic device containing a radar target clustering and identification device according to an embodiment of the present invention. (Except for...) Figure 9 In addition to the processor, memory, network interface, and non-volatile memory shown, the electronic device in the embodiment may also include other hardware, such as a forwarding chip responsible for processing packets. Taking software implementation as an example, such as... Figure 10 As shown, a device in a logical sense is formed by the CPU of its host electronic device reading the corresponding computer program from non-volatile memory into memory and running it. This embodiment provides a radar target clustering and identification device, including:

[0064] The first acquisition unit 1001 is used to acquire multiple sample pairs, each sample pair including the target high-resolution one-dimensional distance image HRRP and the corresponding deep dimensionality reduction features;

[0065] The mapping clustering unit 1002 is used to input the deep dimensionality reduction features of each sample pair into the manifold mapping to obtain the manifold mapping features, and to cluster the manifold mapping features to obtain the target clustering result;

[0066] The second acquisition unit 1003 is used to acquire the target deep dimensionality reduction features of the radar target to be classified (HRRP), and input the target deep dimensionality reduction features into the manifold mapping to obtain the target manifold mapping features;

[0067] The clustering identification unit 1004 is used to cluster the target manifold mapping features according to the target clustering results, so as to determine the clustering label of the radar target HRRP.

[0068] In one embodiment of the present invention, the device may further include: a training unit, used to train a deep autoencoder network by taking the target HRRP in the sample pair as input and the deep dimensionality reduction features in the sample pair as output, so as to obtain a trained deep network architecture.

[0069] The second acquisition unit is specifically used to input the HRRP of the radar target to be classified into the deep network architecture, and to obtain the deep dimensionality reduction features of the target based on the output of the deep network architecture.

[0070] In one embodiment of the present invention, the device may further include:

[0071] The testing unit is used to input the test sample set into the trained deep autoencoder network to obtain deep dimensionality reduction features X; the deep dimensionality reduction features X are input into the manifold mapping to obtain the manifold mapping features Y; the test samples in the test sample set are target HRRPs, and each target HRRP is labeled with its class label; the manifold mapping features Y are clustered according to the target clustering results to output the clustering label of each test sample and the clustering accuracy of all test samples; the deep autoencoder network is determined to be well trained based on the clustering accuracy.

[0072] It is understood that the structures illustrated in the embodiments of the present invention do not constitute a specific limitation on a radar target clustering and identification device. In other embodiments of the present invention, a radar target clustering and identification device may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0073] The information interaction and execution process between the modules in the above-mentioned device are based on the same concept as the method embodiment of the present invention, and the specific details can be found in the description of the method embodiment of the present invention, and will not be repeated here.

[0074] This invention also provides an electronic device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements a radar target clustering and identification method according to any embodiment of this invention.

[0075] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform a radar target clustering and identification method according to any embodiment of this invention.

[0076] Specifically, a system or apparatus equipped with a storage medium may be provided, on which software program code implementing the functions of any of the embodiments described above is stored, and the computer (or CPU or MPU) of the system or apparatus may read and execute the program code stored in the storage medium.

[0077] In this case, the program code read from the storage medium can itself implement the function of any of the above embodiments, and therefore the program code and the storage medium storing the program code constitute part of the present invention.

[0078] Examples of storage media used to provide program code include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, program code can be downloaded from a server computer via a communication network.

[0079] Furthermore, it should be clear that not only can the program code read by the computer be executed, but also the operating system or other components operating on the computer can be instructed based on the program code to perform some or all of the actual operations, thereby realizing the function of any of the embodiments described above.

[0080] Furthermore, it is understood that the program code read from the storage medium is written to the memory set in the expansion board inserted into the computer or to the memory set in the expansion module connected to the computer. Then, based on the instructions of the program code, the CPU or other components installed on the expansion board or expansion module execute some and all of the actual operations, thereby realizing the function of any of the above embodiments.

[0081] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0082] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as ROM, RAM, magnetic disk, or optical disk.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A radar target clustering and identification method, characterized in that, include: Acquire multiple sample pairs, each of which includes the target high-resolution one-dimensional distance image (HRRP) and the corresponding deep dimensionality reduction features; The deep dimensionality reduction features of each sample pair are input into the manifold mapping to obtain the manifold mapping features, and the manifold mapping features are clustered to obtain the target clustering result; The target HRRP in the sample pair is used as input, and the deep dimensionality reduction features in the sample pair are used as output to train the deep autoencoder network, resulting in a trained deep network architecture. The HRRP of the radar target to be classified is input into the deep network architecture. Based on the output of the deep network architecture, the deep dimensionality reduction feature of the target is obtained. The deep dimensionality reduction feature of the target is then input into the manifold mapping to obtain the target manifold mapping feature. Based on the target clustering results, the target manifold mapping features are clustered to determine the clustering label of the radar target HRRP; After training the deep autoencoder network and before obtaining the trained deep network architecture, the method further includes: inputting the test sample set into the trained deep autoencoder network to obtain deep dimensionality reduction features X. The deep dimensionality reduction feature X is input into the manifold mapping to obtain the manifold mapping feature Y; the test samples in the test sample set are target HRRPs, and each target HRRP is labeled with its class label; the manifold mapping feature Y is clustered according to the target clustering results to output the clustering label of each test sample and the clustering accuracy of all test samples; the deep autoencoder network is determined to be well trained based on the clustering accuracy.

2. The method according to claim 1, characterized in that, The hyperparameters of the deep autoencoder network include the number of neurons in each layer, the ideal activity of hidden layer neurons, the weight parameters of the neural network, the weight descent parameters, and the sparsity penalty factor weights. The hyperparameters of a trained deep network architecture include the number of neurons in each layer and the network weight parameters of the neurons.

3. The method according to any one of claims 1-2, characterized in that, The manifold mapping is a t-SNE manifold mapping.

4. A radar target clustering and identification device, characterized in that, include: The first acquisition unit is used to acquire multiple sample pairs, each sample pair including the target high-resolution one-dimensional distance image HRRP and the corresponding deep dimensionality reduction features; The mapping clustering unit is used to input the deep dimensionality reduction features of each sample pair into the manifold mapping to obtain the manifold mapping features, and to cluster the manifold mapping features to obtain the target clustering result; The second acquisition unit is used to acquire the target deep dimensionality reduction features of the radar target to be classified (HRRP), and input the target deep dimensionality reduction features into the manifold mapping to obtain the target manifold mapping features; A clustering identification unit is used to cluster the target manifold mapping features according to the target clustering results, so as to determine the clustering label of the radar target HRRP; It also includes: a training unit, which takes the target HRRP in the sample pair as input and the deep dimensionality reduction features in the sample pair as output to train the deep autoencoder network and obtain the trained deep network architecture. The second acquisition unit is specifically used to input the HRRP of the radar target to be classified into the deep network architecture, and to obtain the deep dimensionality reduction features of the target based on the output of the deep network architecture; It also includes: a testing unit, used to input a test sample set into the trained deep autoencoder network to obtain deep dimensionality reduction features X; the deep dimensionality reduction features X are input into the manifold mapping to obtain manifold mapping features Y; the test samples in the test sample set are target HRRPs, and each target HRRP is labeled with its class label; the manifold mapping features Y are clustered according to the target clustering results to output the clustering label of each test sample and the clustering accuracy of all test samples; and the deep autoencoder network is determined to be well trained based on the clustering accuracy.

5. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor, when executing the computer program, implements the method as described in any one of claims 1-3.

6. A computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method of any one of claims 1-3.

Images