A flight target identification system based on swarm intelligence cooperation

Abstract

The application discloses a kind of flying target identification systems based on swarm intelligence cooperation, including multi-band radar array, database and host computer;Multi-band radar array, database and host computer are sequentially connected, the monitored airspace is irradiated by the multi-band radar array, and radar echo data is stored to the database, and the host computer includes data preprocessing module, graph data reconstruction module, characteristic vector aggregation module and target identification module.The application constructs multi-frequency radar array using multi-band multi-aspect radar, realizes the construction of characteristic vector of multi-band radar echo information based on graph data reconstruction and node aggregation of graph neural network, then obtains multi-target cluster based on clustering algorithm, clusters fitting target characteristic signal, and finally achieves high-precision, high-reliability multi-target detection and identification.

Description

Technical Field

[0001] This invention relates to the field of radar data processing, and in particular to a flight target identification system and method based on swarm intelligence collaboration. Background Technology

[0002] Radar detects and determines the spatial location of targets by utilizing the reflection of electromagnetic waves from them. Radar emits high-power electromagnetic pulses, which are radiated into space by the antenna. When these pulses encounter a target, some of the electromagnetic waves are reflected back, forming an echo signal. The radar's receiving antenna collects these echo signals, and through signal analysis and processing, information such as the target's position, heading, and speed can be obtained.

[0003] Radars operating in different frequency bands possess different characteristics and features. High-frequency radars offer higher resolution, enabling more precise target identification, but suffer from poor penetration and are susceptible to interference from atmospheric conditions, rain, snow, and fog. Low-frequency radars exhibit strong penetration, allowing detection of stealthy or low-flying targets, but their lower resolution makes it difficult to distinguish target details. Therefore, single-frequency radars often fall short of meeting the system's requirements for accurate target localization and identification.

[0004] To overcome the limitations of single-band radar, integrating multi-band radar arrays for target identification is an effective method. Multi-band radar arrays can leverage the advantages of different frequency bands, combining echo signal data from various bands to improve the accuracy and reliability of target identification. Furthermore, multi-band radar arrays can perform omnidirectional, multi-angle, and multi-scale scanning of the airspace, increasing the probability and coverage of target detection while reducing the risk of target escape and interference.

[0005] However, multi-band radar arrays also face several challenges: how to effectively process and analyze massive, heterogeneous, and complex echo signal data; how to extract and fuse effective features from radar echo signals of different frequency bands; and how to achieve collaborative processing and optimization of echo signal data from multiple radar stations. Therefore, realizing swarm intelligence collaborative detection of multi-band radar arrays is not only of significant theoretical and practical value, but also a key focus and challenge in radar collaborative detection both domestically and internationally. Summary of the Invention

[0006] To overcome the shortcomings of single-band radar in achieving both high resolution and high penetration, this invention provides a multi-band, multi-angle, high-precision, high-reliability, and highly intelligent flight target identification system based on swarm intelligence collaboration.

[0007] The technical solution adopted by this invention to solve its technical problem is:

[0008] A flight target recognition system based on swarm intelligence collaboration includes a multi-band radar array, a database, and a host computer, which are connected in sequence.

[0009] A multi-band radar array scans the monitored airspace and stores the echo signal data of each frequency band in the database. The multi-band radar array consists of M radars distributed at different locations, each transmitting a different frequency band B. m (B m =[f L f H Modulation signal of ]) And receive the corresponding frequency band B m echo signal These radars cover the high and low frequency bands of commonly used radars, which is beneficial for obtaining scan echo signals at different frequency bands and acquiring the characteristics of aircraft in the monitored airspace at each frequency band. Here, m refers to the corresponding radar number (m = 1, ..., M, and M ≥ 2). and These refer to the modulated signal transmitted by the m-th radar in the i-th transmission and the radar echo signal received in the i-th transmission, respectively.

[0010] The database aggregates and stores echo signals from M radars located at different locations. The signals from each radar were time-aligned to ensure that the echo signals corresponding to the same index i from different radars were in the same time period, which is beneficial for the feature fusion of echo signals later.

[0011] The host computer includes a data preprocessing module, a graph data reconstruction module, a feature vector aggregation module, and a target recognition module.

[0012] The data preprocessing module is used to preprocess radar group echo data, and the process is as follows:

[0013] ① Collect echo signals from M radars within the same time period i from the database. Subsequently, each echo signal was normalized to obtain the normalized echo signal:

[0014]

[0015] in and Represent The minimum and maximum values.

[0016] ② For the normalized echo signal of each radar Perform a Fourier transform (FFT) to obtain the frequency domain signal.

[0017]

[0018] FFT stands for Fourier Transform, and k represents the frequency index. Performing Fourier Transform on the echo signal transforms it from the time domain to the frequency domain for analysis, which is beneficial for segmenting information in each frequency band.

[0019] ③ For the frequency domain signal of each radar Power spectral density (PSD) estimation is performed to obtain the power spectrum:

[0020]

[0021] Where N represents the number of sampling points. The purpose of power spectral density estimation is to quantify the power level of the echo signal at various frequencies in order to identify different frequency bands B. m The main frequency components in radar echo signals.

[0022] ④ Power spectrum for each radar After filtering, the power spectrum is obtained:

[0023]

[0024] Where H m (k) represents the frequency response function of the filter corresponding to the m-th radar, for each H m (k), we have:

[0025]

[0026] in, The center frequency of the filter Q is the quality factor of the filter:

[0027]

[0028]

[0029]

[0030] Among them, f s k is the sampling rate. L and k H Let B be the frequency variables corresponding to the lower and upper cutoff frequencies of the Butterworth filter, and B be the -3dB bandwidth. The Butterworth filter has a flat amplitude-frequency response and higher smoothness within the passband, meaning that all frequencies within the passband can pass through the filter with a relatively consistent gain. This helps maintain the original amplitude characteristics of the signal, avoids introducing additional amplitude variations, and also reduces frequency response ripple and discontinuities, ensuring that the filtered echo signal is as undistorted as possible.

[0031] ⑤ The filtered power spectrum for each radar Performing an inverse Fourier transform (IFFT) yields the frequency-selective and denoised radar echo signal:

[0032]

[0033] Where IFFT stands for Inverse Fourier Transform.

[0034] The graph data reconstruction module is used to construct graph-structured data from multi-band radar echo signals, and this is accomplished through the following process:

[0035] ① For each radar station, the pre-processed and denoised echo data Representing it as radar array echo map data G i Nodes in The constructed set of nodes is represented as:

[0036]

[0037] This constructs a node set V containing M nodes. i Each node can be represented as a vector, i.e. Where m represents the radar station number and i represents the sampling point number in the time series.

[0038] ② The cosine similarity between each node is calculated as follows:

[0039]

[0040] ③ Construct the set of edges E between nodes i :

[0041] E_i={(m,n)|si^mn≥σ} (12)

[0042] Where σ is the similarity threshold, σ>0, σ<1. When the cosine similarity of the radar echo data between two radar stations m and n is greater than the similarity threshold σ, the two radar stations m and n are considered to have a strong correlation. Therefore, an edge is constructed between the two nodes on the graph.

[0043] ④ Based on the node set and edge set constructed in ① and ③, construct the radar array echo map data corresponding to the i-th time period:

[0044] G i =(V i E i (13)

[0045] The feature vector aggregation module is used to aggregate the feature vectors of each node in the constructed graph data, which is accomplished through the following process:

[0046] ① For a certain node v on the graph m Calculate its neighbor node v n Relative to the importance of this node:

[0047] e mn =Leaky ReLU(a T [Wv m ||Wv n (14)

[0048] Where W is the trainable weight parameter of the feature transformation of the node in this layer, Leaky ReLU(·) is the Leaky ReLU activation function, and α is the weight coefficient. To ensure that the weight coefficient e mn Normalization is applied to nodes of different degrees:

[0049]

[0050] in Represented as node v m All single-hop neighbor nodes.

[0051] ② After calculating the weight coefficients, perform weighted summation and aggregate node features to obtain the center node v. m The new feature vector:

[0052]

[0053] Where || represents the concatenation of output features. It is the attention mechanism of the k-th layer (a k The weight coefficients W obtained by calculating (·)) k These are the weight parameters for the linear transformation of the node features corresponding to the k-th layer. The multi-head attention mechanism enhances the representation capability of the feature vector, allowing attention to different parts of the input at different scales and maximizing the fusion of multi-band echo information.

[0054] ③ In the final (prediction) layer, a multi-attention mechanism is used, employing averaging instead of concatenation to achieve node aggregation, thereby obtaining the final aggregated node feature vector:

[0055]

[0056] The target recognition module is used to aggregate the feature vectors v' of each node. m The identification process is as follows:

[0057] ① For each radar station m, the aggregate node feature vector v' mThe YOLO target detection algorithm was used to detect targets, and the location information of multiple potential targets in the echo signal was obtained.

[0058] L m =YOLO(v' m (18)

[0059] Where YOLO(·) represents the YOLO object detection algorithm, L m This represents the location of multiple potential targets in the echo signal of the m-th radar station.

[0060] ② For each radar station m, the aggregate node feature vector v' m and the locations L of multiple potential targets detected by it m A cut function is used to truncate the signal, resulting in the ROI (Region of Interest) feature vectors of multiple potential targets in the echo signal.

[0061] R m =CUT(v' m L m (19)

[0062] Where CUT represents the truncation function, R m This represents the eigenvectors (ROIs) of multiple potential targets in the echo signal of the m-th radar station. Let ROI represent the eigenvector of the c-th potential target in the echo signal of the m-th radar station.

[0063] ③ For the set of ROI feature vectors of multiple potential targets detected by all radar stations, R m (m = 1, 2, ..., M), cluster them using the DBSCAN algorithm to obtain the set of target clusters in the spatial domain:

[0064] C target =DBSCAN(R1, R2, ..., R M (20)

[0065] Where DBSCAN represents the DBSCAN clustering algorithm, C target Denotes the set of target clusters in the airspace, where This represents the set of ROIs corresponding to the k-th cluster.

[0066] ④ For each target cluster Calculate the average value of its cluster eigenvectors (ROIs):

[0067]

[0068] in This represents the number of ROIs in the ROI cluster, D{·} represents the radar signal analysis function, and S k It represents information such as the position, heading, speed, and category of target k in the airspace.

[0069] This allows for the detection and identification of targets within the airspace monitored by the multi-band radar array, obtaining precise information such as target location, heading, speed, and category from multi-band fusion.

[0070] The technical concept of this invention is as follows: This invention utilizes multi-band multi-directional radar to construct a multi-frequency radar array, and realizes the construction of feature vectors of multi-band radar echo information based on graph data reconstruction and node aggregation of graph neural networks. Subsequently, it obtains multi-target clusters based on clustering algorithms, fits target feature signals in clusters, and finally achieves high-precision and high-reliability multi-target detection and recognition.

[0071] The beneficial effects of this invention are mainly reflected in: 1. Multi-band radar arrays can simultaneously achieve high resolution and high penetration; 2. Fully utilize the network spatiotemporal information of the radar array to achieve high-precision observation; 3. High intelligence and minimal impact from human factors. Attached Figure Description

[0072] Figure 1 This is a hardware structure diagram of the multi-band radar array proposed in this invention;

[0073] Figure 2 This is a system composition diagram proposed in this invention. Detailed Implementation

[0074] The present invention will be further described below with reference to the accompanying drawings. The embodiments of the present invention are used to explain and illustrate the invention, but not to limit it. Any modifications and alterations made to the present invention within the spirit and scope of the claims fall within the protection scope of the present invention.

[0075] Reference Figure 1 , Figure 2 A smart radar maritime target detection system includes a multi-band radar array 1, a database 2, and a host computer 3, which are sequentially connected. The multi-band radar array scans the monitored airspace and stores the echo signal data of each frequency band in the database.

[0076] like Figure 1 As shown, the multi-band radar array consists of M radars distributed at different locations, each transmitting a different frequency band B. m (B m =[f L f H Modulation signal of ]) The radar station covers a frequency band ranging from 300MHz to 12GHz. The resolution of each radar gradually increases from 300MHz to 12GHz, while the penetration decreases accordingly. Therefore, the frequency bandwidth and center frequency of each radar can be adjusted according to the actual number of radars and monitoring needs. The corresponding receiving frequency band for each radar is B. m echo signal Where m refers to the corresponding radar number (m = 1, ..., M, and M ≥ 2), and These refer to the modulated signal transmitted by the m-th radar in the i-th transmission and the radar echo signal received in the i-th transmission, respectively.

[0077] The database aggregates and stores echo signals from M radars located at different locations. The signals from each radar were time-aligned to ensure that echo signals corresponding to the same index i from different radars occurred within the same time period.

[0078] The host computer includes a data preprocessing module, a graph data reconstruction module, a feature vector aggregation module, and a target recognition module.

[0079] The data preprocessing module is used to preprocess radar group echo data, and the process is as follows:

[0080] ① Collect echo signals from M radars within the same time period i from the database. Subsequently, each echo signal was normalized to obtain the normalized echo signal:

[0081]

[0082] in and Represent The minimum and maximum values.

[0083] ② For the normalized echo signal of each radar Perform a Fourier transform (FFT) to obtain the frequency domain signal.

[0084]

[0085] Where FFT stands for Fourier transform, and k represents the frequency index.

[0086] ③ For the frequency domain signal of each radar Power spectral density (PSD) estimation is performed to obtain the power spectrum:

[0087]

[0088] Where N represents the number of sampling points. The purpose of power spectral density estimation is to quantify the power level of the echo signal at various frequencies, since each radar station corresponds to a specific frequency band B. m Therefore, power spectral density estimation in the frequency domain makes it easier to identify and segment different frequency bands B. m The main frequency components in radar echo signals.

[0089] ④ Power spectrum for each radar After filtering, the power spectrum is obtained:

[0090]

[0091] Where H m (k) represents the frequency response function of the filter corresponding to the m-th radar, for each H m (k), we have:

[0092]

[0093] in, The center frequency of the filter Q is the quality factor of the filter:

[0094]

[0095]

[0096]

[0097] Among them, f s k is the sampling rate. L and k H Here, k represents the frequency variables corresponding to the lower and upper cutoff frequencies of the Butterworth filter. L and k H According to the corresponding radar frequency band B m Determined, that is To correspond to the center of the radar frequency band, B represents a -3dB bandwidth. The Butterworth filter has a flat amplitude-frequency response and higher smoothness within the passband, meaning that all frequencies within the passband can pass through the filter with a relatively consistent gain. This helps maintain the original amplitude characteristics of the radar echo signal, avoiding the introduction of additional amplitude variations, while also reducing frequency response ripples and discontinuities, ensuring that the filtered echo signal is as distortion-free as possible.

[0098] ⑤ The filtered power spectrum for each radar Performing an inverse Fourier transform (IFFT) yields the frequency-selective and denoised radar echo signal:

[0099]

[0100] IFFT stands for Inverse Fourier Transform.

[0101] The graph data reconstruction module is used to construct graph-structured data from multi-band radar echo signals, and this is accomplished through the following process:

[0102] ① For each radar station, the pre-processed and denoised echo data Representing it as radar array echo map data G i Nodes in The constructed set of nodes is represented as:

[0103]

[0104] This constructs a node set V containing M nodes. i Each node can be represented as a vector, i.e. Where m represents the radar station number and i represents the sampling point number in the time series.

[0105] ② The cosine similarity between each node is calculated as follows:

[0106]

[0107] ③ Construct the set of edges E between nodes i :

[0108]

[0109] Where σ is the similarity threshold, σ > 0, σ < 1. The similarity threshold σ can be adjusted according to the actual situation, such as dynamic adjustment based on factors such as target size, distance, and motion state. Here we select σ as 0.8. When the cosine similarity of the radar echo data between two radar stations m and n is greater than the similarity threshold σ, the two radar stations m and n are considered to have a strong correlation. Therefore, an edge is constructed between the two nodes in the graph. The edge of the constructed graph is undirected, indicating that the correlation between the two nodes is equal.

[0110] ④ Based on the node set and edge set constructed in ① and ③, construct the radar array echo map data corresponding to the i-th time period:

[0111] G i =(V i E i (13)

[0112] The feature vector aggregation module is used to aggregate the feature vectors of each node in the constructed graph data, which is accomplished through the following process:

[0113] ① For a certain node v on the graph m Calculate its neighbor node v n Relative to the importance of this node:

[0114] e mn =Leaky ReLU(a T [Wv m ||Wv n (14)

[0115] Where W is the trainable weight parameter of the feature transformation of the node in this layer, Leaky ReLU(·) is the Leaky ReLU activation function, and α is the weight coefficient. To ensure that the weight coefficient e mn Normalization is applied to nodes of different degrees:

[0116]

[0117] in Represented as node v m All single-hop neighbor nodes.

[0118] ② After calculating the weight coefficients, perform weighted summation and aggregate node features to obtain the center node v. m The new feature vector:

[0119]

[0120] Where || represents the concatenation of output features. The weight coefficients W are calculated by the attention mechanism (ak(·)) of the k-th layer. k These are the weight parameters for the linear transformation of the node features corresponding to the k-th layer.

[0121] ③ In the final (prediction) layer, a multi-attention mechanism is used, employing averaging instead of concatenation to achieve node aggregation, thereby obtaining the final aggregated node feature vector:

[0122]

[0123] The target recognition module is used to aggregate the feature vectors v′ of each node. m The identification process is as follows:

[0124] ① For each radar station m, the aggregate node feature vector v' m The YOLO target detection algorithm was used to detect targets, and the location information of multiple potential targets in the echo signal was obtained.

[0125] L m =YOLO(v′) m (18)

[0126] Where YOLO(·) represents the YOLO object detection algorithm. For a detailed implementation, please refer to the paper "You Only Look Once: Unified, Real-Time Object Detection.L" published by REDMON J et al. at CVPR 2016. m This represents the location of multiple potential targets in the echo signal of the m-th radar station.

[0127] ② For each radar station m, the aggregate node feature vector v′ m and the locations B of the multiple potential targets detected by it m A cut function is used to truncate the signal, resulting in the ROI (Region of Interest) feature vectors of multiple potential targets in the echo signal.

[0128] R m =CUT(v' m L m (19)

[0129] Where CUT represents the truncation function, R m This represents the eigenvectors (ROIs) of multiple potential targets in the echo signal of the m-th radar station. Let ROI represent the eigenvector of the c-th potential target in the echo signal of the m-th radar station.

[0130] ③ For the set of ROI feature vectors of multiple potential targets detected by all radar stations, R m (m = 1, 2, ..., M), cluster them using the DBSCAN algorithm to obtain the set of target clusters in the spatial domain:

[0131] C target =DBSCAN(R1, R2, ..., R M (20)

[0132] DBSCAN represents the DBSCAN clustering algorithm, a density-based clustering method that does not require pre-determining the number of clusters, can handle outliers, and is suitable for multi-objective clustering. C target Denotes the set of target clusters in the airspace, where This represents the set of ROIs corresponding to the k-th cluster.

[0133] ④ For each target cluster Calculate the average value of its cluster eigenvectors (ROIs):

[0134]

[0135] in This represents the number of ROIs in the ROI cluster, D{·} represents the radar signal analysis function, and S k It represents information such as the position, heading, speed, and category of target k in the airspace.

[0136] This allows for the detection and identification of targets within the airspace monitored by the multi-band radar array, obtaining precise information such as target location, heading, speed, and category from multi-band fusion.

Claims

1. A flight target recognition system based on swarm intelligence collaboration, characterized in that: It includes a multi-band radar array, a database, and a host computer, wherein the multi-band radar array, the database, and the host computer are connected in sequence. The multi-band radar array scans the monitored airspace and stores the radar group echo data of each frequency band into the database for processing by the host computer. The host computer includes: Data preprocessing module: used to preprocess radar group echo data; Graph Data Reconstruction Module: Used to reconstruct preprocessed radar group echo data into graph-structured data; Feature vector aggregation module: used to aggregate the nodes of the constructed graph structure data to obtain feature vectors; Target recognition module: used to identify the feature vectors of each node obtained by aggregation; The feature vector aggregation module is used to aggregate the nodes of the constructed graph data to obtain feature vectors, and this is accomplished through the following process: ① For a certain node on the graph Calculate its neighbor nodes Relative to the importance of this node: (14) in These are the trainable weight parameters for the feature transformation of this node. yes Activation function These are weighting coefficients; to ensure the weighting coefficients... Normalization is applied to nodes of different degrees: (15) in Represented as nodes All single-hop neighbor nodes; ② After calculating the weight coefficients, the node features are aggregated by weighted summation to obtain the center node. The new feature vector: (16) in This represents the concatenation of output features. It is the first Attention mechanism of layers ( The calculated weight coefficients, It is the first The weight parameters for the linear transformation of the node features corresponding to the layer; ③ In the final (prediction) layer, a multi-attention mechanism is employed, using averaging instead of concatenation to achieve node aggregation, thereby obtaining the final aggregated node feature vector: (17)。 2. The flight target recognition system based on swarm intelligence collaboration according to claim 1, characterized in that: The multi-band radar array consists of components distributed at different locations. It consists of several radars, each transmitting on a different frequency band. modulation signal and receive the corresponding frequency band echo signal ,in This refers to the corresponding radar number ( ,and ), and They refer to the first The radar in the first The modulated signal transmitted in the second transmission and the corresponding first transmission The radar echo signal received this time.

3. The flight target recognition system based on swarm intelligence collaboration according to claim 1, characterized in that: The database aggregates and stores data distributed in different locations. radar echo signal The signals from each radar were time-aligned to ensure that different radars had the same index. The corresponding echo signals are in the same time period.

4. The flight target recognition system based on swarm intelligence collaboration according to claim 1, characterized in that: The data preprocessing module is used to preprocess radar group echo data, and the process is as follows: ① Collect from the database Multiple radars at the same time period echo signal Subsequently, each echo signal was normalized to obtain the normalized echo signal: (1) in and Represent The minimum and maximum values; ② For the normalized echo signal of each radar Perform a Fourier transform to obtain the frequency domain signal. (2) in Indicates Fourier transform, Indicates the frequency sequence number; ③ For the frequency domain signal of each radar Power spectral density estimation is performed to obtain the power spectrum: (3) in This indicates the number of sampling points; the purpose of power spectral density estimation is to quantify the power level of the echo signal at various frequencies in order to identify different frequency bands. The main frequency components in radar echo signals; ④ Power spectrum for each radar After filtering, the power spectrum is obtained: (4) in Indicates the first The frequency response function of the filter corresponding to each radar is given. ,have: (5) in, The center frequency of the filter Q is the quality factor of the filter: (6) (7) (8) in, Sampling rate, and Let be the frequency variables corresponding to the lower and upper cutoff frequencies of the Butterworth filter. -3dB bandwidth; ⑤ The filtered power spectrum for each radar Performing an inverse Fourier transform (IFFT) yields the frequency-selective denoised radar echo signal. (9) in This represents the inverse Fourier transform.

5. The flight target recognition system based on swarm intelligence collaboration according to claim 1, characterized in that: The graph data reconstruction module is used to construct graph-structured data from multi-band radar echo signals, and this is accomplished through the following process: ① For each radar station, the pre-processed and denoised echo data This is represented as radar array echo map data. Nodes in The constructed set of nodes is represented as: (10) This will build a collection A set of nodes Each node is represented as a vector, i.e. ,in Indicates the radar station number. Indicates the sampling point number in the time series; ② Calculate the cosine similarity between each node as follows: (11) ③ Construct the set of edges between nodes : (12) in For similarity threshold, When two radar stations The cosine similarity between the radar echo data is greater than the similarity threshold. At that time, it was believed that these two radar stations They are strongly related, so an edge is constructed between these two nodes on the graph; ④ Based on the node set and edge set constructed in ① and ③, construct the first... Radar array echo data corresponding to the time period: (13)。 6. The flight target recognition system based on swarm intelligence collaboration according to claim 1, characterized in that: The target recognition module is used to aggregate the feature vectors of each node. The identification process is as follows: ① For each radar station aggregate node feature vector ,use The target detection algorithm performs target detection on the signal and obtains the location information of multiple potential targets in the echo signal. (18) in express Object detection algorithms Indicates the first The location of multiple potential targets in the echo signals of a radar station; ② For each radar station aggregate node feature vector and the locations of multiple potential targets detected by it. A cut function is used to truncate the signal, resulting in feature vectors of multiple potential targets in the echo signal. : (19) in This represents the truncation function. Indicates the first Feature vectors of multiple potential targets in the echo signals of a radar station , Indicates the first The first radar station's echo signal The feature vector ROI of each potential target; ③ For the set of ROI feature vectors of multiple potential targets detected by all radar stations, The DBSCAN algorithm was used to cluster the target clusters in the spatial domain, resulting in the following set: (20) in express Clustering algorithms Denotes the set of target clusters in the airspace, where Indicates the first The feature vectors corresponding to each cluster gather; ④ For each target cluster Calculate the average value of the ROI (Region of Interest) within its cluster: (21) in This indicates that In the cluster quantity, Represents the radar signal analysis function. Represents targets in the airspace Position, heading, speed, and category information; This allows for the detection and identification of targets within the airspace monitored by the multi-band radar array, obtaining precise target location, heading, speed, and category information through multi-band fusion.

Images