A near-field spatial spectrum classification and retrieval method based on extreme value position difference features
Through the method based on the extreme position difference characteristics, the problem of Fresnel near-field spatial spectrum is not efficient and poor real-time in the target positioning process, efficient classification and retrieval is achieved, and factors such as multi-objective and frequency offset are adapted to the following factors The impact of the system improves the reliability of the system.
Patent Information
- Application Number
- CN202411239900.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-05
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-09-05
AI Technical Summary
The prior art is susceptible to factors such as target characteristics, transmission delay, frequency offset and multi-objectives in the target positioning process of Fresnel near-field spatial spectrum, resulting in low efficiency and poor real-time performance.
Using near-field spatial spectrum classification and search method based on extreme position difference characteristics, the search space is divided into potential target points through the mathematical model of the radio transceiver array device, the template spatial spectrum is obtained using simulation, and the complex autocorrelation matrix and singular value decomposition are calculated. , select the maximum extreme point of the left singular vector as the search vector, build a search database, and perform corresponding processing on the measurement space spectrum to extract the search subspace.
Improve the classification and retrieval efficiency of spatial spectrum, enhance the robustness of the classifier, reduce the search subspace, improve the real-time nature of target positioning, adapt to the influence of factors such as multi-objectives and frequency offsets, and improve the reliability of the system. sex.
Smart Images

Figure CN118779775B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data processing, and in particular to a near-field spatial spectrum classification and retrieval method based on extreme value position differential features. Background Art
[0002] Under the premise that the spatial spectra in different areas of the Fresnel near field are independent of each other, the spatial spectrum can be used for target positioning. The near-field spatial spectrum obtained through actual testing is called the measured spatial spectrum. The near-field spatial spectrum corresponding to the target position obtained through simulation is called the template spatial spectrum. The target position can be detected by correlating the measured spatial spectrum with all template spatial spectra. Due to the large number of potential target positions and the large number of template spatial spectra, the search space is extremely large and the algorithm has poor real-time performance. Although the algorithm based on extreme value correlation can reduce the amount of calculation, it cannot avoid traversing the space. According to the electromagnetic field characteristics of signal propagation, selecting appropriate feature vectors to represent the near-field spatial spectrum and dividing the entire search space into several smaller subspaces through the retrieval system is an effective means to simplify the target positioning process. The clustering and retrieval method based on extreme values is not efficient in scenarios that are easily affected by factors such as target characteristics, transmission delay, frequency offset, and multiple targets. Summary of the invention
[0003] Aiming at the problem that the near-field spatial spectrum is easily affected by factors such as target characteristics, transmission delay, frequency offset, and multiple targets, the present invention proposes a near-field spatial spectrum classification and retrieval method based on extreme value position differential features.
[0004] The technical solution adopted by the present invention is as follows:
[0005] A near-field spatial spectrum classification and retrieval method based on extreme value position differential features comprises the following steps:
[0006] Step 1: Based on the mathematical model of the radio transceiver array device, the search space is divided into potential target points, a unit radiation source is placed at each potential target point, and the array receiving signal is obtained by simulation; then a sampling matrix composed of independent steering vectors is used The array received signal is weighted to obtain the template space spectrum of each potential target point position, and the template space spectrum is converted into a complex vector ,all Constructing the template library dataset ;in, is the set value, q∈[1,Q] , is the sampling matrix Dimensions;
[0007] Step 2: For each template space spectrum , calculate the complex autocorrelation matrix G =[ g ij ]= f q f q H , then take the triangular matrix and set the main diagonal to zero, and get the matrix ;
[0008] Step 3: For each matrix Perform singular value decomposition: , get the left singular vector matrix U =[ u 1 , u 2 … u P ] ;
[0009] Step 4: For the left singular vector matrix , select the left singular vector The row numbers of the first N largest extreme points i q =[ i 1 q , i 2 q … i N q ] T , as the template space spectrum The search vector of; wherein, N is a set value;
[0010] Step 5: Retrieve vector , forming a search record ,in c q =[ x y z ] T yes The coordinates of
[0011] Step 6: According to the search vector , for all search records Classify and construct and The mapping relationship between them: , forming a search database ;
[0012] Step 7: The radio transceiver array device transmits a coherent signal to the observation space, performs target detection, obtains array reception data, and uses a sampling matrix composed of independent steering vectors. Weight the array received data to obtain the measured spatial spectrum ;
[0013] Step 8: Measure the spatial spectrum , calculate the complex autocorrelation matrix , then take the triangular matrix and set the main diagonal to zero, and get the matrix ;
[0014] Step 9, for the matrix Perform singular value decomposition: , get the left singular vector matrix U ́ =[ u 1 ́ , u 2 ́ … u p ́ ] ;
[0015] Step 10, for the left singular vector matrix , select the left singular vector The row numbers of the first N largest extreme points i e =[ i 1 e , i 2 e … i N e ] T , as the measurement space spectrum The search vector of
[0016] Step 11: Use the mapping relationship , from the search database A search subspace is extracted and measured that is related to the spatial spectrum.
[0017] Due to the adoption of the above technical solution, the beneficial effects of the present invention compared with the prior art are:
[0018] The present invention uses the position difference vector between multiple extreme elements of the spatial spectrum as the classification feature, gets rid of the absolute coordinate constraint, reflects the relationship of the internal structure of the spatial spectrum, captures the change pattern of the spatial spectrum in different directions, is invariant to transformations such as translation and scaling of the signal, is more suitable as a classification feature vector, and can enhance the robustness of the classifier.
[0019] The present invention realizes efficient classification and retrieval of the Fresnel near-field spatial spectrum database through feature extraction, feature dimension reduction and data set classification, and effectively improves the recall rate.
[0020] The classification method of the present invention based on the extreme value position differential feature greatly reduces the search subspace, reduces the complex vector correlation operation, and improves the real-time performance of target positioning.
[0021] The present invention can better adapt to the scenario where the near-field spatial spectrum is easily affected by factors such as target characteristics, transmission delay, frequency offset, multiple targets, etc., and improves the reliability of the classification and retrieval system. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] Figure 1 Schematic diagram of a radio transceiver array device in an embodiment of the present invention.
[0023] Figure 2 Graph showing the spatial spectrum distribution of the Fresnel near field in an embodiment of the present invention.
[0024] Figure 3 Graph showing the distribution of the template spatial spectrum in the Fresnel near-field spatial spectrum in an embodiment of the present invention.
[0025] Figure 4 4 is a processing flow chart of a near-field spatial spectrum classification and retrieval method based on extreme value position differential features in an embodiment of the present invention. DETAILED DESCRIPTION
[0026] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.
[0027] The present invention provides a near-field spatial spectrum classification and retrieval method based on extreme value position differential features, such as Figure 4 As shown, the specific steps include:
[0028] Step 1, based on Figure 1 Mathematical model of radio transceiver array device, by simulating a sampling matrix composed of independent steering vectors Perform spatial scanning, place a unit radiation source at each possible target position, obtain the complex distribution map of the template space spectrum, and build a template library data set.
[0029] Specifically, the near-field spatial spectrum classification and retrieval method based on extreme value position differential features in this embodiment is executed by an electronic device such as a computer or the hardware and / or software therein. Figure 1 As shown in the figure, it is a schematic diagram of the radio transceiver array device in this embodiment, assuming that the three-dimensional aperture of the radio transceiver array device is , 100 radio transceivers are randomly arranged inside, working at L-band 1.25GHz, with a corresponding beam width of 0.00175 degrees. All radio transceivers are ideally synchronized in time, frequency and space, and various phase zero values are accurately corrected, detecting near-field targets in a fully coherent state, Figure 1 Where λ is the operating wavelength of the radio transceiver, and ~ means approximately. Figure 2 As shown in FIG. 1 , it is a schematic diagram of the measurement space spectrum in this embodiment. Figure 3 As shown, in this embodiment, a range of about Q=5.124×10 5 The dotted box represents a template spatial spectrum. The angular spacing between potential target points is 0.002625 degrees, which is greater than 1.5 times the array beam width. This spacing can ensure the independence between the template spatial spectra. The center position of these template spatial spectra is the target position.
[0030] By modeling the radio transceiver array device and placing a unit radiation source at each potential target point , the array receiving signal is obtained by simulation. Sampling matrix It consists of 1024 steering vectors. The angular interval between the steering vectors is greater than 0.002625 degrees to ensure their independence. Weighted, obtain the template space spectrum corresponding to each potential target point position .all Constructing the template library dataset .
[0031] Step 2: For each template space spectrum , calculate the complex autocorrelation matrix G =[ g ij ]= f q f q H ∈ C 1024 × 1024 , The number of elements in is 1,048,576, which is very large and needs feature extraction for dimensionality reduction. For feature dimensionality reduction, only the Hermite matrix The lower triangular matrix of . To prevent The autocorrelation results of the extreme values affect the feature extraction. Set the main diagonal to zero, and get the matrix .
[0032] Step 3: For each matrix Perform singular value decomposition: , get the left singular vector matrix U =[ u 1 , u 2 … u P ] , extract the feature vector.
[0033] Step 4: For the left singular vector matrix , select the left singular vector The first N = 15 maximum extreme points , the row numbers of the first N=15 maximum extreme points are i q =[ i 1 q , i 2 q … i 15 q ] T ,Will As a template space spectrum The retrieval vector is further used to reduce the data dimension.
[0034] Step 5: Retrieve vector , forming a search record ,in c q =[ x y z ] T yes The coordinates of q∈[1 , 5.124 × 10 5 ] .
[0035] Step 6: According to the search vector , for all search records Classify and divide the entire search space into multiple search subspaces to construct and The mapping relationship between them: , forming a search database ,in q∈[1 , 5.124 × 10 5 ] .
[0036] Step 7: The radio transceiver array device transmits a coherent signal to the observation space to detect the target, obtain the array receiving data, and use the data in step 1 to Weight the array received data to obtain the measured spatial spectrum .
[0037] Step 8: Measure the spatial spectrum , calculate the complex autocorrelation matrix , then take the triangular matrix and set the main diagonal to zero, and get the matrix .
[0038] Step 9, for the matrix Perform singular value decomposition: , get the left singular vector matrix U ́ =[ u 1 ́ , u 2 ́ … u p ́ ] .
[0039] Step 10, for the left singular vector matrix , select the left singular vector The row numbers of the first N=15 maximum extreme points of i e =[ i 1 e , i 2 e … i 15 e ] T , as the measurement space spectrum The search vector.
[0040] Step 11: Use the mapping relationship , from the search database A search subspace is extracted and measured that is related to the spatial spectrum.
[0041] Due to the superposition characteristics of electromagnetic fields, the composite Fresnel field of multiple targets will cause the characteristics of the near-field spatial spectrum to change. The retrieval capability of the near-field spatial spectrum feature vector can be evaluated by the measurement spatial spectrum formed by multiple targets. 10,000 random double-target measurement spatial spectra are generated. , and match it with the feature vector of the template spatial spectrum of the target contained in it, and observe the number of features that can be matched to evaluate the target retrieval ability of the feature vector. The singular vector and the extreme value vector are used as feature vectors to measure the spatial spectrum. Template Space Spectrum The comparison of the retrieval capabilities is shown in Table 1. Among them, The representative singular value is used as the feature vector, and the sample distribution number of 10,000 dual-target measurement space spectra under different effective feature numbers; The number of sample distributions of 10,000 dual-target measurement space spectra under different effective feature numbers when the representative extreme value is used as the feature vector; The sample distribution percentage of 10,000 dual-target measurement space spectra under different effective feature numbers when the representative extreme value is used as the feature vector; The representative singular value is used as the feature vector, and the sample distribution number of 10,000 dual-target measurement space spectra under different effective feature numbers.
[0042] In measuring spatial spectrum In the case of dual targets, some elements in each target vector feature will be destroyed to varying degrees. In this case, a good feature vector can retain more undamaged feature elements. Therefore, using this feature vector, more accurate target classification and retrieval can be performed. As can be seen from Table 1, the number of samples with only one feature element retained by the singular vector feature is only 7, accounting for 0.07%. However, the number of samples with only one feature element retained by the extreme value vector feature is 69, accounting for 0.69%, nearly 10 times more. In the comparison of the number of samples with 2 to 11 feature elements retained, the singular vector feature also has a similar obvious advantage. This proves that the singular vector feature has better multi-target retrieval capabilities than the extreme value vector feature.
[0043] Table 1 Comparison of retrieval characteristics of singular vector features and extreme value vector features
[0044]
[0045] In summary, the present invention aims at the problem that the near-field spatial spectrum is easily affected by factors such as path attenuation, target characteristics, and multi-target aliasing, and proposes a near-field spatial spectrum classification and retrieval method based on extreme position differential features, which improves the robustness of the classification retrieval system. It is used to divide the entire search space into several smaller subspaces, construct a classification system for the subspaces, reduce the amount of related calculations, and improve real-time performance.
Claims
1. A near-field spatial spectrum classification and retrieval method based on extreme position differential features, characterized in that: The following steps are involved: Step 1: Based on the mathematical model of the radio transceiver array device, the search space is divided into potential target points, a unit radiation source is placed at each potential target point, and the array receiving signal is obtained by simulation; then a sampling matrix composed of independent steering vectors is used The array received signal is weighted to obtain the template space spectrum of each potential target point position, and the template space spectrum is converted into a complex vector ,all Constructing the template library dataset ;in, is the set value, , is the sampling matrix Dimensions; Step 2: For each template space spectrum , calculate the complex autocorrelation matrix , then take the triangular matrix and set the main diagonal to zero, and get the matrix ; Step 3: For each matrix Perform singular value decomposition: , get the left singular vector matrix ; Step 4: For the left singular vector matrix , select the left singular vector The row numbers of the first N largest extreme points , as the template space spectrum The search vector of; wherein, N is a set value; Step 5: Retrieve vector , forming a search record ,in It is the template space spectrum The coordinates of Step 6: According to the search vector For all search records Classify and construct and The mapping relationship between them: , forming a search database ; Step 7: The radio transceiver array device transmits a coherent signal to the observation space, performs target detection, obtains array reception data, and uses a sampling matrix composed of independent steering vectors. Weight the array received data to obtain the measured spatial spectrum ; Step 8: Measure the spatial spectrum , calculate the complex autocorrelation matrix , then take the triangular matrix and set the main diagonal to zero, and get the matrix ; Step 9, for the matrix Perform singular value decomposition: , get the left singular vector matrix ; Step 10, for the left singular vector matrix , select the left singular vector The row numbers of the first N largest extreme points , as the measurement space spectrum The search vector of Step 11: Use the mapping relationship , from the search database A search subspace is extracted and measured that is related to the spatial spectrum.
