Radiation source signal sorting method based on grid voting and adaptive peak detection

Abstract

The invention discloses a radiation source signal sorting method based on grid voting and adaptive peak detection, which comprehensively considers multi-target and target-free complex environments, designs an improved grid voting and adaptive detection method, and adopts an adaptive multi-target peak detection algorithm. The method can effectively achieve the simultaneous detection of a plurality of targets in a voting scene, can avoid the misjudgment when the target does not exist, and improves the signal sorting performance in a complex electromagnetic environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of radar, and particularly relates to a method for sorting radiation source signals based on grid voting and adaptive peak detection. Background Art

[0002] The sorting of radar radiation source pulse signals is an important link in the field of electronic reconnaissance, and it is an important prerequisite and basis for radar signal feature extraction, recognition, and threat assessment. Traditional signal sorting methods based on PRI (Pulse Repetition Interval) include the cumulative difference histogram method, the sequential difference histogram method, the PRI transform method, and the plane transform method, etc., which have good sorting effects for specific PRI modulation type signals. However, the reconnaissance area of the space-based reconnaissance platform is wide, and the number of radiation sources obtained is huge. With the increase in the number of radar radiation sources, the pulse density increases, the pulse parameter overlap is serious, and the inter-pulse information correlation is weak. The existing PRI-based sorting methods are difficult to accurately sort different radar radiation source pulse signals with similar frequency and pulse width parameters and complex PRI changes. However, for the space-based rotating interferometer system, the grid voting sorting method can effectively improve the limitations of the traditional PRI-based signal sorting method and achieve high-precision sorting of radiation source pulse signals.

[0003] The traditional grid voting sorting method uses a rotating interferometer to measure the phase difference information of radar radiation source pulse signals, generates a voting matrix, and realizes the sorting of radar radiation source pulse signals by finding the voting peak. It has good sorting accuracy in single-target voting scenarios. However, due to the wide coverage area of the space-based reconnaissance satellite, it will face a complex electromagnetic environment and multi-target voting scenarios. Relying solely on the traditional voting sorting method of finding the position with the largest peak will cause the loss of targets or false alarms, resulting in a sharp decline or even failure of the signal sorting performance. Therefore, it is of great significance to effectively sort multi-radiation source signals in a complex electromagnetic environment. Summary of the Invention

[0004] In order to overcome the deficiencies of the prior art, the present invention provides a method for sorting radiation source signals based on grid voting and adaptive peak detection. Considering the complex environment with multiple targets and no targets, an improved grid voting and adaptive detection method is designed, and an adaptive multi-target peak detection algorithm is adopted, which can effectively realize the simultaneous detection of multiple targets in the voting scenario, and at the same time can avoid misjudgment when there are no targets, and improve the signal sorting performance in a complex electromagnetic environment.

[0005] The technical solution adopted by the present invention to solve its technical problems is as follows:

[0006] Step 1: Construction of the voting matrix;

[0007] Step 2: Adaptive multi-target peak detection;

[0008] Step 3: Multi-target association fusion;

[0009] Preferably, the step 1 is specifically:

[0010] Set the initial parameters, divide the observation area evenly into M*M grids, and define the center point of each grid as the geographical location x of the grid; at time t, the radar radiation source pulse signal comes from the m=1, 2, ..., M 2 The phase difference φ when the grid t Satisfies the following relationship:

[0011] φ t =2πη[u t (x)] T b t (1)

[0012] where b t represents the unit vector of the rotating interferometer at time t, which is determined by the spatial position of the two channels of the rotating interferometer; u t (x) represents the unit vector determined by the grid position x and the satellite position at time t; η represents the baseline wavelength ratio;

[0013] The theoretical phase difference value of each grid is calculated by formula (1). If the difference between the theoretical phase difference value and the measured value is less than the preset threshold T, the radar radiation source pulse signal is voted to the grid corresponding to the theoretical phase difference value, otherwise no voting is performed. All time points and grids are traversed to finally generate the voting matrix B.

[0014] Preferably, the step 2 is specifically:

[0015] Step 2-1: When there is no target in the search area, the generated voting matrix distribution is only generated by background noise and approximately obeys an independent and identically distributed Gaussian distribution, that is, B(i,j)~N(μ,σ 2 ), by estimating the mean μ and standard deviation σ of the entire matrix, the voting matrix is standardized to Z(i,j), that is:

[0016]

[0017] Where B(i,j) represents the number of votes in grid (i,j); i,j=1,2,...M represents the row index and column index in the grid voting matrix;

[0018] When Z(i,j)≈0, the grid is noise, and when Z(i,j)>>0, the grid may be a target;

[0019] Extract the maximum value of the standardized voting matrix based on outlier detection and generalized likelihood ratio As a statistic, this statistic measures the deviation between the extreme values in the normalized voting matrix and the background noise; if there is a target in the search area, there will be one or more extreme value deviations, existing in the form of outliers; extract the top K maximum values {Z (1) , Z (2) ,..., Z (K)} of the normalized voting matrix Z(i, j) to construct the statistic:

[0020]

[0021] Use this statistic to measure the overall deviation between the extreme values and the background noise. If the overall deviation exceeds the significance threshold γ, it is determined that there is a target in the search area; otherwise, it is considered that there is no target in the search area.

[0022] Step 2-2: For each grid, set up a protection unit and a reference unit around it. Use B(i, j) to represent the number of votes in the (i, j)th grid, and the reference unit is represented as B ref , and the noise in the reference unit follows a Gaussian distribution μ ref represents the mean of the number of votes in the reference unit, and σ ref represents the variance of the number of votes in the reference unit; conduct statistical analysis on the voting values in the reference unit, and generate a dynamic threshold T(i, j) according to the preset false alarm rate P FA . The specific formula is:

[0023] T(i, j) = μ ref + k * σ ref

[0024]

[0025] where N represents the number of reference units, B ref,k represents the voting value in the kth reference unit, k = Q -1 (P FA ), and Q -1 represents the upper quantile function of the standard normal distribution;

[0026] For each grid (i, j), compare its voting number B(i, j) with the dynamic threshold T(i, j). If the voting value exceeds the dynamic threshold, it is considered that there is a radar radiation source target in this grid. After traversing all grids, a radar radiation source target set V is obtained.

[0027] Preferably, the specific content of step 3 is:

[0028] Associate and fuse the targets in the target set V. Specifically, for any two radar radiation source targets a and b, calculate the Euclidean distance between them:

[0029]

[0030] where x a and y a represent the positions of the radar radiation source target a, and x b and y b represent the positions of the radar radiation source target b;

[0031] Set the association threshold d, traverse all the targets in the target set V. For each pair of radar radiation source targets a and b, calculate the Euclidean distance between the radar radiation source targets and compare it with the association threshold; if D a,b < d, then average the positions of the two targets to fuse them into one radar radiation source target; if D a,b ≥ d, then it is determined that the two targets have sufficient separation in space and belong to two independent targets;

[0032] During the traversal process, continuously update the target set until all targets have completed the association judgment and fusion operation to obtain the final radar radiation source target set.

[0033] Preferably, the significance threshold γ is set to the ±3σ standard, that is, γ = 3.

[0034] A computer program that causes a computer to execute the above-mentioned radiation source signal sorting method.

[0035] An electronic device, including: a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory so that the electronic device executes the above-mentioned radiation source signal sorting method.

[0036] A computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned radiation source signal sorting method is implemented.

[0037] A chip, including: a processor, which is used to call and run a computer program from a memory, so that a device equipped with the chip executes the above-mentioned radiation source signal sorting method.

[0038] A computer program product, the computer program product includes a computer storage medium, the computer storage medium stores a computer program, the computer program includes instructions that can be executed by at least one processor, and when the instructions are executed by the at least one processor, the above-mentioned radiation source signal sorting method is implemented.

[0039] The beneficial effects of the present invention are as follows:

[0040] The present invention performs adaptive multi-objective peak detection on the generated voting matrix, enabling effective sorting of radar emitter signals.

[0041] The invention has the following effects:

[0042] 1. Compared with the traditional PRI-based sorting method, it can effectively sort and locate radar emitter signals with close parameters and complex PRI modulation variations.

[0043] 2. Compared with the traditional voting sorting method, it can effectively avoid false target judgments caused by the appearance of peaks with low votes in a radar emitter target-free environment, improving the robustness of the sorting method.

[0044] 3. In a voting scenario with multiple radar emitters present, it can achieve accurate detection and sorting of wide-area multiple emitters, improving the sorting accuracy in a complex electromagnetic environment. Description of the Drawings

[0045] Figure 1 It is a flowchart of the method of the present invention.

[0046] Figure 2 It is the grid voting and detection results in a target-free environment. (a) is the grid voting result diagram, (b) is the detection result of the present invention, and (c) is the detection result of the traditional method.

[0047] Figure 3 It is the grid voting and detection results in a multi-emitter target environment. (a) is the grid voting result diagram, (b) is the detection result of the present invention, and (c) is the detection result of the traditional method. Detailed Embodiment

[0048] The following further describes the present invention in conjunction with the drawings and embodiments.

[0049] The present invention proposes a method for sorting emitter signals based on grid voting and adaptive peak detection, which can effectively sort multi-emitter signals in a complex electromagnetic environment and also achieve the positioning of radar emitters. The simulation results show that compared with the traditional voting sorting method, the improved grid voting and adaptive detection signal sorting method proposed by the present invention can accurately sort multiple targets simultaneously and avoid false judgments when the target does not exist.

[0050] As Figure 1 shown, the present invention provides a method for sorting emitter signals based on grid voting and adaptive peak detection, mainly including the following steps:

[0051] Step 1: Voting matrix construction. Set initial parameters, evenly divide the observation area into M grids, and define the center point of each grid as the geographical location x of the grid. For the t-th moment, when the radar radiation source pulse signal comes from the m-th (m = 1, 2,..., M) grid, the phase difference satisfies the following relationship:

[0052] φ t = 2πη[u t (x)] T b t

[0053] where b t represents the unit vector of the rotating interferometer at the t-th moment, which is determined by the spatial positions of the two channels of the rotating interferometer. u t (x) represents the unit vector determined by the grid position x and the satellite position at the t-th moment, and η represents the baseline wavelength ratio;

[0054] Based on the calculation of the above formula, the theoretical value of the phase difference of each grid can be calculated. If the difference between the theoretical value and the measured value is less than the preset threshold T, the radar radiation source pulse signal is voted to the grid corresponding to the phase difference, otherwise no voting is performed. Traverse all moments and grids, and finally generate the voting matrix B.

[0055] Step 2-1: Adaptive multi-target peak detection. For the voting matrix B obtained in the above step, peaks will be formed at the grids where the radar radiation sources are located. However, in a complex electromagnetic environment, due to interference factors such as noise, phase ambiguity, and measurement errors of the rotating interferometer, there will be a large number of false peaks in the voting matrix. When there are no radar radiation source targets in the search area, the voting results are as Figure 2 (a) shows, and there are a large number of false peaks with low votes. When there are multiple radar radiation source targets in the search area, the voting results are as Figure 3 (a) shows, and there will be a large number of peaks, far exceeding the number of real targets. If only relying on the maximum value of the peak or the number of peaks to determine the target position, false targets will appear, affecting the final sorting effect. To improve the accuracy of multi-target detection, this patent proposes an adaptive multi-target peak detection algorithm, which can effectively find the positions of real targets among a large number of peaks. The specific implementation steps are as follows:

[0056] Step 2-1: When there are no targets in the search area, the distribution of the generated voting matrix is only generated by background noise and approximately follows an independent and identically distributed Gaussian distribution, that is, B(i,j) ~ N(μ, σ 2 ), by estimating the mean μ and standard deviation σ of the entire matrix, the voting matrix can be standardized to Z(i,j), that is:

[0057]

[0058] Wherein, B(i,j) represents the number of votes of the grid (i,j); i, j (i,j = 1,2,...M) represent the row index and column index in the grid voting matrix;

[0059] When Z(i,j) ≈ 0, the grid (i,j) belongs to noise. When Z(i,j) >> 0, this grid may be a target.

[0060] Based on the idea of outlier detection and generalized likelihood ratio, the maximum value of the normalized voting matrix is extracted As a statistic, this statistic measures the degree of deviation between the extreme values in the matrix and the background noise. If there is a target in this search area, there will be one or more extreme value deviations, existing in the form of outliers. To improve robustness, the top K maximum values {Z (1) , Z (2) ,..., Z (K)} of the normalized matrix Z(i,j) are extracted to construct a statistic:

[0061]

[0062] This statistic is used to measure the overall deviation degree between the extreme values and the background noise, and a significance threshold γ is set. If it exceeds this threshold, it is determined that there is a target in this area, otherwise it is considered that there is no target in this area. The significance threshold γ is usually set to the ±3σ standard, that is, γ = 3.

[0063] Step 2-2: For each grid, protection units and reference units are set up around it. Let B(i,j) represent the number of votes in the (i,j)th grid, and the reference unit is represented as B ref , and the noise in the reference unit follows a Gaussian distribution Statistical analysis is performed on the voting values in the reference unit, and a dynamic threshold T(i,j) is generated according to the preset false alarm rate P FA . The specific formula is:

[0064] T(i,j) = μ ref + k * σ ref

[0065]

[0066] Where N represents the number of reference units, B ref,k represents the voting value in the kth reference unit, k = Q -1 (P FA ), and Q -1 represents the upper quantile function of the standard normal distribution.

[0067] For each grid (i, j), compare its voting count B(i, j) with the threshold T(i, j). If the voting value exceeds the threshold, it is considered that there is a radar radiation source target within the grid. After traversing all grids, a radar radiation source target set V can be obtained.

[0068] Step 3: Multi-target association and fusion. To further improve the detection probability, perform association and fusion on the targets in the target set V obtained in the previous step. Specifically, for any two radar radiation source targets a and b, calculate the Euclidean distance between them:

[0069]

[0070] where, x a and y a represent the position of radar radiation source target a, and x b and y b represent the position of radar radiation source target b;

[0071] Taking into account the measurement accuracy of the radar system and the actual application scenario, set a reasonable association threshold d. Subsequently, traverse all targets in the target set V. For each pair of targets a and b, calculate the Euclidean distance between the targets and compare it with the association threshold. If D a,b < d, then average the positions of the two targets to fuse them into one radar radiation source target. If D a,b ≥ d, it is considered that the two targets have sufficient separation in space and belong to two independent targets. Continuously update the target set during the traversal process until all targets have completed the association judgment and fusion operation to obtain the final radar radiation source target set.

[0072] After completing the above steps, set the voting count of the grids where no radar radiation source targets are detected to zero. For the scenario without radiation source targets, the detection result of the present invention is as shown in Figure 2 (b), without causing false positives. The detection result of the traditional method is as shown in Figure 2 (c), where one target is still detected in the target-free environment, resulting in a false alarm. In the scenario of multiple radiation source targets, the detection result of the present invention is as shown in Figure 3 (b), which can accurately detect five radiation source targets. The detection result of the traditional method is as shown in Figure 3 (c), where only one target is detected. Table 1 shows the simulation results in the environment of multiple radiation source targets, and the actual positions and estimated positions are represented by longitude and latitude. It can be seen that the adaptive multi-target detection algorithm proposed by the present invention has better performance than the traditional detection algorithm in both target-free and multi-target scenarios.

[0073] Table 1 Simulation results in the environment of multiple radiation source targets

[0074] Target serial number Actual position Position estimated by the method of the present invention Position estimated by the traditional method 1 [-31.0576,-53.8444] [-30.9729,-53.7397] - 2 [-32.5813,-57.6548] [-32.5492,-57.6166] - 3 [-33.8922,-56.2647] [-33.9125,-56.2533] - 4 [-29.2553,-53.8757] [-29.2688,-53.9101] - 5 [-34.429,-51.0517] [-34.4237,-51.0557] [-34.4237,-51.0557]

Claims

1. A method for sorting radiation source signals based on grid voting and adaptive peak detection, characterized in that It includes the following steps: Step 1: Voting matrix construction; Step 2: Adaptive multi-object peak detection; Step 3: Multi-object association and fusion.

2. The radiation source signal sorting method based on grid voting and adaptive peak detection according to claim 1, wherein The specific content of Step 1 is as follows: Set the initial parameters, evenly divide the observation area into M*M grids, and define the center point of each grid as the geographical location x of the grid; for the t-th moment, the phase difference φ of the radar radiation source pulse signal when it comes from the m = 1, 2,..., M 2 grids t satisfies the following relationship: φ t = 2πη[u t (x)] T b t (1) where b t represents the unit vector of the rotational interferometer at time t, which is determined by the spatial positions of the two channels of the rotational interferometer; u t (x) represents the unit vector determined by the grid position x and the satellite position at time t; η represents the baseline wavelength ratio; Calculate the theoretical value of the phase difference of each grid by Equation (1). If the difference between the theoretical value of the phase difference and the measured value is less than the preset threshold T, then vote the radar radiation source pulse signal to the grid corresponding to the theoretical value of the phase difference; otherwise, no voting is performed. Traverse all times and grids, and finally generate the voting matrix B.

3. A method for sorting radiation source signals based on grid voting and adaptive peak detection according to claim 2, characterized in that, The specific content of Step 2 is as follows: Step 2-1: When there is no target in the search area, the generated voting matrix distribution is only generated by background noise and approximately follows an independent and identically distributed Gaussian distribution, that is, B(i,j) ∼ N(μ,σ 2 ), by estimating the mean μ and standard deviation σ of the entire matrix, the voting matrix is normalized to Z(i,j), that is: In the formula, B(i,j) represents the voting number of the grid (i,j); i,j = 1,2,...M represent the row index and column index in the grid voting matrix; When Z(i,j)≈0, this grid belongs to noise. When Z(i,j)>>0, this grid may be a target; Extract the maximum value of the normalized voting matrix based on outlier detection and the generalized likelihood ratio As a statistic, this statistic measures the deviation between the extreme values in the normalized voting matrix and the background noise; If there is a target in the search area, there will be one or more extreme value deviations, which exist in the form of outliers; extract the top K maximum values {Z (1) , Z (2) ,..., Z (K)} of the normalized voting matrix Z(i, j) to construct a statistic: Use this statistic to measure the overall deviation degree of the extreme value from the background noise. If the overall deviation degree exceeds the significance threshold γ, it is determined that there is a target in this search area; otherwise, it is considered that there is no target in this search area. Step 2-2: For each grid, set up protection cells and reference cells around it. Let B(i,j) represent the number of votes in the (i,j)-th grid, and the reference cell is denoted as B ref , and the noise in the reference cell follows a Gaussian distribution μ ref represents the mean value of the number of votes in the reference cell, and σ ref represents the variance of the number of votes in the reference cell; perform statistical analysis on the vote values in the reference cell, and generate a dynamic threshold T(i,j) according to the preset false alarm rate P FA The specific formula is: T(i,j) = μ ref + k * σ ref where N represents the number of reference units, B ref,k represents the voting value within the k-th reference unit, k = Q -1 (P FA ) and Q -1 represents the upper quantile function of the standard normal distribution; For each grid (i,j), compare its voting number B(i,j) with the dynamic threshold T(i,j). If the voting value exceeds the dynamic threshold, it is considered that there is a radar radiation source target in this grid. After traversing all grids, a radar radiation source target set V is obtained.

4. A method for sorting radiation source signals based on grid voting and adaptive peak detection according to claim 3, characterized in that The specific content of Step 3 is as follows: Perform association and fusion on the targets in the target set V. Specifically, for any two radar radiation source targets a,b, calculate the Euclidean distance between them: Among them, x a , y a represent the position of radar radiation source target a, and x b , y b represent the position of radar radiation source target b; Set the association threshold d, traverse all targets in the target set V. For each pair of radar radiation source targets a and b, calculate the Euclidean distance between the radar radiation source targets and compare it with the association threshold. If D a,b < d, then average the positions of the two targets to fuse them into one radar radiation source target. If D a,b ≥ d, then it is determined that the two targets have sufficient separation in space and belong to two independent targets; During the traversal process, continuously update the target set until all targets have completed the association judgment and fusion operation, and obtain the final radar radiation source target set.

5. A method for sorting radiation source signals based on grid voting and adaptive peak detection according to claim 4, characterized in that The significance threshold γ is set to the ±3σ standard, that is, γ = 3.

6. A computer program, characterized in that, The computer program enables the computer to execute the method described in any one of claims 1 to 5.

7. An electronic device, characterized in that, It includes: A processor and a memory; The memory is used to store the computer program, and the processor is used to execute the computer program stored in the memory, so that the electronic device executes the method described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method described in any one of claims 1 to 5.

9. A chip, characterized in that, It includes: A processor, which is used to call and run the computer program from the memory, so that the device installed with the chip executes the method described in any one of claims 1 to 5.

10. A computer program product, characterized in that, The computer program product includes a computer storage medium. The computer storage medium stores a computer program. The computer program includes instructions that can be executed by at least one processor. When the instructions are executed by the at least one processor, the method described in any one of claims 1 to 5 is implemented.

Images