Gaussian background distance walking point target CFAR detection method

Abstract

The Gauss background distance walking point target CFAR detection method relates to the field of radar adaptive detection technology, and aims to solve the problem that the existing technology does not consider the distance walking of the target, so that the performance will seriously decrease in the case that the target exists distance walking. The Gauss background distance walking point target CFAR detection method has the CFAR characteristic to the clutter in the Gauss background, can effectively compensate the distance walking effect of the target, reduces the problem that the performance decreases due to the distance walking of the target, and can effectively accumulate the energy of the target. The output of the clutter is flat, the signal-to-noise ratio of the output is the highest, and the target detection is most beneficial. In addition, by utilizing the prior Toeplitz structure information of the clutter covariance matrix in the wide stationary background, the Gauss background distance walking point target CFAR detection method can greatly reduce the number of reference cells required for estimating the covariance matrix, and has good detection performance in the case of small samples.

Description

Technical Field

[0001] This invention relates to the field of radar adaptive detection technology, specifically a CFAR detection method for moving point targets against a Gaussian background distance. Background Technology

[0002] Radar detects targets by emitting electromagnetic waves and receiving their echo signals. Due to the relative motion between the target and the radar, radar echoes take on different forms. To improve radar detection capabilities, modern radar systems often use multiple pulse echoes for joint processing; the time required for one such joint processing is called the coherent processing interval. Compared to processing a single pulse individually, joint processing of multiple pulses can significantly improve the signal-to-noise ratio of the target echo, thereby increasing the probability of target detection and enhancing the radar's overall detection capability.

[0003] Early radars had low range resolution, and targets within the coherent processing interval could be considered to be within a single range resolution cell. The echoes of a target across different pulses were assumed to differ only in phase, without any change in the echo envelope; that is, the target did not move between pulses. Under this assumption, the Moving Target Detection (MTD) method was first proposed. This method assumes that the interference in the echo is only Gaussian white noise and that the target does not move. However, in actual detection, the interference is often not Gaussian white noise, and the target may also move.

[0004] To address the problem of interference suppression, Adaptive Moving Target Detection (AMTD) was proposed. This method can effectively suppress clutter, but its output changes with the clutter parameters, making it impossible to achieve constant false alarm rate (CFAR) detection using a single threshold in practical applications. Subsequently, the Adaptive Matched Filter (AMF) detector was proposed. AMF uses AMTD as the numerator, while the normalization factor in the denominator gives AMF CFAR characteristics for Gaussian-distributed clutter parameters. However, the two methods mentioned above only consider the correlation of clutter and do not account for the target's range movement. Therefore, the performance of these methods will significantly degrade when the target's range movement is present.

[0005] To address the problem of multi-pulse joint processing when the target is moving at range, the Hough transform (HT) was the first proposed method. HT is a non-coherent accumulation method that only utilizes the amplitude information of the target echo, not its phase information. Later, the Keystone transform (KT) was proposed. KT is a coherent accumulation method that can utilize both the amplitude and phase information of the target echo, effectively improving radar detection performance compared to HT. However, KT requires complex interpolation calculations and suffers from Doppler ambiguity. To address the problems of KT, the Radon-Fourier transform (RFT), based on Radon and Fourier transforms, was proposed. This method can be seen as a generalized filter bank, a generalization of the MTD (Multi-Pulse Detection Method) for targets moving at range. However, both RFT and MTD assume a Gaussian white noise background, and correlated clutter can also accumulate energy through the RFT method, thus affecting detection performance. This limits the use of RFT in practical scenarios. To address this problem, Adaptive RFT (ARFT) was proposed. ARFT can be seen as a generalization of AMTD in the case of distance walking. Similar to AMTD, this method can effectively suppress clutter, but the output depends on the parameters of the clutter, meaning that this method is not CFAR.

[0006] Based on the above analysis, proposing a detector that simultaneously possesses the following three properties is of great significance for achieving accurate distance-based moving target detection:

[0007] (1) It can simultaneously utilize the amplitude and phase information of the target echo, thereby enabling effective energy accumulation for moving targets.

[0008] (2) It can suppress related Gaussian clutter, thus enabling its effective application in real clutter scenarios;

[0009] (3) The output has invariance to clutter parameters, that is, it has CFAR characteristics to clutter parameters. This characteristic can ensure that the false alarm rate of the radar detector remains unchanged, which has important practical engineering significance. Summary of the Invention

[0010] The purpose of this invention is to address the problem that existing technologies do not consider the target's distance movement, which leads to a significant performance degradation when the target moves. Therefore, this invention proposes a Gaussian background distance-moving point target CFAR detection method.

[0011] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0012] The Gaussian background distance CFAR detection method for moving point targets includes the following steps:

[0013] Step 1: Acquire radar echo data after A / D sampling, wherein the radar echo data includes data of the unit to be detected and data of the reference unit;

[0014] Step 2: Given the range of parameters for the uniform motion of the target of interest, take values ​​at equal intervals within the range of parameters for the uniform motion of the target, and each value is a parameter point;

[0015] Step 3: Select the i-th parameter point within the range of parameters for the uniform motion of the target. The value corresponding to this parameter point is (r i ,v i ), r i v represents the initial distance of the target relative to the radar. i Represents the target's velocity relative to the radar, and according to (r i ,v i Construct the target echo matrix S;

[0016] Step 4: Obtain the target cell matrix Z and the reference cell matrix Z based on the target cell data and the reference cell data, respectively. L And satisfy Z L =[z1,…,z L ], where z1 represents the first reference element, z L Z represents the Lth reference element, and the reference element matrix is ​​Z. L The horizontal axis represents different distance units, and the vertical axis represents different pulses;

[0017] Step 5: Based on the reference element matrix Z L Estimating the clutter covariance matrix or diagonal matrix And based on the estimated clutter covariance matrix or diagonal matrix By combining the target echo matrix S and the matrix of the unit to be detected Z, the detection statistics are obtained;

[0018] Step 6: Compare the detection statistic with the threshold. If the detection statistic exceeds the threshold, the parameter point is the target; otherwise, it is clutter.

[0019] Step 7: Remove the parameter points selected in Step 3 from the target uniform motion parameter range, and repeat Steps 3 to 6 to complete the judgment of each parameter point within the target uniform motion parameter range.

[0020] Furthermore, the covariance matrix Represented as:

[0021]

[0022] Where L represents the number of reference elements, z i Let (·) represent the i-th reference unit.H This represents the conjugate transpose of a matrix.

[0023] Furthermore, the detection statistics are expressed as follows:

[0024]

[0025] Where tr(·) represents the trace of the matrix, (·) H The conjugate transpose of a matrix is ​​represented by (·). -1 represents the inverse of a matrix, and |·| represents the modulus of a complex number.

[0026] Furthermore, the diagonal matrix Represented as:

[0027]

[0028] Where diag(·) denotes a diagonal matrix, z i Let i represent the i-th reference cell, FFT(·) represents the Fast Fourier Transform, and |·| represents element-wise modulo.

[0029] Furthermore, the detection statistics are expressed as follows:

[0030]

[0031] Where tr(·) represents the trace of the matrix, (·) H The conjugate transpose of a matrix is ​​represented by (·). -1 · represents the inverse of a matrix, FFT(·) represents the Fast Fourier Transform, and |·| represents the modulus of a complex number.

[0032] Furthermore, the target echo matrix S is represented as:

[0033]

[0034] Where m represents the m-th pulse, n represents the n-th sampling point, and α represents the target amplitude. r0 represents the initial distance of the target relative to the radar, ρ r f represents the radar range resolution. d T represents the Doppler frequency of the target. r This indicates the pulse repetition period.

[0035] Furthermore, the radar range resolution ρ r Represented as:

[0036]

[0037] The target's Doppler frequency f d Represented as:

[0038]

[0039] Where B represents bandwidth, c represents speed of light, λ represents wavelength of electromagnetic waves emitted by radar, and v represents velocity of the target relative to radar.

[0040] Furthermore, the threshold value is obtained through a Monte Carlo experiment.

[0041] Furthermore, the elements in the clutter covariance matrix R in the Monte Carlo experiment satisfy:

[0042] R(i,j)=σ 2 ρ |i-j|

[0043] Where, σ 2 Let represent the average power of the clutter, ρ represent the first-order correlation coefficient, and i and j represent the elements in the covariance matrix R.

[0044] The beneficial effects of this invention are:

[0045] This application exhibits CFAR characteristics against clutter in a Gaussian background, while effectively compensating for target range-shifting effects and reducing performance degradation caused by target range-shifting. It effectively accumulates target energy, produces a flat clutter output, and boasts the highest signal-to-noise ratio, making it ideal for target detection.

[0046] In addition, by utilizing the prior Toeplitz structure information of the clutter covariance matrix under a wide stationary background, this application can greatly reduce the number of reference cells required to estimate the covariance matrix, and has good detection performance in the case of small samples. Attached Figure Description

[0047] Figure 1 Here is a flowchart of the GAMF detection process;

[0048] Figure 2 Block diagram of the GAMF detector;

[0049] Figure 3 Here is a flowchart of the T-GAMF detection process;

[0050] Figure 4 Block diagram of the T-GAMF detector;

[0051] Figure 5 A schematic diagram for extracting data from the unit to be detected and the reference unit;

[0052] Figure 6 A schematic diagram of the MTD method output;

[0053] Figure 7 This is a schematic diagram of the RFT method output;

[0054] Figure 8 A schematic diagram of the AMTD method output;

[0055] Figure 9 A schematic diagram of the ARFT method output;

[0056] Figure 10 A schematic diagram of the AMF method output;

[0057] Figure 11 A schematic diagram of the GAMF method output;

[0058] Figure 12 This is a schematic diagram of the false alarm probability curve for GAMF.

[0059] Figure 13 This is a schematic diagram of the false alarm probability curve for T-GAMF.

[0060] Figure 14 Schematic diagram of detection probability curves for different methods;

[0061] Figure 15 A diagram summarizing the different methods is provided. Detailed Implementation

[0062] It should be noted that, where there is no conflict, the various embodiments disclosed in this application can be combined with each other.

[0063] Specific Implementation Method 1: The Gaussian background distance CFAR detection method for moving point targets described in this implementation method includes the following steps:

[0064] Step 1: Acquire radar echo data after A / D sampling; the radar echo data includes data of the unit to be detected and data of the reference unit;

[0065] The clutter follows a Gaussian distribution. The reference cell data only includes the echo of the clutter, while the data of the cell to be detected may include the echo of the target. There must be an echo of the clutter.

[0066] Step 2: Given the range of parameters for the uniform motion of the target of interest, and take values ​​at equal intervals within the range of parameters for the uniform motion of the target;

[0067] The GAMF process for each parameter value is as follows: Figure 1 As shown, the T-GAMF procedure for each parameter value is as follows: Figure 3 As shown,

[0068] Step 3: Based on the values ​​selected at equal intervals in Step 2, select the value corresponding to the i-th parameter point, i.e., (r i ,v i ), r i v is the initial distance between the target and the radar.i Given the target's velocity relative to the radar, construct the target echo matrix S based on the selected value;

[0069] Step 4: Based on the values ​​selected in Step 3, obtain the data matrix Z of the target cell and the data matrix Z of the reference cell, respectively, using the data of the target cell and the data of the reference cell. L And satisfy Z L =[z1,…,z L ], where z1 represents the first reference element, z L This represents the Lth reference cell, where the horizontal axis represents different range cells and the vertical axis represents different pulses. That is, the same row represents the echo of a pulse from different range cells, and the same column represents the echo of a pulse from different pulses from the same range cell.

[0070] Figure 5 A schematic diagram of this step is provided, where the horizontal axis represents different range cells, and the vertical axis represents different pulses. That is, the same row represents the echo of a pulse in different range cells, and the same column represents the echo of a pulse in different range cells. Solid black squares represent the resolvable cells occupied by the target with parameter value (r0, v) in the echo, and black squares represent the resolvable cells occupied by clutter echoes. The dotted-line squares represent the detection window, which occupies K range cells in the range dimension and completely contains the target echo. The data within the detection window is called the target cell data, represented by matrix Z. The dotted-line squares adjacent to the detection window represent the reference window, which satisfies the assumption that it only contains clutter echoes and occupies L range cells in the range dimension. The data within the reference window is called the reference cell data, represented by matrix Z. L To express.

[0071] Step 5: For the GAMF method with non-small samples, estimate the covariance matrix based on the reference cell data. The expression is:

[0072]

[0073] Where L represents the number of reference elements, z i Let (·) represent the i-th reference unit. H Representing the conjugate transpose of a matrix, for the T-GAMF method with small samples, the diagonal matrix is ​​estimated from the reference cell data. The expression is:

[0074]

[0075] Where diag(·) denotes a diagonal matrix, z i Let i represent the i-th reference cell, FFT(·) represents the Fast Fourier Transform, and |·| represents element-wise modulo.

[0076] Step 6: Calculate the detection statistic based on the above calculation results. The calculation process for the GAMF detection statistic is as follows: Figure 2 As shown, the expression is:

[0077]

[0078] Where S is the target echo matrix with parameters (r0, v), Z is the data matrix of the unit to be detected, tr(·) represents the trace of the matrix, and (·) H The conjugate transpose of a matrix is ​​represented by (·). -1 represents the inverse of a matrix, and |·| represents the modulus of a complex number.

[0079] The calculation process of T-GAMF detection statistics is as follows: Figure 4 As shown, the expression is:

[0080]

[0081] Where S is the target echo matrix with parameters (r0, v), Z is the data matrix of the unit to be detected, tr(·) represents the trace of the matrix, and (·) H The conjugate transpose of a matrix is ​​represented by (·). -1 · represents the inverse of a matrix, FFT(·) represents the Fast Fourier Transform, and |·| represents the modulus of a complex number.

[0082] Step 7: Compare the detection statistic with the threshold. If it exceeds the threshold, it is judged as a target; otherwise, it is judged as clutter.

[0083] Step 8: Remove the points selected in Step 3, and repeat Steps 3 to 7 to determine whether the target with the corresponding motion parameters exists.

[0084] The effectiveness of the algorithm was verified using Monte Carlo simulation experiments. The radar pulse width T was set. P =10μs, bandwidth is B=30MHz, sampling frequency f s =B=30MHz, carrier frequency is f c =150MHz, pulse repetition period is T r =0.5ms. The expression for the target echo matrix at this time is as follows:

[0085]

[0086] Where m represents the m-th pulse, n represents the n-th sampling point, and α is the target amplitude. r0 is the initial distance of the target relative to the radar. Let c be the radar range resolution and c be the speed of light. Let be the Doppler frequency of the target, and v be the target's velocity relative to the radar. The initial target distance is set to r0 = 2000 m, and the velocity to v = 680 m / s. The clutter simulation follows a complex Gaussian distribution with a covariance matrix R, and its elements satisfy...

[0087] R(i,j)=σ 2 ρ |i-j|

[0088] Where σ 2 Let ρ be the average power of the clutter, and let ρ be the first-order correlation coefficient, which is related to the bandwidth of the clutter.

[0089] Experiment 1: Comparison of outputs from different methods. The target amplitude α = 3, clutter parameters σ = 1, ρ = 0.9, and echo pulse number M = 64 were set. It was also assumed that the covariance matrix of the clutter was known. The outputs of different methods are as follows: Figure 6-11 As shown, the results indicate that the GAMF method of the present invention can effectively accumulate the energy of the target, and the clutter output is flat, while the output signal-to-noise ratio is the highest, which is most beneficial for target detection.

[0090] Experiment 2: Simulation of False Alarm Characteristics of GAMF and T-GAMF. With the echo pulse number set to M=64 and the number of reference elements to L=150, the false alarm probability as a function of the threshold was simulated under different clutter parameters. The false alarm characteristic curve of GAMF is shown below. Figure 12 As shown, the false alarm characteristic curve of T-GAMF is as follows: Figure 13 As shown in the results, both methods of the present invention have CFAR characteristics for clutter parameters.

[0091] Experiment 3: Detection performance comparison. The echo pulse number was set to M=16, the number of reference units to L=48, and the false alarm rate to 10%. -4 First, through 10 6 The thresholds for different methods were calculated using a series of Monte Carlo experiments, and then passed through 10... 4 The Monte Carlo experiment calculated the detection probability curves of different methods as the signal-to-noise ratio changes, and the experimental results... Figure 14As shown, SCR represents the signal-to-clutter ratio, and GMF represents GAMF under known covariance matrix conditions. The results show that GMF achieves the best detection performance under known covariance matrix conditions, which is better than existing ARFT, RFT, AMF, AMTD, and MTD methods. At the same time, T-GAMF achieves the closest detection performance to GMF under the same number of reference cells, indicating that T-GAMF has the lowest dependence on the number of reference cells. This experiment verifies two points: (1) The proposed method can effectively accumulate energy for distance-moving point targets and has better detection performance than existing methods; (2) T-GAMF has the lowest dependence on the number of reference cells. In a wide-range stable clutter background, T-GAMF can greatly reduce the number of reference cells required to estimate the covariance compared to GAMF.

[0092] Finally, to summarize the various methods, such as Figure 15 As shown, compared with existing methods, the GAMF in this application has the following three characteristics: (1) It can simultaneously utilize the amplitude and phase information of the target echo, thereby enabling effective energy accumulation for range-moving targets; (2) It can suppress related Gaussian clutter, thus enabling effective application in actual clutter scenarios; (3) The output has invariance to clutter parameters, that is, it has CFAR characteristics for clutter parameters. This characteristic can ensure that the false alarm rate of the radar detector remains unchanged, which has important practical engineering significance. In addition, T-GAMF has a lower dependence on reference cells than GAMF, and can be applied to small samples.

[0093] To address the problems of existing methods, this application provides a detector with the following three properties:

[0094] (1) It can simultaneously utilize the amplitude and phase information of the target echo, thereby enabling effective energy accumulation for moving targets.

[0095] (2) It can suppress related Gaussian clutter, thus enabling its effective application in real clutter scenarios;

[0096] (3) The output has invariance to clutter parameters, that is, it has CFAR characteristics to clutter parameters. This characteristic can ensure that the false alarm rate of the radar detector remains unchanged, which has important practical engineering significance.

[0097] Furthermore, in order to further improve the detection performance of the radar, considering the problem of a small number of reference cells and the difficulty in estimating the clutter covariance matrix when the pulse number is large, this invention also proposes a small sample algorithm T-GAMF to solve this problem.

[0098] It should be noted that the specific embodiments are merely explanations and illustrations of the technical solution of the present invention and should not be used to limit the scope of protection. Any modifications made in accordance with the claims and specification of the present invention that are only partial should still fall within the protection scope of the present invention.

Claims

1. A Gaussian background distance CFAR detection method for moving point targets, characterized in that... Includes the following steps: Step 1: Acquire radar echo data after A / D sampling, wherein the radar echo data includes data of the unit to be detected and data of the reference unit; Step 2: Given the range of parameters for the uniform motion of the target of interest, take values ​​at equal intervals within the range of parameters for the uniform motion of the target, and each value is a parameter point; Step 3: Select the i-th parameter point within the range of parameters for the target's uniform motion. The value corresponding to this parameter point is... , This indicates the initial distance of the target relative to the radar. Indicates the target's velocity relative to the radar, and according to Construct the target echo matrix ; Step 4: Obtain the matrix of the cells to be detected based on the data of the cells to be detected and the data of the reference cells. and reference unit matrix and satisfy ,in Indicates the first reference unit. Indicates the first Reference units, reference unit matrix The horizontal axis represents different distance units, and the vertical axis represents different pulses; Step 5: Based on the reference unit matrix Estimating the clutter covariance matrix or diagonal matrix And based on the estimated clutter covariance matrix or diagonal matrix Combined with the target echo matrix and the matrix of units to be detected The detection statistics are obtained. Step 6: Compare the detection statistic with the threshold. If the detection statistic exceeds the threshold, then the parameter point is the target. Otherwise, it is noise; Step 7: Remove the parameter points selected in Step 3 from the range of the target uniform motion parameters, and repeat Steps 3 to 6 to complete the judgment of each parameter point within the range of the target uniform motion parameters. The detection statistics are expressed as follows: in, Represents the trace of a matrix. This represents the conjugate transpose of a matrix. Represents the inverse of a matrix. Represents the modulus of a complex number; The covariance matrix Represented as: in, Indicates the number of reference units. Indicates the first One reference unit, Represents the conjugate transpose of a matrix; Alternatively, the detection statistic can be expressed as: in, Represents the trace of a matrix. This represents the conjugate transpose of a matrix. Represents the inverse of a matrix. Represents the Fast Fourier Transform. Represents the modulus of a complex number; The diagonal matrix Represented as: in, Represents a diagonal matrix. Indicates the first One reference unit, Represents the Fast Fourier Transform. This indicates taking the modulo of the element.

2. The Gaussian background distance CFAR detection method for moving point targets according to claim 1, characterized in that... The target echo matrix Represented as: in, Indicates the first One pulse, Indicates the first One sampling point, Indicates the target range. , This indicates the initial distance of the target relative to the radar. Indicates radar range resolution. Indicates the Doppler frequency of the target. Indicates the pulse repetition period. Indicates an intermediate variable.

3. The Gaussian background distance CFAR detection method for moving point targets according to claim 2, characterized in that... The radar range resolution Represented as: Target Doppler frequency Represented as: in, Indicates bandwidth. Represents the speed of light. The wavelength of the electromagnetic waves emitted by radar. This indicates the target's velocity relative to the radar.

4. The Gaussian background distance CFAR detection method for moving point targets according to claim 1, characterized in that... The threshold value was obtained through Monte Carlo experiments.

5. The Gaussian background distance CFAR detection method for moving point targets according to claim 4, characterized in that... The clutter covariance matrix in the Monte Carlo experiment The elements in satisfy: in, This represents the average power of the clutter. This represents the first-order correlation coefficient. and Represents the covariance matrix The elements in.

Images