Non-coherent multi-frequency multi-scanning radar data processing method

Abstract

The invention discloses a non-coherent multi-frequency multi-scanning radar data processing method. The method comprises the steps of 1, constructing a target echo intensity synthesis model with a confidence coefficient matrix; 2, respectively constructing parameter estimation models for m coefficients in the m coefficient matrix and n coefficients in the n coefficient matrix by adopting a method of combining Bayesian parameter estimation and an RANSAC (Random Sample Consensus) algorithm; 3, estimating parameters; 4, updating parameters; and step 5, radar echo data processing. According to the method, a robust statistical method and a Bayesian estimation method are combined, a target echo intensity confidence coefficient matrix is established, and multiple groups of radar echo intensity data of the same target generated in a multi-frequency multi-scanning working mode are re-synthesized, so that the radar echo stability is effectively improved; radar echo flicker caused by frequency agility, target motion attitude adjustment, environmental factor influence and the like can be handled. In addition, real-time operation iteration can be achieved, and the requirement for operation resources is not high.

Description

Technical Field

[0001] This invention relates to the field of ship monitoring radar data processing, and in particular to a non-coherent multi-frequency multi-scan radar data processing method. Background Technology

[0002] Incoherent multi-frequency multi-scan radar technology is a technique that uses a single radar to detect the same target multiple times at a given moment through frequency agility. The challenges of this operating mode for radar signal processing include dealing with radar echo scintillation caused by frequency agility, target motion attitude adjustments, and environmental factors. Furthermore, it involves determining the reliability of multiple sets of intensity data acquired simultaneously, removing outliers, and ensuring the consistency of random sampling.

[0003] When removing outliers in radar echo intensity, erroneous values ​​should be removed. However, the intensity fluctuations caused by radar echo "flickering" are not erroneous values ​​but genuine intensity fluctuations. These intensity fluctuations are determined by the Swerling model. Since there are many Swerling models, the intensity fluctuations vary greatly between them. For a moving target, different attitudes can cause its Swerling model to change, making it uncontrollable. In existing technologies, the simplest method to avoid data distortion is direct accumulation without any judgment, assigning the same weighting coefficient to all data. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to address the shortcomings of the prior art by providing a noncoherent multi-frequency multi-scan radar data processing method. This noncoherent multi-frequency multi-scan radar data processing method combines Bayesian parameter estimation with the RANSAC algorithm to perform confidence modeling and evaluation of noncoherent multi-frequency multi-scan radar signals, thereby improving the data confidence problem under radar echo scintillation conditions.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0006] A noncoherent multi-frequency multi-scan radar data processing method includes the following steps.

[0007] Step 1: Construct a target echo intensity synthesis model with a confidence coefficient matrix; Assume that the number of scans of the multi-frequency multi-scan radar is n groups, and the number of frequency bands scanned in each group is m. Then the confidence coefficient matrix includes an m-coefficient matrix and an n-coefficient matrix, both of which are diagonal matrices.

[0008] Step 2: Using a combination of Bayesian parameter estimation and RANSAC algorithm, parameter estimation models are constructed for the m coefficients in the m coefficient matrix and the n coefficients in the n coefficient matrix, respectively.

[0009] Step 3, Parameter Estimation: Use a multi-frequency multi-scan radar to continuously acquire radar echo data M times, and use the RANSAC algorithm to calculate the posterior probability of parameter confidence; then substitute the obtained posterior probability of parameter confidence into the corresponding parameter estimation model to obtain the m-coefficient matrix and n-coefficient matrix.

[0010] Step 4, Parameter Update: Obtain the latest radar echo data of the multi-frequency multi-scan radar according to the set time or in real time, and replace the radar echo data of the earliest time in the M radar echo data to form updated M radar echo data; for the updated M radar echo data, use the parameter estimation method in step 3 to obtain the updated m coefficient matrix and n coefficient matrix.

[0011] Step 5, Radar echo data processing: Substitute the updated m-coefficient matrix and n-coefficient matrix into the target echo intensity synthesis model constructed in Step 1 to obtain the updated target echo intensity synthesis model; use the updated target echo intensity synthesis model to synthesize the echo data of the multi-frequency multi-scan radar.

[0012] In step 2, the parameter estimation model includes an m-parameter estimation model and an n-parameter estimation model; the expression for the m-parameter estimation model is:

[0013]

[0014] In the formula, m i Let be the confidence coefficient for the i-th frequency band; where 1 ≤ i ≤ m.

[0015] P mi Let the echo intensity detected in the i-th frequency band be the posterior probability of the target's existence if the target actually exists.

[0016] P fa The false alarm rate of the multi-frequency multi-scan radar is a known set value.

[0017] P ma The false alarm rate of the multi-frequency multi-scan radar is a known set value.

[0018] n j Let be the confidence coefficient of the j-th scan group; where 1≤j≤n.

[0019] P nj Let the composite echo intensity of the j-th scan group be the posterior probability of the target's existence if the target actually exists.

[0020] In step 3, P mi The expression is calculated using the RANSAC algorithm and is as follows:

[0021]

[0022] In the formula, N mi The number of radar echoes that are ultimately determined to be present when performing regression fitting on n sets of scanned radar echo data collected M times in the i-th frequency band using the RANSAC algorithm.

[0023] In step 3, P nj The expression is calculated using the RANSAC algorithm and is as follows:

[0024]

[0025] In the formula, N nj The number of radar echoes that are ultimately determined to exist is determined when performing regression fitting on the synthetic echo intensity radar echo data of the j-th group of M acquisitions using the RANSAC algorithm.

[0026] In step 3, let A be the intensity of the synthesized echo detected by the j-th scan in the a-th acquisition. aj And 1≤a≤M, then A aj The calculation formula is:

[0027]

[0028] In the formula, A j1 A j2 A ji and A jm These are the radar echo data of the first, second, i, and m frequency bands detected by the j-th scan group, respectively.

[0029] m1, m2, m i and m m These are the confidence coefficients for the first frequency band, the second frequency band, the i-th frequency band, and the m-th frequency band, respectively.

[0030] In steps 3 and 4, 2 < m < 10, n < 10, and M ≥ 50.

[0031] In steps 3 and 4, the total number of times the RANSAC algorithm is used to regress and fit the radar echo data from all M acquisitions to calculate the posterior probability of the parameter confidence level is W. The specific formula for calculating W is:

[0032]

[0033] In steps 3 and 4, the total time for calculating the posterior probability of the parameter confidence level using the RANSAC algorithm regression fitting for all radar echo data collected in M ​​times is T, where T is in the millisecond or microsecond range.

[0034] In step 1, the expression for the target echo intensity synthesis model is:

[0035]

[0036] In the formula, n1, n j and n n These are the confidence coefficients for the 1st, jth, and mth scan groups in the n-coefficient matrix, respectively.

[0037] A 11 A 12 …A 1i and A 1m These are radar echo data detected in frequency bands 1 to m in the first scan group.

[0038] A j1 A j2 …A ji and A jm These are the radar echo data detected in frequency bands 1 to m in the j-th scan group, respectively.

[0039] A n1 A n2 …A ni …A nm These are the radar echo data detected in frequency bands 1 to m in the nth scan group.

[0040] m1, m i and m m These are the confidence coefficients for the 1st, 1st, and 1st frequency bands in the m-coefficient matrix, respectively.

[0041] The present invention has the following beneficial effects:

[0042] 1. This invention introduces a Bayesian estimation method, combines prior data to obtain posterior probabilities, constructs a confidence coefficient matrix, and corrects the weighting coefficients.

[0043] 2. This invention specifically employs the RANSAC method, a robust regression method, which is particularly suitable for removing abnormal radar echo scintillation data.

[0044] In summary, this invention combines robust statistical methods with Bayesian estimation methods to establish a target echo intensity confidence coefficient matrix. It resynthesizes multiple sets of radar echo intensity data for the same target generated by multi-frequency, multi-scan operating modes, effectively improving radar echo stability and enabling it to cope with radar echo flicker caused by frequency agility, target motion attitude adjustments, and environmental factors. Furthermore, this invention allows for real-time iterative computation, with low computational resource requirements. Detailed Implementation

[0045] Suppose a multi-frequency multi-scan radar has n scan groups, and each scan group has m frequency bands. Its single-frequency single-scan intensity for a certain target is represented as A. jiWhere 1≤i≤m, 1≤j≤n, in the prior art, the incoherent multi-frequency multi-scan synthesized intensity of this target is usually expressed as:

[0046]

[0047] For a more intuitive representation, it can be shown in matrix form as follows:

[0048]

[0049] In the formula, A 11 A 12 …A 1i and A 1m These are radar echo data detected in frequency bands 1 to m in the first scan group.

[0050] A j1 A j2 …A ji and A jm These are the radar echo data detected in frequency bands 1 to m in the j-th scan group, respectively.

[0051] A n1 A n2 …A ni …A nm These are the radar echo data detected in frequency bands 1 to m in the nth scan group.

[0052] In this embodiment, taking m=4 and n=4 as an example, a detailed explanation is given. After a certain detection is completed, the detection result for a certain point target is obtained.

[0053] Using the existing incoherent multi-frequency multi-scan synthesis intensity calculation formula, it can be expressed as:

[0054]

[0055] However, as described in the background section of this invention, since all data are given the same weighting coefficient, it cannot cope with radar echo flicker caused by frequency agility, target motion attitude adjustment, and environmental factors.

[0056] This invention provides a noncoherent multi-frequency multi-scan radar data processing method, comprising the following steps.

[0057] Step 1: Construct a target echo intensity synthesis model with a confidence coefficient matrix. The confidence coefficient matrix includes an m-coefficient matrix and an n-coefficient matrix, both of which are diagonal matrices. The m-coefficient matrix represents the average weighting coefficients for each of the m frequency bands. The correction is made by applying the average weighted coefficients of the n-coefficient matrix to each of the n-scan groups. All have been corrected.

[0058] The expression for the target echo intensity synthesis model is:

[0059]

[0060] In the formula, n1, n j and n n These are the confidence coefficients for the 1st, jth, and mth scan groups in the n-coefficient matrix, respectively.

[0061] m1, m i and m m These are the confidence coefficients for the 1st, 1st, and 1st frequency bands in the m-coefficient matrix, respectively.

[0062] Step 2: Using a combination of Bayesian parameter estimation and RANSAC algorithm, parameter estimation models are constructed for the m coefficients in the m coefficient matrix and the n coefficients in the n coefficient matrix, respectively.

[0063] Suppose that in a radar detection (where the radar completes one transmission, reception, and sampling cycle), the radar echo intensity of a target is P. Due to radar frequency agility, target motion attitude adjustments, and environmental factors in multi-frequency, multi-scan operation, echo scintillation occurs, causing the value of P to fluctuate by ΔP in each acquisition. According to the Swerling model, the probability density of ΔP follows a chi-square distribution, meaning the value of ΔP can be positive or negative. When ΔP is positive and excessively large, a non-existent target may be detected as present, resulting in a false alarm. The false alarm rate P is then used to quantify this. fa This means that when ΔP is negative and excessively large, a present target may be mistaken for non-existent, resulting in a missed alarm by the radar. The missed alarm rate P is used to represent this. ma In this embodiment, P is set. fa =10 -4 P ma =10%.

[0064] Assuming the radar conducts M consecutive detections, event A represents the detection echo intensity being ultimately determined to indicate the presence of a target; event B represents the actual presence of the target during that detection. According to Bayes' conditional probability formula:

[0065]

[0066] Among them, P(A) i |B) is the probability that the radar-detected echo intensity is determined to be the presence of the target when the target actually exists.

[0067] For each frequency band and each scan, complete the calculation of the posterior probability P(A i By estimating |B), we can obtain the m coefficient and n coefficient.

[0068] In this invention, the prior probability P(A) i The approximate values ​​of the prior probability P(B) and the prior probability P(B) are as follows:

[0069] P(A i ) = 1 - P fa ;

[0070] P(B) = 1 - P ma ;

[0071] Posterior probability P(B|A) i The value of ) corresponds to the following two posterior probabilities of confidence in this invention.

[0072] P mi The echo intensity detected in the i-th frequency band when the target actually exists is determined as the posterior probability of the target's existence.

[0073] P nj The synthetic echo intensity of the j-th scan group is determined as the posterior probability of the target's existence if the target actually exists.

[0074] The above P mi and P nj All use P nj It was calculated using the RANSAC algorithm.

[0075] Radar detection across m frequency bands yielded m target intensity data sets {P1, P2, P3, ..., Pm-1, Pm}. Using the RANSAC regression fitting method, based on RANSAC algorithm theory, X data points satisfying the regression function are ultimately determined to be "interior points" (i.e., the target is identified), meaning these X target intensity data points are considered reliable. The remaining mx data points are considered unreliable. Recording the index of each data intensity indicates whether the echo detected in its corresponding frequency band is considered reliable or unreliable. Using these m frequency bands, 100, 1000, or more detections are conducted, and RANSAC regression fitting is performed. The proportion of detections where each frequency band data is considered reliable out of the total number of detections is statistically analyzed to obtain the posterior probability P(B|A). i ).

[0076] In rapid, multi-band, multi-group scans, we retain all the data. Through RANSAC regression calculations, we obtain a target that, although flickering, is mostly composed of "inliers." We then use an algorithm to adjust the confidence levels of the intensity obtained from different frequency bands. In our actual tests, we found that some frequency bands had very few "inliers," while others had almost no fluctuations and were entirely composed of "inliers." A significant advantage of the RANSAC algorithm is that it does not rely on a data model, making it very suitable for this scenario.

[0077] Furthermore, a drawback of the RANSAC algorithm is that the larger the number of points in the dataset, the greater the computational load, the higher the resource requirements, and the slower the computation. This invention effectively avoids this drawback by using the RANSAC algorithm, because the number of frequency bands and scan groups in an incoherent multi-frequency multi-scan radar typically does not exceed 10, i.e., 2 < m < 10, n < 10. Therefore, the computational load is extremely small, greatly leveraging the advantages of the algorithm.

[0078] Let W be the total number of times W is used to perform RANSAC algorithm regression fitting to calculate the posterior probability of parameter reliability for all radar echo data collected in M ​​times, and let T be the total time. Then, the specific formula for calculating W is:

[0079]

[0080] Taking m=10 and n=2 as an example, to collect M=1000 consecutive probes, only the following steps are required:

[0081]

[0082] Regression operations require one multiplication and two addition / subtraction operations each time. For any modern computing chip, the total computation time T is in the milliseconds or even microseconds.

[0083] In this invention, the constructed parameter estimation model includes an m-parameter estimation model and an n-parameter estimation model; wherein, the preferred expression for the m-parameter estimation model is:

[0084]

[0085] In the formula, m i is the confidence coefficient for the i-th frequency band.

[0086] P fa The false alarm rate of the multi-frequency multi-scan radar is a known set value.

[0087] P ma The false alarm rate of the multi-frequency multi-scan radar is a known set value.

[0088] n j Let be the confidence coefficient of the j-th scan group.

[0089] Step 3, Parameter Estimation: M radar echo data are continuously acquired using a multi-frequency multi-scan radar, and the RANSAC algorithm is used to calculate the posterior probability P of the parameter confidence level. mi and P nj Then, the obtained parameter confidence posterior probabilities are substituted into the corresponding parameter estimation model to obtain the m-coefficient matrix and the n-coefficient matrix.

[0090] In this invention, the above-mentioned P mi and Pnj The expressions are as follows:

[0091]

[0092] In the formula, N mi The number of radar echoes that are ultimately determined to exist is determined when performing regression fitting on n sets of scanned radar echo data collected M times in the i-th frequency band using the RANSAC algorithm.

[0093] N nj This refers to the number of radar echoes ultimately identified as present when performing regression fitting on the synthetic echo intensity radar echo data of the j-th group acquired from M acquisitions using the RANSAC algorithm. Preferably, M ≥ 50.

[0094] In this embodiment, the number of acquisitions is set to M=100, and the radar echoes acquired from the four frequency bands are recorded as An1, An2, An3 and An4 in sequence. Thus, 400 sets of radar echo data are acquired for each frequency band.

[0095] Next, RANSAC regression calculations were performed on 400 sets of radar echo data collected in each frequency band, and the analysis results are as follows:

[0096] An1: 370 internal points.

[0097] An2: 360 internal points.

[0098] An3: 328 internal points.

[0099] An4: 304 internal points.

[0100] Therefore, the four confidence coefficients in the m-coefficient matrix are calculated as follows:

[0101]

[0102] Therefore, the m-coefficient matrix is:

[0103]

[0104] The echo intensity synthesis model after adding the m-coefficient matrix is ​​expressed as:

[0105]

[0106] go through Each regression operation requires one multiplication and two addition / subtraction operations. If the processor's clock speed reaches 1GHz, the multiplication operation takes 12 cycles and the addition / subtraction operation takes 2 cycles. In this embodiment, it takes 76.8µs to complete the calculation of the m coefficient.

[0107] After the m-coefficient is calculated, the n-coefficient is calculated. The radar echo data of the m frequency bands detected by each scan are synthesized once. Let the intensity of the synthesized echo detected by the j-th scan in the a-th acquisition be A. aj And 1≤a≤M, then A aj The calculation formula is:

[0108]

[0109] In the formula, m1, m2, m i and m m These are the confidence coefficients for the first, second, i-th, and m-th frequency bands, respectively. In this example, 100 synthetic echo intensities are obtained for each scan group. RANSAC regression calculations are performed on four sets of 100 n data points. The assumed result is: A a1 For interior point A, with 99th power, a2 For interior point A, the number of times is 92. a3 For interior point A, the process is 88 times. a4 For interior points, the 95th order yields the coefficient of n:

[0110]

[0111] Therefore, the n-coefficient matrix is:

[0112]

[0113] Ultimately, the radar echo composite intensity model was derived from...

[0114]

[0115] Become

[0116]

[0117] go through Each regression operation requires one multiplication and two addition / subtraction operations. If the processor's clock speed reaches 1GHz, the multiplication operation takes 12 cycles and the addition / subtraction operation takes 2 cycles. In this embodiment, it takes 19.2µs to complete the calculation of the n coefficient.

[0118] The calculation of the m and n coefficients took a total of 96 µs. For a radar with a repetition rate of 2000 Hz, this can be completed within 500 µs of a single transmission cycle, making real-time online computation entirely feasible.

[0119] Step 4, Parameter Update: Obtain the latest radar echo data from the multi-frequency multi-scan radar according to a set time or in real-time, and replace the earliest radar echo data in the M radar echo data to form updated M radar echo data. For the updated M radar echo data, use the parameter estimation method from Step 3 to obtain the updated m-coefficient matrix and n-coefficient matrix. The m-coefficient matrix and n-coefficient matrix can be continuously updated in the background by the algorithm, calling historical detection data and inputting it into the parameter estimation model, ensuring that the radar detection performance remains at its optimal state at all times.

[0120] Step 5, Radar echo data processing: Substitute the updated m-coefficient matrix and n-coefficient matrix into the target echo intensity synthesis model constructed in Step 1 to obtain the updated target echo intensity synthesis model; use the updated target echo intensity synthesis model to synthesize the echo data of the multi-frequency multi-scan radar.

[0121] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A method for processing noncoherent multi-frequency multi-scan radar data, characterized in that: Includes the following steps: Step 1: Construct a target echo intensity synthesis model with a confidence coefficient matrix; Assume that the number of scans of the multi-frequency multi-scan radar is n groups, and the number of frequency bands scanned in each group is m. Then the confidence coefficient matrix includes an m-coefficient matrix and an n-coefficient matrix, both of which are diagonal matrices. Step 2: Using a combination of Bayesian parameter estimation and RANSAC algorithm, parameter estimation models are constructed for the m coefficients in the m coefficient matrix and the n coefficients in the n coefficient matrix, respectively. Step 3, Parameter Estimation: Use a multi-frequency multi-scan radar to continuously acquire radar echo data M times, and use the RANSAC algorithm to calculate the posterior probability of parameter confidence; then substitute the obtained posterior probability of parameter confidence into the corresponding parameter estimation model to obtain the m-coefficient matrix and n-coefficient matrix. Step 4, Parameter Update: Obtain the latest radar echo data of the multi-frequency multi-scan radar according to the set time or in real time, and replace the radar echo data of the earliest time in the M radar echo data to form updated M radar echo data; For the updated M radar echo data, use the parameter estimation method in step 3 to obtain the updated m coefficient matrix and n coefficient matrix. Step 5, Radar echo data processing: Substitute the updated m-coefficient matrix and n-coefficient matrix into the target echo intensity synthesis model constructed in Step 1 to obtain the updated target echo intensity synthesis model; use the updated target echo intensity synthesis model to synthesize the echo data of the multi-frequency multi-scan radar.

2. The noncoherent multi-frequency multi-scan radar data processing method according to claim 1, characterized in that: In step 2, the parameter estimation model includes an m-parameter estimation model and an n-parameter estimation model; the expression for the m-parameter estimation model is: In the formula, m i Let be the confidence coefficient for the i-th frequency band; where 1 ≤ i ≤ m; P mi The echo intensity detected in the i-th frequency band is used to determine the posterior probability of the target's existence if the target actually exists. P fa The false alarm rate of the multi-frequency multi-scan radar is known, and the set value is given. P ma The false alarm rate of a multi-frequency multi-scan radar is known, and the set value is given. n j Let be the confidence coefficient of the j-th scan group; where 1≤j≤n; P nj Let the composite echo intensity of the j-th scan group be the posterior probability of the target's existence if the target actually exists.

3. The noncoherent multi-frequency multi-scan radar data processing method according to claim 2, characterized in that: In step 3, P mi The expression is calculated using the RANSAC algorithm and is as follows: In the formula, N mi The number of radar echoes that are ultimately determined to exist is determined when the RANSAC algorithm is used to perform regression fitting on the n sets of scanning radar echo data collected M times in the i-th frequency band.

4. The noncoherent multi-frequency multi-scan radar data processing method according to claim 3, characterized in that: In step 3, P nj The expression is calculated using the RANSAC algorithm and is as follows: In the formula, N nj The number of radar echoes that are ultimately determined to exist is determined when performing regression fitting on the synthetic echo intensity radar echo data of the j-th group of M acquisitions using the RANSAC algorithm.

5. The noncoherent multi-frequency multi-scan radar data processing method according to claim 4, characterized in that: In step 3, let A be the intensity of the synthesized echo detected by the j-th scan in the a-th acquisition. aj And 1≤a≤M, then A aj The calculation formula is: In the formula, A j1 A j2 A ji and A jm These are the radar echo data of the first, second, i, and m frequency bands detected by the j-th scan group, respectively. m1, m2, m i and m m These are the confidence coefficients for the first frequency band, the second frequency band, the i-th frequency band, and the m-th frequency band, respectively.

6. The noncoherent multi-frequency multi-scan radar data processing method according to claim 1, characterized in that: In steps 3 and 4, 2 < m < 10, n < 10, and M ≥ 50.

7. The noncoherent multi-frequency multi-scan radar data processing method according to claim 1, characterized in that: In steps 3 and 4, the total number of times the RANSAC algorithm is used to regress and fit the radar echo data from all M acquisitions to calculate the posterior probability of the parameter confidence level is W. The specific formula for calculating W is:

8. The noncoherent multi-frequency multi-scan radar data processing method according to claim 1, characterized in that: In steps 3 and 4, the total time for calculating the posterior probability of the parameter confidence level using the RANSAC algorithm regression fitting for all radar echo data collected in M ​​times is T, where T is in the millisecond or microsecond range.

9. The noncoherent multi-frequency multi-scan radar data processing method according to claim 1, characterized in that: In step 1, the expression for the target echo intensity synthesis model is: In the formula, n1 and n j and n n These are the confidence coefficients for the 1st, jth, and mth scan groups in the n-coefficient matrix, respectively. A 11 A 12 …A 1i and A 1m These are radar echo data detected in frequency bands 1 to m in the first scan group, respectively. A j1 A j2 …A ji and A jm These are the radar echo data detected in frequency bands 1 to m in the j-th scan group, respectively; A n1 A n2 …A ni …A nm These are the radar echo data detected in frequency bands 1 to m in the nth scan group, respectively; m1, m i and m m These are the confidence coefficients for the 1st, 1st, and 1st frequency bands in the m-coefficient matrix, respectively.

Images