Angular flicker noise pre-processing method based on adaptive amplitude suppression
Patent Information
- Application Number
- CN202610524788.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-20
- Publication Date
- 2026-08-18
AI Technical Summary
[0005]本发明针对现有技术中存在的缺点和不足，提出了一种通过引入幅度抑制制器，分离角闪烁噪声中的大值分量与小值分量，通过幅度缩放的方式将大值分量转换为小值分量，并利用均值滤波器达到减小噪声方差的目的，从而有效降低跟踪滤波器前端输入噪声方差，并在目标机动的情况下保持良好效果的基于自适应幅度抑制的角闪烁噪声预处理方法
[0011] This invention introduces an amplitude suppressor while retaining the original feedback structure. The amplitude suppressor changes the noise amplitude distribution, retains the large-value noise component, and suppresses the large-value noise component, thereby enhancing the tracking accuracy of the algorithm. Simulation experiments show that although there is a slight sacrifice in computation time, it significantly improves the tracking accuracy under angular flicker noise and target maneuvering conditions, and has robustness, making it valuable for engineering applications.
Smart Images
Figure CN122592348A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of radar signal processing technology, specifically to an adaptive amplitude suppression-based corner scintillation noise preprocessing method that can effectively reduce the variance of the input noise at the front end of the tracking filter and maintain good performance under target maneuvering conditions. Background Technology
[0002] When the radar is close to the target, the target can be considered as an extended target composed of multiple scattering centers. When the radar antenna emits electromagnetic waves that illuminate the target, each scattering center generates an echo, and these echoes interfere with each other, causing distortion of the wavefront of the echo reaching the receiving port surface, forming a new apparent center. Ultimately, the angle is measured from this apparent center. When there is relative movement between the radar and the target, the echo at the radar receiving port surface also changes continuously, causing the apparent center of the target to change continuously, ultimately resulting in angular scintillation. Angular scintillation is usually represented by line deviation, which is the distance between the synthesized apparent center and the actual center. For close-range guidance, angular scintillation is one of the main sources of tracking error; in severe cases, it can even cause the missile to completely deviate from the target. Therefore, suppressing angular scintillation noise is of great significance for terminal guidance.
[0003] Since radar architecture is generally difficult to change, researchers often employ signal processing techniques to suppress the impact of angular scintillation on radar tracking accuracy. Signal processing techniques for angular scintillation suppression mainly include angular scintillation preprocessors, improved Kalman filters, and adaptive tracking filters based on online learning.
[0004] Regarding angular scintillation preprocessors, in 2000, Chang et al. replaced the traditional M-estimators with a feedback median filter, achieving good tracking results even with high target velocities. However, due to the performance limitations of the median filter, there is still room for improvement in the model's tracking accuracy. Summary of the Invention
[0005] This invention addresses the shortcomings and deficiencies of existing technologies by proposing an adaptive amplitude suppression-based angular flicker noise preprocessing method. This method introduces an amplitude suppressor to separate large and small components in angular flicker noise, converts large components into small components through amplitude scaling, and uses a mean filter to reduce noise variance. This effectively reduces the noise variance at the front end of the tracking filter and maintains good performance even when the target is maneuvering.
[0006] This invention achieves its purpose through the following measures: An angular scintillation noise preprocessing method based on adaptive amplitude suppression is used for preprocessing target angle measurement data in a radar terminal guidance system. The method is characterized by the following steps: Step 1: Obtain the current batch of angle measurement data sequence. The angle measurement data includes angular scintillation noise modeled by a Gaussian-Laplace mixture distribution. Based on the target position estimate and velocity prediction value of the previous moment, calculate the target position prediction value of the current moment. Step 2: Calculate the deviation between the angle measurement data sequence and the predicted target position to obtain the deviation sequence; Step 3: Perform adaptive amplitude suppression processing on the deviation sequence, specifically including: setting an amplitude threshold, identifying large and small components in the deviation sequence; performing nonlinear scaling on the large component to compress its amplitude to a level close to that of the small component, while keeping the small component unchanged; Step 4: Add the amplitude-suppressed deviation sequence to the target position prediction value to obtain the corrected measurement sequence; Step 5: Apply mean filtering to the corrected measurement sequence and output the preprocessing result. The preprocessing result is then input into the tracking filter to update the target state estimate.
[0007] In step 1 of this invention, the estimated position of the target at the previous moment is obtained through a tracking algorithm. The velocity estimate of the target at the previous moment The estimated acceleration of the target at the previous moment The radar sampling period is The current batch of M consecutive angle measurements are , This represents the M-th sample in the k-th batch. and The preset adaptive parameter vector is used to construct the hierarchical threshold and scaling coefficient, and the current position is predicted by the following formula: (1).
[0008] In step 2 of this invention, calculating the deviation sequence specifically involves subtracting the predicted position from each measured value to obtain the deviation vector. (2), .
[0009] In step 3 of this invention, the extreme values and dynamic range of the deviation are first calculated, that is, the maximum and minimum values of the absolute value of the deviation and the difference between them are extracted. : (3), (4), (5) is used for subsequent adaptive threshold generation, where the more severe the noise, The larger the threshold, the greater the threshold, thus avoiding excessive suppression; Then, generate the stratification thresholds using a preset parameter vector. and dynamic range Construct n layers of decreasing thresholds, as shown in the following formula: (6), This enables soft-grading suppression, where outliers of different magnitudes are graded according to severity, which is more precise than a single threshold. Divide the deviations into subsets: Divide the M deviations into n hierarchical sets based on their absolute values, as shown in the following formula: (7) Adaptive amplitude suppression is achieved through differential processing; hierarchical nonlinear scaling is performed as shown in the following equation: (8), of which , From vector , represents the scaling factor corresponding to the i-th layer. It is the deviation of the i-th layer.
[0010] In step 4 of this invention, the average of all scaled deviations is first calculated, and then compared with the predicted position. Add them together to get the final output. As shown in the following formula: (9).
[0011] This invention introduces an amplitude suppressor while retaining the original feedback structure. The amplitude suppressor changes the noise amplitude distribution, retains the large-value noise component, and suppresses the large-value noise component, thereby enhancing the tracking accuracy of the algorithm. Simulation experiments show that although there is a slight sacrifice in computation time, it significantly improves the tracking accuracy under angular flicker noise and target maneuvering conditions, and has robustness, making it valuable for engineering applications. Attached Figure Description
[0012] Appendix Figure 1 This is a schematic diagram of Gaussian-Laplace mixed-angle flicker noise.
[0013] Appendix Figure 2 This is a flowchart of a feedback median filter.
[0014] Appendix Figure 3 This is a flowchart of the present invention.
[0015] Appendix Figure 4 This is a simulation curve of tracking accuracy in an embodiment of the present invention. Detailed Implementation
[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0017] The probability distribution function of the Gaussian-Laplace mixture distribution can be expressed as: (10) The Gaussian distribution function is: (11), The Laplace distribution function is: (12) The weights are based on the data values proposed by Dunn and Howard. The Gaussian distribution parameters are: The Laplace distribution parameters are: Therefore, the probability distribution of angular flicker noise is a Gaussian-Laplace mixture distribution, and the probability density function is: (13) The angular flicker noise used in the simulation is as follows: Figure 1 As shown, the Gaussian distribution describes the small-value, high-probability component, while the Laplace distribution has a sharper peak and a longer tail, used to describe the large-value, long-tailed component.
[0018] Angle Scintillation Preprocessor Based on Adaptive Amplitude Suppression Traditional classical feedback median filter structure, such as Figure 2 As shown: Let for Let there be n random variables, and arrange these observations in ascending order such that... (14) random variable , This is called The first random variable A series of ordinal statistics. Let... Let be odd number. Let be the set of these random variables. The output of the median filter is defined as... (15); When the pulse repetition rate of modern radar is often higher than the required tracking rate, the tracking system can provide more measurement data than it can process. In this case, a batch of measurement data can be preprocessed, and the results can then be passed to a Kalman filter. Assume that the target dynamics and the natural spherical observation model have been decoupled into independent channels on their respective axes. Consider a one-dimensional tracker and assume that only position measurements are available. The target position and noise position measurements can be described by the following equations: (16) in It is the sampling period. and These are the target's velocity and acceleration, respectively. These are measured values. It measures noise; If we assume the input preprocessor is divided into consecutive batches, to measure values The batch centered on is represented as ,in The median filter output is expressed as The next input batch is... The corresponding output is Without a feedback structure, the median filter output is: (17) Assuming the target has a constant velocity, or if the third term in equation (16) is negligible, then: (18), of which It is easy to prove. ,when At that time, it is assumed that the measurement noise is in the form of a superposition of uncorrelated zero-mean Gaussian noise and zero-mean Laplace distributed noise. The variance is: (19) in The variance is Gaussian noise. Let the variance be the Laplace distribution noise. For weights. When At that time, there were no theoretical results to apply, but it was easy to prove. along with It increases as the value increases. Therefore, it decreases. A feedback structure is introduced.
[0019] The structure of the feedback median filter is as follows: Figure 2 As shown, after introducing the feedback structure: (20) in, Speed prediction The calculation formula is: (21) In practical applications, often Therefore, we can have such expectations, therefore there are , making decline.
[0020] The feedback median filter addresses the issue that its efficiency can significantly decrease when the input samples are not identically distributed, making the algorithm more robust. However, due to the performance limitations of the median filter, there is still room for improvement in tracking accuracy.
[0021] To improve preprocessor performance, this example introduces an amplitude suppressor based on the feedback median filter. The algorithm block diagram is as follows. Figure 3 As shown, an amplitude suppressor is used instead of a median filter. The algorithm predicts the target position at the next moment based on the previous estimation result and feeds it back to the amplitude suppressor. After amplitude suppression, the target passes through a mean filter, and finally, a Kalman filter completes the tracking filtering. The amplitude suppressor processing method is as follows: Table 1 Amplitude Suppressor Flowchart
[0022] Simulation experiment: Experiment 1: Experimental objective: To test the tracking accuracy of the improved method under angular flicker noise and compare it with the classical method.
[0023] The simulation duration was set to 60 seconds, the sampling period to 0.0714 seconds, and the median filter window size to 35. A Kalman filter was used, and the process noise variance was... An angular scintillation line deviation was simulated using a Gaussian-Laplace mixture model, with an initial velocity of... A maneuver occurs between the 60th and 80th seconds, with an acceleration of... Monte Carlo simulation 500 times.
[0024] like Figure 4 As shown, the RMSE of the classic method is approximately 2.8m, while that of the improved method is approximately 2.0m, representing a performance improvement of about 38.5%. This indicates that under the same conditions, the improved method has a smaller tracking error and can maintain stability.
[0025] Experiment 2: (1) Experimental objective: To compare the single-run time of the algorithm before and after the improvement.
[0026] (2) Experimental conditions: The target maneuvering and noise distribution are the same as in Experiment 1. The hardware equipment is CPU: i9-13980HX, memory: 16G, hard disk: 1TB.
[0027] (3) Analysis of experimental results As shown, the improved method has a single runtime of approximately 46.1917 μs, which is slightly longer than the classic algorithm's 33.5381 μs, but it still meets the requirements for engineering applications.
[0028] Table 2 Algorithm Single Run Length
[0029] In summary, an improvement was made to the classic feedback median filter. While retaining the original feedback structure, an amplitude suppressor was introduced. This amplitude suppressor alters the noise amplitude distribution, preserving large-value noise components while suppressing others, thus enhancing the algorithm's tracking accuracy. Simulation results show that although there is a slight sacrifice in computation time, it significantly improves tracking accuracy under angular flicker noise and target maneuvering conditions, and also demonstrates robustness and engineering application value.
Claims
1. A method for preprocessing angular scintillation noise based on adaptive amplitude suppression, used for preprocessing target angle measurement data in a radar terminal guidance system, characterized in that, Includes the following steps: Step 1: Obtain the current batch of angle measurement data sequence. The angle measurement data includes angular scintillation noise modeled by a Gaussian-Laplace mixture distribution. Based on the target position estimate and velocity prediction value of the previous moment, calculate the target position prediction value of the current moment. Step 2: Calculate the deviation between the angle measurement data sequence and the predicted target position to obtain the deviation sequence; Step 3: Perform adaptive amplitude suppression processing on the deviation sequence, specifically including: setting an amplitude threshold, identifying large and small components in the deviation sequence; performing nonlinear scaling on the large component to compress its amplitude to a level close to that of the small component, while keeping the small component unchanged; Step 4: Add the amplitude-suppressed deviation sequence to the target position prediction value to obtain the corrected measurement sequence; Step 5: Apply mean filtering to the corrected measurement sequence and output the preprocessing result. The preprocessing result is then input into the tracking filter to update the target state estimate.
2. The corner flicker noise preprocessing method based on adaptive amplitude suppression according to claim 1, characterized in that, In step 1, the target's position estimate from the previous moment is obtained using a tracking algorithm. The velocity estimate of the target at the previous moment The estimated acceleration of the target at the previous moment The radar sampling period is The current batch of M consecutive angle measurements are , This represents the M-th sample in the k-th batch. and The preset adaptive parameter vector is used to construct the hierarchical threshold and scaling coefficient, and the current position is predicted by the following formula: （1）。 3. The corner scintillation noise preprocessing method based on adaptive amplitude suppression according to claim 1, characterized in that, Step 2, specifically calculating the deviation sequence, involves subtracting the predicted position from each measured value to obtain the deviation vector. (2), .
4. The corner flicker noise preprocessing method based on adaptive amplitude suppression according to claim 1, characterized in that, In step 3, the extreme values and dynamic range of the deviation are first calculated, that is, the maximum and minimum values of the absolute value of the deviation and the difference between them are extracted. : （3）， （4）， (5) is used for subsequent adaptive threshold generation, where the more severe the noise, The larger the threshold, the greater the threshold, thus avoiding excessive suppression; Then, generate the stratification thresholds using a preset parameter vector. and dynamic range Construct n layers of decreasing thresholds, as shown in the following formula: (6), This enables soft-grading suppression, where outliers of different magnitudes are graded according to severity, which is more precise than a single threshold. Divide the deviations into subsets: Divide the M deviations into n hierarchical sets based on their absolute values, as shown in the following formula: (7) Adaptive amplitude suppression is achieved through differential processing; hierarchical nonlinear scaling is performed as shown in the following equation: (8), of which , From vector , represents the scaling factor corresponding to the i-th layer. It is the deviation of the i-th layer.
5. The angle flicker noise preprocessing method based on adaptive amplitude suppression according to claim 1, characterized in that, In step 4, the average of all scaled deviations is first calculated, and then compared with the predicted position. Add them together to get the final output. As shown in the following formula: （9）。