A method for estimating sea clutter distribution parameters
By windowing the sea clutter echo samples and combining local characteristics, a sea clutter Pareto distribution parameter estimation method based on variable intervals is adopted. This solves the problem of inaccurate parameter estimation in traditional methods and improves the target detection performance in the context of sea clutter.
Patent Information
- Application Number
- CN202210181604.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-02-26
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2042-02-26
AI Technical Summary
In existing technologies, the traditional moment estimation method ignores the local characteristics of sea clutter in the parameter estimation of sea clutter amplitude distribution models, resulting in poor parameter estimation performance, especially affecting target detection performance under different sea conditions.
By windowing the sea clutter echo samples and combining local characteristics, a Pareto distribution parameter estimation method based on variable intervals for sea clutter is adopted, which improves the estimation accuracy of the distribution parameters.
The target detection performance under sea clutter background is improved. By improving the parameter estimation method, the fitting effect of the sea clutter distribution model is enhanced, and the interference of sea clutter on target detection is suppressed.
Smart Images

Figure CN114646935B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of radar signal processing technology, and in particular to a method for estimating parameters of a sea clutter amplitude distribution model. The present invention can be used for sea warning or search radars, and by analyzing the statistical characteristics of radar sea clutter units and fully considering the local distribution characteristics of sea clutter samples, the parameter estimation accuracy of the sea clutter distribution model is improved, and technical support is provided for the accurate design of radar target detection methods. Background Art
[0002] In the application of radar target detection against the sea surface background, sea radars are usually faced with the interference of large-scale sea clutter, which seriously restricts the detection performance of weak targets at sea. The characteristics of sea clutter are complex and changeable. Among them, the statistical characteristics of sea clutter amplitude are the basis for the design of radar target detection algorithms. The accurate identification of sea clutter statistical models is the key factor to improve target detection performance. Therefore, it is of great significance to accurately model the sea clutter amplitude distribution model and estimate its parameters.
[0003] Since sea clutter is not only affected by natural environmental factors such as wind speed, wind direction, and sea conditions, but also by parameters such as radar operating frequency, incident angle, and working bandwidth, sea clutter exhibits strong spatiotemporal correlation, non-stationary, and non-Gaussian characteristics. The generalized Pareto distribution fully considers the spatiotemporal correlation and tailing characteristics of sea clutter, and can not only accurately match the sea clutter amplitude distribution model, but also has a relatively concise probability density function expression. It is widely used in the field of sea clutter modeling and target detection. Therefore, the present invention mainly estimates parameters for the generalized Pareto distribution. However, the conventional generalized Pareto distribution parameter estimation uses the moment estimation method, which only considers the amplitude sequence of the entire sea clutter, but ignores the local characteristics of sea clutter in different amplitude intervals under different sea conditions; at the same time, the measured sea clutter data usually contains abnormal scattering units, and the amplitude of these abnormal scattering units is high, which will affect the effect of parameter estimation, resulting in poor results of the moment estimation method. Summary of the invention
[0004] In order to overcome the shortcomings of the prior art, the present invention provides a method for estimating sea clutter distribution parameters. The purpose of the present invention is to address the lack of robustness of the existing sea clutter amplitude distribution model parameter estimation method, and propose a sea clutter amplitude distribution model parameter estimation method based on a variable interval, which improves the accuracy of distribution parameter estimation by windowing sea clutter echo samples, thereby improving target detection performance under sea clutter background.
[0005] The present invention overcomes the deficiency of parameter estimation of generalized Pareto distribution based on traditional moment estimation method, combines traditional moment estimation theory with local characteristics of sea clutter, performs windowing processing on sea clutter sequence, and proposes a parameter estimation method of sea clutter Pareto distribution based on variable interval, which improves the result of goodness of fit test and helps to improve target detection under sea clutter background.
[0006] The technical solution adopted by the present invention to solve the technical problem includes the following steps:
[0007] (1) Radar echo sequence sorting;
[0008] The radar transmitter transmits a pulse signal and the radar receiver receives the time domain echo signal X={x k , k=1,2,3..N} as the amplitude sample sequence of sea clutter data, and sort the obtained original amplitude sample sequence X in ascending order to obtain the increasing sample sequence X'={x (k) , k=1,2,3...N}, where x (k) represents the kth sample element in the increasing sample sequence X', and N represents the length of the sea clutter sequence X';
[0009] (2) Calculation of sea clutter sequence amplitude histogram;
[0010] According to the increasing order sea clutter sample data X'={x (k) , k=1,2,3...N}, calculate the sea clutter amplitude histogram;
[0011] (3) Calculation of generalized Pareto distribution parameters;
[0012] (3.1) Selection of sea clutter sequence: determine the selection criteria of sea clutter sequence;
[0013] In the case of high sea conditions, the sea clutter sequences with larger amplitudes dominate the sea clutter sequences. When performing parameter estimation, some sea clutter sequences with smaller amplitudes are discarded, and the sequences with larger amplitudes are selected for parameter estimation. Similarly, in the case of low sea conditions, the sea clutter sequences with smaller amplitudes dominate the sea clutter sequences. When performing parameter estimation, some sea clutter sequences with larger amplitudes are discarded, and the sea clutter sequences with smaller amplitudes are selected for parameter estimation.
[0014] (3.2) Interval second-order moment calculation;
[0015] According to the probability density function of the generalized Pareto distribution and the increasing sample sequence X' obtained in step (1), combined with the selection criteria of the sea clutter sequence in (3.1), the interval is selected Integrate the echo data within the interval to obtain The second moment of the interval within;
[0016] (3.3) Interval zero-order moment calculation;
[0017] According to the probability density function of the generalized Pareto distribution and the radar echo sequence sorted in step (2), the interval (x (L1) ,x (L2) )The zero-order moment of the data within;
[0018] (3.4) Parameter calculation;
[0019] According to steps (3.1) and (3.2), the results of the interval second-order moment and the percentage of the data in the interval to the total data are combined to solve the parameters λ and a of the generalized Pareto distribution. The equation is:
[0020]
[0021] (4) Goodness-of-fit test calculation;
[0022] (4.1) Calculate the mean square error (Msd);
[0023] The mean square error is one of the most commonly used methods for goodness of fit testing. The mean square error function is defined as:
[0024]
[0025] where f r (·) and f k (·) represents the probability density function of sample data and the probability density function of fitting model, k represents the serial number of sea clutter sequence, and N represents the length of sea clutter sequence;
[0026] (4.2) Calculate the modified mean square error (Mmsd);
[0027] The modified mean square error focuses on testing the fitting effect of the tail part, which is the description effect of the sea peaks in the sea clutter echo sequence. These sea peaks seriously affect the detection of targets by maritime radar. Taking this into account, the improved mean square error is defined as:
[0028]
[0029]
[0030] where f r (·) represents the probability density function of the sample data, f t (·) represents the probability density function of the fitting model, Represents a set of tailing samples, taking the smallest is the detection threshold, function is constant false alarm detection, Pf is the false alarm probability;
[0031] Msd and Mmsd are used to measure the accuracy of parameter estimation. The smaller the value of Msd and Mmsd, the smaller the error between the probability density function of the sample data and the probability density function of the fitting model, and the more accurate the parameter estimation.
[0032] In step 3.2, the interval second-order moment is defined as follows based on the upper and lower boundaries of the interval:
[0033]
[0034]
[0035]
[0036] Among them, L1 and L2 are two quantiles, and 0<L1<L2<1, and Respectively represent the amplitude values of the sea clutter sequence at the quantiles L1 and L2, Represents the calculation result of the interval second-order moment;
[0037] The meaning of interval estimation in actual clutter sequence is: N Increasing sorting is performed. According to the differences in sea clutter echo data in different sea conditions and different distance units, combined with the selection criteria of sea clutter sequence in 3.1, the second-order origin moment of the data in the interval is calculated according to the following formula:
[0038]
[0039] In step 3.3, the zero-order moment of the interval is the percentage of the data in the calculated interval to the total data. According to the upper and lower boundaries of the interval, the zero-order moment of the data in the interval is defined as:
[0040]
[0041] in Represents the calculation result of the interval zero-order moment.
[0042] The beneficial effect of the present invention is that a sea clutter Pareto distribution parameter estimation method based on a variable interval is proposed according to the difference in sea clutter echo data of different sea conditions and different distance units, interval parameter estimation is completed by windowing processing, the fitting effect of the generalized Pareto distribution probability density function is improved, and it is helpful to improve target detection under the sea clutter background.
[0043] The present invention uses the second-order moment and zero-order moment of the data in the interval to perform parameter estimation, wherein the calculation of the zero-order moment is essentially the percentage of the used data in the total data. Compared with using traditional high-order moments, the proposed method can improve the accuracy of parameter estimation. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 It is a flow chart of the algorithm implementation of the present invention.
[0045] Figure 2 It is the probability density function of sample data and the probability density function of fitting model.
[0046] Figure 3 It is a graph of the mean square error between the probability density function of the sample data and the probability density function of the fitted model.
[0047] Figure 4 It is a graph of the improved mean square error between the probability density function of the sample data and the probability density function of the fitted model. DETAILED DESCRIPTION
[0048] The present invention is further described below in conjunction with the accompanying drawings and embodiments.
[0049] Reference Figure 1 The specific implementation process of the present invention is as follows:
[0050] Step 1: Sorting the radar echo sequence. Use the radar transmitter to send a pulse signal and send the time domain sample echo signal X = {x k , k = 1, 2, 3.. N} as the sample sequence of sea clutter data, sort the sample echo sequence in ascending order, and obtain the increasing sample sequence X' = {x (k) ,k=1,2,3...N}.
[0051] Step 2: Statistical analysis of the amplitude histogram of the sea clutter sequence. According to the increasing order of the sea clutter sample echo sequence X = {x (k) , k=1,2,3...N}, calculate the sea clutter amplitude histogram.
[0052] Step 3: Calculation of generalized Pareto distribution parameters;
[0053] 3.1) Calculate the probability density function expression of the generalized Pareto distribution.
[0054] The receiving vector of the generalized Pareto distribution is expressed as:
[0055]
[0056] Among them, c is the radar receiver receiving vector, τ is the slowly varying texture component, and u is the rapidly varying speckle component. The texture component is a random variable that obeys the inverse gamma distribution, and its probability density expression is:
[0057]
[0058] Where a is the scale parameter; λ is the shape parameter; Γ(·) is the gamma function;
[0059] The speckle component obeys a complex Gaussian random variable with a mean of 0. The amplitude of the clutter vector is expressed as:
[0060]
[0061] Where c(i) represents the i-th component of the clutter vector. According to the expression of the clutter vector x, when the texture component τ is known, the clutter amplitude x is the modulus of two Gaussian distributions with the same mean and variance, and obeys the Rayleigh distribution. The conditional probability density function expression of the clutter amplitude x is:
[0062]
[0063] According to the total probability formula, the probability density expression of the sea clutter amplitude x is derived as follows:
[0064]
[0065] make Substituting into the above formula we get:
[0066]
[0067] Where x is the clutter amplitude, λ is the shape parameter, and a is the scale parameter. λ determines the non-Gaussian nature of sea clutter. When λ→∞, the generalized Pareto distribution degenerates into the Rayleigh distribution; when λ→0, the generalized Pareto distribution has a slightly heavier tail.
[0068] 3.2) For the increasing order echo sequence obtained in step 3.1), calculate the expression of interval second order moment estimation and obtain the equation:
[0069]
[0070] 3.3) For the increasing order echo sequence obtained in step 3.1), calculate the zero-order moment of the data in the interval and obtain the equation:
[0071] (1+aλ -1 x (L1) 2 ) -λ -(1+aλ -1 x (L2) 2 )-λ =L2-L1
[0072] 3.4) For the equations listed in 3.2) and 3.3), solve the system of equations simultaneously and obtain the parameters a and λ.
[0073] Step 4, goodness of fit test;
[0074] 4.1) For the parameters λ and a of the Pareto distribution estimated in step 3, plot the probability density function of the sample data and the probability density function of the fitted model to obtain Figure 2 .
[0075] 4.2) Perform a goodness of fit test on the curve graph drawn in 4.1) and calculate the mean square error (MSD) between the probability density function of the sample data and the probability density function of the fitting model. The calculation formula is:
[0076]
[0077] where f r (·) and f t (·) represents the probability density function of sample data and the probability density function of fitting model, k represents the sequence number of sea clutter sequence, and N represents the length of sea clutter sequence.
[0078] 4.2) Perform a goodness of fit test on the curve graph drawn in 4.1) and calculate the modified mean square error (mmsd) between the probability density function of the sample data and the probability density function of the fitted model. The calculation formula is:
[0079]
[0080]
[0081] f r (·) and f k The definition of (·) is the same as that of the mean square error, f r (·) represents the probability density function of the sample data, f k (·) represents the probability density function of the fitting model, Represents a set of tailing samples. Take the smallest is the detection threshold, function is constant false alarm detection, P f is the false alarm probability.
[0082] The effect of the present invention can be further illustrated by the following measured data experiments:
[0083] Experimental data: The measured sea clutter data set file name used in this experiment is 19980126_222936_ANTSTEP.CDF. The number of range units is 28, the number of pulses in each range unit is 60,000, and the total number of data is 28*60,000. The detailed radar parameters are shown in Table 1:
[0084] Table 1 Test X-band radar parameter index table
[0085] Parameters Parameter Value IF 150MHz Radar resolution 30m Antenna height 30m Pulse Width 200us Antenna gain 45.7dB Polarization HH, HV, VH, VV
[0086] Experimental conditions: This experiment selects sea clutter data of the 1st, 2nd, 3rd, 4th, 6th, 7th, 8th, 9th, 12th, and 13th distance units. This experiment uses the generalized Pareto distribution to model the measured sea clutter data, and uses the biquantile estimation method, truncated moment estimation method, and interval moment estimation to estimate the parameters.
[0087] Test content and results:
[0088] Figure 3 is the result graph of the mean square error between the probability density function of the sample data and the probability density function of the fitted model. Figure 3 It can be found that the mean square error (MSD) between the probability density function of the fitting model obtained by interval moment estimation and the probability density function of the sample data is the smallest.
[0089] Figure 4 is the result graph of the improved mean square error between the probability density function of the sample data and the probability density function of the fitted model. Figure 4 It can be found that the modified mean square error (mmsd) between the probability density function of the fitting model obtained by interval moment estimation and the probability density function of the sample data is the smallest.
[0090] Goodness of fit test performance analysis: In this section, the mean square error (MSD) and modified mean square error (MMSD) between the probability density function of the ten distance unit fitting models and the probability density function of the sample data obtained by the present invention are averaged. Table 2 shows the comparison of the error values of different parameter estimation methods.
[0091] Table 2 Comparison of goodness-of-fit test performance of different methods
[0092] Parameter Estimation Methods msd mmsd The proposed method <![CDATA[4.3952×10 -4 ]]> <![CDATA[1.0352×10 -9 ]]> Truncated moment estimate <![CDATA[4.7805×10 -4 ]]> <![CDATA[2.4614×10 -9 ]]> Bi-Quartile Estimation <![CDATA[1.2×10 -3 ]]> <![CDATA[2.3783×10 -9 ]]>
[0093] It can be intuitively found from Table 2 that the parameter estimation method of the present invention can focus on improving the fitting effect of the tail part of the probability density function on the basis of improving the fitting effect of the probability density function of the fitting model as a whole, which helps to suppress the interference of large-amplitude clutter on target detection and improve the performance of target detection.
Claims
1. A method for estimating sea clutter distribution parameters, characterized in that The steps include: (1) Radar echo sequence sorting; The radar transmitter transmits a pulse signal and the radar receiver receives the time domain echo signal X={x k , k = 1, 2, 3.. N} as the amplitude sample sequence of sea clutter data, and the obtained original amplitude sample sequence X is sorted in ascending order to obtain the increasing sample sequence X' = {x (k) , k=1,2,3...N}, where x (k) represents the kth sample element in the increasing sample sequence X', and N represents the length of the sea clutter sequence X'; (2) Calculation of sea clutter sequence amplitude histogram; According to the increasing order sea clutter sample data X'={x (k) , k=1,2,3...N}, calculate the sea clutter amplitude histogram; (3) Calculation of generalized Pareto distribution parameters; (3.1) Selection of sea clutter sequence: determine the selection criteria of sea clutter sequence; (3.2) Interval second-order moment calculation; According to the probability density function of the generalized Pareto distribution and the increasing sample sequence X' obtained in step (1), combined with the selection criteria of the sea clutter sequence in (3.1), the interval is selected Integrate the echo data within the interval to obtain The second moment of the interval within; (3.3) Interval zero-order moment calculation; According to the probability density function of the generalized Pareto distribution and the radar echo sequence sorted in step (2), the interval (x (L1) ,x (L2) )The zero-order moment of the data within; (3.4) Parameter calculation; According to steps (3.1) and (3.2), the results of the interval second-order moment and the percentage of the data in the interval to the total data are combined to solve the parameters λ and a of the generalized Pareto distribution. The equation is: (4) Goodness-of-fit test calculation; (4.1) Calculate the mean square error Msd; The mean square error function is defined as: where f r (·) and f k (·) represents the probability density function of sample data and the probability density function of fitting model, k represents the serial number of sea clutter sequence, and N represents the length of sea clutter sequence; (4.2) Calculate the corrected mean square error Mmsd; The improved mean square error is defined as: where f r (·) represents the probability density function of the sample data, f t (·) represents the probability density function of the fitting model, Represents a set of tailing samples, taking the smallest is the detection threshold, function is constant false alarm detection, P f is the false alarm probability; Msd and Mmsd are used to measure the accuracy of parameter estimation. The smaller the value of Msd and Mmsd, the smaller the error between the probability density function of the sample data and the probability density function of the fitting model, and the more accurate the parameter estimation.
2. The method for estimating sea clutter distribution parameters according to claim 1, characterized in that: In step 3.2, the interval second-order moment is defined as follows based on the upper and lower boundaries of the interval: Among them, L1 and L2 are two quantiles, and 0<L1<L2<1, and Respectively represent the amplitude values of the sea clutter sequence at the quantiles L1 and L2, Represents the calculation result of the interval second-order moment; The meaning of interval estimation in actual clutter sequence is: N Increasing sorting is performed. According to the differences in sea clutter echo data in different sea conditions and different distance units, combined with the selection criteria of sea clutter sequence in step 3.1, the second-order origin moment of the data in the interval is calculated according to the following formula:
3. The method for estimating sea clutter distribution parameters according to claim 1, characterized in that: In step 3.3, the zero-order moment of the interval is the percentage of the data in the calculated interval to the total data. According to the upper and lower boundaries of the interval, the zero-order moment of the data in the interval is defined as: in Represents the calculation result of the interval zero-order moment.
