Rain clutter modeling method based on dual-parameter composite Gaussian model

Abstract

This invention discloses a rain clutter modeling method based on a dual-parameter composite Gaussian model. This method primarily addresses the problem that existing technologies cannot accurately describe the amplitude distribution characteristics of multi-pulse rain clutter in the millimeter wave band. Its implementation scheme comprises the following steps: 1) constructing four distribution models of millimeter wave rain clutter amplitude using the four dual-parameter models in the composite Gaussian model; 2) fitting a set of rain clutter datasets X with each of the four constructed millimeter wave rain clutter amplitude distribution models; 3) calculating the KS distance and KL divergence of each of the four millimeter wave rain clutter amplitude distribution models; 4) constructing an optimal amplitude distribution model selection metric; and 5) using the optimal amplitude distribution model selection metric to select the millimeter wave rain clutter amplitude distribution model with the best fitting performance for the rain clutter dataset X. This method accurately describes the amplitude distribution characteristics of millimeter wave rain clutter and can be used to construct a constant false alarm detector whose threshold adaptively changes with rain clutter characteristics, thereby improving target detection performance.

Description

Technical Field

[0001] The present invention belongs to the field of radar signal processing, and in particular relates to a rain clutter modeling method, which can be used for target detection by millimeter-wave radar in a rainy environment. Background Art

[0002] When using radar to detect and identify targets such as vehicles and aircraft in adverse weather conditions, the echo signals received by the radar contain not only the scattered signals from the targets but also dynamic clutter such as rain, snow, and fog. In the millimeter wave band, the backscattered echo power of rain clutter is high and has Doppler shift and Doppler bandwidth. The presence of rain clutter severely impacts detection performance, whether detecting targets along the range dimension or the Doppler channel. Accurately modeling rain clutter facilitates the use of adaptive detection algorithms tailored to the clutter characteristics, improving radar target detection performance in rainy environments.

[0003] In the early days, statistical modeling of rain clutter mainly used Rayleigh distribution. As radar resolution increases, rain clutter has stronger non-Gaussian characteristics and deviates greatly from the Rayleigh distribution. The lognormal distribution model and the Weibull distribution model can describe the amplitude distribution of rain clutter in high-resolution radars, but these two models can only fit a single pulse and cannot describe the spatiotemporal correlation of rain clutter. Liu Ruiping and Shen Fumin published a method of simulating rain clutter using zero memory nonlinearity (ZNML) in a paper published in Fire Control Radar Technology, 2005(01):43-46. It first generates a correlated Gaussian distribution random sequence, and then converts the correlated Gaussian random sequence into a correlated lognormal random sequence through nonlinear transformation. The amplitude characteristics of the generated simulated data are close to the measured rain clutter in the millimeter wave band. However, since this method requires the design of a linear filter based on the clutter power spectrum characteristics, the conversion of the correlation characteristics is completed simultaneously with the conversion of the amplitude probability density function. The amplitude and correlation of the clutter cannot be controlled independently. Therefore, it is only suitable for the simulation of rain clutter of a single pulse and cannot be used to describe the rain clutter received by a multi-pulse radar.

[0004] To address the shortcomings of the lognormal and Weibull distribution models, which cannot describe the inter-pulse correlation of clutter, the composite Gaussian model (CGM) has been proposed in the field of sea clutter modeling. This model has excellent fitting results for high-resolution sea clutter in X-band radars and fully accounts for the spatiotemporal correlation of clutter, making it suitable for sea clutter modeling in multi-pulse radars. This model models clutter as the product of a texture component and a speckle component. The speckle component follows a complex Gaussian distribution, while the texture component can follow different distribution types. In recent years, researchers have proposed a variety of texture distributions and corresponding composite Gaussian models. Four texture distributions are commonly used: the gamma distribution, the inverse gamma distribution, the inverse Gaussian distribution, and the lognormal distribution. These texture distributions respectively give rise to the K distribution model, the generalized Pareto distribution model, the inverse Gaussian texture composite Gaussian IG-CG model, and the lognormal texture composite Gaussian CG-LNT model. However, the application of the composite Gaussian model has primarily focused on sea clutter, with the development of various optimal and near-optimal coherent detection algorithms. However, it has not yet been applied to rain clutter modeling. Summary of the Invention

[0005] The purpose of the present invention is to address the shortcomings of the above-mentioned existing rain clutter modeling methods and propose a rain clutter modeling method based on a dual-parameter composite Gaussian model to accurately describe the amplitude distribution characteristics of rain clutter in the millimeter wave band and realize modeling of multi-pulse rain clutter received by millimeter wave radar.

[0006] To achieve the above object, the technical solution of the present invention is as follows:

[0007] (1) Using the four two-parameter models in the composite Gaussian model, we construct the millimeter-wave band rain clutter amplitude distribution model KRM based on K distribution, the millimeter-wave band rain clutter amplitude distribution model GPRM based on generalized Pareto distribution, the millimeter-wave band rain clutter amplitude distribution model IGRM based on IGCG distribution, and the millimeter-wave band rain clutter amplitude distribution model LNRM based on CGLN distribution.

[0008] (2) The optimal amplitude distribution model selection measure is constructed using the KS distance KSD(X;·) between the cumulative distribution function of any rain clutter model and the empirical cumulative distribution function of the dataset X and the KL divergence KLD(X;·) between the probability density function of any rain clutter model and the empirical probability density function of the dataset X:

[0009]

[0010] Among them, Model Best (X) represents the optimal amplitude distribution model of the rain clutter dataset X, It represents the millimeter wave band rain clutter amplitude distribution model that minimizes the value in the brackets.

[0011] (3) The four millimeter-wave band rain clutter amplitude distribution models are used to fit a set of measured rain clutter data sets X. The probability density function and standard cumulative distribution function of each model fitting data set X are obtained. The optimal amplitude distribution model selection measure is used to evaluate the fitting effect of each model. The model with the best fitting performance for the data set is selected as the final millimeter-wave band rain clutter amplitude distribution model.

[0012] Compared with the prior art, the present invention has the following advantages:

[0013] First, the present invention uses a two-parameter composite Gaussian model to construct a millimeter-wave band rain clutter amplitude distribution model, and describes the rain clutter sequence as the product of a texture component and a speckle component that obeys a complex Gaussian distribution. The texture component reflects the non-Gaussian nature of the rain clutter, and the speckle component reflects the inter-pulse correlation of the rain clutter. Therefore, the proposed millimeter-wave band rain clutter amplitude distribution model can accurately describe the amplitude distribution characteristics of multi-pulse rain clutter received by the millimeter-wave radar, and can realize constant false alarm detection of the millimeter-wave radar in a rainy environment.

[0014] Second, the present invention constructs an optimal amplitude distribution model selection measure, and the four proposed millimeter-wave band rain clutter amplitude distribution models are respectively applicable to different rainfall intensities. This allows the radar to determine the millimeter-wave band rain clutter amplitude distribution model with the best fitting performance corresponding to a set of rain clutter data received in various environments such as light rain, moderate rain, or heavy rain, thereby improving target detection performance under various rainfall intensities. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Flowchart for the implementation of the present invention;

[0016] Figure 2 This is a comparison chart of the change curves of the empirical probability density of the measured rain clutter amplitude and the fitted probability density of the millimeter wave band rain clutter amplitude distribution model KRM based on the K distribution of the present invention;

[0017] Figure 3 This is a comparison chart of the change curves of the empirical probability density of the measured rain clutter amplitude and the fitted probability density of the millimeter wave band rain clutter amplitude distribution model GPRM based on the generalized Pareto distribution of the present invention;

[0018] Figure 4 This is a comparison chart of the empirical probability density of the measured rain clutter amplitude and the fitted probability density of the millimeter wave band rain clutter amplitude distribution model IGRM based on the IGCG distribution of the present invention;

[0019] Figure 5 This is a comparison chart of the change curves of the empirical probability density of the measured rain clutter amplitude and the fitted probability density of the millimeter wave band rain clutter amplitude distribution model LNRM based on the CGLN distribution of the present invention. DETAILED DESCRIPTION

[0020] The embodiments and effects of the present invention are further described in detail below with reference to the accompanying drawings.

[0021] Reference Figure 1 , the implementation steps of this example are as follows:

[0022] Step 1: Use the four two-parameter models in the composite Gaussian model to construct four distribution models of rain clutter amplitude in the millimeter wave band.

[0023] (1.1) The millimeter-wave band rain clutter amplitude distribution model KRM based on the K distribution is expressed in the form of the K distribution model in the composite Gaussian model to represent the amplitude distribution characteristics of the millimeter-wave band rain clutter under the background of sporadic light rain. Its probability density function PDF(r;KRM) is expressed as follows:

[0024]

[0025] Where Γ() is the gamma function, The order is 1 / λ K The second kind of modified Bessel function, r represents the amplitude sequence of rain clutter, b represents the average power of rain clutter, λ K Indicates the non-Gaussian nature of rain clutter in the background of sporadic light rain, λ K The larger the value, the stronger the non-Gaussianity of the rain clutter;

[0026] (1.2) The millimeter-wave band rain clutter amplitude distribution model GPRM based on the generalized Pareto distribution model in the composite Gaussian model is used to represent the amplitude distribution characteristics of millimeter-wave band rain clutter under heavy rain and rainstorm backgrounds. Its probability density function PDF(r;GPRM) is expressed as follows:

[0027]

[0028] Where r represents the amplitude sequence of rain clutter, b represents the average power of rain clutter; λ p Indicates the non-Gaussian nature of rain clutter in heavy rain and rainstorm background, λ p The larger the value, the stronger the non-Gaussianity of the rain clutter;

[0029] (1.3) The millimeter-wave band rain clutter amplitude distribution model IGRM based on the IGCG distribution is expressed in the form of the inverse Gaussian texture composite Gaussian model IG-CG in the composite Gaussian model to represent the amplitude distribution characteristics of the millimeter-wave band rain clutter under moderate rain background. Its probability density function PDF(r;IGRM) is expressed as follows:

[0030]

[0031] in, is an intermediate variable, r represents the amplitude sequence of rain clutter, b represents the average power of rain clutter; λ I Indicates the non-Gaussian nature of rain clutter in moderate rain background, λ I The larger the value, the stronger the non-Gaussianity of the rain clutter;

[0032] (1.4) The millimeter-wave band rain clutter amplitude distribution model LNRM of the CGLN distribution is expressed in the form of the lognormal texture composite Gaussian model CG-LNT in the composite Gaussian model to represent the amplitude distribution characteristics of the millimeter-wave band rain clutter under the background of light rain. Its probability density function PDF(r;LNRM) is expressed as follows:

[0033]

[0034] Where τ represents an independent and identically distributed positive random sequence that obeys a two-parameter lognormal distribution r represents the amplitude sequence of rain clutter, b represents the average power of rain clutter; λ L Indicates the non-Gaussian nature of rain clutter in light rain background, λ L The larger the value, the stronger the non-Gaussianity of the rain clutter.

[0035] Step 2: Use the millimeter wave band rain clutter amplitude distribution model to fit the measured rain clutter dataset X.

[0036] This step uses the four millimeter-wave band rain clutter amplitude distribution models constructed in step 1 to fit a set of measured rain clutter datasets X, respectively, and obtains the probability density function and standard cumulative distribution function of each model-fitted dataset X. The specific implementation is as follows:

[0037] (2.1) Modulo all the data in the dataset X and arrange them in ascending order to obtain the rain clutter amplitude sequence r;

[0038] (2.2) Fit the rain clutter dataset X using the K-distribution-based rain clutter amplitude distribution model KRM;

[0039] (2.2.1) Estimating the biquantile parameters of the K-distributed rain clutter amplitude series r yields the estimated scale parameters of the K-distributed rain clutter amplitude distribution model KRM. and shape parameter estimates

[0040] (2.2.2) and Substitute the probability density function PDF(r;KRM) of the K-distributed millimeter wave band rain clutter amplitude distribution model KRM to obtain the probability density function of the model fitting the rain clutter data set X.

[0041] (2.2.3) Yes The standard cumulative distribution function of the rain clutter data set X fitted by the model is obtained by integrating the rain clutter amplitude sequence r in

[0042]

[0043] Among them, Γ() is the gamma function, K 1 / λ () is the second kind of modified Bessel function with order 1 / λ, r represents the amplitude sequence of rain clutter, is the estimated value of the scale parameter, which represents the average power of rain clutter in the dataset X; is the estimated value of the shape parameter, which indicates the non-Gaussian nature of the rain clutter in the dataset X. The larger the value, the stronger the non-Gaussianity of the rain clutter.

[0044] (2.3) The rain clutter amplitude distribution model GPRM based on the generalized Pareto distribution is used to fit the rain clutter dataset X:

[0045] (2.3.1) Estimating the biquantile parameters of the generalized Pareto distribution for the rain clutter amplitude series r yields the estimated scale parameters of the rain clutter amplitude distribution model GPRM based on the generalized Pareto distribution. and shape parameter estimates

[0046] (2.3.2) and Substitute the probability density function PDF(r;GPRM) of the millimeter wave band rain clutter amplitude distribution model GPRM based on the generalized Pareto distribution to obtain the probability density function of the model fitting the rain clutter data set X

[0047] (2.3.3) Yes The standard cumulative distribution function of the rain clutter data set X fitted by the model is obtained by integrating the rain clutter amplitude sequence r in

[0048]

[0049] Where r represents the amplitude sequence of rain clutter, is the estimated value of the scale parameter, which represents the average power of rain clutter in the dataset X; is the estimated value of the shape parameter, which indicates the non-Gaussian nature of the rain clutter in the dataset X. The larger the value, the stronger the non-Gaussianity of the rain clutter.

[0050] (2.4) Use the rain clutter amplitude distribution model IGRM based on IGCG distribution to fit the rain clutter dataset X:

[0051] (2.4.1) Estimating the biquantile parameters of the IGCG distribution for the rain clutter amplitude series r yields the estimated scale parameters of the millimeter-wave band rain clutter amplitude distribution model IGRM based on the IGCG distribution. and shape parameter estimates

[0052] (2.4.2) and Substitute the probability density function PDF (r; IGRM) of the millimeter wave band rain clutter amplitude distribution model IGRM based on the IGCG distribution to obtain the probability density function of the model fitting the rain clutter data set X

[0053] (2.4.3) Yes The standard cumulative distribution function of the rain clutter data set X fitted by the model is obtained by integrating the rain clutter amplitude sequence r in

[0054]

[0055] Where r represents the amplitude sequence of rain clutter, is the estimated value of the scale parameter, which represents the average power of rain clutter in the dataset X; is the estimated value of the shape parameter, which indicates the non-Gaussian nature of the rain clutter in the dataset X. The larger the value, the stronger the non-Gaussianity of the rain clutter.

[0056] (2.5) Use the rain clutter amplitude distribution model LNRM based on CGLN distribution to fit the rain clutter dataset X:

[0057] (2.5.1) Estimating the biquantile parameters of the CGLN distribution for the rain clutter amplitude series r yields the estimated scale parameters of the millimeter wave band rain clutter amplitude distribution model LNRM based on the CGLN distribution. and shape parameter estimates

[0058] (2.5.2) and Substitute the probability density function PDF(r;LNRM) of the millimeter wave band rain clutter amplitude distribution model LNRM based on CGLN distribution to obtain the probability density function of the model fitting the rain clutter data set X

[0059] (2.5.3) Yes The standard cumulative distribution function of the rain clutter data set X fitted by the model is obtained by integrating the rain clutter amplitude sequence r in

[0060]

[0061] Where τ represents an independent and identically distributed random sequence, which obeys the lognormal distribution r represents the amplitude sequence of rain clutter, is the estimated value of the scale parameter, which represents the average power of rain clutter in the dataset X; is the estimated value of the shape parameter, which indicates the non-Gaussian nature of the rain clutter in the dataset X. The larger the value, the stronger the non-Gaussianity of the rain clutter.

[0062] Step 3: Calculate the KS distance and KL divergence of the rain clutter amplitude distribution model in the four millimeter wave bands respectively.

[0063] (3.1) Obtain the rain clutter amplitude sequence r according to step (2.1);

[0064] (3.2) Generate the empirical probability density function EPDF(r;X) and empirical cumulative distribution function ECDF(r;X) of the rain clutter amplitude sequence r respectively:

[0065]

[0066]

[0067] r (n) represents the nth element of the rain clutter amplitude sequence r in ascending order, and N represents the number of elements in the sequence r;

[0068] (3.3) Calculate the KS distances of the four millimeter-wave band rain clutter amplitude distribution models: KSD(X; KRM), KSD(X; GPRM), KSD(X; IGRM), and KSD(X; LNRM). The KS distance of any model is expressed as KSD(X; ·). The calculation formula is as follows:

[0069]

[0070] Among them, r (n) represents the nth element of the rain clutter amplitude sequence, ECDF(r (n) ; X) represents the empirical cumulative distribution function of the data set X, represents the scale parameter and shape parameter obtained by using the biquantile estimation when fitting the rain clutter model to the dataset X, represents the estimated value of the parameter used by any rain clutter model The obtained standard cumulative distribution function;

[0071] (3.4) Calculate the KL divergence of the four millimeter-wave rain clutter amplitude distribution models: KLD(X; KRM), KLD(X; GPRM), KLD(X; IGRM), and KLD(X; LNRM). The KL divergence of any model is expressed as KLD(X; ·). The calculation formula is as follows:

[0072]

[0073] Among them, EPDF(r (n) ; X) represents the empirical probability density function of the data set X, represents the estimated value of the parameter used by any rain clutter model The obtained probability density function.

[0074] Step 4: Select the millimeter wave band rain clutter amplitude distribution model with the best fitting performance for the rain clutter dataset X.

[0075] (4.1) The optimal amplitude distribution model selection measure is constructed using the KS distance KSD(X;·) and KL divergence KLD(X;·) of the rain clutter amplitude distribution model in any millimeter wave band:

[0076]

[0077] Among them, Model Best (X) represents the optimal amplitude distribution model of the rain clutter dataset X, represents the millimeter wave band rain clutter amplitude distribution model that minimizes the value in the brackets;

[0078] (4.2) Bring the four KS distances obtained in step (3.3) and the four KL divergences obtained in step (3.4) into the optimal amplitude distribution model selection measure Model Best (X), the model with the best fitting performance for the rain clutter dataset X is obtained and used as the final millimeter-wave band rain clutter amplitude distribution model to complete the construction of the rain clutter model.

[0079] When a millimeter-wave radar detects a target in rainfall, it receives strong rain clutter along with the target's reflected signal, which interferes with target detection. The proposed millimeter-wave rain clutter amplitude distribution model can be used to fit the rain clutter received by the radar, deriving its non-Gaussian and average power characteristics. This allows the use of a constant false alarm detector (CFAR) whose threshold adaptively changes with the rain clutter characteristics, improving target detection performance in rainfall environments.

[0080] The effects of the present invention can be further illustrated by the following experimental results:

[0081] 1. Experimental data:

[0082] Set the millimeter-wave radar's carrier frequency to 26.9 GHz, pulse repetition frequency to 16,000 Hz, range resolution to 1.22 meters, number of range cells to 1,024, and number of pulses to 160. Place the radar on the rooftop with its beam pointing toward mid-air.

[0083] The millimeter-wave radar was used to collect rain clutter data with moderate to heavy rainfall intensity. Strong rain clutter existed in the 1-300 distance units in the near area of ​​the data.

[0084] 2. Experimental content:

[0085] All pulses within the 1-30 range units of the measured rain clutter data collected by the millimeter-wave radar are intercepted as rain clutter dataset 1. The rain clutter amplitude distribution characteristics in this dataset are the same. The four millimeter-wave band rain clutter amplitude distribution models constructed by the present invention are used to fit rain clutter dataset 1 to obtain probability density curves. These are then compared with the empirical probability density curve of dataset 1 to observe the fitting effects of the four models, where:

[0086] The fitting effect of the millimeter wave band rain clutter amplitude distribution model KRM based on K distribution is as follows: Figure 2 ,

[0087] The fitting effect of the millimeter wave band rain clutter amplitude distribution model GPRM based on the generalized Pareto distribution is shown as follows: Figure 3 ,

[0088] The fitting effect of the millimeter wave band rain clutter amplitude distribution model IGRM based on IGCG distribution is as follows: Figure 4 ,

[0089] The fitting effect of the millimeter wave band rain clutter amplitude distribution model LNRM based on CGLN distribution is as follows: Figure 5 .

[0090] In the figure, the horizontal axis represents the rain clutter amplitude, the vertical axis represents the probability density of the rain clutter amplitude, the solid line represents the empirical probability density curve of the rain clutter data in dataset 1, and the dotted line represents the probability density curve of dataset 1 fitted by the rain clutter model of the present invention.

[0091] Depend on Figure 2 and Figure 5 It can be seen that the probability density curves of KRM and LNRM deviate greatly from the empirical probability density curve, indicating that the fitting effects of the two models are poor.

[0092] Depend on Figure 4 It can be seen that the probability density curve of IGRM has a small deviation from the empirical probability density curve, which indicates that the fitting effect of the model is good;

[0093] Depend on Figure 3It can be seen that the probability density curve of GPRM has a very small deviation from the empirical probability density curve, which is the smallest among the four figures, indicating that the model has the best fitting effect on data set 1.

[0094] The optimal amplitude distribution model selection measure constructed by the present invention is used to calculate the measurement values ​​of the four models: KRM is 0.0380, GPRM is 0.0207, IGRM is 0.0222, and LNRM is 0.0384. The smaller the value, the better the fitting effect. The model with the best fitting performance selected by this measure is GPRM, which is consistent with the result obtained by Figure 2 、 3 , 4, and 5. Therefore, for dataset 1, the final millimeter-wave band rain clutter amplitude distribution model is selected as the millimeter-wave band rain clutter amplitude distribution model GPRM based on the generalized Pareto distribution.

Claims

1. A rain clutter modeling method based on a dual-parameter composite Gaussian model, characterized in that: These include: (1) Using the four two-parameter models in the composite Gaussian model, we construct the millimeter-wave band rain clutter amplitude distribution model KRM based on K distribution, the millimeter-wave band rain clutter amplitude distribution model GPRM based on generalized Pareto distribution, the millimeter-wave band rain clutter amplitude distribution model IGRM based on IGCG distribution, and the millimeter-wave band rain clutter amplitude distribution model LNRM based on CGLN distribution. (2) The optimal amplitude distribution model selection measure is constructed using the KS distance KSD(X;·) between the cumulative distribution function of any rain clutter amplitude distribution model and the empirical cumulative distribution function of the dataset X and the KL divergence KLD(X;·) between the probability density function of any rain clutter amplitude distribution model and the empirical probability density function of the dataset X: Among them, Model Best (X) represents the optimal amplitude distribution model of the rain clutter dataset X, represents the millimeter wave band rain clutter amplitude distribution model that minimizes the value in the brackets; The KSD(X;·) is expressed as follows: Where r represents the rain clutter amplitude sequence obtained by taking the modulus of all values ​​in the rain clutter dataset X and arranging them in ascending order, r (n) represents the nth element of the rain clutter amplitude sequence, ECDF(r (n) ; X) represents the empirical cumulative distribution function of the data set X, represents the estimated value of the parameter used by any rain clutter model The obtained standard cumulative distribution function is represents the scale parameter and shape parameter obtained by using the biquantile estimation when fitting the rain clutter model to the dataset X; The KLD(X;·) is expressed as follows: Where N is the number of elements in the sequence r; r is the rain clutter amplitude sequence obtained by taking the modulus of all values ​​in the rain clutter dataset X and arranging them in ascending order, r (n) represents the nth element of the rain clutter amplitude sequence, EPDF(r (n) ; X) represents the empirical probability density function of the data set X, represents the estimated value of the parameter used by any rain clutter model The obtained probability density function is represents the scale parameter and shape parameter obtained by using the biquantile estimation when fitting the rain clutter model to the dataset X; (3) The four millimeter-wave band rain clutter amplitude distribution models are used to fit a set of measured rain clutter data sets X. The probability density function and standard cumulative distribution function of each model fitting data set X are obtained. The optimal amplitude distribution model selection measure is used to evaluate the fitting effect of each model. The model with the best fitting performance for the data set is selected as the final millimeter-wave band rain clutter amplitude distribution model.

2. The method according to claim 1, characterized in that The millimeter wave band rain clutter amplitude distribution model KRM based on K distribution is constructed in (1). The millimeter wave band rain clutter amplitude distribution model KRM based on K distribution is expressed in the form of the K distribution model in the composite Gaussian model to represent the amplitude distribution characteristics of the millimeter wave band rain clutter under the background of sporadic light rain. Its probability density function PDF(r;KRM) is expressed as follows: Where Γ() is the gamma function, The order is 1 / λ K The second kind of modified Bessel function, r represents the amplitude sequence of rain clutter, b represents the average power of rain clutter, λ K Indicates the non-Gaussian nature of rain clutter in the background of sporadic light rain, λ K The larger the value, the stronger the non-Gaussianity of the rain clutter.

3. The method according to claim 1, characterized in that The millimeter wave band rain clutter amplitude distribution model GPRM based on the generalized Pareto distribution is constructed in (1). The millimeter wave band rain clutter amplitude distribution model GPRM based on the generalized Pareto distribution is expressed in the form of the generalized Pareto distribution model in the composite Gaussian model to represent the amplitude distribution characteristics of the millimeter wave band rain clutter under the background of heavy rain and rainstorm. Its probability density function PDF(r;GPRM) is expressed as follows: Where r represents the amplitude sequence of rain clutter, b represents the average power of rain clutter; λ p Indicates the non-Gaussian nature of rain clutter in heavy rain and rainstorm background, λ p The larger the value, the stronger the non-Gaussianity of the rain clutter.

4. The method according to claim 1, wherein The millimeter wave band rain clutter amplitude distribution model IGRM based on the IGCG distribution is constructed in (1). The millimeter wave band rain clutter amplitude distribution model IGRM based on the IGCG distribution is expressed in the form of the inverse Gaussian texture composite Gaussian model IG-CG in the composite Gaussian model to represent the amplitude distribution characteristics of the millimeter wave band rain clutter under the background of moderate rain. Its probability density function PDF(r;IGRM) is expressed as follows: in, is an intermediate variable, r represents the amplitude sequence of rain clutter, b represents the average power of rain clutter; λ I Indicates the non-Gaussian nature of rain clutter in moderate rain background, λ I The larger the value, the stronger the non-Gaussianity of the rain clutter.

5. The method according to claim 1, characterized in that The millimeter wave band rain clutter amplitude distribution model LNRM based on CGLN distribution is constructed in (1). The millimeter wave band rain clutter amplitude distribution model LNRM based on CGLN distribution is expressed in the form of the log-normal texture composite Gaussian model CG-LNT in the composite Gaussian model to represent the amplitude distribution characteristics of millimeter wave band rain clutter under light rain background. Its probability density function PDF(r;LNRM) is expressed as follows: Where τ represents an independent and identically distributed positive random sequence that obeys a two-parameter lognormal distribution r represents the amplitude sequence of rain clutter, b represents the average power of rain clutter; λ L Indicates the non-Gaussian nature of rain clutter in light rain background, λ L The larger the value, the stronger the non-Gaussianity of the rain clutter.

6. The method according to claim 1, characterized in that In (3), a millimeter-wave band rain clutter amplitude distribution model based on K distribution is used to fit a set of measured rain clutter data sets X, which is implemented as follows: (3a) Modulo all the data in the dataset X and arrange them in ascending order to obtain the rain clutter amplitude sequence r; (3b) The K-distributed biquantile parameter estimation is performed on the rain clutter amplitude sequence r to obtain the scale parameter estimation value of the millimeter wave band rain clutter amplitude distribution model KRM based on the K distribution. and shape parameter estimates (3c) and Substitute the probability density function PDF(r;KRM) of the K-distributed millimeter wave band rain clutter amplitude distribution model KRM to obtain the probability density function of the model fitting the rain clutter data set X. (3d) Yes The standard cumulative distribution function of the rain clutter data set X fitted by the model is obtained by integrating the rain clutter amplitude sequence r in Where Γ() is the gamma function, The order is The second kind of modified Bessel function, r represents the amplitude sequence of rain clutter, is the estimated value of the scale parameter, which represents the average power of rain clutter; is the estimated value of the shape parameter, which indicates the non-Gaussian nature of the rain clutter in the dataset X. The larger the value, the stronger the non-Gaussianity of the rain clutter.

7. The method according to claim 1, characterized in that In (3), a millimeter-wave band rain clutter amplitude distribution model based on generalized Pareto distribution is used to fit a set of measured rain clutter data sets X, which is implemented as follows: (3e) Modulo all the data in the dataset X and arrange them in ascending order to obtain the rain clutter amplitude sequence r; (3f) The biquantile parameter estimation of the generalized Pareto distribution is performed on the rain clutter amplitude sequence r, and the scale parameter estimation value of the millimeter wave band rain clutter amplitude distribution model GPRM based on the generalized Pareto distribution is obtained. and shape parameter estimates (3g) and Substitute the probability density function PDF (r; GPRM) of the millimeter wave band rain clutter amplitude distribution model GPRM of the generalized Pareto distribution to obtain the probability density function of the model fitting the rain clutter data set X (3h) The standard cumulative distribution function of the rain clutter data set X fitted by the model is obtained by integrating the rain clutter amplitude sequence r in Where r represents the amplitude sequence of rain clutter, is the estimated value of the scale parameter, which represents the average power of rain clutter; is the estimated value of the shape parameter, which indicates the non-Gaussian nature of the rain clutter in the dataset X. The larger the value, the stronger the non-Gaussianity of the rain clutter.

8. The method according to claim 1, characterized in that In (3), a millimeter-wave band rain clutter amplitude distribution model based on IGCG distribution is used to fit a set of measured rain clutter data sets X, which is implemented as follows: (3i) Modulo all the data in the dataset X and arrange them in ascending order to obtain the rain clutter amplitude sequence r; (3j) Estimation of the biquantile parameters of the IGCG distribution of the rain clutter amplitude series r is performed to obtain the estimated scale parameters of the millimeter wave band rain clutter amplitude distribution model IGRM based on the IGCG distribution. and shape parameter estimates (3k) and Substitute the probability density function PDF (r; IGRM) of the millimeter wave band rain clutter amplitude distribution model IGRM based on the IGCG distribution to obtain the probability density function of the model fitting the rain clutter data set X (3l) Yes The standard cumulative distribution function of the rain clutter data set X fitted by the model is obtained by integrating the rain clutter amplitude sequence r in Where r represents the amplitude sequence of rain clutter, is the estimated value of the scale parameter, which represents the average power of rain clutter; is the estimated value of the shape parameter, which indicates the non-Gaussian nature of the rain clutter in the dataset X. The larger the value, the stronger the non-Gaussianity of the rain clutter.

9. The method according to claim 1, characterized in that In (3), a millimeter-wave band rain clutter amplitude distribution model based on CGLN distribution is used to fit a set of measured rain clutter data sets X, which is implemented as follows: (3m) Modulo all the data in the dataset X and arrange them in ascending order to obtain the rain clutter amplitude sequence r; (3n) Estimation of the biquantile parameters of the CGLN distribution of the rain clutter amplitude series r is performed to obtain the estimated scale parameters of the millimeter wave band rain clutter amplitude distribution model LNRM based on the CGLN distribution. and shape parameter estimates (3o) and Substitute the probability density function PDF(r;LNRM) of the millimeter wave band rain clutter amplitude distribution model LNRM based on CGLN distribution to obtain the probability density function of the model fitting the rain clutter data set X (3p) The standard cumulative distribution function of the rain clutter data set X fitted by the model is obtained by integrating the rain clutter amplitude sequence r in Where τ represents an independent and identically distributed positive random sequence that obeys a two-parameter lognormal distribution r represents the amplitude sequence of rain clutter, is the estimated value of the scale parameter, which represents the average power of rain clutter; is the estimated value of the shape parameter, which indicates the non-Gaussian nature of the rain clutter in the dataset X. The larger the value, the stronger the non-Gaussianity of the rain clutter.

10. The method according to claim 1, characterized in that In (3), the model with the best fitting performance for the data set is selected through the optimal amplitude distribution model selection measure as the final millimeter wave band rain clutter amplitude distribution model, which is implemented as follows: (3r) Modulo all the data in the dataset X and arrange them in ascending order to obtain the rain clutter amplitude sequence r; (3s) Generate the empirical probability density function EPDF(r;X) and empirical cumulative distribution function ECDF(r;X) of the rain clutter amplitude sequence r: Where h = median (r (n+1) -r (n) ,n=1,2,…,N-1), r (n) represents the nth element of the rain clutter amplitude sequence r in ascending order, and N represents the number of elements in the sequence r; (3t) Using the empirical cumulative distribution function ECDF(r;X) and the standard cumulative distribution function of the four models fitting the data set X, calculate the KS distances of the four models: KSD(X;KRM), KSD(X;GPRM), KSD(X;IGRM), and KSD(X;LNRM); (3u) Using the empirical probability density function EPDF(r;X) and the probability density functions of the four models fitting the data set X, calculate the KL divergence of the four models: KLD(X;KRM), KLD(X;GPRM), KLD(X;IGRM), and KLD(X;LNRM); (3v) Substitute the four KS distances obtained in (3t) and the four KL divergences obtained in (3u) into the optimal amplitude distribution model selection measure Model Best (X), the model with the best fitting performance for the rain clutter dataset X is obtained as the final millimeter wave band rain clutter amplitude distribution model.

Images