Sea clutter generation method based on characteristic parameter depth cognition

Abstract

The invention discloses a sea clutter generation method based on characteristic parameter depth cognition. The sea clutter generation method comprises the following steps: step 1, constructing a model training data set based on actually measured sea clutter data; step 2, constructing a sea clutter characteristic parameter prediction network; and step 3, generating space-time correlation K distribution sea clutter. The invention provides a sharing-multitask neural network, deep cognition of sea clutter characteristic parameters is realized, and a foundation is laid for accurate generation of sea clutters. On the basis of deep cognition of sea clutter characteristic parameters, a space-time correlation K distribution sea clutter generation method is combined, and the whole process from radar working parameters and marine environment parameters to sea clutter data generation is achieved.

Description

Technical Field

[0001] The invention belongs to the field of radar system detection performance evaluation, and in particular relates to a method for realizing end-to-end sea clutter generation based on radar operating parameters and ocean environment parameters in this field. Background Art

[0002] The main purpose of marine radar is to detect targets. However, during the target detection process, the radar echo also contains echo signals reflected from the sea surface, which is called sea clutter. Sea clutter will have a serious impact on the target detection performance of the radar system. The radar system has strict requirements for detection performance and requires a lot of testing. The environment of radar field testing is often unstable and easily restricted and affected by weather and other conditions. The test cost is high and the time period is long. Therefore, it is very meaningful to generate high-fidelity sea clutter data to assist in the evaluation of radar system detection performance.

[0003] Regardless of the sea clutter generation method, making the generated sea clutter close to the measured data is highly dependent on the in-depth understanding of the sea clutter characteristics. After many years of research on sea clutter characteristics, many aspects have been developed, such as amplitude distribution characteristics, Doppler spectrum characteristics, amplitude mean characteristics and spatial correlation. The current sea clutter generation methods are all based on artificially assumed sea clutter characteristic parameters, which cannot be combined with the actual radar working parameters and ocean environment parameters, and cannot achieve sea clutter generation based on environmental parameter description, which is exactly the most needed ability for sea clutter generation for radar system evaluation.

[0004] In the research on the cognition of sea clutter characteristics, each sea clutter characteristic is often studied separately, without considering the coupling effect between different characteristic parameters, and failing to achieve a unified analysis and study of sea clutter characteristics. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a sea clutter generation method based on deep cognition of characteristic parameters, which can realize the whole process from measuring parameters to sea clutter generation.

[0006] The present invention adopts the following technical solution:

[0007] A sea clutter generation method based on deep cognition of characteristic parameters, the improvement of which is that it comprises the following steps:

[0008] Step 1: Construct a model training dataset based on measured sea clutter data:

[0009] Step 11, record the measured sea clutter data received by the radar as Y(m,n), where Y is a complex matrix, m is the number of range units of the sea clutter data, and n is the number of pulses of the sea clutter data;

[0010] Step 12, based on Y, select the data of 5 range gates near the beam center for superposition, and calculate the shape parameters and scale parameters of their amplitude distribution that obey the K distribution. The K distribution function is as follows:

[0011]

[0012] In the above formula, Γ(·) is the Gamma function, K v-1 (·) is the modified v-1 order Bessel function, v is the shape parameter, a is the scale parameter, and y is the clutter amplitude;

[0013] For the sea clutter amplitude sample {y i ,i=1,2,…,5n}, the calculation formula for the q-order origin moment is:

[0014]

[0015] For the K distribution function, its q-order moment expression is:

[0016]

[0017] The shape parameters and scale parameters of the distribution model are estimated using the first-order moment and second-order moment estimation methods. The calculation formula is as follows:

[0018]

[0019] In the above formula, is the first-order origin moment, is the second-order origin moment, is the estimated value of the shape parameter, is the estimated value of the scale parameter;

[0020] Step 13, based on Y, select the data of 5 range gates near the beam center, calculate the frequency shift parameters and broadening parameters of their Doppler spectra that obey the Gaussian distribution, and take the average value:

[0021] The Burg method is used to analyze the single range gate sea clutter data {y i ,i=1,2,…,n} spectrum is estimated, the AR model order is set to 50, the FFT points are set to 1024, the sampling rate is the pulse repetition frequency of the radar, and the expression of the power spectrum density function S(f) is:

[0022]

[0023] In the above formula, E p represents the variance, f is the sea clutter Doppler frequency, a m Represents AR model parameters;

[0024] The Doppler spectrum characteristic parameters of each range gate data are obtained by solving the following formula:

[0025]

[0026]

[0027] In the above formula, f d represents the Doppler shift, B w represents Doppler broadening;

[0028] Step 14: Select the data of 5 range gates near the center of the beam based on Y, calculate their scattering coefficients respectively, and take the average value:

[0029] For pulse compression radar, the scattering coefficient σ of each range gate of data Y 0 Calculated by the following formula:

[0030]

[0031] In the above formula, R m Indicates the distance from the sea surface to the radar antenna corresponding to the current range gate, P r Represents radar received power, P t Represents radar transmission power, G m Indicates the current range gate radar antenna gain value, λ indicates the radar wavelength, A m Indicates the area of ​​the sea surface illuminated by the current range gate radar beam, ρ = τ t / τ c is the pulse compression ratio, τ t is the pulse width before pulse pressure, τ c =1 / B is the pulse width after pulse compression, B is the bandwidth;

[0032] Step 15, calculate the spatial attenuation coefficient of the correlation between different range gates based on Y:

[0033] According to the sea clutter data Y, the correlation coefficient between the i-th range gate data Y(i) and the i+l-th range gate data Y(i+l) is calculated:

[0034]

[0035] In the above formula, represents the square of the modulus of the sea clutter sequence, It represents the mean square value of the sea clutter sequence, which is the average power. The superscript * represents the complex conjugate.

[0036] Fit the spatial correlation coefficient between the first range gate and other range gates with an exponential decay function to obtain the spatial attenuation coefficient:

[0037] R(m)=γm

[0038] In the above formula, γ represents the spatial attenuation coefficient;

[0039] Step 16, matching the calculated six sea clutter characteristic parameters with radar operating parameters and ocean environment parameters to construct a model training data set;

[0040] Step 2: Build a sea clutter characteristic parameter prediction network:

[0041] Step 21, select the following parameters as input to the model:

[0042] Radar operating parameters: bandwidth, polarization mode, ground grazing angle;

[0043] Marine environmental parameters: effective wave height, average wave period, wave direction angle, maximum wind speed, wind direction angle, current speed, current direction angle, water temperature;

[0044] Step 22, using a shared-multi-task neural network model, the shared layer network adopts the ResNet18 network model, and each subtask network adopts the DNN5 network model;

[0045] Step 23, introduce a multi-task loss function:

[0046]

[0047] In the above formula, are the mean square errors of different feature parameter prediction tasks, σ1,…,σ6 are the weights of different tasks learned during training, and logσ1,…,logσ6 are regularizers;

[0048] Step 24: When training the model, select the ReLu function as the activation function and the Adam function as the optimization method;

[0049] Step 3: Generate spatiotemporal correlated K-distributed sea clutter:

[0050] Step 31, generate complex data that conforms to the standard Gaussian distribution for each distance unit m, the length n of the complex data represents the number of pulses, design a Gaussian filter H(f) based on the predicted Doppler spectrum characteristic parameters, perform Fourier transform on the complex data of the standard Gaussian distribution to the frequency domain, multiply it by the filter H(f), and then perform inverse Fourier transform on the time domain to obtain complex data g with the same time correlation m (k);

[0051] Step 32, by recursive formula The speckle component z of each range gate is obtained in turn m (k), z1(k)=g1(k);

[0052] Step 33, based on the predicted shape parameters and scale parameters, a sequence of Chi distribution is generated whose amplitude distribution satisfies the following formula to obtain a texture component;

[0053]

[0054] In the above formula, x i ,i=1,…,n is the amplitude of the texture component for generating sea clutter data;

[0055] Step 34, multiplying the texture component and the speckle component to obtain the final generated K-distributed sea clutter data with temporal and spatial correlation.

[0056] Furthermore, according to the radar equation, the reflection modulation function expressions of different range gate units are as follows:

[0057]

[0058] In the above formula, is the ground rubbing angle corresponding to the door at different distances, σ 0 is the scattering coefficient, s m is the area of ​​the gate at different distances, is the antenna transmission power gain of different range gates, X(m,n) is the normalized spatiotemporal correlation K-distributed sea clutter data generated in step 34, and the matrix A(m,n) is data including pulse dimension and range dimension.

[0059] The beneficial effects of the present invention are:

[0060] The present invention proposes a shared-multitask neural network, which realizes the deep cognition of sea clutter characteristic parameters and lays the foundation for the accurate generation of sea clutter. Based on the deep cognition of sea clutter characteristic parameters, combined with the time-space related K distribution sea clutter generation method, the whole process from radar working parameters and ocean environment parameters to the generation of sea clutter data is realized.

[0061] The present invention realizes for the first time a sea clutter generation method based on the description of radar operating parameters and ocean environment parameters, establishes a shared-multitask sea clutter characteristic parameter deep cognitive model, and combines the time-space correlation K distribution sea clutter generation method. The correlation between the generated sea clutter data and the measured sea clutter data in statistical characteristics exceeds 93%. The method proposed in the present invention can effectively support the evaluation of the detection performance of the radar system. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic flow diagram of the method of the present invention;

[0063] Figure 2 It is a schematic diagram of the overall structure of the shared-multi-task neural network model;

[0064] Figure 3 It is the principle diagram of the time-space related K distribution sea clutter simulation method;

[0065] Figure 4 It is a comparison diagram of sea clutter data generated by the method of the present invention and measured sea clutter data;

[0066] Figure 5 It is a pulse dimension amplitude comparison diagram of sea clutter data generated by the method of the present invention and measured sea clutter data;

[0067] Figure 6 It is a characteristic comparison diagram of sea clutter data generated by the method of the present invention and measured sea clutter data. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0069] Embodiment 1, this embodiment discloses a method for generating sea clutter based on deep cognition of characteristic parameters, the core content of which is to achieve deep cognition of sea clutter characteristic parameters based on radar operating parameters and ocean environment parameters. To achieve the above objectives, the present invention constructs a shared-multitask neural network model in accordance with the requirements of deep cognition of sea clutter characteristic parameters, and combines the time-space related K distribution sea clutter generation method to achieve end-to-end sea clutter generation. Figure 1 As shown, the specific steps include:

[0070] Step 1: Construct a model training dataset based on measured sea clutter data:

[0071] Step 11, record the measured sea clutter data received by the radar as Y(m,n), where Y is a complex matrix, m is the number of range units of the sea clutter data, and n is the number of pulses of the sea clutter data;

[0072] Step 12, based on Y, select the data of 5 range gates near the beam center for superposition, and calculate the shape parameters and scale parameters of their amplitude distribution that obey the K distribution. The K distribution function is as follows:

[0073]

[0074] In the above formula, Γ(·) is the Gamma function, K v-1 (·) is the modified v-1 order Bessel function, v is the shape parameter, a is the scale parameter, and y is the clutter amplitude;

[0075] For the sea clutter amplitude sample {y i,i=1,2,…,5n}, the calculation formula for the q-order origin moment is:

[0076]

[0077] For the K distribution function, its q-order moment expression is:

[0078]

[0079] The shape parameters and scale parameters of the distribution model are estimated using the first-order moment and second-order moment estimation methods. The calculation formula is as follows:

[0080]

[0081] In the above formula, is the first-order origin moment, is the second-order origin moment, is the estimated value of the shape parameter, is the estimated value of the scale parameter;

[0082] Therefore, the shape parameters and scale parameters of the amplitude distribution obeying the K distribution can be estimated based on the sea clutter data samples.

[0083] Step 13, based on Y, select the data of 5 range gates near the beam center, calculate the frequency shift parameters and broadening parameters of their Doppler spectra that obey the Gaussian distribution, and take the average value:

[0084] The Burg method is used to analyze the single range gate sea clutter data {y i ,i=1,2,…,n} spectrum is estimated, the AR model order is set to 50, the FFT points are set to 1024, the sampling rate is the pulse repetition frequency of the radar, and the expression of the power spectrum density function S(f) is:

[0085]

[0086] In the above formula, E p represents the variance, f is the sea clutter Doppler frequency, a m Represents AR model parameters;

[0087] Assuming that the average Doppler spectrum of each range gate data of sea clutter data Y is a Gaussian spectrum model, the Doppler spectrum characteristics of sea clutter can be characterized by two parameters: Doppler frequency shift and broadening. Then the Doppler spectrum characteristic parameters of each range gate data are obtained by solving the following formula:

[0088]

[0089] In the above formula, f d represents the Doppler shift, B w represents Doppler broadening;

[0090] Step 14: Based on Y, select the data of 5 range gates near the beam center, calculate their scattering coefficients (which can reflect the scattering intensity of sea clutter), and take the average value:

[0091] For pulse compression radar, the scattering coefficient σ of each range gate of data Y 0 Calculated by the following formula:

[0092]

[0093] In the above formula, R m Indicates the distance from the sea surface to the radar antenna corresponding to the current range gate, P r Represents the radar receiving power, which is the square of the data modulus value, P t Represents radar transmission power, G m Indicates the current range gate radar antenna gain value, λ indicates the radar wavelength, A m Indicates the area of ​​the sea surface illuminated by the current range gate radar beam, ρ = τ t / τ c is the pulse compression ratio, τ t is the pulse width before pulse pressure, τ c =1 / B is the pulse width after pulse compression, B is the bandwidth;

[0094] Step 15, calculate the spatial attenuation coefficient of the correlation between different range gates based on Y:

[0095] According to the sea clutter data Y, the correlation coefficient between the i-th range gate data Y(i) and the i+l-th range gate data Y(i+l) is calculated:

[0096]

[0097] In the above formula, represents the square of the modulus of the sea clutter sequence, It represents the mean square value of the sea clutter sequence, which is the average power. The superscript * represents the complex conjugate.

[0098] The commonly used spatial correlation function is the exponential decay function. The spatial correlation coefficient between the first range gate and other range gates is fitted with the exponential decay function to obtain the spatial attenuation coefficient:

[0099] R(m)=γ m

[0100] In the above formula, γ represents the spatial attenuation coefficient;

[0101] Step 16, matching the calculated six sea clutter characteristic parameters with radar operating parameters and ocean environment parameters to construct a model training data set;

[0102] Step 2: Build a sea clutter characteristic parameter prediction network:

[0103] Step 21, select the following parameters as input to the model:

[0104] Radar operating parameters: bandwidth, polarization mode, ground grazing angle;

[0105] Marine environmental parameters: effective wave height, average wave period, wave direction angle, maximum wind speed, wind direction angle, current speed, current direction angle, water temperature;

[0106] Step 22, using a shared-multi-task neural network model, the shared layer network uses the ResNet18 network model, and each subtask network uses the DNN5 network model; the overall structure of the model is as follows Figure 2 As shown in the figure, by sharing the parameters of the underlying feature modules and making the subtask networks independent, the network's performance indicators for predicting sea clutter characteristic parameters are improved.

[0107] Step 23, in order to balance the differences between the six feature parameter prediction task losses, a multi-task loss function is introduced:

[0108]

[0109] In the above formula are the mean square errors of different feature parameter prediction tasks, σ1,…,σ6 are the weights of different tasks learned during training, and logσ1,…,logσ6 are regularizers;

[0110] Step 24: The number of iterations and batch size during model training are both hyperparameters that need to be adjusted. When training the model, select the ReLu function as the activation function and the Adam function as the optimization method;

[0111] In order to quantitatively evaluate the accuracy of model prediction, the mean absolute error (MAE) indicator is used as an example to evaluate the prediction results. The MAE of each characteristic parameter prediction is recorded, as shown in the following table. The smaller the MAE value, the smaller the prediction error, and the higher the prediction accuracy of the model.

[0112] The average error of the model characteristic parameter prediction

[0113]

[0114] Step 3: Generate spatiotemporal correlated K-distributed sea clutter:

[0115] Based on the spherically invariant random process (SIRP), a K-distributed sea clutter generation method with time-space correlation is obtained. The principle is as follows: Figure 3 shown.

[0116] Step 31, generate complex data that conforms to the standard Gaussian distribution for each range gate m, the length n of the complex data represents the number of pulses, design a Gaussian filter H(f) based on the predicted Doppler spectrum characteristic parameters, perform Fourier transform on the complex data of the standard Gaussian distribution to the frequency domain, multiply it by the filter H(f), and then perform inverse Fourier transform on the time domain to obtain complex data g with the same time correlation m (k);

[0117] Step 32, by recursive formula The speckle component z of each range gate is obtained in turn m (k), z1(k)=g1(k); and introduce the spatial correlation between different range gates, γ is the predicted spatial attenuation coefficient;

[0118] Step 33, based on the predicted shape parameters and scale parameters, a sequence of Chi distribution is generated whose amplitude distribution satisfies the following formula to obtain a texture component;

[0119]

[0120] In the above formula, x i ,i=1,…,n is the amplitude of the texture component for generating sea clutter data;

[0121] Step 34, multiplying the texture component and the speckle component to obtain the final generated K-distributed sea clutter data with temporal and spatial correlation.

[0122] Consider the modulation effect of the scattering coefficient reflecting the scattering intensity of sea clutter and the radar antenna pattern on the sea clutter data. According to the radar equation, the reflection modulation function expression of different range gate units is as follows:

[0123]

[0124] In the above formula, is the floor rubbing angle corresponding to the door at different distances, P t is the radar transmission power, λ is the wavelength of the transmitted radar signal, σ 0 is the scattering coefficient, s m is the area of ​​the gate at different distances, R m is the distance between the sea area illuminated by the current range gate and the radar, is the antenna transmission power gain of different range gates, X(m,n) is the normalized spatiotemporal correlation K-distributed sea clutter data generated in step 34, and the matrix A(m,n) is data including pulse dimension and range dimension.

[0125] The effect of the method of the present invention can be further illustrated by the following experimental results: Based on the measured sea clutter data of a Ku-band radar, the characteristic parameter values ​​and the predicted characteristic parameter values ​​are calculated and shown in the following table.

[0126] Calculated and predicted values ​​of clutter characteristic parameters

[0127]

[0128] According to the predicted characteristic parameter values, Figure 4 It is a comparison diagram of sea clutter data generated by the method of the present invention and measured sea clutter data. The number of range gates of the sea clutter data is 200 and the pulse length is 10,000.

[0129] like Figure 5 As shown, the data of the 100th range gate is selected and the sea clutter data amplitude diagram of its pulse dimension is plotted.

[0130] like Figure 6 As shown in the figure, the comparison diagrams of the measured sea clutter data and the generated sea clutter data are drawn from four aspects: the probability density distribution and cumulative probability density distribution of amplitude distribution, Doppler spectrum and distance correlation. It can be seen from the figure that the generated sea clutter data and the measured sea clutter data have the same change trend and have a high degree of similarity in the distribution of each characteristic.

[0131] In order to quantitatively evaluate the effect of the proposed method in generating sea clutter, the Pearson Correlation Coefficient is introduced to evaluate the correlation between the generated sea clutter data and the measured sea clutter data in terms of statistical characteristics. The Pearson Correlation Coefficient is calculated as follows:

[0132]

[0133] In the above formula, X, Y represent the distribution functions of generated sea clutter data and measured sea clutter data respectively; σ X Represents the standard deviation of the distribution function value of the generated sea clutter data; σ Y Represents the standard deviation of the measured sea clutter data distribution function value.

[0134] The characteristic parameters are predicted based on the test set data, and sea clutter data is generated according to the predicted characteristic parameters. The similarity between the measured sea clutter data and the generated sea clutter data is calculated using the above indicators from the statistical characteristics, and the records are shown in the following table.

[0135] Similarity between the distribution of generated sea clutter data and measured sea clutter data

[0136]

[0137] The closer the Pearson correlation coefficient is to 1, the better the effect of fitting the generated data to the measured data. From the evaluation results, it can be concluded that the sea clutter data generated by the method of the present invention is very close to the measured sea clutter data, which can provide strong support for the evaluation of the detection performance of the radar system.

Claims

1. A sea clutter generation method based on deep cognition of characteristic parameters, characterized in that: The steps include: Step 1: Construct a model training dataset based on measured sea clutter data: Step 11, record the measured sea clutter data received by the radar as Y(m,n), where Y is a complex matrix, m is the number of range units of the sea clutter data, and n is the number of pulses of the sea clutter data; Step 12, based on Y, select the data of 5 range gates near the beam center for superposition, and calculate the shape parameters and scale parameters of their amplitude distribution that obey the K distribution. The K distribution function is as follows: In the above formula, Γ(·) is the Gamma function, K v-1 (·) is the modified v-1 order Bessel function, v is the shape parameter, a is the scale parameter, and y is the clutter amplitude; For the sea clutter amplitude sample {y i ,i=1,2,…,5n}, the calculation formula for the q-order origin moment is: For the K distribution function, its q-order moment expression is: The shape parameters and scale parameters of the distribution model are estimated using the first-order moment and second-order moment estimation methods. The calculation formula is as follows: In the above formula, is the first-order origin moment, is the second-order origin moment, is the estimated value of the shape parameter, is the estimated value of the scale parameter; Step 13, based on Y, select the data of 5 range gates near the beam center, calculate the frequency shift parameters and broadening parameters of their Doppler spectra that obey the Gaussian distribution, and take the average value: The Burg method is used to analyze the single range gate sea clutter data {y i ,i=1,2,…,n} spectrum is estimated, the AR model order is set to 50, the FFT points are set to 1024, the sampling rate is the pulse repetition frequency of the radar, and the expression of the power spectrum density function S(f) is: In the above formula, E p represents the variance, f is the sea clutter Doppler frequency, a m Represents AR model parameters; The Doppler spectrum characteristic parameters of each range gate data are obtained by solving the following formula: In the above formula, f d represents the Doppler shift, B w represents Doppler broadening; Step 14: Select the data of 5 range gates near the center of the beam based on Y, calculate their scattering coefficients respectively, and take the average value: For pulse compression radar, the scattering coefficient σ of each range gate of data Y 0 Calculated by the following formula: In the above formula, R m Indicates the distance from the sea surface to the radar antenna corresponding to the current range gate, P r Represents radar received power, P t Represents radar transmission power, G m Indicates the current range gate radar antenna gain value, λ indicates the radar wavelength, A m Indicates the area of ​​the sea surface illuminated by the current range gate radar beam, ρ = τ t / τ c is the pulse compression ratio, τ t is the pulse width before pulse pressure, τ c =1 / B is the pulse width after pulse compression, B is the bandwidth; Step 15, calculate the spatial attenuation coefficient of the correlation between different range gates based on Y: According to the sea clutter data Y, the correlation coefficient between the i-th range gate data Y(i) and the i+l-th range gate data Y(i+l) is calculated: In the above formula, represents the square of the modulus of the sea clutter sequence, It represents the mean square value of the sea clutter sequence, which is the average power. The superscript * represents the complex conjugate. Fit the spatial correlation coefficient between the first range gate and other range gates with an exponential decay function to obtain the spatial attenuation coefficient: R(m)=γ m In the above formula, γ represents the spatial attenuation coefficient; Step 16, matching the calculated six sea clutter characteristic parameters with radar operating parameters and ocean environment parameters to construct a model training data set; Step 2: Build a sea clutter characteristic parameter prediction network: Step 21, select the following parameters as input to the model: Radar operating parameters: bandwidth, polarization mode, ground grazing angle; Marine environmental parameters: effective wave height, average wave period, wave direction angle, maximum wind speed, wind direction angle, current speed, current direction angle, water temperature; Step 22, using a shared-multi-task neural network model, the shared layer network adopts the ResNet18 network model, and each subtask network adopts the DNN5 network model; Step 23, introduce a multi-task loss function: In the above formula, are the mean square errors of different feature parameter prediction tasks, σ1,…,σ6 are the weights of different tasks learned during training, and logσ1,…,logσ6 are regularizers; Step 24: When training the model, select the ReLu function as the activation function and the Adam function as the optimization method; Step 3: Generate spatiotemporal correlated K-distributed sea clutter: Step 31, generate complex data that conforms to the standard Gaussian distribution for each distance unit m, the length n of the complex data represents the number of pulses, design a Gaussian filter H(f) based on the predicted Doppler spectrum characteristic parameters, perform Fourier transform on the complex data of the standard Gaussian distribution to the frequency domain, multiply it by the filter H(f), and then perform inverse Fourier transform on the time domain to obtain complex data g with the same time correlation m (k); Step 32, by recursive formula The speckle component z of each range gate is obtained in turn m (k), z1(k)=g1(k); Step 33, based on the predicted shape parameters and scale parameters, a sequence of Chi distribution is generated whose amplitude distribution satisfies the following formula to obtain a texture component; In the above formula, x i ,i=1,…,n is the amplitude of the texture component for generating sea clutter data; Step 34, multiplying the texture component and the speckle component to obtain the final generated K-distributed sea clutter data with temporal and spatial correlation.

2. The sea clutter generation method based on deep cognition of characteristic parameters according to claim 1 is characterized in that: According to the radar equation, the reflection modulation function expressions of different range gate units are as follows: In the above formula, is the ground rubbing angle corresponding to the door at different distances, σ 0 is the scattering coefficient, s m is the area of ​​the gate at different distances, is the antenna transmission power gain of different range gates, X(m,n) is the normalized spatiotemporal correlation K-distributed sea clutter data generated in step 34, and the matrix A(m,n) is data including pulse dimension and range dimension.

Images