Iteration-based space-time adaptive processing rapid solution implementation method

Abstract

The invention belongs to the technical field of signal and information processing, and relates to an iteration-based space-time adaptive processing rapid solution implementation method, which comprises the following steps of: 1, receiving echo data by a radar array; 2, estimating a clutter and noise covariance matrix of the detection range gate according to echo data received by a radar array; step 3, obtaining a target steering vector, preprocessing a covariance matrix and initializing a weight vector wk, k = 0; 4, setting the number of iterations K according to the clutter influence degree of the detection area; 5, iteratively calculating a weight vector and updating the weight vector; 6, if k is smaller than or equal to K, k is equal to k + 1, the fifth step is executed, and otherwise the sixth step is executed; step 7, performing normalization processing on the final wk to obtain a processor weight vector w; and 8, performing target detection on each range gate according to the processor weight vector.

Description

Technical Field

[0001] The invention belongs to the technical field of signal and information processing, and relates to a method for realizing fast solution of space-time adaptive processing based on iteration. Background Art

[0002] In the field of radar target detection, the fast space-time adaptive processing STAP technology has attracted widespread attention due to its ability to effectively suppress clutter. STAP technology makes full use of the multiple spatial information provided by multi-channel radar and the time domain information provided by coherent pulses. Through two-dimensional adaptive joint filtering in the spatial and time domains, it can effectively suppress the influence of clutter in target detection problems. However, the conventional solution method involves high-dimensional matrix inversion operations, which has high computational complexity and cannot be implemented in real time, becoming one of the main obstacles to the implementation of STAP engineering.

[0003] The existing fast STAP implementation methods mainly use dimensionality reduction processing methods, that is, reducing the matrix dimension by designing a dimensionality reduction matrix, thereby reducing the computational complexity, such as the mDT method or the JDL method. However, this type of method will sacrifice certain clutter suppression performance, and still involves direct matrix inversion operations, lacking flexibility, so the application effect in actual engineering is poor. Summary of the invention

[0004] The purpose of the present invention is to avoid the high computational complexity problem caused by the inversion operation of high-dimensional matrix in the traditional STAP algorithm, and to provide a fast STAP implementation method based on iteration for the shortcoming that the existing dimension reduction STAP method lacks flexibility. After obtaining the high-dimensional matrix by conventional methods, the present invention regards the problem of solving the STAP filter weight vector as an unconstrained quadratic convex optimization problem, and directly iterates and solves the weight vector by using the pre-processed conjugate gradient method. When the clutter influence in the detection area is small, a small number of iterations can be set, and when the clutter influence in the detection area is large, the number of iterations can be increased accordingly, so that the clutter suppression performance and the computational time consumption can be flexibly interchangeable. The present invention not only reduces the burden of large computational complexity in the traditional STAP method, but can also be extended to any other STAP processing method that requires the inversion of a high-dimensional matrix to achieve fast and flexible clutter suppression.

[0005] The technical solution of the present invention:

[0006] A fast solution implementation method based on iterative space-time adaptive processing includes the following steps:

[0007] Step 1: The radar array receives echo data;

[0008] Step 2: Estimate the detection range gate clutter plus noise covariance matrix based on the radar array received echo data;

[0009] Step 3: Output the target orientation vector, preprocess the covariance matrix and initialize the weight vector w k , k = 0;

[0010] Step 4: Set the number of iterations K according to the degree of clutter influence in the detection area;

[0011] Step 5: Start iterative calculation of weight vector and update weight vector;

[0012] Step 6: If k≤K, then k=k+1 and return to step 5, otherwise go to step 6;

[0013] Step 7: Change the last w k Perform normalization to obtain the processor weight vector w;

[0014] Step 8: Perform target detection for each range gate based on the processor weight vector.

[0015] Further, in step 1, the echo data of the echo data that does not contain the target in the lth range gate is,

[0016]

[0017] Among them, σ i is the random complex clutter echo intensity received by the i-th clutter scattering unit, which satisfies the normal distribution; n is Gaussian white noise, N c represents the number of clutter scattering units, v(f si ,f di ) is the space-time steering vector of the i-th clutter scattering unit, expressed as:

[0018]

[0019] in represents the Kronecker product, N is the number of radar array elements, M is the number of pulses received by each radar array element, and f si and f di are the spatial frequency and Doppler frequency of the i-th clutter scattering unit, respectively, expressed as:

[0020]

[0021] where θ i and are the azimuth and elevation angles of the clutter scattering unit relative to the radar array, λ is the wavelength, v is the speed of the clutter scattering unit relative to the radar array, and f PRF is the pulse repetition frequency.

[0022] Furthermore, in step 2, the two range gate data before and after the detection range gate are taken as protection units, and a total of L = 2NM sample data other than the target range gate and the protection unit are selected to estimate the noise covariance matrix R formula as follows:

[0023]

[0024] The superscript H stands for conjugate transpose.

[0025] Furthermore, in step 3, the target orientation vector can be expressed as:

[0026]

[0027] where f st and f dt are the spatial frequency and Doppler frequency of the target, respectively, expressed as:

[0028]

[0029] where and θ t and are the azimuth and elevation angles of the target relative to the radar array, v t is the speed of the target relative to the radar array

[0030] The preprocessing operation is to use the data in the first row and first column of the covariance matrix to normalize all other data. The formula is as follows:

[0031] R′=R. / abs[R(1,1)]

[0032] R′ is the covariance matrix after preprocessing;

[0033] The initialized weight vector is expressed as follows: w0 = ones(NM, 1).

[0034] Further, in step 5, the iterative calculation process is as follows:

[0035] Calculate the current gradient:

[0036]

[0037] Compute the current conjugate gradient:

[0038]

[0039] in,

[0040] Calculate the iteration step size:

[0041]

[0042] Further, in step 5, the formula for updating the weight vector is as follows:

[0043] w k =w k-1 +α k p k .

[0044] Furthermore, in step seven, the weight vector normalization formula is as follows:

[0045]

[0046] Further, in step eight, in step eight, the target detection formula for each range gate is as follows according to the processor weight vector:

[0047] y l =w H x l .

[0048] Beneficial effects of the present invention:

[0049] The present invention designs a fast solution implementation method for space-time adaptive processing based on iteration. The method can converge quickly, solves the high computational complexity problem of directly inverting high-dimensional matrices in traditional STAP processing, and can adjust the number of iterations according to the influence of clutter in the detection area. It has the characteristics that clutter suppression performance and calculation time can be flexibly interchanged. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A flowchart of a fast solution implementation method for space-time adaptive processing based on iteration;

[0051] Figure 2 It is a simulation diagram of the convergence of the method of the present invention with the number of iterations;

[0052] Figure 3 Simulation diagram of improvement factor corresponding to different numbers of iterations;

[0053] Figure 4 Output simulation graphs of target detectability corresponding to different iteration numbers;

[0054] Figure 5 This is a comparison chart of the amount of calculation between the method of the present invention and the direct inversion method. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] like Figure 1 As shown, the process of the present invention is: first, the noise covariance matrix of the detection area samples is calculated and preprocessed according to the echo data received by the radar system, the iteration parameters are initialized and the number of iterations is set according to the impact of the detection area, and finally the normalized STAP filter weight vector is obtained by iterative calculation. Specifically, the two range gate data before and after the detection range gate are used as protection units, a certain number of range gate sample data are used to estimate the noise covariance matrix of the detection area and the matrix is ​​preprocessed, the initial weight vector and the initial iteration parameters are set, and the number of iterations is set according to the size of the clutter impact in the detection area. If the clutter impact is small, a small number of iterations is set, and if the clutter impact is large, the number of iterations is increased, and then iterative calculation is performed until the iteration ends, and finally the normalized STAP filter weight vector is obtained.

[0057] Step 1: The radar array receives echo data;

[0058] The radar echo data of the lth range gate that does not contain the target is:

[0059]

[0060] where σ i In order to satisfy a certain distribution of random complex clutter echo intensity, it is assumed here that it satisfies the log-normal distribution. n is Gaussian white noise, v(f si ,f di ) is the space-time steering vector of the i-th clutter scattering unit and can be expressed as:

[0061]

[0062] in represents the Kronecker product, f si and f di They are spatial frequency and Doppler frequency respectively and can be expressed as:

[0063]

[0064] where θ i and are the azimuth and elevation angles of the clutter scattering unit relative to the radar array, respectively.

[0065] Step 2: Estimate the detection range gate clutter plus noise covariance matrix based on radar received data;

[0066] The two range gate data before and after the detection range gate are regarded as protection units, and a total of L = 2NM sample data excluding the target range gate and the protection unit are selected to estimate the noise covariance matrix

[0067]

[0068] Step 3: Give the target orientation vector, preprocess the noise covariance matrix and initialize the iteration parameters.

[0069] The target orientation vector can be expressed as:

[0070]

[0071] Preprocessing covariance matrix

[0072] R′=R. / abs[R(1,1)]

[0073] Set the initial STAP filter weight vector to

[0074] w0=ones(NM,1)

[0075] Step 4: Set the number of iterations K according to the degree of clutter influence in the detection area and start iterative calculation of the weight vector.

[0076] The specific iterative steps are:

[0077] Find the current gradient,

[0078]

[0079] Find the current conjugate gradient,

[0080]

[0081] where β k The solution formula is:

[0082]

[0083] Calculate the iteration step size,

[0084]

[0085] Step 5: Update the STAP weight vector.

[0086] w k =w k-1 +α k p k

[0087] Step 6: Repeat steps 4 and 5 until the number of iterations reaches K.

[0088] Step 7: Add the final w k Perform normalization to obtain the STAP processor weight vector w.

[0089]

[0090] Figure 2It shows the convergence of the method of the present invention under different iteration numbers. It can be seen that when the iteration number is greater than 10, the clutter suppression performance basically converges. This shows that the method has a fast convergence speed and can effectively reduce the computational complexity. Figure 3 By setting the number of iterations to 5, 10, and 50, the corresponding improvement factor graphs are obtained, which further illustrates that the method can converge quickly and achieve ideal clutter suppression performance with a small number of iterations.

[0091] Step 8: Perform STAP target detection for each range gate:

[0092] y l =w H x l

[0093] Example:

[0094] Assume that an airborne radar system is a uniform linear array with N = 32 channels, the array element spacing is half a wavelength, M = 16 pulses are emitted in a coherent pulse processing interval, it is in a positive side view working state, the radar operating frequency is 450MHz, the pulse repetition frequency is 1200Hz, and the signal bandwidth is 5MHz. The airborne radar platform height H = 12km, the platform speed V = 120m / s. Assume that the detection environment noise ratio CNR = 40dB, the target speed is 20m / s, and it is located 50km away from the radar system, that is, the 480th range gate, and the signal-to-noise ratio SCR = 10dB. The clutter scattering unit is modeled as N equally spaced at each range gate from 0 to 180 degrees in azimuth. c = 180 scattering points.

[0095] Figure 4 The STAP target detection power output corresponding to the setting of 5, 10, and 50 iterations is shown. It can be seen that if the clutter influence in the detection area is small, setting a small number of iterations can obtain ideal clutter suppression performance. In this experiment, setting the number of iterations to 5 can better detect the target.

[0096] Figure 5 The comparison of the computational complexity of this method with direct inversion shows that this method can significantly reduce the computational complexity, and the higher the matrix dimension, the more obvious the effect of reducing the computational complexity.

[0097] The above is only a specific embodiment of the present invention, and the present invention is described in detail. The unexplained part is a conventional technology. However, the protection scope of the present invention is not limited to this. Any changes or substitutions that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. The protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A fast solution implementation method based on iterative space-time adaptive processing, characterized by: The following steps are involved: Step 1: The radar array receives echo data; Step 2: Estimate the detection range gate clutter plus noise covariance matrix based on the radar array received echo data; Step 3: Output the target orientation vector, preprocess the covariance matrix and initialize the weight vector w k , k = 0; Step 4: Set the number of iterations K according to the degree of clutter influence in the detection area; Step 5: Start iterative calculation of weight vector and update weight vector; Step 6: If k≤K, then k=k+1 and return to step 5, otherwise go to step 6; Step 7: Change the last w k Perform normalization to obtain the processor weight vector w; Step 8: Perform target detection for each range gate based on the processor weight vector.

2. The method according to claim 1, characterized in that: In step 1, the echo data of the lth range gate that does not contain the target is, Among them, σ i is the random complex clutter echo intensity received by the i-th clutter scattering unit, which satisfies the normal distribution; n is Gaussian white noise, N c represents the number of clutter scattering units, v(f si ,f di ) is the space-time steering vector of the i-th clutter scattering unit, expressed as: in represents the Kronecker product, N is the number of radar array elements, M is the number of pulses received by each radar array element, and f si and f di are the spatial frequency and Doppler frequency of the i-th clutter scattering unit, respectively, expressed as: where θ i and are the azimuth and elevation angles of the clutter scattering unit relative to the radar array, λ is the wavelength, v is the speed of the clutter scattering unit relative to the radar array, and f PRF is the pulse repetition frequency.

3. The method according to claim 2, characterized in that: In step 2, the two range gate data before and after the detection range gate are taken as protection units, and a total of L = 2NM sample data excluding the target range gate and the protection unit are selected to estimate the noise covariance matrix R formula as follows: The superscript H stands for conjugate transpose.

4. The method according to claim 3, characterized in that: In step 3, the target orientation vector can be expressed as: where f st and f dt are the spatial frequency and Doppler frequency of the target, respectively, expressed as: where and θ t and are the azimuth and elevation angles of the target relative to the radar array, v t is the speed of the target relative to the radar array; The preprocessing operation is to use the data in the first row and first column of the covariance matrix to normalize all other data. The formula is as follows: R′=R. / abs[R(1,1)] R′ is the covariance matrix after preprocessing; The initialized weight vector is expressed as follows: k =ones(NM,1),k=0.

5. The method according to claim 4, characterized in that: In step 5, the iterative calculation process is as follows: Calculate the current gradient: Compute the current conjugate gradient: in, Calculate the iteration step size:

6. The method according to claim 5, characterized in that: In step 5, the formula for updating the weight vector is as follows: w k =w k-1 +α k p k 。 7. The method according to claim 6, characterized in that: In step 7, the weight vector normalization formula is as follows:

8. The method according to claim 7, characterized in that: In step eight, the target detection formula for each range gate is as follows according to the processor weight vector: y l =w H x l 。

Images