Three-dimensional imaging method under grid mismatch based on segmented random sparsity

Abstract

The invention relates to the technical field of grid mismatch, in particular to a three-dimensional imaging method under grid mismatch based on segmented random sparsity, which comprises a time domain signal; processing the time domain signal by using a segmented random sparse strategy to obtain sparse data; setting radar three-dimensional imaging model initial parameters and target initial parameters by taking the random sparse data as residual error initial values; in the iteration process, a column with the maximum relevancy with the residual vector in the observation matrix is searched, and a selected support set is stored; using nonlinear least square to simultaneously estimate a complex scattering coefficient and a three-dimensional space coordinate; performing residual error updating by using the estimated values of the complex scattering coefficient and the three-dimensional space coordinates; and setting the number of iterations, and performing loop termination judgment by using the updated residual error and the threshold value. According to the method, the problems that the NLS-OMP method loses some frequency band information and further loses distance direction information, and the two-dimensional imaging target recognition precision is poor, the time overhead is large and defocusing is caused by grid mismatching cannot be solved are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of grid mismatch, and in particular to a three-dimensional imaging method under grid mismatch based on piecewise random sparseness. Background Art

[0002] MIMO radar sparse imaging technology supported by Compressed Sensing (CS) theory has developed rapidly.

[0003] Wang Wei et al. proposed an adaptive off-grid correction method for MIMO radar 3D imaging. This method constructs a Bayesian probability density function based on the target's sparse echo model and then uses the maximum a posteriori probability method to achieve robust CS imaging with mismatch errors. However, this method requires first-order Taylor approximation processing and assumes that the target follows a prior distribution. However, in real scenarios, targets do not necessarily conform to these prior distributions.

[0004] Yang et al.'s Random-Frequency SAR Imaging Based on Compressed Sensing (RFS) method uses random sparse sampling to reconstruct target images by transmitting a small number of random frequency points. When multiple targets are separated by less than one resolution unit, the small number of frequency points may cause some frequency bands to be discarded entirely, resulting in loss of corresponding range information and a significant degradation in CS imaging quality. Summary of the Invention

[0005] Aiming at the shortcomings of existing methods, the present invention solves the problem that the NLS-OMP method loses certain frequency band information and then loses range information, and cannot solve the problems of poor two-dimensional imaging target recognition accuracy, high time overhead and defocusing caused by grid mismatch.

[0006] The technical solution adopted by the present invention is: a three-dimensional imaging method under grid mismatch based on piecewise random sparseness includes the following steps:

[0007] Step 1: Obtain MIMO radar time domain signal;

[0008] As a preferred embodiment of the present invention, MIMO different array echo signals are used to form a time domain signal of a planar array.

[0009] As a preferred embodiment of the present invention, the center frequency of the MIMO radar is 3 GHz and the bandwidth is 1 GHz.

[0010] Step 2: Use the piecewise random sparse strategy to process the time domain signal to obtain sparse data;

[0011] As a preferred embodiment of the present invention, sparse data acquisition includes:

[0012] Divide the vector of length MNLV into equal parts i seg Segments, and then randomly select an element from each segment to form a length of i seg Piecewise random sparse data y seg ; Where V is the total number of array positions, L is the total number of frequency points, M is the number of transmitting array elements, and N is the number of receiving array elements.

[0013] Step 3: Construct a piecewise random sparse 3D imaging model, and use the random sparse data as the residual initial value to set the initial parameters of the radar 3D imaging model and the initial parameters of the target;

[0014] As a preferred embodiment of the present invention, the initial target parameters include: azimuth, distance and altitude.

[0015] Step 4: During the iteration process, search for the column in the observation matrix that has the greatest correlation with the residual vector and save the selected support set;

[0016] Step 5: Use nonlinear least squares to simultaneously estimate the complex scattering coefficient and three-dimensional spatial coordinates;

[0017] As a preferred embodiment of the present invention, the formula of nonlinear least squares estimation is:

[0018]

[0019] in, and They are the current support set Λ [i] The corresponding complex scattering coefficient and three-dimensional space coordinates.

[0020] Step 6: Update the residual using the estimated values of the complex scattering coefficient and the three-dimensional space coordinates;

[0021] As a preferred embodiment of the present invention, the formula for residual update is:

[0022]

[0023] Among them, R tr (m,n,k) is the two-way distance from the mth transmitting array element to the kth target and from the nth receiving array element to the kth grid point; f' l is the center frequency of the lth pulse, and d(k) is the complex scattering coefficient of the kth grid point.

[0024] Step 7: Set the number of iterations and use the updated residual and threshold to determine the loop termination;

[0025] As a preferred embodiment of the present invention, a three-dimensional imaging system under grid mismatch based on piecewise random sparseness includes: a memory for storing instructions executable by a processor; and a processor for executing the instructions to implement a three-dimensional imaging method under grid mismatch based on piecewise random sparseness.

[0026] As a preferred embodiment of the present invention, a computer readable medium stores computer program code, and when the computer program code is executed by a processor, a three-dimensional imaging method under grid mismatch based on piecewise random sparse is implemented.

[0027] Beneficial effects of the present invention:

[0028] 1. Without increasing hardware costs, the array is moved to form a planar array to acquire MIMO radar time-domain signals, achieving three-dimensional imaging of the scene ahead. This avoids the traditional MIMO linear array radar's two-dimensional image, which projects the three-dimensional scene onto a two-dimensional plane in the range-azimuth direction, losing the target's elevation information and being insufficient to fully describe the target's three-dimensional structural characteristics.

[0029] 2. Compared to the random sparse sampling method proposed by Yang et al., which reconstructs the target image by transmitting a small number of random frequency points, the present method uses a segmented random sparse strategy to construct sparse data. This segmentation ensures that certain frequency bands in the signal are not completely discarded, resulting in the loss of corresponding range information. This segmented random sparse strategy can reduce computational complexity while fully utilizing range information to improve imaging resolution. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a flow chart of the three-dimensional imaging method under grid mismatch based on piecewise random sparseness of the present invention;

[0031] Figure 2 is a general random sparse three-dimensional imaging result map;

[0032] Figure 3 This is the segmented random sparse three-dimensional imaging result diagram of the present invention. DETAILED DESCRIPTION

[0033] The present invention will be further described below in conjunction with the accompanying drawings and embodiments. This figure is a simplified schematic diagram, which only illustrates the basic structure of the present invention in a schematic manner, and therefore only shows the components related to the present invention.

[0034] like Figure 1 As shown, a three-dimensional imaging method under grid mismatch based on piecewise random sparseness includes the following steps:

[0035] Step 1: Use array movement to form a planar array to obtain MIMO radar time domain signals;

[0036] The time domain signal y of the array is composed of the echo signals at different array positions during the movement of the MIMO array. y is a vector of MNLV×1, where V is the total number of array positions, L is the total number of frequency points, M is the number of transmitting array elements, and N is the number of receiving array elements.

[0037] Step 2: Use the segmented random sparse strategy to construct sparse data;

[0038] Divide the vector of length MNLV into equal parts i seg Segments, and then randomly select an element from each segment to form a length of i seg Piecewise random sparse data y seg ;

[0039] For example, the general random sparse strategy randomly selects 720 data points from the entire available 2×9×5×256 data points, which is equivalent to selecting only 1 / 32 of the data; while the segmented random strategy of the present invention divides the entire available 2×9×5×256 data points into 720 segments, and then randomly selects a data point in each segment, which is also equivalent to selecting only 1 / 32 of the data;

[0040] Step 3: Construct a piecewise random sparse 3D imaging model and set the initial parameters of the radar 3D imaging model and the initial parameters of the target;

[0041] The parameters of the piecewise random sparse 3D imaging model include: residual, support set, and complex scattering coefficient; the target initial parameters are the initial values of the target's 3D spatial coordinates;

[0042] The three-dimensional spatial coordinates are the azimuth, range, and altitude position information of the target position in the radar image coordinate system;

[0043] The initial value of the residual is r [0] =y seg , the initial value of the support set is The initial value of the complex scattering coefficient is d [0] =[], the initial value of the azimuth is x [0] =[], the initial value of the distance is y [0] =[], the initial value of height is z [0] =[], the initial value of the number of iterations i=1;

[0044] Step 4: Under the condition of setting the initial parameters of the radar segmented random sparse imaging model, in the i-th iteration, search the observation matrix A seg The column with the largest correlation with the residual vector r;

[0045] In the i-th iteration, the observation matrix A is searched seg The column j with the largest correlation with the residual vector r [i] ; and save the selected support set Λ[i] ;

[0046] The formulas obtained by the maximum column and support set can refer to the NLS-OMP method.

[0047] Step 5: In the i-th iteration, use nonlinear least squares to simultaneously estimate the complex scattering coefficient and the three-dimensional space coordinates. The current support set Λ [i] The formula for the complex scattering coefficient and three-dimensional space coordinate value is:

[0048]

[0049] in, and They are the current support set Λ [i] The corresponding complex scattering coefficient and three-dimensional space coordinates;

[0050] Complex scattering coefficient d k is a complex number, and the target three-dimensional space coordinate (x k ,y k ,z k ) are all real numbers and cannot be estimated simultaneously using the gradient method; the NLS minimization problem is solved by jointly estimating the complex scattering coefficient and three-dimensional space coordinates of the target; the NLS problem is divided into real and imaginary parts and rewritten as:

[0051]

[0052] The optimization problem is converted into an NLS problem only related to real numbers, and the NLS problem is solved.

[0053] Step 6: In the i-th iteration, the residual is updated using the estimated results of the complex scattering coefficient and the three-dimensional space coordinates. The formula is:

[0054]

[0055] Among them, R tr (m,n,k) is the two-way distance from the mth transmitting array element to the kth target and from the nth receiving array element to the kth grid point; f' l is the center frequency of the lth pulse, and d(k) is the complex scattering coefficient of the kth grid point.

[0056] Step 7: Set the number of iterations i=i+1 and return to step 4; when the residual is less than the threshold, terminate the loop; and give the complex scattering coefficient and three-dimensional coordinates of the target.

[0057] The present invention adopts a segmented random sparse strategy from two-dimensional imaging to three-dimensional imaging, fully utilizes distance information, reduces calculation complexity, and ensures imaging quality.

[0058] Simulation conditions

[0059] The center frequency of the MIMO radar transmission signal is set to 3 GHz, the bandwidth is 1 GHz, the transmission signal adopts a stepped frequency signal system, the total number of system frequency points is 256, the number of transmitting array elements is 2, the number of receiving array elements is 9, and the number of array positions is 5;

[0060] The number of scene orientation points is 41, and the number of scene distance points is 41.

[0061] Table 1 Target parameters and estimation results

[0062]

[0063] Ten targets are set in the imaging scene. The specific information of the target complex scattering coefficient and three-dimensional position distribution is shown in Table 1. The amplitude of the target obeys the Swerling4 fluctuation model, and the phase is randomly distributed between (-π, π). For the convenience of comparison, the complex scattering coefficients of the targets with the same horizontal and vertical coordinates in the upper and lower layers are set to be the same. The distance between target 1 and targets 2, 3, 4, and 5 is set to 1 / 2 resolution unit. Similarly, the distance between target 6 and targets 7, 8, 9, and 10 is also set to 1 / 2 resolution unit.

[0064] Figure 2 The imaging results of the general random sparse NLS-OMP (patent: a radar robust compressed sensing imaging method under grid mismatch) are given. In this case, the amplitude and three-dimensional spatial coordinates of the target are wrong, and there are many false targets. Figure 3 The present invention presents the results of piecewise random sparse imaging, in which the amplitudes and three-dimensional spatial coordinates of all targets are accurately reconstructed. On a computer with an Intel™ 2.4 GHz CPU, the algorithms run in 28 seconds using NLS-OMP and 20 seconds using the present invention. The present invention significantly reduces the running time of the piecewise random sparse strategy and produces significantly better imaging results than NLS-OMP.

[0065] With the above-described preferred embodiments of the present invention as a guide, and with reference to the above description, relevant personnel are fully capable of making various changes and modifications without departing from the technical scope of this invention. The technical scope of this invention is not limited to the contents of the specification and must be determined according to the scope of the claims.

Claims

1. A three-dimensional imaging method under grid mismatch based on piecewise random sparseness, characterized in that: The following steps are involved: Step 1: Obtain MIMO radar time domain signal; Step 2: Use the piecewise random sparse strategy to process the time domain signal to obtain sparse data; Step 3: Construct a piecewise random sparse 3D imaging model, and use the random sparse data as the residual initial value to set the initial parameters of the radar 3D imaging model and the initial parameters of the target; Step 4: During the iteration process, search for the column in the observation matrix that has the greatest correlation with the residual vector and save the selected support set; Step 5: Use nonlinear least squares to simultaneously estimate the complex scattering coefficient and three-dimensional spatial coordinates; Step 6: Update the residual using the estimated values of the complex scattering coefficient and the three-dimensional space coordinates; Step 7: Set the number of iterations and use the updated residual and threshold to determine the loop termination.

2. The 3D imaging method under grid mismatch based on piecewise random sparseness according to claim 1, characterized in that: Sparse data acquisition includes: Divide the vector of length MNLV into equal parts i seg Segments, and then randomly select an element from each segment to form a length of i seg Piecewise random sparse data y seg ; Where V is the total number of array positions, L is the total number of frequency points, M is the number of transmitting array elements, and N is the number of receiving array elements.

3. The 3D imaging method under grid mismatch based on piecewise random sparseness according to claim 1, characterized in that: The initial target parameters include: azimuth, distance and altitude.

4. The 3D imaging method under grid mismatch based on piecewise random sparseness according to claim 1, characterized in that: The formula for nonlinear least squares estimation is: in, and They are the current support set Λ [i] The corresponding complex scattering coefficient and three-dimensional space coordinates.

5. The three-dimensional imaging method under grid mismatch based on piecewise random sparseness according to claim 1, characterized in that: The formula for residual update is: Among them, R tr (m,n,k) is the two-way distance from the mth transmitting array element to the kth target and from the nth receiving array element to the kth grid point; f' l is the center frequency of the lth pulse, and d(k) is the complex scattering coefficient of the kth grid point.

6. The three-dimensional imaging method under grid mismatch based on piecewise random sparseness according to claim 1, characterized in that: The time domain signal of the area array is composed of MIMO different array echo signals.

7. The three-dimensional imaging method under grid mismatch based on piecewise random sparseness according to claim 6, characterized in that: The center frequency of the MIMO radar is 3 GHz and the bandwidth is 1 GHz.

8. A three-dimensional imaging system under grid mismatch based on piecewise random sparseness, characterized in that: include: a memory for storing instructions executable by the processor; A processor, configured to execute instructions to implement the three-dimensional imaging method under grid mismatch based on piecewise random sparse according to any one of claims 1 to 7.

9. A computer-readable medium storing computer program code, characterized in that When the computer program code is executed by a processor, the computer program code implements the three-dimensional imaging method under grid mismatch based on piecewise random sparseness as claimed in any one of claims 1 to 7.

Images