Multi-angle radar image-based three-dimensional imaging method for multiple moving targets

Abstract

The invention discloses a multi-angle radar image-based three-dimensional imaging method for multiple moving targets, and relates to the technical field of radar three-dimensional imaging. The invention aims to solve the problem that the existing multi-angle three-dimensional imaging method cannot be directly applied to a moving target. According to the invention, echo signal separation and parameter estimation of a plurality of moving targets are realized, focusing two-dimensional radar imaging of each target is realized, and speed estimation of each target is realized. Furthermore, a multi-angle moving target projection model is established, image matching of multi-angle radar images is realized based on the multi-angle moving target projection model, and finally three-dimensional imaging of a plurality of moving targets is realized. The method can be applied to multi-angle radar moving target three-dimensional imaging.

Description

Technical Field

[0001] The invention belongs to the technical field of radar three-dimensional imaging. Background Art

[0002] Radar systems, with their all-weather, all-day remote sensing capabilities, are widely used in ship detection, imaging, and identification. However, traditional monostatic radar imaging systems only produce two-dimensional images in range and azimuth, lacking altitude information. Radar three-dimensional imaging technology can capture three-dimensional information about a target, effectively resolving geometric distortions such as overlap, camber, and shadowing common in traditional two-dimensional radar imaging. Three-dimensional imaging technology can more accurately depict the target's geometric structure, providing a crucial basis for target classification and identification, thus effectively supporting target monitoring. Consequently, radar three-dimensional imaging technology has become a current research hotspot.

[0003] Currently, the main technologies for 3D radar imaging include tomography, interferometry, and multi-angle technology. Tomography achieves 3D resolution by forming a height-dimensional synthetic aperture through multiple observations or multiple antennas, while interferometry extracts height information using the phase difference of radar images acquired at different locations. Both technologies require multi-baseline interferometric data, which significantly increases the difficulty of implementation and system complexity. In contrast, multi-angle technology is an incoherent 3D imaging method that extracts height features through the amplitude information of multi-angle images. This technology has the advantages of low system requirements and high implementation efficiency. With the improvement of radar imaging system resolution and the development of multi-view radar imaging technology, the high-precision realization of multi-angle 3D imaging technology has gained strong support, making it a more ideal choice for 3D imaging.

[0004] The principle of multi-angle 3D imaging is based on the range-Doppler model proposed by Curlander, “Location of spaceborne sar imagery” (IEEE Transactions on Geoscience and Remote Sensing, vol. GE-20, no. 3, pp. 359–364, 1982). Subsequent studies have improved this model to enhance the accuracy of altitude estimation and reduce computational complexity.

[0005] Capaldo et al. used this model to perform a 3D reconstruction of the Merano area in Italy with an altitude accuracy better than 3.0 meters in “High-resolution SAR radargrammetry: A first application with cosmoskymed spotlight imagery” (IEEE Geoscience and Remote Sensing Letters, vol. 8, no. 6, pp. 1100–1104, 2011).

[0006] Li et al. improved the traditional method by considering the effect of target height on the incident angle in "3D point cloud reconstruction of observation scene without prior information based on the single-channel single-pass WAS-SAR system" (IEEE Transactions on Geoscience and Remote Sensing, 2024).

[0007] In recent years, research on three-dimensional radar measurement of small man-made targets has also made progress: Feng et al. proposed a method based on inverse mapping and voting in "3D point cloud reconstruction using inversely mapping and voting from single pass CSAR images" (Remote Sensing, vol. 13, no. 17, 2021), which achieved three-dimensional reconstruction of vehicle targets.

[0008] Zhang et al. used a probabilistic model to weaken the influence of target anisotropy on height estimation in “A probabilistic approach for stereo 3D point cloud reconstruction from airborne single-channel multiaspect SAR image sequences” (Remote Sensing, vol. 14, no. 22, 2022) and completed the 3D reconstruction of the terminal building of Taiyuan Yaocheng Airport.

[0009] In summary, existing research is all focused on stationary targets, and its principles cannot be directly applied to moving targets, making it difficult to meet the demand for three-dimensional imaging of moving targets. Summary of the Invention

[0010] The present invention aims to solve the problem that existing multi-angle three-dimensional imaging methods cannot be directly applied to moving targets. A three-dimensional imaging method based on multi-angle radar images for multiple moving targets is now provided.

[0011] This application provides a three-dimensional imaging method for multiple moving targets based on multi-angle radar images, including:

[0012] The echo signals of multiple moving targets at two different viewing angles are collected in the radar observation area, and the echo signals are compressed to obtain the echo data of multiple moving targets at two different viewing angles.

[0013] Perform multi-scale two-dimensional fractional Fourier transform on each echo data to obtain a multi-scale two-dimensional fractional Fourier transform spectrum, and solve the Doppler parameter estimation value of each moving target under two different viewing angles;

[0014] intercepting the energy concentration area of ​​the peak point of the multi-scale two-dimensional fractional-order Fourier transform spectrum and performing a multi-scale two-dimensional fractional-order Fourier inverse transform to obtain a separated moving target echo signal;

[0015] performing range migration correction and azimuth pulse compression on the separated moving target echo signals based on Doppler parameter estimation values ​​of each moving target at the two different viewing angles, to obtain two-dimensional focused radar images of each moving target at the two different viewing angles;

[0016] Estimate the moving target's velocity vector based on the center point difference of the two-dimensional focused radar images of the same moving target at two different viewing angles;

[0017] Construct a multi-angle moving target projection matrix based on the radar's motion state, working state and the velocity vector of each moving target;

[0018] Matching the two-dimensional radar images of each target at two different viewing angles based on the multi-angle moving target projection matrix to obtain the same-name points in the imaging images of the same target at two different viewing angles;

[0019] A multi-angle three-dimensional imaging equation group is established and solved based on the multi-angle moving target projection matrix and the same-name points, thereby obtaining a three-dimensional imaging result of each target.

[0020] In one possible design, performing distance compression on each echo signal to obtain echo data of multiple moving targets at multiple different viewing angles includes:

[0021] The echo data is obtained according to the following formula:

[0022]

[0023] in, is the echo data of the mth moving target under the ith viewing angle, B is the echo signal bandwidth, sinc(·) is the Singer function, t r is the fast time, c is the speed of light, t a is the slow time, λ is the carrier wavelength, is a constant, and These are the Doppler parameters of moving targets.

[0024] In one possible design, performing a multi-scale two-dimensional fractional Fourier transform on each echo data to obtain a multi-scale two-dimensional fractional Fourier transform spectrum includes:

[0025] Perform a multi-scale two-dimensional fractional Fourier transform on the echo data according to the following formula:

[0026]

[0027] in, is the multi-scale two-dimensional fractional Fourier transform spectrum, S(·) is the input echo data, and are respectively the fast time delay and kernel function with slow time as the variable,

[0028]

[0029] is the rotation angle, σ is the delay factor, ζ is the fractional-order domain variable, λ is the carrier wavelength, c is the speed of light, t r For fast time, t a is the slow time, n is a positive integer, and δ(·) is the impulse function.

[0030] In one possible design, solving the Doppler parameter estimation value of each moving target at each viewing angle includes:

[0031] Let σ = ζ and calculate the Doppler parameter estimation value of each moving target at each viewing angle according to the following formula:

[0032]

[0033]

[0034] in, and are the estimated Doppler parameters of the mth moving target under the ith viewing angle.

[0035] In one possible design, intercepting the energy concentration region of the peak point of the multi-scale two-dimensional fractional Fourier transform spectrum and performing a multi-scale two-dimensional fractional Fourier inverse transform to obtain the separated moving target echo signal includes:

[0036] Let σ = ζ, and use the window function to intercept The energy concentration area at the peak point;

[0037] The multi-scale two-dimensional fractional Fourier inverse transform is performed on the intercepted separation area to obtain the separated moving target echo signal.

[0038] In one possible design, estimating the motion velocity vector of the moving target based on the center point difference of the two-dimensional focused radar images of the same moving target at two different viewing angles includes:

[0039] The imaging projection points of the moving targets in each two-dimensional focused radar image are extracted using a speckle extraction algorithm;

[0040] Calculate the centroid position of each moving target in the two-dimensional focused radar image based on the imaging projection point and convert it into a global coordinate system;

[0041] Use the global nearest neighbor algorithm to match the centroid positions of the same target under two different perspectives;

[0042] The motion velocity vector of the moving target is calculated based on the difference in the center of mass position of the same target under two different perspectives.

[0043] In one possible design, constructing a multi-angle moving target projection matrix based on the radar's motion state, operating state, and the motion velocity vector of each moving target includes:

[0044] The intersection of the radar beam center and the ground is the coordinate origin O, the XOY plane coincides with the ground plane, and the Y axis is parallel to Q c⊥0 The vector pointing to O constructs the global coordinate system O-XYZ, where Qc⊥0 Q c0 At the projection point on the ground, Q c0 is the position of the radar at time t = 0, and the coordinates of any scattering point P on the moving target in the global coordinate system O-XYZ at time t = 0 are l p =[x p ,y p ,z p ];

[0045] Establish the local coordinate system OX for the i-th perspective i Y i Z, any scattering point P on the moving target in the local coordinate system OX i Y i The coordinates in Z are And the scattering point P is at X i OY i Projection point P on the plane i The coordinates are

[0046] make and Construct a multi-angle moving target projection model:

[0047]

[0048] Among them, E i is the projection matrix of the multi-angle moving target, and its expression is:

[0049]

[0050] α i is the rotation angle under the i-th viewing angle, θ i is the lower viewing angle under the i-th viewing angle, t i is the central moment of the echo recorded at the i-th perspective, is the relative velocity vector between the radar and the moving target at the i-th viewing angle, The moving target is in the local coordinate system OX i Y i Speed ​​in Z.

[0051] In one possible design, matching the two-dimensional radar images of each target at two different viewing angles based on the multi-angle moving target projection matrix to obtain the same-name points in the imaging images of the same target at the two different viewing angles includes:

[0052] The Blob algorithm is used to extract the projection point set of scattering points on the moving target in the radar image at any viewing angle;

[0053] For each projection point, the possible height value range of the moving target is traversed to obtain the candidate points with the same name corresponding to each projection point;

[0054] The normalized cross-correlation operator is used to evaluate the correlation between each point in the projected point set and each point in its corresponding candidate homonymous point set, and the candidate homonymous point with the highest correlation is taken as the homonymous point of the projected point.

[0055] In a possible design, for each projection point, the possible height value range of the moving target is traversed to obtain a set of candidate points with the same name corresponding to each projection point, including:

[0056] Assume the projection point set Candidate homonymous point set Decompose the multi-angle moving target projection matrix into E i =[E iA |e iB ], E iA is a 3×3 reversible matrix, e iB is a 3×1 vector;

[0057] Use the following formula to traverse the possible height range of the moving target:

[0058]

[0059] in, is the p-th projection point in the projection point set, is the qth candidate homonymous point in the candidate homonymous point set, p = 1, 2, ..., P, q = 1, 2, ..., Q, P is the number of projection points in the projection point set, and Q is the number of candidate homonymous points in the candidate homonymous point set.

[0060] In one possible design, establishing and solving a multi-angle three-dimensional imaging equation set based on the multi-angle moving target projection matrix and the same-name points to obtain a three-dimensional imaging result for each target includes:

[0061] Expand the multi-angle moving target projection matrix

[0062] Expand the multi-angle moving target projection matrix E i =[e i1 ,e i2 ,e i3 ] T ;

[0063] When candidate homonyms is the projection point The same-name points are used to construct the multi-angle three-dimensional imaging equations:

[0064]

[0065] The SVD decomposition algorithm is used to solve the multi-angle three-dimensional imaging equation group to obtain the three-dimensional coordinates of the target scattering points, thereby realizing three-dimensional imaging of the target.

[0066] Beneficial effects of this application:

[0067] 1. The present invention can realize three-dimensional reconstruction of multiple moving targets by using radar imaging results at two angles.

[0068] 2. The present invention can effectively realize the separation and parameter estimation of target echo signals of multiple moving target scenes, and further realize two-dimensional focused imaging of the target.

[0069] 3. The present invention can use multi-angle information to realize the speed estimation of multiple moving targets.

[0070] 4. The present invention can utilize multi-angle imaging projection relationships to perform multi-angle radar image matching, which can effectively improve matching efficiency and matching accuracy.

[0071] 5. The present invention can accurately estimate the height information of the target and realize three-dimensional imaging of multiple moving targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 A flowchart of a three-dimensional imaging method based on multi-angle radar images for multiple moving targets;

[0073] Figure 2 Schematic diagram of geometric relationship of multi-view observation;

[0074] Figure 3 is a schematic diagram of the projection geometric relationship of the i-th perspective;

[0075] Figure 4 Schematic diagram of the scattering point model of the simulation target;

[0076] Figure 5(a) is a three-dimensional imaging result of target 1;

[0077] Figure 5(b) is the three-dimensional imaging result of target 2;

[0078] Figure 5(c) is the three-dimensional imaging result of target three. DETAILED DESCRIPTION

[0079] 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 making creative work are within the scope of protection of the present invention. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other in the absence of conflict.

[0080] Specific implementation method 1: refer to Figure 1 Specifically describing this embodiment, the three-dimensional imaging method for multiple moving targets based on multi-angle radar images described in this embodiment includes:

[0081] The invention collects echo signals of multiple moving targets at two different viewing angles in the radar observation area, and performs distance compression on each echo signal to obtain echo data of multiple moving targets at two different viewing angles; performs multi-scale two-dimensional fractional Fourier transform on each echo data to obtain a multi-scale two-dimensional fractional Fourier transform spectrum, and solves the Doppler parameter estimation value of each moving target at two different viewing angles; intercepts the energy concentration area of ​​the peak point of the multi-scale two-dimensional fractional Fourier transform spectrum and performs multi-scale two-dimensional fractional Fourier inverse transform to obtain the separated moving target echo signal; performs distance migration on the separated moving target echo signal based on the Doppler parameter estimation value of each moving target at the two different viewing angles. The invention adopts a novel method for the present invention to obtain a two-dimensional focused radar image of each moving target at two different viewing angles by performing dynamic correction and azimuth pulse compression; estimate the motion velocity vector of the moving target based on the center point difference of the two-dimensional focused radar images of the same moving target at two different viewing angles; construct a multi-angle moving target projection matrix based on the motion state, working state and motion velocity vector of each moving target; match the two-dimensional radar images of each target at two different viewing angles based on the multi-angle moving target projection matrix to obtain the same-name points in the imaging images of the same target at two different viewing angles; establish and solve a multi-angle three-dimensional imaging equation group based on the multi-angle moving target projection matrix and the same-name points to obtain the three-dimensional imaging result of each target.

[0082] In one embodiment, performing distance compression on each echo signal to obtain echo data of multiple moving targets at multiple different viewing angles includes:

[0083] The echo data is obtained according to the following formula:

[0084]

[0085] in, is the echo data of the mth moving target under the ith viewing angle, B is the echo signal bandwidth, sinc(·) is the Singer function, t r is the fast time, c is the speed of light, t a is the slow time, λ is the carrier wavelength, is a constant, and These are the Doppler parameters of moving targets.

[0086] In one embodiment, performing a multi-scale two-dimensional fractional Fourier transform on each echo data to obtain a multi-scale two-dimensional fractional Fourier transform spectrum includes:

[0087] Perform a multi-scale two-dimensional fractional Fourier transform on the echo data according to the following formula:

[0088]

[0089] in, is the multi-scale two-dimensional fractional Fourier transform spectrum, S(·) is the input echo data, and are respectively the fast time delay and kernel function with slow time as the variable,

[0090]

[0091] is the rotation angle, σ is the delay factor, ζ is the fractional-order domain variable, λ is the carrier wavelength, c is the speed of light, t r For fast time, t a is the slow time, n is a positive integer, and δ(·) is the impulse function.

[0092] In one embodiment, solving the Doppler parameter estimation value of each moving target at each viewing angle includes:

[0093] Let σ = ζ and calculate the Doppler parameter estimation value of each moving target at each viewing angle according to the following formula:

[0094]

[0095] in, and are the estimated Doppler parameters of the mth moving target under the ith viewing angle.

[0096] In one embodiment, intercepting the energy concentration region of the peak point of the multi-scale two-dimensional fractional Fourier transform spectrum and performing a multi-scale two-dimensional fractional Fourier inverse transform to obtain the separated moving target echo signal includes:

[0097] Let σ = ζ, and use the window function to intercept The energy concentration area at the peak point;

[0098] The multi-scale two-dimensional fractional Fourier inverse transform is performed on the intercepted separation area to obtain the separated moving target echo signal.

[0099] In one embodiment, estimating the motion velocity vector of the moving target based on the center point difference of the two-dimensional focused radar images of the same moving target at two different viewing angles includes:

[0100] The imaging projection points of the moving targets in each two-dimensional focused radar image are extracted using a speckle extraction algorithm;

[0101] Calculate the centroid position of each moving target in the two-dimensional focused radar image based on the imaging projection point and convert it into a global coordinate system;

[0102] Use the global nearest neighbor algorithm to match the centroid positions of the same target under two different perspectives;

[0103] The motion velocity vector of the moving target is calculated based on the difference in the center of mass position of the same target under two different perspectives.

[0104] In one embodiment, constructing a multi-angle moving target projection matrix according to the motion state, working state, and motion velocity vector of each moving target of the radar includes:

[0105] The intersection of the radar beam center and the ground is the coordinate origin O, the XOY plane coincides with the ground plane, and the Y axis is parallel to Q c⊥0 The vector pointing to O constructs the global coordinate system O-XYZ, where Q c⊥0 Q c0 At the projection point on the ground, Q c0 is the position of the radar at time t = 0 (the radial velocity of the radar relative to the center of the target area is zero at this time), and the coordinates of any scattering point P on the moving target in the global coordinate system O-XYZ at time t = 0 are l p =[x p ,y p ,z p ];

[0106] Establish the local coordinate system OX for the i-th perspective i Y i Z, any scattering point P on the moving target in the local coordinate system OX i Y i The coordinates in Z are And the scattering point P is at X i OY i Projection point P on the plane i The coordinates are

[0107] make and Construct a multi-angle moving target projection model:

[0108]

[0109] Among them, E i is the projection matrix of the multi-angle moving target, and its expression is:

[0110]

[0111] α i is the rotation angle under the i-th viewing angle, θ i is the lower viewing angle under the i-th viewing angle, t i is the central moment of the echo recorded at the i-th perspective, is the relative velocity vector between the radar and the moving target at the i-th viewing angle, The moving target is in the local coordinate system OX i Y i Speed ​​in Z.

[0112] In one embodiment, matching the two-dimensional radar images of each target at two different viewing angles based on the multi-angle moving target projection matrix to obtain the same-name points in the images of the same target at the two different viewing angles includes:

[0113] The Blob algorithm is used to extract the projection point set of scattering points on the moving target in the radar image at any viewing angle;

[0114] For each projection point, the possible height value range of the moving target is traversed to obtain the candidate points with the same name corresponding to each projection point;

[0115] The normalized cross-correlation operator is used to evaluate the correlation between each point in the projected point set and each point in its corresponding candidate homonymous point set, and the candidate homonymous point with the highest correlation is taken as the homonymous point of the projected point.

[0116] In one embodiment, traversing the possible height value range of the moving target for each projection point to obtain a set of candidate points with the same name corresponding to each projection point includes:

[0117] Assume the projection point set Candidate homonymous point set Decompose the multi-angle moving target projection matrix into E i =[E iA |e iB ], E iA is a 3×3 reversible matrix, e iB is a 3×1 vector;

[0118] Use the following formula to traverse the possible height range of the moving target:

[0119]

[0120] in, is the p-th projection point in the projection point set, is the qth candidate homonymous point in the candidate homonymous point set, p = 1, 2, ..., P, q = 1, 2, ..., Q, P is the number of projection points in the projection point set, and Q is the number of candidate homonymous points in the candidate homonymous point set.

[0121] In one embodiment, the establishing and solving a multi-angle three-dimensional imaging equation group based on the multi-angle moving target projection matrix and the same-name points to obtain a three-dimensional imaging result of each target includes:

[0122] Expand the multi-angle moving target projection matrix E i =[e i1 ,e i2 ,e i3 ] T ;

[0123] Expand the multi-angle moving target projection matrix E i =[e i1 ,e i2 ,e i3 ] T ;

[0124] When candidate homonyms is the projection point The same-name points are used to construct the multi-angle three-dimensional imaging equations:

[0125]

[0126] The SVD decomposition algorithm is used to solve the multi-angle three-dimensional imaging equation group to obtain the three-dimensional coordinates of the target scattering points, thereby realizing three-dimensional imaging of the target.

[0127] To further introduce the embodiments of this application, Figure 1 A three-dimensional imaging method for multiple moving targets based on multi-angle radar images is provided, including steps 1 to 6. The numbering of each step does not necessarily limit the order in which they are performed. Each step is described in detail below:

[0128] Reference Figure 1 Specifically describing this embodiment, the three-dimensional imaging method for multiple moving targets based on multi-angle radar images described in this embodiment specifically includes the following steps:

[0129] Step 1: Use the radar to collect the echo signals of multiple uniform linear motion targets in its observation area at two different viewing angles, perform distance compression on each echo signal, and obtain the echo data after distance compression.

[0130] The echo signal reflected by the mth uniform linear motion target at the i-th viewing angle is processed by range compression to obtain the echo data The expression is:

[0131]

[0132] Where B is the signal bandwidth, sinc(·) is the Singer function, and t r is the fast time, c is the speed of light, is a constant, and are the Doppler parameters of the moving target, t a is the slow time, and λ is the carrier wavelength.

[0133] Step 2: Use the multi-scale two-dimensional fractional Fourier transform (MTFrFT) to separate the echo signals of the moving targets and estimate the Doppler parameters of each moving target, and further achieve two-dimensional focused imaging of each moving target from two perspectives. The specific process is as follows:

[0134] Step 2.1 Perform a multi-scale two-dimensional fractional Fourier transform on the echo data to obtain the MTFrFT spectrum of the echo data. MTFrFT can gather the signal energy of the same target, and its formula is:

[0135]

[0136] Wherein, S(·) represents the input echo data. and They represent the fast time delay and kernel function with slow time as the variable, and the expressions are:

[0137]

[0138] is the rotation angle, σ is the delay factor, ζ is the fractional-order domain variable, δ(·) is the impulse function, and n is a positive integer.

[0139] Let σ = ζ, we can get

[0140] Step 2: Assumptions A peak point in Corresponding to the mth target, according to the position of the peak point, the Doppler parameter estimation value of the mth target can be solved using the following formula:

[0141]

[0142] in, and are the estimated Doppler parameters of the mth moving target under the ith viewing angle.

[0143] Step 2 and 3: Construct a window function that can cover the peak energy concentration area:

[0144]

[0145] Among them, W ζ Window length.

[0146] The energy concentration area of ​​each peak point The signal is intercepted and inverse MTFrFT is performed to obtain the separated moving target echo signal. The formula for inverse multi-scale two-dimensional fractional Fourier is:

[0147]

[0148] Step 24: Use the Doppler parameter estimation value obtained in step 22 to perform range migration correction and azimuth pulse compression on the separated moving target echo signals obtained in step 23 to obtain two-dimensional focused radar images of each moving target at two different viewing angles.

[0149] Step 3: Estimate the velocity vector of each target based on the center point difference of the target's two-dimensional radar imaging images at two angles. The specific process is as follows:

[0150] Step 3. Calculate the centroid position of each target in the radar image at each viewing angle.

[0151] The Blob extraction algorithm is used to extract the imaging projection points of the moving targets in the radar image, and the center of mass position of each target in the radar image at each viewing angle is calculated:

[0152]

[0153] in, is the center of mass position of the mth target at the ith viewing angle, N is the total number of scattering points, is the extracted imaging projection point.

[0154] Step 3.2: Position the center of mass Convert to the global coordinate system, the conversion formula is:

[0155]

[0156] Step 3. Use the global nearest neighbor algorithm (GNN) to match the center of mass position of the same target in the radar images under two viewing angles, and calculate the motion velocity vector of the moving target based on the change in the center of mass position.

[0157]

[0158] Among them, v m is the velocity vector of the mth target in the global coordinate system, and Δt represents the time interval between the two viewing angles of the radar observation target.

[0159] Step 4: Based on the motion state of the radar at the two viewing angles, the radar's working state, and the target's speed value obtained in step 3, a multi-angle moving target projection matrix is ​​established for each target at each viewing angle.

[0160] Multi-view geometric relationships such as Figure 2 As shown in Figure 2, the radar system observes the moving target from two different perspectives, uses the ground as the imaging projection plane, and generates radar images respectively. Assuming that the motion state of the radar platform is known, using Q c (t) represents the radar position, where t represents the system time. The moment when the radial velocity of the radar relative to the center of the target area is zero is defined as t = 0, and the radar position at this time is recorded as Q c0 , its projection point on the ground is denoted as Q c⊥0 .

[0161] Establish a global coordinate system O-XYZ: the coordinate origin O is the intersection of the radar beam center and the ground, the XOY plane coincides with the ground plane, and the Y axis is parallel to the Q c⊥0 The vector pointing to O determines the direction of the X and Z axes. For the i-th perspective (i=1,2), establish the local coordinate system OX i Y i Z, such as Figure 3 As shown, let t i It represents the central moment of the echo under the i-th observation angle. The radar position and its projection point on the ground are respectively recorded as Q c (t i ) and Q c⊥ (t i ). Let Y i Axis and Q c⊥ (t i ) points to the same direction as the vector O. The global and local coordinate systems share the Z axis, and the two can be rotated around the Z axis by an angle α i Achieve coordinate system alignment.

[0162] Assume that the target moves in a straight line at a uniform speed within the radar observation area, and its velocity vector is recorded as v p =[v px ,v py,0]. For any scattering point P on the target, its position in the O-XYZ coordinate system at time t = 0 is l p =[x p ,y p ,z p ], then at t i The position of this point in the global coordinate system at this moment can be expressed as l p +v p t i , corresponding to the local coordinate system OX i Y i The coordinates in Z are express.

[0163] The local coordinate system OX at the i-th perspective i Y i The projection geometry model under Z is as follows Figure 3 As shown. Radar motion can usually be regarded as uniform linear motion, and its velocity vector in the local coordinate system is expressed as H i and θ i Respectively represent t i Radar position Q at the moment c (t i ) altitude and radar viewing angle. Definition Target at OX i Y i The velocity in the Z coordinate system, where use Represents the relative velocity vector between the radar and the target.

[0164] Scattering point P at OX i Y i The coordinates in the Z coordinate system are Its IPP plane (X i OY i Projection point P on the plane i Coordinates make and The relationship between the two can be expressed as a multi-angle moving target projection model:

[0165]

[0166] Among them, the matrix E i It is called the multi-angle moving target projection matrix, which describes the projection transformation of the moving target from the global coordinate system to the local coordinate system of the i-th perspective. The expression is:

[0167]

[0168] α iis the rotation angle under the i-th viewing angle, θ i is the lower viewing angle under the i-th viewing angle, t i is the central moment of the echo recorded at the i-th perspective, is the relative velocity vector between the radar and the moving target at the i-th viewing angle, The moving target is in the local coordinate system OX i Y i Speed ​​in Z.

[0169] Step 5: Based on the projection matrix generated in step 4, perform image matching on the two-dimensional radar images of each target under two viewing angles to obtain the same-name points of the same target imaged under two viewing angles. The specific process is as follows:

[0170] Step 51: Set the radar image at the first viewing angle as the main image, and set the radar image at the second viewing angle as the secondary image.

[0171] Step 52: Use Blob algorithm to extract the projection point set of target scattering points in the main image Wherein, P is the number of projection points in the projection point set, p=1, 2, ..., P.

[0172] Step 5.3: For each projection point, use formula (9) to traverse the possible height value range of the moving target and obtain the candidate point set with the same name. Wherein, Q is the number of candidate homonymous points in the candidate homonymous point set, q=1,2,...,Q.

[0173]

[0174] Among them, E i is a 3×4 matrix, which can be decomposed into E i =[E iA |e iB ].E iA is a 3×3 reversible matrix, e iB is a 3 × 1 vector. Combining the two perspectives and performing a simple transformation yields formula (9).

[0175] Step 54: Use the normalized cross correlation (NCC) operator to evaluate the p-th projection point in the projection point set S1 The correlation between each candidate homonymous point and its corresponding candidate homonymous point set S2. The candidate homonymous point with the highest correlation is considered to be The points of the same name.

[0176] The formula of the NCC operator is:

[0177]

[0178] in, is the qth candidate homonymous point in the candidate homonymous point set S2, and For and The rectangular area in the master and slave images centered at μ1 and μ2 are and The average gray value.

[0179] Step 6: Based on the multi-angle moving target projection matrix obtained in step 4 and the same-name points of the two-view radar images obtained in step 5, the multi-angle projection matrix is ​​expanded into E i =[e i1 ,e i2 ,e i3 ] T , we can establish the multi-angle three-dimensional imaging equations:

[0180]

[0181] in, yes By solving the equations using the SVD decomposition algorithm, we can obtain the three-dimensional coordinates of the target scattering points, that is, the three-dimensional imaging results of the target.

[0182] In order to verify the beneficial effects of this embodiment, the following simulation experiments are carried out:

[0183] The target scattering point model used in the simulation is as follows: Figure 4 As shown in Table 1, there are three targets in total. The positions and velocities of the targets in the reference coordinate system are shown in Table 1. The simulation parameters of the radar imaging system are shown in Table 2, and the parameters of the two viewing angles are shown in Table 3.

[0184] Table 1 Target speed and position

[0185]

[0186] Table 2 Radar system simulation parameters

[0187]

[0188] Table 3 Simulation parameters of two viewing angles

[0189]

[0190] Figure 5(a)-Figure 5(c)The following are the results of 3D imaging of three targets using the multi-angle imaging method for multiple moving targets described in the present invention. The diamonds represent the actual target scattering point locations, the circles represent the estimated target scattering point locations, and their colors represent their height values. It can be seen that the 3D imaging results of the targets closely resemble the distribution of the actual target scattering points. The average nearest distance (ANND) is used to evaluate the similarity between the 3D imaging results and the actual targets. p and Ω q The distribution represents the set of scattering points p of three-dimensional imaging and the set of real target scattering points q. The expression of ANND is:

[0191]

[0192] Among them, N p The ANND of the three-dimensional imaging results of the three targets are calculated to be 1.07m, 1.01m and 0.83m respectively, which proves the effectiveness of the present embodiment.

[0193] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely illustrative of the principles and applications of the invention. It should therefore be understood that many modifications may be made to the illustrative embodiments, and that other arrangements may be devised, without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in ways other than those described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be used in other described embodiments.

Claims

1. A three-dimensional imaging method for multiple moving targets based on multi-angle radar images, characterized in that: include: The echo signals of multiple moving targets at two different viewing angles are collected in the radar observation area, and the echo signals are compressed to obtain the echo data of multiple moving targets at two different viewing angles. Perform multi-scale two-dimensional fractional Fourier transform on each echo data to obtain a multi-scale two-dimensional fractional Fourier transform spectrum, and solve the Doppler parameter estimation value of each moving target under two different viewing angles; intercepting the energy concentration area of ​​the peak point of the multi-scale two-dimensional fractional Fourier transform spectrum and performing a multi-scale two-dimensional fractional Fourier inverse transform to obtain a separated moving target echo signal; performing range migration correction and azimuth pulse compression on the separated moving target echo signals based on Doppler parameter estimation values ​​of each moving target at the two different viewing angles, to obtain two-dimensional focused radar images of each moving target at the two different viewing angles; Estimate the moving target's velocity vector based on the center point difference of the two-dimensional focused radar images of the same moving target at two different viewing angles; Construct a multi-angle moving target projection matrix based on the radar's motion state, working state and the velocity vector of each moving target; Matching the two-dimensional radar images of each target at two different viewing angles based on the multi-angle moving target projection matrix to obtain the same-name points in the imaging images of the same target at two different viewing angles; A multi-angle three-dimensional imaging equation group is established and solved based on the multi-angle moving target projection matrix and the same-name points, thereby obtaining a three-dimensional imaging result of each target.

2. The three-dimensional imaging method for multiple moving targets based on multi-angle radar images according to claim 1, characterized in that: The method of performing distance compression on each echo signal to obtain echo data of multiple moving targets at multiple different viewing angles includes: The echo data is obtained according to the following formula: in, is the echo data of the mth moving target under the ith viewing angle, B is the echo signal bandwidth, sinc(·) is the Singer function, t r is the fast time, c is the speed of light, t a is the slow time, λ is the carrier wavelength, is a constant, and These are the Doppler parameters of moving targets.

3. The three-dimensional imaging method for multiple moving targets based on multi-angle radar images according to claim 1, characterized in that: The step of performing a multi-scale two-dimensional fractional Fourier transform on each echo data to obtain a multi-scale two-dimensional fractional Fourier transform spectrum includes: Perform a multi-scale two-dimensional fractional Fourier transform on the echo data according to the following formula: in, is the multi-scale two-dimensional fractional Fourier transform spectrum, S(·) is the input echo data, and are respectively the fast time delay and kernel function with slow time as the variable, is the rotation angle, σ is the delay factor, ζ is the fractional-order domain variable, λ is the carrier wavelength, c is the speed of light, t r For fast time, t a is the slow time, n is a positive integer, and δ(·) is the impulse function.

4. The three-dimensional imaging method for multiple moving targets based on multi-angle radar images according to claim 3, characterized in that: The method of solving the Doppler parameter estimation value of each moving target at each viewing angle includes: Let σ = ζ and calculate the Doppler parameter estimation value of each moving target at each viewing angle according to the following formula: in, and are the estimated Doppler parameters of the mth moving target under the ith viewing angle.

5. The three-dimensional imaging method for multiple moving targets based on multi-angle radar images according to claim 3, characterized in that: The method of intercepting the energy concentration region of the peak point of the multi-scale two-dimensional fractional Fourier transform spectrum and performing a multi-scale two-dimensional fractional Fourier inverse transform to obtain the separated moving target echo signal includes: Let σ = ζ, and use the window function to intercept The energy concentration area at the peak point; The multi-scale two-dimensional fractional Fourier inverse transform is performed on the intercepted separation area to obtain the separated moving target echo signal.

6. The three-dimensional imaging method for multiple moving targets based on multi-angle radar images according to claim 1, characterized in that: The method of estimating the motion velocity vector of the moving target based on the center point difference of the two-dimensional focused radar images of the same moving target at two different viewing angles includes: The imaging projection points of the moving targets in each two-dimensional focused radar image are extracted using a speckle extraction algorithm; Calculate the center of mass position of each moving target in the two-dimensional focused radar image based on the imaging projection point, and convert it into a global coordinate system; Use the global nearest neighbor algorithm to match the centroid positions of the same target under two different perspectives; The motion velocity vector of the moving target is calculated based on the difference in the center of mass position of the same target under two different perspectives.

7. The three-dimensional imaging method for multiple moving targets based on multi-angle radar images according to claim 1, characterized in that: The method of constructing a multi-angle moving target projection matrix according to the motion state, working state and motion velocity vector of each moving target of the radar includes: The intersection of the radar beam center and the ground is the coordinate origin O, the XOY plane coincides with the ground plane, and the Y axis is parallel to Q c⊥0 The vector pointing to O constructs the global coordinate system O-XYZ, where Q c⊥0 Q c0 At the projection point on the ground, Q c0 is the position of the radar at time t = 0, and the coordinates of any scattering point P on the moving target in the global coordinate system O-XYZ at time t = 0 are l p =[x p ,y p ,z p ]; Establish the local coordinate system OX for the i-th perspective i Y i Z, any scattering point P on the moving target in the local coordinate system OX i Y i The coordinates in Z are And the scattering point P is at X i OY i Projection point P on the plane i The coordinates are make and Construct a multi-angle moving target projection model: Among them, E i is the projection matrix of the multi-angle moving target, and its expression is: α i is the rotation angle under the i-th viewing angle, θ i is the lower viewing angle under the i-th viewing angle, t i is the central moment of the echo recorded at the i-th perspective, is the relative velocity vector between the radar and the moving target at the i-th viewing angle, The moving target is in the local coordinate system OX i Y i Speed ​​in Z.

8. The three-dimensional imaging method for multiple moving targets based on multi-angle radar images according to claim 7, characterized in that: The matching of the two-dimensional radar images of each target at two different viewing angles based on the multi-angle moving target projection matrix to obtain the same-name points in the imaging images of the same target at two different viewing angles includes: The Blob algorithm is used to extract the projection point set of scattering points on the moving target in the radar image at any viewing angle; For each projection point, the possible height value range of the moving target is traversed to obtain the candidate points with the same name corresponding to each projection point; The normalized cross-correlation operator is used to evaluate the correlation between each point in the projected point set and each point in its corresponding candidate homonymous point set, and the candidate homonymous point with the highest correlation is taken as the homonymous point of the projected point.

9. The three-dimensional imaging method for multiple moving targets based on multi-angle radar images according to claim 8, characterized in that: The method of traversing the possible height value range of the moving target for each projection point to obtain a set of candidate points with the same name corresponding to each projection point includes: Assume the projection point set Candidate homonymous point set Decompose the multi-angle moving target projection matrix into E i =[E iA |e iB ], E iA is a 3×3 reversible matrix, e iB is a 3×1 vector; Use the following formula to traverse the possible height range of the moving target: in, is the p-th projection point in the projection point set, is the qth candidate synonym point in the candidate synonym point set, p = 1, 2, ..., P, q = 1, 2, ..., Q, P is the number of projection points in the projection point set, and Q is the number of candidate synonym points in the candidate synonym point set.

10. The three-dimensional imaging method for multiple moving targets based on multi-angle radar images according to claim 8, characterized in that: The method of establishing and solving a multi-angle three-dimensional imaging equation group based on the multi-angle moving target projection matrix and the same-name points to obtain a three-dimensional imaging result of each target includes: Expand the multi-angle moving target projection matrix When candidate homonyms is the projection point The same-name points are used to construct the multi-angle three-dimensional imaging equations: The SVD decomposition algorithm is used to solve the multi-angle three-dimensional imaging equation group to obtain the three-dimensional coordinates of the target scattering points, thereby realizing three-dimensional imaging of the target.

Images