A method for two-dimensional super-resolution imaging of a deep space probe hovering downward
By establishing an azimuth pitch echo model and weighted least squares algorithm, the problem of blind spots and insufficient resolution of the deep space detector down view area in the prior art is solved, and a high-resolution deep space detector hover down view is realized Two-dimensional super-resolution imaging provides more accurate detection and landing guidance.
Patent Information
- Application Number
- CN202210777768.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-07-04
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2042-07-04
AI Technical Summary
Existing synthetic aperture radar imaging technology and Doppler beam sharpening technology have imaging blind spots in the visual area under deep space detectors, making it impossible to achieve high-resolution imaging. The azimuth and pitch angle resolution of traditional real aperture radars is limited by the antenna aperture. , accurate deep space detector landing guidance cannot be provided.
By establishing an azimuth pitch echo model and using a weighted least squares algorithm, the two-dimensional scanning geometric configuration of a fixed RAR platform is derived, matrixed and vectorized transformation is carried out, the weighted least squares cost function is constructed, and the target backscattering is iteratively estimated. The coefficient significantly improves the resolution of azimuth and pitch angles.
It significantly improves the azimuth and pitch angle resolution of the real-aperture radar, realizes two-dimensional super-resolution imaging of the hovered down view of the deep space detector, and provides more accurate detection and landing guidance capabilities.
Smart Images

Figure CN115079174B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of radar signal processing, and in particular relates to a two-dimensional super-resolution imaging method for a deep space probe hovering and looking downward. Background Art
[0002] In deep space exploration missions, the safe landing of the probe is of vital importance. Due to the large changes in natural light illumination in the deep space environment, the landing method based on optical sensors is greatly affected by sandstorms and clouds on the surface of the planet, making it difficult to provide the probe with all-day and all-weather downward detection. Radar has the advantages of strong penetration and no restrictions on light, and is an effective means to achieve high-resolution downward detection of deep space probes. Due to mechanical defects, the existing synthetic aperture radar imaging technology and Doppler beam sharpening technology have imaging blind spots in the downward-looking area and cannot be applied to the downward-looking imaging of deep space probes. Real aperture radar (RAR) is an important remote sensing method, widely used in civil and military occasions such as terrain mapping and target perception. Traditional RAR obtains a series of stable observation echoes through scanning, but its azimuth and elevation resolution are severely limited by the size of the antenna aperture.
[0003] In order to solve the problem of downward resolution of deep space probes, the paper "Foessel-Bunting A, Whittaker W. MMW-scanning radar for descent guidance and landing safeguard [C], Proceedings of the 6th International Symposium on Artificial Intelligence, Robotics, and Automation in Space. 2001." proposed a fixed platform RAR downward detection method based on millimeter waves. This method can be used for real-time downward detection of Mars probes. However, the resolution of this method is limited by the antenna aperture of the deep space probe and cannot provide accurate guidance for the probe landing. In order to improve the resolution, the paper "Super-resolution imaging for real aperture radar by two-dimensional inverse filtering, in 2019 6th Asia-Pacific Conference on Synthetic Aperture Radar (APSAR). IEEE, 2019." proposed a two-dimensional super-resolution imaging method based on fast Fourier transform (FFT) to improve the two-dimensional resolution of RAR. This method transforms the two-dimensional deconvolution problem into an inverse filtering problem in the frequency domain. In this process, a threshold parameter is introduced to control the noise amplification in the inverse filtering process. This method can indeed be used for two-dimensional super-resolution imaging, but its improvement in azimuth and elevation two-dimensional resolution is limited. Summary of the invention
[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a deep space probe hovering downward two-dimensional super-resolution imaging method that improves the azimuth and elevation two-dimensional angular resolution of RAR through azimuth and elevation echo model conversion and weighted least squares algorithm.
[0005] The objective of the present invention is achieved through the following technical solution: A method for two-dimensional super-resolution imaging of a deep space probe hovering downward comprises the following steps:
[0006] Step 1: Establish the echo expression of the fixed RAR platform: Assume that the target P is located at R p ,θ p , Represent the distance, azimuth, and elevation of the target respectively; the area of interest is Ω, and the scanning radar scans along the azimuth and elevation directions at a scanning speed of ω a and ωp Scan; for a certain point P, the RAR range-azimuth-elevation three-dimensional echo is expressed as:
[0007]
[0008] Where θ represents the azimuth variable, represents the pitch angle variable, τ represents the fast time variable; κ is the amplitude constant related to the radar system parameters, σ p is the effective scattering coefficient at point P, a(·) is the two-dimensional antenna pattern function, rect(·) is the rectangular envelope function, T p is the time width of the transmitted signal, K r is the frequency modulation slope of the transmitted signal, c is the propagation speed of the electromagnetic wave;
[0009] Step 2: pre-process the echo in the range direction;
[0010] Step 3: Construct an azimuth-elevation two-dimensional convolution model;
[0011] Step 4: Discretize the echo model;
[0012] Step 5: Calculate the target scattered power diagonal matrix;
[0013] Step 6: Calculate the autocorrelation matrix R;
[0014] Step 7: Calculate R -1 ;
[0015] Step 8: Iteratively estimate the target backscatter coefficient.
[0016] Furthermore, the specific implementation method of step 2 is: perform fast Fourier transform on the distance direction of equation (1), and multiply it by the pulse compression matching function in the frequency domain to obtain a spatial three-dimensional echo with high distance resolution expressed as:
[0017]
[0018] in represents a target point in the airspace Ω, Indicates that it is located is the radar backscatter coefficient of the target at , sinc(·) represents the envelope of the pulse compression response function, and B is the bandwidth of the transmitted signal.
[0019] Furthermore, the specific implementation method of step 3 is as follows: for a fixed range slice, only different azimuth and elevation scattering are considered; the echo data of the range unit is regarded as the convolution of the target scattering coefficient and the two-dimensional antenna pattern in the azimuth and elevation two-dimensional scene; in order to obtain a more concise expression, assuming that κ is a constant of 1 and only the azimuth and elevation changes are considered, then formula (2) is expressed as:
[0020]
[0021] Considering additive white noise, it can be further expressed as:
[0022]
[0023] Where * represents a two-dimensional convolution operation, Equivalent to a two-dimensional convolution kernel.
[0024] Furthermore, the specific implementation method of step 4 is: simplifying the two-dimensional deconvolution model (4) into the following matrix multiplication form:
[0025] Y=AXB T +N (5)
[0026] in, represents the echo data matrix, represents a complex number set, M and N represent the number of sampling points of the echo in the azimuth and elevation directions respectively; represents the target scattering coefficient matrix, K1 and K2 represent the dimensions of the target scattering matrix in azimuth and elevation directions respectively; represents the noise matrix, T represents the matrix transpose operation, and Respectively represent the azimuth and elevation antenna pattern modulation matrices; where A and B are specifically:
[0027]
[0028]
[0029] in, represents the sampling point of the azimuth antenna pattern, represents the sampling point of the antenna pattern in the elevation direction; L1 and L2 are the number of sampling points of the antenna pattern in the azimuth and elevation directions; the sampling points M, N, K1, K2, L1, and L2 satisfy the relationship between them:
[0030]
[0031] Using the properties of the Kronecker product, model (5) is simplified to:
[0032] y=Fx+n (9)
[0033] Among them, y, F, x, and n satisfy the following relations respectively,
[0034]
[0035] Among them, vec(·) represents vectorization, represents the Kronecker product.
[0036] Furthermore, the specific implementation method of step 5 is as follows: let P be a K1K2×K1K2 dimensional diagonal matrix, whose diagonal elements include the power of each target point of the target scattering coefficient matrix X, then P is:
[0037]
[0038] Where diag(·) represents the diagonalization operation, P k =|x k | 2 , k=1,...,K1K2,x k Represents the kth element of x.
[0039] Furthermore, the specific implementation method of step six is: define the autocorrelation matrix as:
[0040]
[0041] Where H represents the conjugate transpose operation; R is loaded diagonally:
[0042] R = FPF H +λI (13)
[0043] where I represents the identity matrix and λ represents the regularization parameter that balances noise and resolution.
[0044] Furthermore, the specific implementation method of step seven is: divide R into blocks:
[0045]
[0046] Among them, C, Then calculate
[0047]
[0048] Among them C -1 = inverse(C) directly uses the traditional method to calculate. inverse(C) means solving the inverse matrix of C by the direct inversion method; then calculating S by the traditional method -1 And divide it into blocks:
[0049]
[0050] in
[0051] Then calculate
[0052] W1=C -1 +C -1 GMGH C -1 (17)
[0053] and
[0054] W2=-C -1 G[MN] (18)
[0055] Finally, we get R -1 It is expressed as:
[0056]
[0057] Furthermore, the specific implementation method of step eight is: constructing a weighted least squares cost function:
[0058]
[0059] Which meets in f k represents the kth column of F; minimizing J gives:
[0060]
[0061] Traverse k = 1, 2, ..., K1K2 to obtain Then, by The two-dimensional target scattering estimate is obtained by rearranging as follows:
[0062]
[0063] The beneficial effects of the present invention are as follows: the present invention derives an azimuth-elevation two-dimensional echo model of a fixed distance slice according to the two-dimensional scanning geometric configuration of a fixed RAR platform. Then, the echo model is converted into a discrete form by matrixing and vectoring. Finally, a weighted least squares cost function is constructed, and the matrix inverse lemma is used to iteratively estimate the target backscatter coefficient. Compared with the traditional method, the present method significantly improves the azimuth and elevation resolution of the RAR. BRIEF DESCRIPTION OF THE DRAWINGS
[0064] Figure 1 is a flow chart of the two-dimensional super-resolution imaging method of the present invention;
[0065] Figure 2 Schematic diagram of geometric imaging of the RAR system of the present invention;
[0066] Figure 3 is the original scene graph of this embodiment;
[0067] Figure 4 is the two-dimensional antenna pattern (side view) of the present invention;
[0068] Figure 5 This is a comparison diagram of imaging results between the method of the present invention and the traditional method;
[0069] Figure 6 This is a cross-sectional diagram of the azimuth super-resolution result of the present invention;
[0070] Figure 7 This is a cross-sectional view of the super-resolution result in the pitch direction of the present invention. DETAILED DESCRIPTION
[0071] The solution of the present invention is to improve the azimuth and elevation two-dimensional angular resolution of RAR through azimuth and elevation echo model conversion and weighted least squares algorithm. First, the echo model of the range slice R is derived according to the two-dimensional scanning geometry of the fixed RAR platform. Then, the echo model is converted into a discrete form through matrixization and vectorization. Finally, a weighted least squares cost function is constructed, and the matrix inverse lemma is used to iteratively estimate the target backscatter coefficient. Compared with the traditional method, this method significantly improves the azimuth and elevation resolution of RAR.
[0072] The present invention is mainly verified by simulation experiment, and all steps and conclusions are verified to be correct on Matlab 2019. The technical solution of the present invention is further explained below in conjunction with the accompanying drawings.
[0073] like Figure 1 As shown, a method for two-dimensional super-resolution imaging of a deep space probe hovering downwards of the present invention comprises the following steps:
[0074] Step 1: Establish the echo expression of the fixed RAR platform: For a fixed platform two-dimensional scanning RAR imaging system, its geometric model is as follows Figure 2 As shown. Assume that there are four point targets distributed in a certain distance unit, such as Figure 3 The simulation parameter settings of the radar platform are shown in Table 1.
[0075] Table 1
[0076] parameter symbol Numeric Working frequency band <![CDATA[f c ]]> 10.5GHz Signal bandwidth B 15MHz Sampling frequency <![CDATA[f s ]]> 40MHz Pulse Width <![CDATA[T p ]]> 2μs Azimuth scanning speed <![CDATA[ω a ]]> 10° / s Azimuth scanning range <![CDATA[Ω a ]]> -10°10° Azimuth beamwidth <![CDATA[Δθ a ]]> 3° Pitch scanning speed <![CDATA[ω p ]]> 10° / s Pitch scanning range <![CDATA[Ω p ]]> -10°10° Elevation beam width <![CDATA[Δθ p ]]> 3° Pulse repetition frequency PRF 100Hz
[0077] Assume that the target P is located at R p ,θ p , Respectively represent the distance, azimuth, and elevation of the target; Ω a =-10°~10° and Ω p =-10°~10°, the scanning radar scans along the azimuth and elevation directions at a scanning speed of ω a =10° / s and ω p =10° / s for scanning; for a certain point P, the range-azimuth-elevation three-dimensional echo of RAR is expressed as:
[0078]
[0079] Where θ represents the azimuth variable, represents the pitch angle variable, τ represents the fast time variable; κ is the amplitude constant related to the radar system parameters, σ p is the effective scattering coefficient at point P, a(·) is the two-dimensional antenna pattern function, rect(·) is the rectangular envelope function, T p =2μs is the time width of the transmission signal, K r is the frequency modulation slope of the transmitted signal, c = 3 × 10 8 m / s is the propagation speed of electromagnetic waves.
[0080] Step 2: Preprocess the echo in the range direction. The specific implementation method is: perform fast Fourier transform on the range direction of equation (22), and multiply it by the pulse compression matching function in the frequency domain to obtain a spatial three-dimensional echo with high range resolution expressed as:
[0081]
[0082] in represents a target point in the airspace Ω, Indicates that it is located is the radar backscatter coefficient of the target at , sinc(·) represents the envelope of the pulse compression response function, and B=15MHz is the bandwidth of the transmission signal.
[0083] Step 3: Construct an azimuth-elevation two-dimensional convolution model. The specific implementation method is as follows: for a fixed range slice, only different azimuth and elevation scattering are considered; the echo data of the range unit is regarded as the convolution of the target scattering coefficient and the two-dimensional antenna pattern in the azimuth and elevation two-dimensional scene; to obtain a more concise expression, assume that κ is a constant of 1 and only consider azimuth and elevation changes, then equation (23) is expressed as:
[0084]
[0085] Considering additive white noise, it can be further expressed as:
[0086]
[0087] Where * represents a two-dimensional convolution operation, is equal to the two-dimensional convolution kernel. The amplitude distribution of the two-dimensional antenna pattern is as follows Figure 4 As shown, the azimuth main lobe width is Δθ a =3°, the elevation main lobe width is Δθ p =3°.
[0088] Step 4: Discretize the echo model; the specific implementation method is: Mathematically speaking, the two-dimensional deconvolution model (25) is a two-dimensional ill-posed inverse problem, which can be simplified to the following matrix multiplication form after discretization:
[0089] Y=AXB T +N (26)
[0090] in, represents the echo data matrix, represents a complex number set, M and N represent the number of sampling points of the echo in the azimuth and elevation directions respectively, and the simulation sets M=256 and N=256; represents the target scattering coefficient matrix, K1 and K2 represent the dimensions of the target scattering matrix in azimuth and elevation directions respectively; represents the noise matrix, T represents the matrix transpose operation, and Respectively represent the azimuth and elevation antenna pattern modulation matrices; where A and B are specifically:
[0091]
[0092]
[0093] in, represents the sampling point of the azimuth antenna pattern, represents the sampling point of the antenna pattern in the elevation direction; L1 and L2 are the number of sampling points of the antenna pattern in the azimuth and elevation directions; the sampling points M, N, K1, K2, L1, and L2 satisfy the relationship between them:
[0094]
[0095] Using the properties of the Kronecker product, model (26) is simplified to:
[0096] y=Fx+n (30)
[0097] Among them, y, F, x, and n satisfy the following relations respectively,
[0098]
[0099] Among them, vec(·) represents vectorization, represents the Kronecker product.
[0100] Step 5: Calculate the target scattering power diagonal matrix; the specific implementation method is: let P be a K1K2×K1K2 dimensional diagonal matrix, whose diagonal elements include the power of each target point of the target scattering coefficient matrix X, then P is:
[0101]
[0102] Where diag(·) represents the diagonalization operation, P k =|x k | 2 , k=1,...,K1K2,x k Represents the kth element of x.
[0103] Step 6: Calculate the autocorrelation matrix R; the specific implementation method is: define the autocorrelation matrix as:
[0104]
[0105] Where H represents the conjugate transpose operation; due to the particularity of the structure of the matrix F, R is a singular matrix, which will lead to R -1 Therefore, the noise interference will be amplified when solving the target backscattering. To ensure accurate angle measurement, R is loaded diagonally:
[0106] R = FPF H +λI (34)
[0107] where I represents the identity matrix and λ represents the regularization parameter that balances noise and resolution.
[0108] Step 7: Calculate the inverse R of the autocorrelation matrix -1 ; The specific implementation method is: According to the cyclic structure of F and the diagonal structure of P, we know that R is a triangular matrix, so R is divided into blocks:
[0109]
[0110] Among them, C, Then calculate
[0111]
[0112] Among them C -1 = inverse(C) directly uses the traditional method to calculate. inverse(C) means solving the inverse matrix of C by the direct inversion method; then calculating S by the traditional method -1 And divide it into blocks:
[0113]
[0114] in
[0115] Then calculate
[0116] W1=C -1 +C -1 GMG H C -1 (38)
[0117] and
[0118] W2=-C -1 G[MN] (39)
[0119] Finally, we get R -1 It is expressed as:
[0120]
[0121] Step 8: Iteratively estimate the target backscatter coefficient; the specific implementation method is: construct a weighted least squares cost function:
[0122]
[0123] Which meets in f k represents the kth column of F; minimizing J gives:
[0124]
[0125] Traverse k = 1, 2, ..., K1K2 to obtain Then, by The two-dimensional target scattering estimate is obtained by rearranging as follows:
[0126]
[0127] In fact, in each direction k All of them can be calculated in parallel, which makes this method more suitable for parallel computing hardware platforms.
[0128] For real aperture radar two-dimensional super-resolution imaging, after 6 iterations, the two-dimensional resolution performance of azimuth and elevation can be significantly improved.
[0129] Figure 5 (a) is the result of traditional real beam imaging, where it can be seen that the four point targets are completely indistinguishable. Figure 5 (b) is the result of two-dimensional deconvolution imaging based on the Bayesian method. Compared with traditional real beam imaging, its resolution is improved. However, this method is very sensitive to noise and has limited improvement on imaging performance. Figure 5 (c) is the result of the FFT-based frequency domain 2D deconvolution slice imaging. It can be seen that this method roughly distinguishes the targets. However, due to the problem of noise amplification, many false targets appear in the image. In addition, this method also requires manual adjustment of the threshold for different signal-to-noise ratio conditions. Figure 5 (d) is the imaging result of the method proposed in the present invention. Obviously, the four point targets are completely distinguished and the side lobes are well suppressed.
[0130] At the same time, in order to more intuitively verify the performance advantages of the proposed method, the azimuth profile results of the method of the present invention are Figure 6 The pitch profile results are given in Figure 7 It can be seen that the real beam echoes overlap in the azimuth and elevation directions, and two targets with the same azimuth or elevation angle cannot be distinguished at all. The method proposed in the present invention has the best resolution performance.
[0131] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.
Claims
1. A method for two-dimensional super-resolution imaging of a deep space probe hovering downward, characterized in that: The following steps are involved: Step 1: Establish the echo expression of the fixed RAR platform: Assume that the target P is located at R p ,θ p , Represent the distance, azimuth, and elevation of the target respectively; the area of interest is Ω, and the scanning radar scans along the azimuth and elevation directions at a scanning speed of ω a and ω p Scan; for a certain point P, the RAR range-azimuth-elevation three-dimensional echo is expressed as: Where θ represents the azimuth variable, represents the pitch angle variable, τ represents the fast time variable; κ is the amplitude constant related to the radar system parameters, σ p is the effective scattering coefficient at point P, a(·) is the two-dimensional antenna pattern function, rect(·) is the rectangular envelope function, T p is the time width of the transmitted signal, K r is the frequency modulation slope of the transmitted signal, c is the propagation speed of the electromagnetic wave; Step 2: Preprocess the echo in the range direction. The specific implementation method is: perform fast Fourier transform on the range direction of equation (1), and multiply it by the pulse compression matching function in the frequency domain to obtain a spatial three-dimensional echo with high range resolution, which is expressed as: in represents a target point in the airspace Ω, Indicates that it is located The radar backscatter coefficient of the target at , sinc(·) represents the envelope of the pulse compression response function, and B is the bandwidth of the transmission signal; Step 3: Construct an azimuth-elevation two-dimensional convolution model. The specific implementation method is as follows: for a fixed range slice, only different azimuth and elevation scattering are considered; the echo data of this range unit is regarded as the convolution of the target scattering coefficient and the two-dimensional antenna pattern in the azimuth and elevation two-dimensional scene; to obtain a more concise expression, assume that κ is a constant of 1 and only consider azimuth and elevation changes, then equation (2) can be expressed as: Considering additive white noise, it can be further expressed as: Where * represents a two-dimensional convolution operation, Equal to the two-dimensional convolution kernel; Step 4: Discretize the echo model; the specific implementation method is: simplify the two-dimensional deconvolution model (4) into the following matrix multiplication form: Y=AXB T +N (5) in, represents the echo data matrix, represents a complex number set, M and N represent the number of sampling points of the echo in the azimuth and elevation directions respectively; represents the target scattering coefficient matrix, K1 and K2 represent the dimensions of the target scattering matrix in azimuth and elevation directions respectively; represents the noise matrix, T represents the matrix transpose operation, and Respectively represent the azimuth and elevation antenna pattern modulation matrices; where A and B are specifically: in, represents the sampling point of the azimuth antenna pattern, represents the sampling point of the antenna pattern in the elevation direction; L1 and L2 are the number of sampling points of the antenna pattern in the azimuth and elevation directions; the sampling points M, N, K1, K2, L1, and L2 satisfy the relationship between them: Using the properties of the Kronecker product, model (5) is simplified to: y=Fx+n (9) Among them, y, F, x, and n satisfy the following relations respectively, Among them, vec(·) represents vectorization, represents the Kronecker product; Step 5: Calculate the target scattering power diagonal matrix; the specific implementation method is: let P be a K1K2×K1K2 dimensional diagonal matrix, whose diagonal elements include the power of each target point of the target scattering coefficient matrix X, then P is: Where diag(·) represents the diagonalization operation, P k =|x k | 2 , k=1,…,K1K2,x k represents the kth element of x; Step 6: Calculate the autocorrelation matrix R; the specific implementation method is: define the autocorrelation matrix as: Where H represents the conjugate transpose operation; R is loaded diagonally: R=FPF H +λI (13) Where I represents the unit matrix, λ represents the regularization parameter that balances noise and resolution; Step 7: Calculate R -1 ; The specific implementation method is: divide R into blocks: in Then calculate Among them C -1 = inverse(C) directly uses the traditional method to calculate. inverse(C) means solving the inverse matrix of C by the direct inversion method; then calculating S by the traditional method -1 And divide it into blocks: in Then calculate W1=C -1 +C -1 WYD H C -1 (17) and W2=-C -1 G[MN] (18) Finally, we get R -1 It is expressed as: Step 8: Iteratively estimate the target backscatter coefficient; the specific implementation method is: construct a weighted least squares cost function: Which meets in f k represents the kth column of F; minimizing J gives: Traverse k = 1, 2, ..., K1K2 to obtain Then, by The two-dimensional target scattering estimate is obtained by rearranging as follows:
Citation Information
Patent Citations
Scanning radar super-resolution imaging method
CN103412305A
Real beam scanning radar super-resolution imaging method
CN103869311A
Deconvolution based radar angle super-resolution imaging method
CN104122549A
Real-beam scanning radar angular super-resolution method under clutter background
CN105137424A
Airborne radar real-time Doppler beam sharpening super-resolution imaging method
CN109613532A
