A coupled space-time baseline optimization method for high orbit differential layer analysis SAR

Abstract

The application discloses a coupled space-time baseline optimization method suitable for high-orbit differential layer analysis SAR, which models the space-time baseline optimization of a GEO SAR as a combination optimization problem, takes minimizing the space-time correlation coefficient and maximizing the unambiguous height as optimization objectives based on the Cramér-Rao lower bound (CRLB) of GEO D-TomoSAR parameter estimation and the baseline characteristics of the GEO SAR, and searches for the space-time baseline data with the best imaging effect in all space-time baseline data under multiple constraint conditions by combining an NSGA-II algorithm.

Description

Technical Field

[0001] This invention belongs to the field of synthetic aperture radar technology, and particularly relates to a method for optimizing coupled spatiotemporal baselines suitable for high-orbit-difference analytical SAR. Background Technology

[0002] Synthetic Aperture Radar (SAR) Differential SAR Tomography (D-TomoSAR), a multi-baseline, multi-temporal SAR interferometry technique, combines the advantages of differential interferometric SAR and tomographic SAR. It can acquire elevation and average deformation velocity information of scatterers, enabling 3D structural reconstruction and deformation monitoring of urban infrastructure. Currently, most data used for differential tomographic imaging is obtained from Low Earth Orbit SAR (LEO SAR). LEO SAR is limited by orbital altitude, resulting in a small coverage area and long observation cycles (several days to over ten days). Creating a SAR image set suitable for D-TomoSAR imaging typically takes about a year. This excessively long time span introduces significant target and atmospheric decorrelation effects, reducing scene correlation and impacting differential tomographic performance. Geosynchronous SAR (GEO SAR) satellites operate at an altitude of approximately 36,000 km. Compared to LEO SAR, the reorbit time is significantly shortened (only 1 day), and the coverage area is wide, enabling effective large-scale, near-continuous observation.

[0003] However, due to differences in operating altitude and the varying perturbation forces affecting GEO SAR and LEO SAR satellite platforms, their spatiotemporal baselines exhibit significant differences. LEO SAR has a smaller spatial baseline span and a random spatiotemporal baseline distribution; while GEO SAR's spatial baseline can reach hundreds of kilometers, and the spatiotemporal baseline relationship shows certain regularity, with a coupling phenomenon in the spatiotemporal baseline distribution. The geometric distribution of the spatiotemporal baselines constrains the imaging accuracy of D-TomoSAR, but existing research on D-TomoSAR mainly focuses on improving imaging algorithms based on low-orbit SAR datasets and introducing complex deformation models, without addressing the issue of spatiotemporal baseline distribution and selection. Furthermore, GEO SAR can acquire sufficient observation data in a short time, but not all of this data is valid for D-TomoSAR imaging, requiring selection of existing data. Summary of the Invention

[0004] To address the aforementioned issues, this invention provides a coupled spatiotemporal baseline optimization method suitable for high-orbit-difference analytical SAR. This method models the spatiotemporal baseline optimization of GEO SAR as a combinatorial optimization problem. Based on the Cramér-Rao lower bound (CRLB) estimated by GEO D-TomoSAR parameters and the characteristics of GEO SAR baselines, the optimization objectives are to minimize the spatiotemporal correlation coefficient and maximize the unambiguous height. Under multiple constraints, the method combines the NSGA-II algorithm to search for the spatiotemporal baseline data that achieves the best imaging effect among all spatiotemporal baseline data.

[0005] The present invention provides a method for optimizing coupled spatiotemporal baselines for high-orbit-difference analytical SAR. It includes:

[0006] Step 1: Acquire all spatiotemporal baseline data observed by the SAR platform. This mainly includes the scene center location P. t The positions of all double orbits on the SAR platform are S = [S1, S2, ..., S...]. N The reorbiting times of all SAR platform orbits are D = [D1, D2, ..., D]. N ], where P t and S i ,i=1,2,…N all represent three-dimensional coordinates (the coordinate system adopts the Earth-fixed system), and N is the total number of observations.

[0007] Let R be the slant range between the SAR platform and the scene at the aperture center time, where R = SP. t Calculate the incident angle Φ at the target location for each observation. The calculation formula is as follows:

[0008]

[0009] Where i = 1, 2, ... N, the symbol · represents the inner product of two vectors, and |||| represents the modulus of a vector.

[0010] Given a signal bandwidth of BW, a light speed of c, and a wavelength of λ, calculate the limiting baseline corresponding to the i-th observation.

[0011]

[0012] Step 2: Set the relevant parameters and initialize the population at the start of the algorithm. The specific steps are as follows:

[0013] Step 2-1, the initial parameter settings of the algorithm mainly include: population size Np, maximum number of iterations MaxGen, and crossover probability P. c Probability of mutation P m commutation probability P e Relevance decision execution flag G, signal-to-noise ratio (SNR), minimum uncorrelated threshold T h The proportion of the largest uncorrelated number Tc The maximum standard deviation threshold T of the high-order estimate s The maximum standard deviation threshold T for deformation rate estimation v The minimum number of observations N required for imaging s The Pareto solution set decision weights are ω1 and ω2.

[0014] Step 2-2: Randomly generate a chromosome of length N using binary encoding, denoted as X = [x1, x2, ..., x...]. N ], where x i x represents the genes on each chromosome. i x has only two possible values: 0 or 1 i =1 indicates that the i-th observation is selected, x i =0 indicates that the i-th observation is not selected. Additionally, each chromosome also represents an individual, generating a total of Np individuals to form the initial population. Each individual represents a possible baseline selection scheme, such as... Figure 3 As shown.

[0015] Step 3: Crossover, mutation, and population merging. Crossover uses a uniform crossover strategy, and mutation follows a basic position mutation strategy. The offspring populations resulting from crossover and mutation are merged with the parent populations to form a new population of size 2Np. To ensure chromosome diversity in the merged population, duplicate individuals are first removed, and then randomly generated chromosomes are added to maintain the population size of 2Np.

[0016] Step 4: Perform a relevance determination every G iterations. This is based on the location of genes with a value of 1 on each chromosome. n represents the number of observations selected according to the selection strategy, satisfying n≤N. The corresponding elements are selected from the double-track positions S to form a new double-track position matrix. The corresponding new heavy orbit time matrix and the corresponding limiting baseline Construct the correlation coefficient matrix Γ for each chromosome under signal-to-noise ratio decorrelation and baseline decorrelation, and use b i,j S represents Ω The spatial vertical baseline formed by the i-th position and the j-th position, where i,j=1,2,…,n. express The i-th limiting baseline value. γ i，j Let the baseline correlation between the i-th element and the j-th element in Ω be represented by the following expression:

[0017]

[0018] When the signal-to-noise ratio correlation is approximately ρ = (1 + SNR) -1 ) -1The correlation coefficient matrix formed is

[0019]

[0020] Use n i This indicates that the element in the i-th row of the correlation coefficient matrix Γ is less than T. h The number of elements can be represented as n. i =∑ j [P(Γ ij <T h ], where j = 1, 2, ..., n, P(Γ ij <T h () indicates a conditional judgment, judging the element Γ ij Is it less than T? h If yes, then the value is 1; otherwise, it is 0. i =n i / n represents the proportion of elements in the i-th row that do not meet the requirements, which must satisfy r. i <T c ,.like When the observation corresponding to the i-th element in Ω contributes little to the entire imaging process, the gene on the chromosome corresponding to the i-th element in Ω is changed from 1 to 0.

[0021] Step 5, function value calculation and selection of the next generation population, the specific steps are as follows:

[0022] Step 5-1, according to S Ω and D Ω This yields a spatial baseline matrix B of size 1×n. Ω and time baseline matrix T Ω Calculate the spatiotemporal correlation coefficient μ for each individual. b,t Maximum unambiguous height H max The standard deviation σ of the high-order estimate s The standard deviation σ of the deformation rate estimate v And the selected number of observations, n. The calculation formulas for each parameter are as follows:

[0023]

[0024]

[0025]

[0026]

[0027] Where, Δb Ω For B Ω The average vertical baseline interval is denoted by , where r is the distance between the reference observation satellite position and the scene for each individual. If the first selected observation is used as the reference, then... Λ s and Λ v It is the spatial frequency and time frequency matrix, both of which are diagonal matrices, as shown in equation (9). The m-th element on the diagonal is... W is the coefficient factor.

[0028]

[0029] Step 5-2, Initial Screening. Based on the standard deviation σ of the height estimate for each individual obtained in Step 5-1. s The standard deviation σ of the deformation rate estimate v And the selected observation number n, to perform preliminary screening of all individuals in the population. If an individual in the population satisfies σ s >T s or σ v >T v or n <N s Remove it from the current population.

[0030] Step 5-3: Fast non-dominated sorting, calculate crowding distance, and select the next offspring. Based on the NSGA-II concept, this aims to minimize the spatiotemporal correlation coefficient μ. b,t Maximize the maximum unambiguous height H max With the goal of obtaining the frontier numbers of all individuals in the population based on the calculated values ​​in step 5-1, the crowding distance of individuals in the population is calculated, and an elite strategy is adopted to generate the next generation of parents based on the frontier level and crowding distance of the individuals.

[0031] Step 6: Determine if the termination condition is met. Check if the current iteration count is greater than the maximum iteration count. If not, reset the iteration count Gen = Gen + 1 and return to step 3. If so, perform a fast non-dominated sort on the last obtained parent, select individuals at level 1, and form the Pareto solution set.

[0032] Step 7: Obtain the preferred solution. After solving the multi-objective problem, a Pareto solution set is obtained. The spatiotemporal correlation coefficient and maximum unambiguous height are calculated for each solution in the Pareto solution set to form the objective function matrix. Sort the two objective function values ​​in order and record the position [pq] of the median of each function in matrix V. If p = q, then the p-th solution in the Pareto solution set is the final selected solution. If p ≠ q, then calculate the ρ value from the p-th solution to the q-th solution in the Pareto solution set according to equation (10), and take the solution with the smallest ρ value as the final selected solution.

[0033]

[0034] in, and Let be the elevation and deformation rate resolutions corresponding to the i-th solution in the Pareto solution set, respectively. B i and T i These are the total span of the spatial baseline and the total span of the temporal baseline corresponding to the i-th solution. and It represents the maximum value of each resolution within the solution set.

[0035] This completes all the steps.

[0036] The beneficial effects of this invention are as follows:

[0037] This invention optimizes spatiotemporal baseline data from all GEO SAR observations, selecting a set of spatiotemporal baseline distributions that enhance GEOD-TomoSAR 3D reconstruction and deformation inversion performance. This effectively reduces the spatiotemporal baseline coupling in imaging and decreases computational complexity. Attached Figure Description

[0038] Figure 1 The flowchart of the spatiotemporal baseline optimization algorithm based on the NSGA-II algorithm;

[0039] Figure 2 For D-TomoSAR imaging geometry;

[0040] Figure 3 A schematic diagram of the binary encoding of the population;

[0041] Figure 4 This is a simulation scene diagram of GEO D-TomoSAR;

[0042] Figure 5 Spatiotemporal baseline distribution maps before and after baseline optimization;

[0043] Figure 6 Preferred baseline aberration tomography imaging results;

[0044] Figure 7 The baseline is optimized by differential tomography imaging results. Detailed Implementation

[0045] The invention will now be described in detail with reference to the accompanying drawings.

[0046] The flowchart of this invention is as follows Figure 1 As shown, achieving optimal spatiotemporal baseline selection for GEO D-TomoSAR requires GEO satellites to conduct multiple observations of the same target, such as... Figure 2 As shown. The specific steps of this invention are as follows:

[0047] Step 1: Acquire all spatiotemporal baseline data observed by the SAR platform. This mainly includes the scene center location P. t The positions of all double orbits on the SAR platform are S = [S1, S2, ..., S...]. N The reorbiting times of all SAR platform orbits are D = [D1, D2, ..., D]. N ], where P t and S i ,i=1,2,…N all represent three-dimensional coordinates (the coordinate system is the Earth-Fixed System), and N is the total number of observations.

[0048] Let R be the slant range between the SAR platform and the scene at the aperture center time, where R = SP. t Calculate the incident angle Φ at the target location for each observation. The calculation formula is as follows:

[0049]

[0050] Where i = 1, 2, ... N, the symbol · represents the inner product of two vectors, and |||| represents the modulus of a vector.

[0051] Given a signal bandwidth of BW, a light speed of c, and a wavelength of λ, calculate the limiting baseline corresponding to the i-th observation.

[0052]

[0053] Step 2: Set the relevant parameters and initialize the population at the start of the algorithm. The specific steps are as follows:

[0054] Step 2-1, the initial parameter settings of the algorithm mainly include: population size Np, maximum number of iterations MaxGen, and crossover probability P. c Probability of mutation P m commutation probability P e Relevance decision execution flag G, signal-to-noise ratio (SNR), minimum uncorrelated threshold T h The proportion of the largest uncorrelated number T c The maximum standard deviation threshold T of the high-order estimate s The maximum standard deviation threshold T for deformation rate estimation v The minimum number of observations N required for imaging s Pareto solution set decision weights ω1 and ω2, initial iteration number Gen = 0.

[0055] Step 2-2, Population Initialization. Generate chromosomes of length N randomly using binary encoding, denoted as X = [x1, x2, ..., x...]. N ], where x i x represents the genes on each chromosome. i x has only two possible values: 0 or 1 i =1 indicates that the i-th observation is selected, xi =0 indicates that the i-th observation is not selected. Additionally, each chromosome also represents an individual; the initial population generates Np individuals, each representing a possible baseline selection scheme, such as... Figure 3 As shown.

[0056] Step 3: Crossover, Mutation, and Population Merging. Parents within the population generate new chromosomes through crossover and mutation. The crossover operation employs a uniform crossover strategy, based on the crossover probability P. c Select individuals for the crossover operation. For each individual's chromosome, according to the chromosomal gene order, sequentially extract genes from corresponding positions on the two chromosomes for crossover. Each gene exchange requires comparing the randomly generated probability with the given exchange probability. The mutation operation is based on the fundamental position mutation concept, according to the mutation probability P. m Chromosomes are randomly selected, and the number of mutated genes is determined by a certain ratio based on the chromosome length. Mutation locations are randomly selected, and the binary values ​​of these locations are inverted. The offspring population resulting from crossover mutation is merged with the parent population to form a new population of size 2Np. To ensure chromosome diversity in the merged population, duplicate individuals are first removed, and then chromosomes are randomly generated to maintain the population size of 2Np.

[0057] Step 4: Perform a relevance determination every G iterations. This is based on the location of genes with a value of 1 on each chromosome. Select corresponding elements from the heavy rail positions S to form a new heavy rail position matrix. The corresponding new heavy orbit time matrix and the corresponding limiting baseline Based on this, considering both signal-to-noise ratio correlation and baseline correlation, a corresponding correlation coefficient matrix is ​​constructed. Using b... i,j S represents Ω The spatial vertical baseline formed by the i-th position and the j-th position, where i,j=1,2,…,n. express The i-th limiting baseline value. γ i,j Let the baseline correlation between the i-th element and the j-th element in Ω be represented by the following expression:

[0058]

[0059] When the signal-to-noise ratio correlation is approximately ρ = (1 + SNR) -1 ) -1 The correlation coefficient matrix formed is

[0060]

[0061] Use n iThis indicates that the element in the i-th row of the correlation coefficient matrix Γ is less than T. h The number of elements can be represented as n. i =∑ j [P(Γ ij <T h ], where P(Γ ij <T h () indicates a conditional judgment, judging the element Γ ij Is it less than T? h If yes, then the value is 1; otherwise, it is 0. i =n i / n represents the proportion of elements in the i-th row that do not meet the requirements, which must satisfy r. i <T c If r i >T c When the value of 1 is less than 0, it means that the observation corresponding to the i-th element in Ω contributes little to the entire imaging process. In this case, the gene on the chromosome corresponding to the i-th element in Ω is changed from 1 to 0.

[0062] Step 5, function value calculation and selection of the next generation population, the specific steps are as follows:

[0063] Step 5-1, according to S Ω and D Ω This yields a spatial baseline matrix B of size 1×n. Ω and time baseline matrix T Ω Calculate the spatiotemporal correlation coefficient μ for each individual. b,t Maximum unambiguous height H max The standard deviation σ of the high-order estimate s The standard deviation σ of the deformation rate estimate v The selected number of observations n, and the calculation formulas for each parameter are as follows:

[0064]

[0065]

[0066]

[0067]

[0068] Where, Δb Ω For B Ω The average vertical baseline interval is denoted by , where r is the distance between the reference observation satellite position and the scene for each individual. If the first selected observation is used as the reference, then... Equations (17) and (18) are derived from CRLB based on parameter estimation, and the specific principles are as follows.

[0069] Assume that the number of SAR images used for SAR differential analysis is N. It is the k-th pixel in the SAR image set. The corresponding probability density function is

[0070]

[0071] For a single-point target, construct the probability density function for vector parameters θ = [s, v] according to equation (19). T The likelihood function is obtained.

[0072]

[0073] The covariance C can be decomposed into the product of the magnitude and phase matrices, expressed as follows:

[0074] C = Φ * ΓΦ (21)

[0075]

[0076]

[0077] According to the above formula, it is obvious that |Φ * ΓΦ|=|Φ * |×|Γ|×|Φ|=|Φ * |×|Φ|×|Γ|=|Γ| is independent of s,v, (Φ * ΓΦ) -1 =Φ -1 Γ -1 (Φ * ) -1 =Φ * Γ -1 Φ. The CRLB of the vector parameter θ can be obtained by the inverse of the Fisher information matrix (FIM).

[0078]

[0079] in,

[0080] Based on the properties of matrix partial derivatives and traces, the FIM is obtained as follows:

[0081]

[0082] Among them, Λ s and Λ v These are the spatial frequency and time frequency matrices, both of which are diagonal matrices, as shown in the equation. Each element on the diagonal...

[0083]

[0084] Inverting equation (25) yields

[0085]

[0086] Among them, coefficient tr(×) represents finding the trace.

[0087] Therefore, after removing the influence of relevant factors, the variance of the elevation and mean deformation rate estimates is:

[0088]

[0089] Step 5-2, Initial Screening. Based on the standard deviation σ of the height estimate for each individual obtained in step 5-1. s The standard deviation σ of the deformation rate estimate v And the selected observation number n, to perform preliminary screening of all individuals in the population. If an individual in the population satisfies σ s >T s or σ v >T v or n <N s Remove the individual from the population, where N s It is determined by the compressed sensing constrained isometric property (RIP).

[0090] Step 5-3: Fast non-dominated sorting, calculate crowding distance, and select the next offspring. Based on the NSGA-II concept, this aims to minimize the spatiotemporal correlation coefficient μ. b,t Maximize the maximum unambiguous height H max To achieve the objective, based on the calculated values ​​from step 5-1, a fast non-dominated sort is performed on all individuals in the population to obtain the frontier numbers of all individuals. The crowding distance of each individual is then calculated. Based on the frontier level and crowding distance of each individual, an elitist strategy is used to generate the next generation's parent. The principle is explained in the reference: K. Deb, A. Pratap, S. Agarwal, and T. Meyarivan, "Afast and elitist multiobjective genetic algorithm: NSGA-II," IEEE Transactions on Evolutionary Computation, vol. 6, no. 2, pp. 182-197, 2002.

[0091] Step 6: Determine if the termination condition is met. Check if the current iteration count is greater than the maximum iteration count. If not, the iteration count Gen = Gen + 1, and return to step 3. If it is met, perform a fast non-dominated sort on the last obtained parent, select individuals at level 1, and form the Pareto solution set.

[0092] Step 7: Obtain the preferred solution. After solving the multi-objective problem, a Pareto solution set is obtained. The spatiotemporal correlation coefficient and maximum unambiguous height are calculated for each solution in the Pareto solution set to form the objective function matrix. Sort the two objective function values ​​in order and record the position [pq] of the median of each function in matrix V. If p = q, then the p-th solution in the Pareto solution set is the final selected solution. If p ≠ q, then calculate the ρ value from the p-th solution to the q-th solution in the Pareto solution set according to equation (10), and take the solution with the smallest ρ value as the final selected solution.

[0093]

[0094] in, and Let be the elevation and deformation rate resolutions corresponding to the i-th solution in the Pareto solution set, respectively. B i and T i These are the total span of the spatial baseline and the total span of the temporal baseline corresponding to the i-th solution. and This represents the maximum resolution of each element within the solution set.

[0095] The following provides an implementation example with specific parameters.

[0096] In this example, we designed the GEO SAR satellite orbit elements as shown in Table 1, and the imaging scene is as follows. Figure 4 The ground plane and the radar-facing facade and roof of the high-rise building are shown. Assuming that the satellite observes this scene once every day, a total of 60 orbits of observation data are obtained.

[0097] Table 1

[0098] orbital parameters numerical values Semi-long wheelbase (km) 42163.9 Eccentricity 0 Track inclination angle (°) 16 Perigee depression angle (°) 0 Longitude of the ascending node (°) 90 Wavelength (m) 0.03 Bandwidth (MHz) 300

[0099] First, following step 1, obtain the satellite coordinates and velocity, and the scene coordinates (by setting the orbital elements in the Satellite Tool Kit software, determining the satellite's double orbit and the scene position, and then reading them through MATLAB); then calculate the slant distance between the satellite and the scene, the incident angle at the scene, and the limiting baseline length for each observation.

[0100] Execute step 2 to set the relevant parameters at the start of the algorithm, as shown in Table 2. After population initialization, the population is represented as a 100×60 0-1 matrix.

[0101] Table 2

[0102] Parameter name numerical values Population size Np 100 Maximum number of iterations (MaxGen) 5000 <![CDATA[Cross probability P c > 0.8 <![CDATA[Mutation probability P m > 0.04 <![CDATA[Gene exchange probability P e > 0.7 <![CDATA[Minimum uncorrelated threshold T h > 0.15 Signal-to-noise ratio (SNR) (dB) 10 <![CDATA[Maximum uncorrelated number ratio T c > 0.7 <![CDATA[Maximum standard deviation threshold T for height estimation s > 0.5 <![CDATA[Maximum standard deviation threshold T for strain rate estimation v > <![CDATA[10 -4 ]]> <![CDATA[The minimum number of observations N required for imaging s > 22 <![CDATA[Decision-making weight ω1]]> 0.8 <![CDATA[Decision-making weight ω2]]> 0.5

[0103] In step 3, each individual is represented by only one chromosome, with a chromosome length equal to the total number of observed data (60). Crossover, mutation, and merging of parent and offspring populations are performed according to the probabilities shown in Table 2, resulting in a new population of 200 individuals.

[0104] In step 4, to ensure chromosome diversity, a correlation determination is performed every 10 iterations. The spatiotemporal correlation coefficient matrix Γ for each chromosome under each observation is constructed. The proportion of the number of unrelated elements corresponding to each gene of each chromosome to the total number of selected genes is calculated. Based on the constraints, the gene values ​​of each chromosome are corrected.

[0105] Step 5 is executed, and the function value calculation results are stored as a 200×5 function value matrix. Each column represents the spatiotemporal correlation coefficient, maximum unambiguous height, standard deviation of height estimation, standard deviation of deformation rate estimation, and total number of selected observations. Individuals in the population whose standard deviations of height estimation, deformation rate estimation, and total number of selected observations do not meet the accuracy requirements are removed. All individuals are classified into levels, and their respective crowding distances are calculated. Finally, individuals with smaller levels and larger crowding distances within the same level are selected to form a new generation parent population of 100 individuals.

[0106] Perform step 6 repeatedly to obtain the final Pareto solution set.

[0107] Perform step 7 to make a final decision from the Pareto solution set and select the optimal selection scheme for the individual, such as... Figure 5 As shown.

[0108] Of course, the present invention may have other various embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and modifications according to the present invention, but these corresponding changes and modifications should all fall within the protection scope of the appended claims.

Claims

1. A coupled space-time baseline optimization method suitable for high orbit differential layering SAR, characterized in that, The method comprises the following steps: Step 1, obtaining all the spatio-temporal baseline data observed by a SAR platform; Step 2, setting relevant parameters at the beginning of the algorithm and initializing a population; Step 3, crossover, mutation and population merging; Step 4, performing a correlation decision once every G iterations; Step 5, function value calculation and new generation population selection: according to the NSGA-II idea, the time-space correlation coefficient is minimized , and the maximum unambiguous height is maximized as the goal, according to the function value, all individuals in the population are quickly non-dominated sorted, the front surface number of all individuals in the population is obtained, the crowding distance of the individuals in the population is calculated, and the next generation of parents is generated by using the elite strategy according to the level of the front surface where the individual is located and the crowding distance. Step 6, judging whether a termination condition is met; Step 7, obtaining a preferred solution.

2. A coupled space-time baseline preferred method for high orbit differential tomographic SAR as claimed in claim 1, characterized in that The data acquired in step 1 includes the scene center position All heavy rail positions of the SAR platform All heavy rail times of the SAR platform Where And Both represent three-dimensional coordinates, with the coordinate system being the Earth-fixed system, Is the total number of observations.

3. A coupled space-time baseline optimization method for high orbit differential layering SAR as claimed in claim 1, characterized in that The step 1, record scene center position , SAR platform all heavy rail position , Aperture center time SAR platform and scene between the oblique distance is , then ; Calculate the angle of incidence at the target for each observation The formula is: ； wherein , the symbol denotes the inner product of two vectors, denotes the norm of a vector; With a signal bandwidth of The speed of light is , wavelength is Calculate the first The limiting baseline corresponding to this observation is: 。 4. A coupled space-time baseline preferred method for high orbit differential tomographic SAR as claimed in claim 1, characterized in that In step 2, the initial parameter settings of the algorithm include: population size , maximum iteration number , crossover probability , mutation probability , exchange probability , correlation decision execution flag G, signal-to-noise ratio , minimum uncorrelated threshold , maximum uncorrelated number ratio , maximum standard deviation threshold of height estimation , maximum standard deviation threshold of deformation rate estimation , minimum number of observations required for imaging , decision weight of Pareto solution set , and .

5. A coupled space-time baseline optimization method for high orbit differential layering SAR as claimed in claim 1 characterized by The step 2 is to randomly generate chromosomes with length of in binary encoding, denoted as where represents the genes on each chromosome, only has two values of 0 or 1, represents the selection of the th observation, represents the non-selection of the th observation.

6. A coupled space-time baseline preferred method for high orbit differential tomographic SAR as claimed in claim 1, characterized in that The step 4, from the heavy rail position The corresponding elements are selected to form a new heavy rail position matrix , the corresponding new heavy rail time matrix , and the corresponding limit baseline ; construct the correlation coefficient matrix corresponding to each chromosome under the conditions of signal-to-noise ratio decorrelation and baseline decorrelation , where represents the spatial vertical baseline formed by the th position and the th position in , where , represents the th limit baseline value in ; represents the baseline correlation under observation corresponding to the th element and the th element in . ； In the signal-to-noise ratio correlation is approximately , the correlation coefficient matrix is: ； Use to represent the correlation coefficient matrix The number of elements in the row that are less than can be represented as where , is the minimum uncorrelation threshold, represents a conditional judgment, judging whether the element is less than , if yes, then 1, otherwise 0; is the proportion of the number of elements in the row that do not meet the requirements, which needs to satisfy where is the maximum uncorrelation number proportion; if , it means that the th element in the row has a smaller contribution to the entire imaging process, and the gene on the chromosome corresponding to the th element in the row is changed from 1 to 0.

7. A coupled space-time baseline preferred method for high orbit differential tomographic SAR as claimed in claim 1, characterized in that In step 5, the method of preliminary screening is: if an individual in the population satisfies or or it is removed from the current population, wherein is the standard deviation of the height estimate, is the standard deviation of the rate of change estimate, is the number of observations selected, is the maximum standard deviation threshold for the height estimate, is the maximum standard deviation threshold for the rate of change estimate, is the minimum number of observations required for imaging.

8. A coupled space-time baseline optimization method for high orbit differential layering SAR as claimed in claim 1 characterized by In step 7, the spatiotemporal correlation coefficient and maximum unambiguous height for each scheme in the Pareto solution set are calculated to form the objective function matrix. Sort the two objective function values ​​in order and record the median of each function in the matrix. The position in the middle ,if Then it corresponds to the Pareto solution set of the th The solution is the final selected scheme; if Then, according to the calculation of the Pareto solution set from the th trial, The solution to the first A solution Value, take The solution corresponding to the minimum value is the final selected scheme: ； wherein, and are the elevation and deformation rate resolution of the th solution in the Pareto solution set, respectively, , , and are the total spatial baseline span and total temporal baseline span corresponding to the th solution, respectively, and are the maximum values of each resolution in the solution set.

Images