A spaceborne InSAR calibrator deployment method based on genetic algorithm and penalty mechanism

Abstract

This invention discloses a method for deploying spaceborne InSAR calibrators based on genetic algorithms and penalty mechanisms. Specifically, the method involves: first, acquiring SAR single-look complex image pairs and performing interferometric processing; then, defining the coordinate matrix of each calibrator to be deployed in the pixel coordinate system as decision variables, randomly selecting decision variables within the study area, calculating the scoring functions of various optimization variables, and adding constraints to form a comprehensive penalty function. A weighted convex combination is then used to form a comprehensive fitness function; a genetic algorithm is used for global optimization, decoding the optimal individual into a list of calibrator pixel coordinates, simultaneously calculating the scores of various sub-objectives and the constraint satisfaction status, generating a deployment map and an index report; finally, calibrators are deployed in the study area, and subsequent baseline calibration and elevation inversion are completed to obtain a high-precision DEM product. This invention can provide an optimal deployment scheme for a specified number of calibrators while ensuring compliance with engineering application requirements.

Description

Technical Field

[0001] This invention belongs to the field of Interferometric Synthetic Aperture Radar (InSAR) data processing, specifically involving a method for deploying spaceborne InSAR calibrators based on genetic algorithms and penalty mechanisms. Background Technology

[0002] Spaceborne InSAR possesses all-weather, all-day Earth observation capabilities. Based on this capability, repeated orbit or multi-baseline pattern complex image pairs can be obtained through interferometric processing to achieve large-scale, high-precision Digital Elevation Model (DEM) inversion. The core of this approach lies in using interferometric phase analysis to extract the slant range difference of ground targets, and then calculating elevation information based on the spatial geometric relationship between the primary and secondary antennas and the ground targets.

[0003] Among these factors, the accuracy of the baseline parameters directly determines the accuracy of the interferometric phase-to-elevation conversion. Any baseline estimation error will lead to a systematic elevation estimation bias. Taking a typical spaceborne InSAR system as an example, a 1% relative error in the baseline length may cause elevation accuracy errors of several meters to tens of meters. Therefore, high-precision baseline calibration is the core of high-precision topographic mapping tasks.

[0004] Currently, using ground-based calibrators to calibrate spaceborne InSAR baselines is an effective means of improving elevation measurement accuracy. Based on different research approaches, calibrator deployment schemes can be divided into two main categories: numerical analysis methods based on sensitivity matrix optimization and qualitative design based on spatial geometric configuration optimization.

[0005] The first type of method takes minimizing the condition number of the sensitivity matrix as the optimization objective and focuses on the mathematical suppression of matrix ill-conditioning. It reduces the sensitivity of baseline solution to observation noise by optimizing the scaler distribution. However, this type of method has difficulty in establishing a physical mapping relationship between the scaler spatial configuration and specific error sources, and it does not adequately consider the observation geometric changes caused by terrain undulations, which limits the applicability of the optimized structure in complex terrain areas.

[0006] The second type of method is based on spatial intersection geometry theory and focuses on qualitative design criteria such as increasing the intersection angle and optimizing spatial distribution. It emphasizes that calibrators should be distributed to form a robust geometric configuration. However, this type of method lacks a systematic model of the quantitative relationship between deployment density, geometric configuration and baseline error correction effectiveness. It is difficult to generalize and optimize the deployment scheme under different terrain conditions and observation geometry, and often relies on expert experience for manual adjustment.

[0007] The two types of research mentioned above have relatively singular focuses. The former focuses on numerical stability while ignoring physical mechanisms, while the latter focuses on geometric intuition but lacks quantitative modeling. Neither of them explores the problem of optimal resource allocation under the constraint of high scaling costs.

[0008] As manually deployed ground control points, calibrators are costly to manufacture, install, and maintain over long periods. Large-scale, dense deployment faces the dual pressures of cost and maintenance. Achieving an effective balance between baseline correction accuracy and deployment cost within a limited number of calibrators, ensuring that calibrator deployment maximizes baseline estimation accuracy while also considering the economic feasibility of engineering implementation, has become a critical issue that urgently needs to be addressed.

[0009] Therefore, there is a need to develop a globally optimal calibrator placement optimization method to achieve high-precision baseline calibration under cost constraints. Summary of the Invention

[0010] To address the aforementioned problems, this invention provides a spaceborne InSAR calibrator deployment method based on genetic algorithms and penalty mechanisms. It employs a genetic algorithm as a global optimization framework and combines it with a multi-objective weighted comprehensive evaluation method to achieve optimal calibrator deployment.

[0011] The specific steps are as follows:

[0012] Step 1: Acquire SAR single-look complex image pairs and perform interferometric processing to obtain the spatial coherence coefficient, line-of-sight geometric parameters, and inverted elevation data for each pixel;

[0013] The gaze geometry parameters include the gaze vector corresponding to each pixel. Primary star slant distance and angle of incidence wait;

[0014] coordinate Spatial coherence coefficient corresponding to the pixel The formula is:

[0015]

[0016] in , To match the complex values ​​in the corresponding windows of the registered master and slave images. Indicates to Take its conjugate.

[0017] The inverted elevation data is generated from the three-dimensional coordinates of the ground point corresponding to each pixel, thus creating a numerical elevation model of the study area. ;

[0018] Step 2: Construct a comprehensive fitness function using the spatial coherence coefficient of each pixel, line-of-sight geometry parameters, and inverted elevation data. ;

[0019] The specific steps include:

[0020] Step 2.1: For the area to be deployed Each scaler is defined as a decision variable, and the coordinate matrix of each scaler in the pixel coordinate system is defined as the following:

[0021]

[0022] Among them, the first calibrator The column coordinates and row coordinates are respectively .

[0023] Step 2.2: Randomly select within the study area Each pixel position forms a set of decision variables. Calculate the scoring function for each optimization variable;

[0024] (1) Calculate the overall terrain suitability score function using the mean score of each point within the group. :

[0025]

[0026] in For pixels Using spatial coherence coefficient and elevation gradient information Defined terrain suitability score; calculation formula is:

[0027]

[0028] in This represents the terrain roughness tolerance threshold.

[0029] (2) Calculation of intra-group based on InSAR imaging geometric model The sensitivity matrix formed by pixels And by using its condition number to construct a continuous piecewise function, the condition number scoring function is obtained. :

[0030]

[0031] in Sensitivity matrix condition number.

[0032] (3) For any two pixels in a group, construct a uniformity scoring function by using the coefficient of variation to measure the uniformity of the point set distribution. :

[0033]

[0034] in The minimum spacing is set empirically based on the size of the study area. , The Euclidean distance between any two pixels within the group. This is the minimum spacing constraint threshold set based on the size of the study area and experience. For all Euclidean distances within the group The coefficient of variation is calculated as the ratio of the mean to the standard deviation.

[0035] (4) Taking into account the visibility and coherence of the primary and secondary binary stars of all pixels in the group, as well as the distance factor, calculate the visibility scoring function. :

[0036]

[0037] in For the first in the group The three-dimensional coordinate information corresponding to each pixel. , The first The orbital positions of the primary and secondary satellites corresponding to each pixel. For line-of-sight geometric visibility, For the first Spatial coherence coefficient of each pixel This is the slant distance factor.

[0038] Step 2.3: Formulate a comprehensive penalty function for each optimization variable and add constraints;

[0039] The comprehensive penalty function is:

[0040]

[0041] in , and The penalty weight coefficient is set according to the importance of each constraint, with an upper bound of 0.5 to ensure that the fitness is non-negative.

[0042] The specific constraints are as follows:

[0043] 1) To ensure the calibrator is within the effective monitoring area, a penalty function is defined regarding the image boundary constraints:

[0044]

[0045] in pixels within a group Euclidean distance to the boundary of the monitoring area The boundary safety distance threshold is set based on experience.

[0046] 2) To avoid interference between scalers, a penalty function is defined for the minimum spacing constraint:

[0047]

[0048] 3) For areas unsuitable for deployment due to low coherence or steep terrain, define a penalty function for physical performance threshold constraints:

[0049]

[0050] in For the set coherence threshold, For pixels within a group The corresponding elevation gradient value, This represents the maximum permissible slope.

[0051] Step 2.4: Combine the scoring functions of each optimization variable with the comprehensive penalty function through a weighted convex combination to form the comprehensive fitness function;

[0052]

[0053] Where the weight coefficients satisfy ;

[0054] Step 3: Use a genetic algorithm to globally optimize the comprehensive fitness function, decode the optimal individual into a list of calibrator pixel coordinates, calculate the scores of each sub-objective and the constraint satisfaction, and generate a deployment map and index report.

[0055] Step 4: According to the pixel coordinate list of the calibrator, complete the placement of the calibrator in the study area, and complete the subsequent baseline calibration and elevation inversion to obtain a high-precision DEM product.

[0056] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0057] (1) This invention takes into account factors such as terrain, geometry, distribution and visibility, and through multi-objective weighted fusion and the introduction of static penalty function to handle actual constraints, it can provide an optimal deployment scheme for a specified number of calibrators while ensuring that the requirements of engineering applications are met.

[0058] (2) The present invention uses a genetic algorithm for global optimization to avoid getting trapped in local optima, and is suitable for the deployment planning of calibrators in large-scale and complex terrain areas. Attached Figure Description

[0059] Figure 1 This is a flowchart of the spaceborne InSAR calibrator deployment method based on genetic algorithm and penalty mechanism of the present invention;

[0060] Figure 2 This is a Ku-band inversion elevation map generated through simulation in this embodiment of the invention;

[0061] Figure 3 This is the fitness evolution curve of the genetic algorithm in this embodiment of the invention;

[0062] Figure 4 This is a demonstration of the final deployment scheme of the calibrator on the DEM topographic map in this embodiment of the invention;

[0063] Figure 5 This is the inversion elevation map obtained after calibrating the baseline parameters in this embodiment of the invention. Detailed Implementation

[0064] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. All equivalent substitutions, improvements, and modifications made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0065] This invention provides a spaceborne InSAR calibrator deployment method based on genetic algorithms and penalty mechanisms. By weighting, it integrates multiple mutually restrictive factors such as terrain suitability, geometric configuration stability, spatial distribution uniformity, and satellite visibility into a unified fitness function. It leverages the advantages of genetic algorithms in performing efficient global optimization in a complex nonlinear and nonconvex solution space while avoiding getting trapped in local optima. This method can provide an optimal deployment scheme for a specified number of calibrators under the premise of multi-index optimization.

[0066] like Figure 1 As shown, the InSAR data is first processed to obtain basic parameters such as geometric parameters and coherence, as well as inverted elevation data. Then, four optimization parameters are calculated: terrain suitability, ill-conditioned sensitivity equation, deployment uniformity, and satellite visibility. Next, penalty function terms such as boundary constraints, spacing constraints, and physical constraints are introduced to construct a comprehensive fitness function as the objective function. Finally, a genetic algorithm is used for global search to obtain the optimal deployment scheme with a specified number of calibrators. The specific steps are as follows:

[0067] Step 1: Acquire SAR single-look complex image pairs and perform interferometric processing to obtain the spatial coherence coefficient, line-of-sight geometric parameters, and inverted elevation data for each pixel;

[0068] Geometric parameter calculation: Based on orbital data and SAR imaging geometry, calculate the line-of-sight vector corresponding to each pixel. Primary star slant distance and angle of incidence wait;

[0069] Based on the registered master-slave composite data, the following is adopted: Sliding window calculation of spatial coherence coefficient; coordinates Spatial coherence coefficient corresponding to the pixel Also known as coherence, the formula is:

[0070]

[0071] in , To match the complex values ​​in the corresponding windows of the registered master and slave images. Indicates to Take its conjugate.

[0072] The inverted elevation data is generated from the three-dimensional coordinates of the ground point corresponding to each pixel, thus creating a numerical elevation model of the study area. The specific steps include:

[0073] 1. Based on the unwrapped phase, combined with the slant distance of the primary star Angle of incidence Using line-of-sight geometric parameters and satellite orbit data such as the position and velocity of the main and auxiliary antennas, a set of nonlinear equations describing the relationship between the three-dimensional coordinates of the target point and the observed values ​​is constructed.

[0074] This set of equations mainly includes the interference phase equation, the slant range equation, and the Doppler equation;

[0075] 2. The above nonlinear equations are solved using the Newton-Raphson iteration method. The initial value of the iteration is assumed to be the reference terrain or flat land. The three-dimensional coordinates of the target point are gradually corrected through iteration until the preset accuracy requirements are met.

[0076] 3. Repeat the above solution process for all pixels to obtain the three-dimensional coordinates of the ground point corresponding to each pixel, and then generate a numerical elevation model of the study area. .

[0077] Step 2: Construct a comprehensive fitness function using the spatial coherence coefficient of each pixel, line-of-sight geometry parameters, and inverted elevation data. ;

[0078] The specific steps include:

[0079] Step 2.1: For the area to be deployed Each scaler is defined as a decision variable, and the coordinate matrix of each scaler in the pixel coordinate system is defined as the following:

[0080]

[0081] Among them, the first calibrator The column coordinates and row coordinates are respectively The range of values ​​is limited by the image size.

[0082] Step 2.2: Randomly select within the study area Each pixel position forms a set of decision variables. Calculate the scoring function for each optimization variable;

[0083] 1) Using coherence coefficient parameters and elevation gradient map Define the terrain suitability score:

[0084]

[0085] in This is the terrain roughness tolerance threshold (usually set to 10~30m / km), used to control the decay rate of gradient effects.

[0086] For a set of calibrators, the overall terrain suitability score function is calculated using the mean score of each point within the set. :

[0087]

[0088] in For pixels Using spatial coherence coefficient and elevation gradient information Define the terrain suitability score for each pixel;

[0089] 2) For each scaler Its sensitivity vector The model can be constructed primarily considering the impact of the baseline on imaging accuracy. The simplified model used in this invention is as follows:

[0090]

[0091] in This is a functional relationship obtained by modifying the auxiliary radar range equation according to the trigonometric cosine theorem under the InSAR imaging geometric model, and then considering the three-dimensional components of the baseline vector in the geocentric surface coordinate system. , , Find the partial derivatives of 3D matrix, within groups The sensitivity matrix G formed by the pixels is further calculated, and its condition number is then calculated. :

[0092]

[0093] in , These are the maximum and minimum singular values, respectively. The ill-conditioning of the geometry was quantified, and the baseline solution became extremely sensitive to observation noise as its value approached infinity.

[0094] To transform the minimization problem into fitness maximization, a continuous piecewise function is constructed using its condition number, resulting in the condition number scoring function. :

[0095]

[0096] (3) The coefficient of variation is used to measure the uniformity of the point set distribution. Let the pixel point set be... The Euclidean distance matrix is The Euclidean distance between any two pixels within a group Further definition:

[0097]

[0098] coefficient of variation The uniformity score is then:

[0099]

[0100] in The minimum spacing is set empirically based on the size of the study area. , This is the minimum spacing constraint threshold set based on the size of the study area and experience. For all Euclidean distances within the group The coefficient of variation is calculated as the ratio of the mean to the standard deviation.

[0101] (4) From the ground point To satellite The line-of-sight vector is Its mold length Sampling along the line of sight Each testing point For each DEM elevation obtained by nearest neighbor interpolation Calculate the line-of-sight height:

[0102]

[0103] Terrain obstruction is defined as the maximum superelevation ratio:

[0104]

[0105] Geometric visibility is modeled using an exponential decay model (considering diffraction and partial occlusion):

[0106]

[0107] Taking into account the visibility and coherence of the primary and secondary binary stars of all pixels in the group, as well as the distance factor, the visibility scoring function is calculated. :

[0108]

[0109] in For the first in the group The three-dimensional coordinate information corresponding to each pixel. , The first The orbital positions of the primary and secondary satellites corresponding to each pixel. For line-of-sight geometric visibility, For the first Spatial coherence coefficient of each pixel This is the slant range factor, used to quantify the attenuation effect of slant range on signal quality.

[0110] Step 2.3: Formulate a comprehensive penalty function for each optimization variable and add constraints;

[0111] Since the optimization variables are comprehensive scores, it is very likely that the placement method will not be reasonable because the comprehensive higher score does not comply with mandatory requirements such as the minimum spacing. Therefore, constraints will be added to the comprehensive penalty function formed by the optimization variables to form hard restrictions.

[0112] For constrained optimization problems, a static penalty function is used to map the degree of constraint violation to fitness decay, and a constraint violation vector is defined. The comprehensive penalty function is:

[0113]

[0114] Here ,Right now:

[0115] in The penalty weight coefficient is set according to the importance of each constraint, with an upper bound of 0.5 to ensure that the fitness is non-negative.

[0116] The specific constraints are as follows:

[0117] (1) Image boundary constraints: To ensure that the calibrator is located within the effective monitoring area, the pixels are defined. Euclidean distance to the boundary:

[0118]

[0119] The boundary penalty function is then:

[0120]

[0121] in pixels within a group Euclidean distance to the boundary of the monitoring area A boundary safety distance threshold is set based on experience to prevent the calibrator from getting too close to the image edge.

[0122] (2) Minimum Spacing Constraint: To avoid mutual interference between calibrators, the distance between any two points of the selected calibrators must not be less than [a certain value]. Define a penalty function for the minimum spacing constraint:

[0123]

[0124] in This is the minimum spacing set empirically based on the size of the study area.

[0125] (3) For areas unsuitable for deployment due to low coherence or steep terrain, a threshold can be set for filtering, and a penalty function for physical performance threshold constraints can be defined:

[0126]

[0127] in For the set coherence threshold, For pixels within a group Spatial coherence coefficient, For pixels within a group The corresponding elevation gradient value, This represents the maximum permissible slope.

[0128] 2.4: The scoring functions of each optimization variable and the comprehensive penalty function are combined using a weighted convex combination to form the comprehensive fitness function;

[0129]

[0130] Where the weight coefficients satisfy ; for the comprehensive penalty function A maximum value limit is imposed to ensure that the fitness is not less than 50% of the original value.

[0131] Step 3: Use a genetic algorithm to globally optimize the comprehensive fitness function, decode the optimal individual into a list of calibrator pixel coordinates, calculate the scores of each sub-objective and the constraint satisfaction, and generate a deployment map and index report.

[0132] Step 4: Based on the optimal individual obtained by the genetic algorithm in the monitoring area, i.e., the list of pixel coordinates of a specified number of calibrators, the calibrators are deployed in the study area. Then, based on the data provided by these calibrators, the subsequent baseline calibration and elevation inversion are completed to obtain a high-precision DEM product.

[0133] Example:

[0134] Step 1: Acquire InSAR data and its orbital parameters from synthetic aperture radar;

[0135] Interferometric processing is performed on SAR single-look complex images, including interferometric image pair registration, reference terrain interferometric phase removal / flat terrain interferometric phase removal, interferometric phase filtering, interferometric phase unwrapping, and elevation inversion, ultimately obtaining the inverted elevation data.

[0136] In the process of interferometry, the inversion of elevation may be significantly affected by errors in various parameters such as geometric parameters and satellite orbit data, which can greatly affect the accuracy of the final elevation calculation. Therefore, a calibration process needs to be added to improve the calculation accuracy.

[0137] Step 2: Construct the objective function, the specific steps are as follows:

[0138] Step 2.1: Define decision variables;

[0139] The inversion elevation data obtained by interferometric processing of InSAR data is based on a two-dimensional region. Therefore, the main objective when studying calibrator deployment schemes is to obtain the two-dimensional coordinates of the calibrators within the monitoring range. Setting this as the final output has the greatest engineering value.

[0140] Step 2.2: In actual calibration, the calibrator is generally a corner reflector placed at a designated location within the monitoring area. To reduce deployment costs, the rationality of deployment is comprehensively evaluated from four aspects: terrain suitability, degree of ill-formation, deployment uniformity, and satellite visibility.

[0141] (1) First, it is necessary to select an area with flat terrain, stable surface, and good scattering characteristics to ensure the long-term reliability and phase quality of the calibrator, combined with the slope value calculated by DEM. Surface roughness threshold and coherence The terrain suitability function is calculated from the information obtained:

[0142]

[0143] (2) The basic task of calibration is to use a calibrator with a known location to establish a mathematical model, i.e., a sensitivity matrix, between the system parameters that may have errors and the inverted elevation. The system parameters are then calibrated by using the error between the calibrator's determined elevation value and the inverted elevation value. The condition number is the ratio of the singular values ​​of the sensitivity matrix; the larger this value, the more unstable the parameter solution becomes, and it may even be impossible to solve. Among the various system parameters, the baseline error has a significant impact on the accuracy of altimetry. Based on the auxiliary satellite distance equation, a deterministic equation that the baseline vector and the elevation value should satisfy can be derived:

[0144]

[0145] in Slant distance of the main star This represents the decomposition of the baseline vector in the ECEF coordinate system. The decomposition value of the principal star's position vector in the ECEF coordinate system. The elevation data is obtained by inversion within the monitoring range. From this, the elevation data for each calibrator can be calculated. sensitivity vector Further obtained 3D sensitivity matrix Calculate its condition number And construct a piecewise continuous scoring function, i.e., a condition number evaluation function;

[0146] (3) The calibrators are evenly distributed in space to avoid clustering or sparseness, which can improve the sampling capability and the robustness of parameter estimation in the monitoring area.

[0147] The coefficient of variation is used to measure the uniformity of the distribution of a point set. Let the pixel point set be... The Euclidean distance matrix is The second term is the hard constraint softening factor.

[0148] (4) Ensure that the calibrator has a good line of sight to both the primary and secondary satellites to guarantee the quality and continuity of the interferometric data acquisition.

[0149] Step 2.3: Constraint modeling;

[0150] For constrained optimization problems, a static penalty function is used to map the degree of constraint violation to fitness decay. The specific constraint terms are as follows:

[0151] (1) Image boundary constraints; (2) Minimum spacing constraints; (3) Physical performance threshold constraints;

[0152] Step 2.4: Calculate the comprehensive fitness function;

[0153] The above optimization terms and constraint terms are combined using a weighted convex combination to form a comprehensive fitness:

[0154]

[0155] Where the weight coefficients satisfy At the same time, the maximum penalty is limited to ensure that the fitness is not lower than 50% of the original value.

[0156] Step 3: Solve using a genetic algorithm. The specific steps are as follows:

[0157] (1) Use real number encoding for a specified quantity Pixel coordinates of the scaler It stipulates that each individual is... It consists of real numbers and restricts the coordinate range to the valid image area;

[0158] (2) Randomly generated Each individual ensures that the initial solution satisfies the boundary constraints. For individuals that do not meet the minimum spacing requirement, adjustments can be made through random reset or local fine-tuning. However, a certain degree of violation is allowed in the initial population, which will be handled by the penalty function later.

[0159] (3) Calculate the overall fitness for each individual. And record the amount of constraint violation;

[0160] (4) A tournament selection method is used, with random selection each time. For each individual, the individual with the highest fitness is selected to enter the next generation. Simultaneously, an elite retention strategy is introduced, directly replicating the best individual from each generation to the next to prevent the loss of the optimal solution.

[0161] (5) Use arithmetic crossover, based on probability. Crossing offspring individuals from parent generations produces offspring. For real-number encoding, arithmetic crossover is defined as: ;in It is a random number;

[0162] (6) Employ polynomial mutation, with probability Randomly perturb the coordinates of the individual, for each coordinate The mutated value is ;in The random numbers are generated by a multinomial distribution to ensure that the variation range is controllable.

[0163] (7) When the maximum number of generations of evolution is reached or the fitness has not improved significantly for several consecutive generations, the evolution is terminated and the current best individual, i.e. the optimal placement scheme of the specified number of calibrators, is output.

[0164] Experimental results

[0165] To demonstrate the effectiveness of this invention, this embodiment uses InSAR data for experimental verification.

[0166] The experiment was conducted using SAR primary and secondary images of the Ku band obtained by simulating the system parameters shown in Table 1, and the simulation data was processed by interferometry according to the steps in the specific implementation method.

[0167] Table 1

[0168]

[0169] Figure 2The inversion elevation map obtained after interferometric processing of the Ku-band generated in the simulation is based on SRTM (Shuttle Radar Topography Mission) data. Therefore, to simulate the errors in the system parameters in the actual situation, the prior DEM data obtained by this interferometric processing is calculated by adding fixed errors of +2.00cm, +2.00cm, and -2.00cm to the baseline vectors decomposed in the ECEF coordinate system of the SRTM data, plus a random error with a maximum value of 0.02cm.

[0170] Assuming that the required number of calibrators is 20, the scores of the four optimization indicators and their set weight coefficients are calculated based on various parameters and formulas, as shown in Table 2 below.

[0171] Table 2

[0172]

[0173] The optimization parameters set in the genetic algorithm are shown in Table 3 below.

[0174] Table 3

[0175]

[0176] Figure 3 The graph shows the fitness evolution curves of the genetic algorithm. The blue line represents the optimal fitness curve, which indicates the quality of the best solution the algorithm can achieve in the current generation. The red line represents the average fitness curve, which reflects the overall quality level of the population. Both lines maintain an upward trend, indicating that the population is progressing as a whole and maintaining good diversity. As the number of iterations increases, the two lines tend to overlap, indicating that the population has converged and reached an optimal state.

[0177] Figure 4 To demonstrate the optimal deployment scheme of the calibrator on the DEM topographic map, the specific coordinate parameter values ​​of the optimal deployment calibrator output by this invention are shown in Table 4 below.

[0178] Table 4

[0179]

[0180] Figure 5 This is a DEM data map obtained by re-inverting the elevation after calibrating the baseline parameters based on the deployment of the calibrator.

[0181] The following compares the optimal placement schemes of 20 calibrators obtained by our method, which comprehensively considers genetic algorithms and penalty mechanisms, with those obtained by existing methods that only consider the placement of calibrators at both the near and far ends and those that only consider ill-conditioned problems. After calibrating the baseline parameter values, the inverted elevation data obtained is compared with the true elevation data. The final elevation error values ​​of the three methods are shown in Table 5 below.

[0182] Table 5

[0183]

[0184] The comparison shows that all three calibrator placement schemes can improve elevation accuracy. However, the elevation error after calibration by the method of the present invention is significantly smaller than the elevation error after considering only the placement on both sides of the far end or the calibration after minimizing the condition number, which proves the superiority of the method of the present invention.

[0185] The above embodiments are only used to illustrate the design concept and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. The protection scope of the present invention is not limited to the above embodiments. Therefore, all equivalent changes or modifications made based on the principles and design ideas disclosed in the present invention are within the protection scope of the present invention.

Claims

1. A space-borne InSAR calibrator deployment method based on genetic algorithm and penalty mechanism, characterized in that, The specific steps are as follows: Firstly, the SAR single-view complex image pairs are acquired and interfered to obtain the spatial coherence coefficient, the line-of-sight geometric parameter and the inversion height data of each pixel, and a comprehensive fitness function is constructed ; The specific steps include: Step 2.1: For the area to be deployed Each scaler is defined as a decision variable, and the coordinate matrix of each scaler in the pixel coordinate system is defined as the following: Among them, the first calibrator The column coordinates and row coordinates are respectively ; Step 2.2: Randomly select within the study area Each pixel position forms a set of decision variables. Calculate the scoring function for each optimization variable; Includes: Overall terrain suitability scoring function Condition number scoring function Uniformity scoring function and visibility scoring function ; Step 2.3: Formulate a comprehensive penalty function for each optimization variable and add constraints; The comprehensive penalty function is: ； in , and The penalty weight coefficient is set according to the importance of each constraint, with an upper bound of 0.5 to ensure that the fitness is non-negative; Let be the penalty function for image boundary constraints; Let be the penalty function for the minimum spacing constraint; This is the penalty function for physical performance threshold constraints; Step 2.4: Combine the scoring functions of each optimization variable with the comprehensive penalty function through a weighted convex combination to form the comprehensive fitness function; ； Where the weight coefficients satisfy ; Then, the genetic algorithm is used to globally optimize the comprehensive fitness function, decode the best individual into a list of calibrator pixel coordinates, and calculate the scores of each sub-objective and the constraint satisfaction status to generate a deployment map and index report. Finally, according to the pixel coordinate list of the calibrator, the calibrator was deployed in the study area, and subsequent baseline calibration and elevation inversion were completed to obtain a high-precision DEM product.

2. The method as described in claim 1, characterized in that, In step one, the gaze geometry parameters include the gaze vector corresponding to each pixel. Primary star slant distance and angle of incidence ; coordinate Spatial coherence coefficient corresponding to the pixel The formula is: ； in , To match the complex values ​​in the corresponding windows of the registered main and auxiliary images. Indicates to Take its conjugate; The inverted elevation data is generated from the three-dimensional coordinates of the ground point corresponding to each pixel, thus creating a numerical elevation model of the study area. .

3. The method as described in claim 2, characterized in that, The overall terrain suitability scoring function Calculate using the mean of the scores within the group: ； in For pixels Using spatial coherence coefficient and elevation gradient information Defined terrain suitability score; The calculation formula is: ； in This represents the terrain roughness tolerance threshold.

4. The method as described in claim 1, characterized in that, The condition number scoring function Yes: Calculation within the InSAR imaging geometric model group The sensitivity matrix formed by pixels And using its condition number, a continuous piecewise function is constructed to obtain: ； in Sensitivity matrix condition number.

5. The method as described in claim 1, characterized in that, The uniformity scoring function For any two pixels within a group, the coefficient of variation is used to measure the uniformity of the point set distribution to construct the following: ； in The minimum spacing is set empirically based on the size of the study area. , For any two pixels within the group and The Euclidean distance between them This is the minimum spacing constraint threshold set based on the size of the study area and experience. For all Euclidean distances within the group The coefficient of variation is calculated as the ratio of the mean to the standard deviation.

6. The method as described in claim 1, characterized in that, The visibility scoring function Taking into account the visibility and coherence of the primary and secondary binary stars of all pixels in the group, as well as the distance factor, the following calculations were performed: ； in For the first in the group The three-dimensional coordinate information corresponding to each pixel. , The first The orbital positions of the primary and secondary satellites corresponding to each pixel. For line-of-sight geometric visibility, For the first Spatial coherence coefficient of each pixel This is the slant distance factor.

7. The method as described in claim 2, characterized in that, The specific constraints of the comprehensive penalty function are as follows: (1) To ensure that the calibrator is located within the effective monitoring area, a penalty function is defined regarding the image boundary constraints: ； in pixels within a group Euclidean distance to the boundary of the monitoring area The boundary safety distance threshold is set based on experience; (2) To avoid interference between scalers, a penalty function for the minimum spacing constraint is defined: ； (3) For areas unsuitable for deployment due to low coherence or steep terrain, define a penalty function for physical performance threshold constraints: ； in For the set coherence threshold, For pixels within a group The corresponding elevation gradient value, This represents the maximum permissible slope.

Images