Clustering multi-baseline phase unwrapping algorithm combined with iterative local search strategy
By combining the clustered multi-baseline phase detangling algorithm with iterative local search strategy, the local extreme value problem of the multi-baseline phase detangling algorithm when solving the intercept term constraint equation is solved, achieving more efficient and accurate phase detangling.
Patent Information
- Application Number
- CN202411879169.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-19
- Publication Date
- 2025-03-07
AI Technical Summary
Existing multi-baseline phase detangling algorithms are prone to local extreme values when solving intercept term constraint equations, which affects the detangling accuracy and efficiency.
A clustered multi-baseline phase unwinding algorithm combined with iterative local search strategy is adopted to separate different fuzzy vectors through histogram clustering, construct integer constraint equations of intercept terms, and solve the solution set that satisfies the minimization objective function through iterative local search.
It effectively avoids local extreme value phenomena, enhances the algorithm's global search ability, optimizes the understanding of entanglement effect, and improves the accuracy and efficiency of phase entanglement.
Smart Images
Figure CN119575381A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of synthetic aperture radar interferometry, and in particular to a clustered multi-baseline phase unwrapping algorithm combined with an iterative local search strategy. Background Art
[0002] InSAR (Interferometric Synthetic Aperture Radar) technology is a remote sensing technology based on synthetic aperture radar (SAR) to obtain ground deformation information. It obtains two or more SAR images from different times and locations for interferometric analysis, thereby accurately measuring the tiny deformation of the surface, which is particularly suitable for monitoring surface movements such as earthquakes, volcanoes, landslides, and urban subsidence.
[0003] Phase unwrapping is the core step in processing interferometric phase in synthetic aperture radar interferometry (InSAR), and the accuracy of its algorithm will directly affect the quality of the final elevation product. At present, according to the number of baselines in the InSAR system, phase unwrapping algorithms can be divided into single-baseline phase unwrapping algorithms and multi-baseline phase unwrapping algorithms. Among them, single-baseline phase unwrapping algorithms are more common in practical applications because of their low data requirements and simple implementation. Traditional single-baseline phase unwrapping algorithms achieve absolute phase recovery by constructing phase gradients and minimizing a certain distance (such as energy function or error function), such as branch cutting method, mass-guided method and least squares method. However, this type of algorithm is often based on the assumption of phase continuity. In areas with large terrain undulations or phase discontinuities, this assumption is often difficult to meet, resulting in reduced unwrapping performance. In order to overcome the defects of single-baseline phase unwrapping algorithms in the face of complex terrain, multi-baseline phase unwrapping algorithms have emerged. Multi-baseline phase unwrapping algorithms expand the range of phase ambiguity coefficients by combining interferograms of different baseline lengths, thereby getting rid of the limitations of the phase continuity assumption.
[0004] The traditional multi-baseline phase unwrapping algorithm uses the Chinese remainder theorem as one of its basic theories. However, due to its poor noise robustness, the algorithm is difficult to apply in practice. Based on this, the clustering-based multi-baseline phase unwrapping algorithm uses the clustering characteristics of multi-baseline data to effectively reduce the interference of noise on the unwrapping process. For example, Yue Jiawei et al. proposed a multi-baseline InSAR phase unwrapping method based on a discrete optimization framework in the Journal of Geodaetics and Cartography (2024, Vol. 53, No. 3, pp. 473-481). In order to solve the intercept constraint equation, this method first transforms the algorithm model into a discrete optimization problem, and then uses the integer programming method to globally solve the ambiguity vector. By solving the integer solution that satisfies the constraint equation, the absolute phase is reconstructed. However, with the increase in the number of interference pairs, the dimension of the ambiguity solution space will increase significantly, resulting in an exponential increase in computational complexity. In practical applications, the complexity of the solution space not only increases the time cost of the solution, but also makes the algorithm more susceptible to interference from noise and local constraints during the search process, thereby falling into local extremes. The generation of local extreme values will directly affect the accuracy of the ambiguity vector, causing the disentanglement result to deviate from the global optimal solution, thereby weakening the reliability of disentanglement. Summary of the invention
[0005] The present invention aims to provide a clustered multi-baseline phase unwrapping algorithm combined with an iterative local search strategy to solve the problem that in the step of solving the intercept term constraint equation, local extreme values may occur in the discrete optimization process involved in the current algorithm, thereby affecting the phase unwrapping accuracy and efficiency.
[0006] To achieve the above object, the present invention adopts the following technical solution: a clustering multi-baseline phase unwrapping algorithm combined with an iterative local search strategy, comprising: step 1: combining interference patterns of different baseline lengths by means of histogram clustering, and separating combinations of different fuzzy vectors; Step 2: Separate all the peaks on the envelope obtained after clustering, and obtain the horizontal coordinate corresponding to each type of peak as the intercept value of that type; Step 3: Construct the integer constraint equation of the intercept term; Step 4: Establish the objective function and transformation constraints; Step 5: Iterate local search to find a solution set that satisfies the minimum objective function , as the fuzzy coefficient corresponding to each type of pixel; the iterative local search solution specifically includes: first setting the initial step size and the acceptability threshold of the decision solution , in the initial search range Step size Complete the global search and determine whether there is an integer value less than the threshold in the interval. If there is a point that meets this condition, mark it as a potential optimal solution; if an area containing a potential optimal solution is found, update the step size to Perform a local fine search to obtain the target solution, and the updated step size in this step will not be less than the minimum step size ; Repeated iterative search and step size adjustment, evaluate the effect of each iteration, and finally find the solution set that makes the objective function optimal; Step 6: Calculate the absolute phase and recover the unwrapped phase of each interferogram.
[0007] Preferably, in step 5, in the global search, if all results are greater than the acceptable threshold , then in Expand the search range on the plane to , and search again.
[0008] Preferably, when the original boundary range is expanded in step 5, a cell is expanded at the boundary and adjacent point pairs are exchanged, and then an iterative search is performed.
[0009] Preferably, if no better solution is obtained after expanding the boundary range, the original solution in step 5 is accepted unchanged, and the state space corresponding to the solution is updated.
[0010] Preferably, in step 5, when the step length in the fine search reaches If the optimal solution is still not found, expand the search range to , and reset the step size to , and repeat the process from global search to local search again.
[0011] Preferably, the search and solution of the intercept term constraint equation in step 5 is regarded as a A fully connected undirected graph consisting of vertices Select The points constitute the solution set ; The elements in the solution set will make the original objective function optimal. is a fully connected undirected graph, where is a set of vertices, is an edge set, and , then each vertex It can be expressed in two-dimensional coordinates, and its mass Defined as: in for To the intercept term constraint equation If we define the set The quality of each point in makes the set have the best characteristics, then the following objective function should be minimized: .
[0012] This scheme proposes an improved multi-baseline phase unwrapping algorithm based on clustering analysis. This method mainly improves the local extreme value phenomenon that may occur in the discrete optimization process when searching and solving the constructed intercept term constraint equation in the clustering-based multi-baseline phase unwrapping algorithm. By dynamically adjusting the step size and the solution space range, the global search capability of the algorithm is enhanced, thereby optimizing the unwrapping effect. In view of the incomplete solution space of the algorithm in the boundary area, the neighborhood operation is introduced to expand the original boundary, and the state space of the corresponding solution at the boundary is adjusted through the point pair exchange method to ensure the integrity of the solution space. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] Figure 1 Schematic diagram of a flow chart of an embodiment of the present invention.
[0014] Figure 2 A schematic diagram of the process of iterative local search solution according to an embodiment of the present invention.
[0015] Figure 3 Schematic diagram of global search and range expansion according to an embodiment of the present invention.
[0016] Figure 4 Schematic diagram of neighborhood operation according to an embodiment of the present invention.
[0017] Figure 5 Schematic diagram of local search according to an embodiment of the present invention.
[0018] Figure 6 This is a long baseline interferogram of an embodiment of the present invention.
[0019] Figure 7 This is a short baseline interference diagram of an embodiment of the present invention.
[0020] Figure 8 This is a phase unwrapping effect diagram of a multi-baseline InSAR phase unwrapping method based on a discrete optimization framework in the background technology of an embodiment of the present invention.
[0021] Fig. 9 This is a diagram showing the phase unwrapping effect of the method according to an embodiment of the present invention.
[0022] Fig.10 It is a reference topographic map of the embodiment of the present invention.
[0023] Fig.11 It is a clustering histogram of an embodiment of the present invention.
[0024] Fig.12 In the embodiment of the present invention, each group of fuzzy vectors corresponds to the equation line Floor plan distribution diagram. DETAILED DESCRIPTION
[0025] The following is further described in detail through specific implementation methods: A clustering multi-baseline phase unwrapping algorithm combined with an iterative local search strategy is shown in the attached Figure 1 As shown, the following steps are included: Step 1: Combine the interference patterns of different baseline lengths by histogram clustering and separate the combinations of different blur vectors.
[0026] Phase unwrapping is the process of estimating the absolute phase from the wrapped phase. The measured wrapped phase has the following relationship with the absolute phase: (1) From formula (1), we can know that the absolute phase The essence of is to solve the corresponding fuzzy coefficient , that is, eliminate phase ambiguity to obtain the true phase value. For the same area, the interferograms corresponding to different baselines should have the same elevation. In InSAR technology, the relationship between elevation and absolute phase can be expressed as: (2) Taking the dual baseline system as an example, combining equation (1) and equation (2) yields: (3) Formula (3) is the basic formula for multi-baseline phase unwrapping. By simplifying them as independent variables and dependent variables, we can get the relationship between different baseline lengths, fuzzy coefficients and winding phases as follows: (4) For a pixel at the same position, in two interference images with different baseline lengths, its blur coefficient can be expressed by the blur vector indicates that, and Should satisfy equation (4). Ideally, the equation for all pixels with the same blur vector is The planes should overlap each other, such as Fig.11 and Fig.12 When there is noise interference, To keep the system parameters unchanged, only the intercept term in equation (4) The value of is disturbed, causing the straight line to The plane is offset, but there is still clustering phenomenon for lines with the same fuzzy vector. This scheme uses the straight graph clustering method to cluster the lines corresponding to each pixel. Perform clustering.
[0027] Step 2: Separate all the peaks on the envelope obtained after clustering, and obtain the horizontal coordinate corresponding to each type of peak as the intercept value of that type.
[0028] Step 3: Construct the integer constraint equation for the intercept term.
[0029] by is the slope, construct its The intercept term constraint equation on the plane, then (4) can be reconstructed as: (5) in is the value of the intercept term corresponding to the class after clustering, and the fuzzy vector It should also satisfy equation (5). Since the fuzzy vector The phase integer fuzzy number represented must be an integer, so that the solution space of the constructed equation system presents discrete characteristics. Therefore, the solution of equation (5) can be regarded as an integer programming problem.
[0030] Step 4: Establish the objective function and transformation constraints.
[0031] The solution of the equation system corresponding to formula (5) constructed for all categories can be regarded as a system composed of A fully connected undirected graph consisting of vertices Select The points constitute the solution set ; The elements in the solution set will make the original objective function optimal.
[0032] set up is a fully connected undirected graph, where is a set of vertices, is an edge set, and , then each vertex It can be expressed in two-dimensional coordinates, and its mass Defined as: (6) in for To the intercept term constraint equation If we define the set The quality of each point in makes the set have the best characteristics, then the following objective function should be minimized: In the formula, represents the integer solution of the intercept constraint equation (5) corresponding to a certain class, represents the sum of the distances of all ideal solutions to the constraint equations of the corresponding category. The goal is to optimize the solution set Make all solutions to its corresponding constraint equation The sum of distance functions is minimized.
[0033] Step 5: Iterate local search to find a solution set that satisfies the minimum objective function , as the fuzzy coefficient corresponding to each type of pixel.
[0034] In this step, the intercept constraint equation is searched and solved, as shown in the attached figure. Figure 2 As shown, specifically including: Step 51: First set the initial step size and the acceptability threshold of the decision solution , in the initial search range Step size Complete the global search and determine whether there is an integer value less than the threshold in the interval. If there is a point that meets the condition, mark it as a potential optimal solution.
[0035] The empirical parameters are verified by experiments and represent the maximum deviation range allowed between the solution and the objective function constraints. and Respectively and The initial search range on the axis, thus constructing a rectangular region for global search, and Should meet To ensure that the density of discrete sampling points in the initial search space is high enough.
[0036] Step 52: In the global search, if all results are greater than the acceptable threshold , as attached Figure 3 As shown, in Expand the search range on the plane to , and search again. To ensure the integrity of the solution space, when expanding the original boundary range, as shown in the attached Figure 4 As shown, a cell is expanded at the boundary and adjacent pairs of points are exchanged, and then an iterative search is performed. , after point pair exchange, becomes Substitute it into the above formula to verify whether it fully makes the objective function optimal. In general, when processing the boundary, the neighborhood operation that makes the quality of the entire solution better will be given priority. If no better solution is obtained, the original solution will be accepted unchanged, but the state space corresponding to the solution will be updated.
[0037] Step 53: If a region containing a potential optimal solution is found, the update step size is Perform local fine search to obtain the target solution, as shown in the attached Figure 5 As shown, the updated step size in this step will not be less than the minimum step size ; When the step size reaches If the optimal solution is still not found, expand the search range to , and reset the step size to , and repeat the process from global search to local search again.
[0038] Step 54: Repeat the iterative search and step size adjustment, evaluate the effect of each iteration, and finally find the solution set that makes the objective function optimal.
[0039] The overall search process of iterative local search in this scheme is divided into two parts: greedy descent strategy and adding disturbance. First, an initial solution is constructed using the greedy strategy and optimized through local search. In the greedy descent process, if a better solution is found, the solution set is updated and iteration continues; if the solution is not improved, the number of unimproved times is recorded and the perturbation step size is increased. When the number of unimproved times reaches the preset threshold, the algorithm restarts the state and enters the iterative search process with the maximum perturbation intensity. Through this solution method, the original algorithm can explore different areas of the solution space, increasing the possibility of finding the global optimal solution.
[0040] Step 6: Calculate the absolute phase and recover the unwrapped phase of each interferogram.
[0041] Solve the solution set that satisfies the minimization objective function by iterative local search , as the fuzzy coefficient corresponding to each type of pixel, so as to solve the absolute phase corresponding to each type of pixel.
[0042] The iterative local search in this scheme is a pseudo-heuristic algorithm for solving discrete optimization problems. The core advantage of this algorithm is that it combines the high efficiency of the greedy strategy with the diversity of the perturbation mechanism. It improves the local extreme value phenomenon that may occur in the discrete optimization process, and enhances the global search ability of the algorithm by dynamically adjusting the step size and the solution space range. It conducts multiple local searches in the solution space and uses the perturbation mechanism to avoid falling into the local optimum and find the global optimal solution, thereby optimizing the disentanglement effect. In view of the incomplete solution space in the boundary area of the algorithm, the neighborhood operation is introduced to expand the original boundary, and the state space of the corresponding solution at the boundary is adjusted by the point pair exchange method to ensure the integrity of the solution space.
[0043] The following is an example of phase unwrapping of the present application method: Sentinel-1 data from the Tom Price Mountains in central Western Australia near 23°S, 117°E were selected as verification. The reference topographic map of this area is as follows: Fig.10 As shown, the different baseline interference patterns obtained by measurement are as follows Figure 6 and Figure 7As shown, the region includes areas with poor coherence such as rivers and lakes. Phase unwrapping is performed using a multi-baseline InSAR phase unwrapping method based on a discrete optimization framework in the background technology and the method of the present application. Figure 8 is a diagram of the disentanglement effect of the method in the background technology, Fig. 9 This is a diagram of the disentanglement effect of the present application method.
[0044] Depend on Figure 8 , Fig. 9 and Fig.10 It can be seen that the traditional algorithm produces some phase separation similar to gentle slope terrain in the upper area, but there is no such terrain in this area in the reference terrain, which indicates that the original algorithm has errors here. In addition, in the lower right area, this method shows a smoother regional transition without obvious jumps or discontinuities, while the original algorithm has more phase discontinuities in this area, resulting in larger errors. Overall, the unwrapping results of this algorithm are smoother than those of the original algorithm, while the unwrapping results corresponding to the original algorithm have more bright spots and irregular textures, and there are some burrs at the edges. Therefore, the unwrapping results of this algorithm on this set of data are better than those of the original algorithm.
[0045] The above is only an embodiment of the present invention, and the common knowledge such as the known specific technical schemes and / or characteristics in the scheme is not described in detail here. It should be pointed out that for those skilled in the art, several variations and improvements can be made without departing from the technical scheme of the present invention. In the present invention, unless otherwise clearly specified and limited, the terms "install", "connect", "connect", "fix" and the like should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be directly connected, or indirectly connected through an intermediate medium, or it can be the internal connection of two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances. The scope of protection claimed in this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.
Claims
1. A clustering multi-baseline phase unwrapping algorithm combined with an iterative local search strategy, characterized by: Step 1: Combine the interference patterns of different baseline lengths by histogram clustering and separate the combinations of different blur vectors; Step 2: Separate all the peaks on the envelope obtained after clustering, and obtain the horizontal coordinate corresponding to each type of peak as the intercept value of that type; Step 3: Construct the integer constraint equation of the intercept term; Step 4: For the integer programming problem, establish the objective function and transformation constraints; Step 5: Iterate local search to find a solution set that satisfies the minimum objective function , as the fuzzy coefficient corresponding to each type of pixel; the iterative local search solution specifically includes: first setting the initial step size and the acceptability threshold of the decision solution , in the initial search range Step size Complete the global search and determine whether there is an integer value less than the threshold in the interval. If there is a point that meets this condition, mark it as a potential optimal solution; if an area containing a potential optimal solution is found, update the step size to Perform a local fine search to obtain the target solution, and the updated step size in this step will not be less than the minimum step size ; Repeated iterative search and step size adjustment, evaluate the effect of each iteration, and finally find the solution set that makes the objective function optimal; Step 6: Calculate the absolute phase and recover the unwrapped phase of each interferogram.
2. The clustered multi-baseline phase unwrapping algorithm combined with iterative local search strategy according to claim 1, characterized in that: In the global search in step 5, if all results are greater than the acceptable threshold , then in Expand the search range on the plane to , and search again.
3. The clustered multi-baseline phase unwrapping algorithm combined with iterative local search strategy according to claim 2, characterized in that: In step 5, when the original boundary range is expanded, a cell is expanded at the boundary and adjacent point pairs are exchanged, and then an iterative search is performed.
4. The clustered multi-baseline phase unwrapping algorithm combined with iterative local search strategy according to claim 3, characterized in that: If no better solution is obtained after expanding the boundary range, the original solution in step 5 is accepted unchanged and the state space corresponding to the solution is updated.
5. The clustered multi-baseline phase unwrapping algorithm combined with iterative local search strategy according to claim 1, characterized in that: In step 5, when the step length in the fine search reaches If the optimal solution is still not found, expand the search range to , and reset the step size to , and repeat the process from global search to local search again.
6. The clustered multi-baseline phase unwrapping algorithm combined with iterative local search strategy according to claim 1, characterized in that: The search and solution of the intercept term constraint equation in step 5 is regarded as a A fully connected undirected graph consisting of vertices Select The points constitute the solution set ; The elements in the solution set will make the original objective function optimal. is a fully connected undirected graph, where is a set of vertices, is an edge set, and , then each vertex It can be expressed in two-dimensional coordinates, and its mass Defined as: in for To the intercept term constraint equation If we define the set The quality of each point in makes the set have the best characteristics, then the following objective function should be minimized: 。