Dam deformation partition identification method and related device

Abstract

The application provides a dam deformation partition identification method and related device, comprising: acquiring multi-source data, the multi-source data comprising first InSAR data and first GNSS data; performing data quality processing on the first InSAR data and the first GNSS data to obtain second InSAR data and second GNSS data; the data quality processing comprises one or more of data alignment, data cleaning and assignment; constructing a global three-dimensional deformation field data according to the second InSAR data and the second GNSS data, the global three-dimensional deformation field comprising a plurality of monitoring points; constructing a first comprehensive feature vector for each monitoring point according to the second InSAR data, the second GNSS data and the global three-dimensional deformation field data; inputting the first comprehensive feature vector into a partition model to obtain a target partition corresponding to each monitoring point.

Description

Technical Field

[0001] This application belongs to the field of computer system technology with specific computational models, and specifically relates to a method and related device for identifying dam deformation zones. Background Technology

[0002] Currently, with the expansion of large-scale water conservancy projects, long-term deformation monitoring of dams and surrounding slopes has become a core aspect of ensuring the safe operation of these projects. The accumulation of even minor deformations can trigger major accidents such as dam leakage and slope landslides. This is especially true in mountainous reservoirs with complex geological conditions, where deformation mechanisms are diverse and spatially heterogeneous, placing higher demands on the accuracy and zonal analytical capabilities of monitoring methods.

[0003] Traditional classification methods rely solely on a single deformation index (such as vertical rate), ignoring the heterogeneity and spatial continuity of measurement errors. This leads to severe fragmentation of zones, making it difficult to match the engineering structural logic (such as the deformation differences between dam foundations and dam abutments), and thus hindering safety assessment decisions. Summary of the Invention

[0004] This application provides a method and related apparatus for identifying dam deformation zones, aiming to integrate multi-source data to achieve probabilistic output of dam and surrounding area zones, providing accurate support for engineering safety assessment and risk identification.

[0005] Firstly, this application provides a method for identifying dam deformation zones, including:

[0006] Acquire multi-source data, including first InSAR data and first GNSS data;

[0007] The first InSAR data and the first GNSS data are subjected to data quality processing to obtain second InSAR data and second GNSS data; the data quality processing includes one or more of data alignment, data cleaning and assignment.

[0008] A global three-dimensional deformation field is constructed based on the second InSAR data and the second GNSS data, wherein the global three-dimensional deformation field includes multiple monitoring points;

[0009] A first comprehensive feature vector is constructed for each monitoring point based on the second InSAR data, the second GNSS data, and the global three-dimensional deformation field data.

[0010] The first comprehensive feature vector is input into the partitioning model to obtain the target partition corresponding to each monitoring point.

[0011] In conjunction with the first aspect, in one possible embodiment, the plurality of monitoring points include a first InSAR monitoring point and a first GNSS monitoring point. The first InSAR data includes the InSAR pixel coordinates of the first InSAR monitoring point, and the first GNSS data includes the GNSS station coordinates of the first GNSS monitoring point. The step of performing data quality processing on the first InSAR data and the first GNSS data to obtain second InSAR data and second GNSS data includes: projecting the InSAR pixel coordinates and the GNSS station coordinates onto the same plane coordinate system; determining a first observation time of the first InSAR data and a second observation time of the first GNSS data; if the first observation time and the second observation time are different, determining the average deformation rate during the overlapping period of the first InSAR data and the first GNSS data; performing time interpolation on the first InSAR data or GNSS data according to the average deformation rate to time-align the first InSAR data and the first GNSS data; removing monitoring points with abnormal InSAR data according to a first outlier removal method to obtain the second InSAR data; and removing monitoring points with abnormal GNSS data according to a second outlier removal method to obtain the second GNSS data.

[0012] In conjunction with the first aspect, in one possible embodiment, the plurality of monitoring points include a plurality of first InSAR monitoring points and a plurality of first GNSS monitoring points; the step of constructing global three-dimensional deformation field data based on the second multi-source data includes: determining K adjacent first GNSS monitoring points for each first InSAR monitoring point; determining a first weight for the K first GNSS monitoring points; and determining the target three-dimensional deformation rate and target three-dimensional error of the first InSAR monitoring points based on the first weights to obtain global three-dimensional deformation field data.

[0013] In conjunction with the first aspect, in one possible embodiment, the multi-source data further includes prior engineering structure data and terrain data; the step of constructing a first comprehensive feature vector for each monitoring point based on the second InSAR data, the second GNSS data, and the global three-dimensional deformation field data includes: determining the predicted line-of-sight rate and residual; constructing a second comprehensive feature vector for each first InSAR monitoring point based on the predicted line-of-sight rate, the residual, the global three-dimensional deformation field data, the prior engineering structure data, and the terrain data; and standardizing the second comprehensive feature vector to obtain a first comprehensive feature vector, wherein the first comprehensive feature vector includes the predicted line-of-sight rate, the residual, the target three-dimensional deformation rate, the target three-dimensional error, the terrain data, and the prior engineering structure data.

[0014] In conjunction with the first aspect, in one possible embodiment, the step of inputting the first comprehensive feature vector into the partitioning model to obtain the target partition corresponding to each monitoring point includes: determining the target three-dimensional variance of each first InSAR monitoring point based on the target three-dimensional error of each first InSAR monitoring point; determining a second weight for each first InSAR monitoring point based on the coherence coefficient, the first error, and the target three-dimensional variance of each first InSAR monitoring point; normalizing the second weight to obtain a third weight; determining the initial parameters of the partitioning model; inputting the first comprehensive feature vector into the partitioning model; and iterating the partitioning model based on the first comprehensive feature vector, the third weight, and the initial parameters to obtain the target partition.

[0015] In conjunction with the first aspect, in one possible embodiment, determining the initial parameters of the partitioning model includes: acquiring prior data of the engineering structure; determining multiple basic partitions based on the structural regions in the prior data of the engineering structure; dividing the multiple first InSAR monitoring points into the multiple basic partitions to obtain multiple preset groups; determining the mean and covariance of the multiple preset groups; and determining the proportion of the first InSAR monitoring points in each preset group to obtain a fourth weight.

[0016] In conjunction with the first aspect, in one possible embodiment, the step of iterating the partitioning model based on the first comprehensive feature vector, the third weight, and the initial parameters to obtain the target partition includes: determining a first probability that each first InSAR monitoring point belongs to one of the plurality of basic partitions based on the first comprehensive feature vector, the third weight, the mean, the covariance, and the fourth weight; determining a first label based on the first probability, the first label being used to indicate the basic partition to which the corresponding first InSAR monitoring point belongs; determining the number of valid samples; updating the mean, the covariance, and the fourth weight based on the number of valid samples to obtain a new mean, a new covariance, and a new fourth weight; determining an objective function based on the new mean, the new covariance, and the new fourth weight; repeating the above steps until the rate of change of the objective function is less than a preset value or equal to zero, then stopping the iteration and outputting the current target partition.

[0017] In conjunction with the first aspect, in one possible embodiment, the method further includes: after B iterations, performing a partition rationality check to obtain target iteration data; the rationality check includes splitting check, merging check, outlier handling, and outlier partition handling; generating partitioning results based on the target partitions and the corresponding target iteration data; the partitioning results include the partition label of each first InSAR monitoring point and the probability of belonging to each partition; statistically analyzing the deformation characteristics of each partition, generating a partitioning interpretation report, the partitioning interpretation report including residual variance reduction rate, spatial coherence, structural region consistency, and outlier boundary proportion.

[0018] Secondly, this application provides a dam deformation zoning identification device, comprising:

[0019] An acquisition unit is used to acquire multi-source data, the multi-source data including first InSAR data and first GNSS data;

[0020] A first processing unit is configured to perform data quality processing on the first InSAR data and the first GNSS data to obtain second InSAR data and second GNSS data; the data quality processing includes one or more of data alignment, data cleaning, and assignment.

[0021] A construction unit is configured to construct global three-dimensional deformation field data based on the second InSAR data and the second GNSS data, wherein the global three-dimensional deformation field includes multiple monitoring points; and to construct a first comprehensive feature vector for each monitoring point based on the second InSAR data, the second GNSS data and the global three-dimensional deformation field data.

[0022] The determining unit is used to input the first comprehensive feature vector into the partitioning model to obtain the target partition corresponding to each monitoring point.

[0023] Thirdly, this application provides an electronic device including a processor, a memory, a communication interface, and one or more programs, said one or more programs being stored in the memory and configured to be executed by the processor, said programs including instructions for performing the steps of the first or second aspect of this application.

[0024] Fourthly, this application provides a computer-readable storage medium storing a computer program for electronic data interchange, wherein the computer program causes a computer to perform some or all of the steps described in the first or second aspect of this application.

[0025] Fifthly, this application provides a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps described in the first or second aspect of this application. The computer program product may be a software installation package.

[0026] As can be seen, this application first acquires multi-source data, including first InSAR data and first GNSS data; then, it performs data quality processing on the first InSAR data and the first GNSS data to obtain second InSAR data and second GNSS data; the data quality processing includes one or more of data alignment, data cleaning, and assignment; based on the second InSAR data and the second GNSS data, it constructs global three-dimensional deformation field data, which includes multiple monitoring points; based on the second InSAR data, the second GNSS data, and the global three-dimensional deformation field data, it constructs a first comprehensive feature vector for each monitoring point; and it inputs the first comprehensive feature vector into a partitioning model to obtain the target partition corresponding to each monitoring point. In this way, the fusion of multi-source data enables probabilistic output of partitioning of the dam and surrounding areas, providing accurate support for engineering safety assessment and risk identification. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a schematic block diagram of the structure of a data processing system provided in an embodiment of this application;

[0029] Figure 2 This is a schematic block diagram of the structure of the first dam deformation multi-data fusion processing system provided in the embodiments of this application;

[0030] Figure 3 This is a schematic block diagram of the structure of the second type of dam deformation multi-data fusion processing system provided in the embodiments of this application;

[0031] Figure 4 This is a flowchart illustrating a method for identifying dam deformation zones according to an embodiment of this application;

[0032] Figure 5 This is a schematic diagram of the dam engineering structure provided in the embodiments of this application;

[0033] Figure 6 This is a schematic block diagram of the structure of the first type of dam deformation zoning identification device provided in the embodiments of this application;

[0034] Figure 7 This is a schematic block diagram of the structure of the second type of dam deformation zoning identification device provided in the embodiments of this application;

[0035] Figure 8 This is a schematic block diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0036] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0037] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, systems, products, or apparatuses.

[0038] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0039] The following is a brief introduction to the relevant terminology used in this application.

[0040] GMM stands for Gaussian Mixture Model. Essentially, it uses a weighted combination of multiple Gaussian distributions (normal distributions) to fit the probability distribution of complex data, and it is a classic model in unsupervised learning.

[0041] InSAR (Interferometric Synthetic Aperture Radar): The core of InSAR is a remote sensing technology that uses the phase difference between two (or repeated observations from the same satellite) synthetic aperture radar (SAR) images to generate interferograms and invert three-dimensional topographic, deformation, and elevation information of the Earth's surface. It is a core means of surveying and mapping and geological disaster monitoring.

[0042] GNSS (Global Navigation Satellite System): Its core is to provide receivers at any location in the world with three-dimensional coordinates (longitude, latitude, and elevation), velocity, and time (PVT) information through multiple navigation satellites orbiting the Earth. It is the fundamental core technology for surveying, navigation, and positioning.

[0043] Currently, with the expansion of large-scale water conservancy projects, long-term deformation monitoring of dams and surrounding slopes has become a core aspect of ensuring the safe operation of these projects. The accumulation of even minor deformations can trigger major accidents such as dam seepage and slope landslides. This is especially true in mountainous reservoirs with complex geological conditions, where deformation mechanisms are diverse and spatially heterogeneous, placing higher demands on the accuracy and zoning analysis capabilities of monitoring methods. Traditional classification methods rely solely on a single deformation index (such as vertical velocity), ignoring the heterogeneity and spatial continuity of measurement errors. This leads to severe zoning fragmentation, failing to match the structural logic of the engineering structure (such as the deformation differences between the dam foundation and abutments), and thus making it difficult to support safety assessment decisions.

[0044] To address the aforementioned issues, this application provides a method and related apparatus for identifying dam deformation zones. This method can be applied to scenarios involving the identification of dam observation point zones. It is applicable to various scenarios, including but not limited to the applications mentioned above.

[0045] The system architecture involved in the embodiments of this application is described below.

[0046] Please see Figure 1This application provides a data processing system, including a receiving device and a data processing device. The receiving device receives data from multiple sources, i.e., multi-source data. Multi-source data includes InSAR data, GNSS data, prior engineering structure data, and terrain data. The data processing device then acquires the multi-source data from the receiving device. The data processing device performs data quality processing on the first InSAR data and the first GNSS data to obtain second InSAR data and second GNSS data. The data quality processing includes one or more of data alignment, data cleaning, and assignment. A global three-dimensional deformation field data is constructed based on the second InSAR data and the second GNSS data. The global three-dimensional deformation field includes multiple monitoring points. A first comprehensive feature vector is constructed for each monitoring point based on the second InSAR data, the second GNSS data, and the global three-dimensional deformation field data. The first comprehensive feature vector is input into a zoning model to obtain the target zoning corresponding to each monitoring point. In this way, the fusion of multi-source data enables probabilistic output of zoning for the dam and its surrounding area, providing accurate support for engineering safety assessment and risk identification.

[0047] Please see Figure 2 and Figure 3 This application provides a multi-source data fusion processing system for dam deformation, including a data acquisition device and a data processing system. The data processing system includes a receiving device and a data processing device. The data acquisition device includes multiple satellites, which respectively acquire InSAR data, GNSS data, prior engineering structure data, and terrain data. Then, the data processing device acquires multi-source data from the receiving device. The data processing device performs data quality processing on the first InSAR data and the first GNSS data to obtain second InSAR data and second GNSS data. The data quality processing includes one or more of data alignment, data cleaning, and assignment. Based on the second InSAR data and the second GNSS data, a global three-dimensional deformation field data is constructed, which includes multiple monitoring points. Based on the second InSAR data, the second GNSS data, and the global three-dimensional deformation field data, a first comprehensive feature vector is constructed for each monitoring point. The first comprehensive feature vector is input into a partitioning model to obtain the target partition corresponding to each monitoring point. In this way, the fusion of multi-source data realizes the probabilistic output of dam and surrounding area partitioning, providing accurate support for engineering safety assessment and risk identification.

[0048] The specific methods will be described in detail below.

[0049] Please see Figure 4 This application also provides a method for identifying dam deformation zones, including:

[0050] S401. Acquire multi-source data, wherein the multi-source data includes first InSAR data and first GNSS data.

[0051] S402. Perform data quality processing on the first InSAR data and the first GNSS data to obtain the second InSAR data and the second GNSS data.

[0052] The data quality processing includes one or more of the following: data alignment, data cleaning, and assignment.

[0053] S403. Construct global three-dimensional deformation field data based on the second InSAR data and the second GNSS data.

[0054] The global three-dimensional deformation field includes multiple monitoring points.

[0055] S404. Construct a first comprehensive feature vector for each monitoring point based on the second InSAR data, the second GNSS data, and the global three-dimensional deformation field data.

[0056] S405. Input the first comprehensive feature vector into the partitioning model to obtain the target partition corresponding to each monitoring point.

[0057] In practice, the multi-source data includes first InSAR data, first GNSS data, prior engineering structure data, and terrain data.

[0058] The first InSAR data includes line-of-sight mean deformation rate (LOS), line-of-sight standard deviation (σ_LOS), and coherence coefficient (coh); the first GNSS data includes three-dimensional deformation rates and observation errors (σ) for the eastward (Vx), northward (Vy), and vertical (Vz) directions of a limited number of stations; the prior engineering structural data includes zoning labels for engineering structural areas such as dam foundation, dam abutment, and upstream and downstream slopes; and the topographic data includes elevation (H), slope (S), and aspect (SlopeDir).

[0059] As shown in Table 1, the following is a specific example to illustrate the multi-source data of this embodiment. Table 1 provides data corresponding to 10 first InSAR monitoring points and two first GNSS monitoring points. It should be understood that the data in Table 1 is only for illustrative purposes and is not intended to be unique. Wherein, LOS represents the InSAR line-of-sight observation data; σ_LOS is the standard deviation of the line-of-sight observation data; coh is the coherence coefficient of the InSAR data, used to evaluate the reliability of the deformation measurement results for each first InSAR monitoring point. Vx, Vy, and Vz are the first three-dimensional deformation rates in the first GNSS data, and σ is the error of the first three-dimensional deformation rate, i.e., the first three-dimensional error. The contents of Table 1 are as follows:

[0060] Table 1 Multi-source data

[0061]

[0062] Multi-source data is acquired through a data processing device, and then quality processing is performed on each source data. Quality processing includes one or more of the following: data alignment, data cleaning, and data assignment. This process unifies the coordinates and time base of the multi-source data, performs quality screening and cleaning, removes low-confidence points, and assigns a quality weight to each point to improve the overall data quality.

[0063] Optionally, the satellite parameters can be selected as incident angle θ = 30° (0.523 radians) and orbital azimuth angle α = 45° (0.785 radians); then the line-of-sight unit vector can be calculated:

[0064]

[0065] Specifically, the plurality of monitoring points includes a first InSAR monitoring point and a first GNSS monitoring point. The first InSAR data includes the InSAR pixel coordinates of the first InSAR monitoring point, and the first GNSS data includes the GNSS station coordinates of the first GNSS monitoring point. The step of performing data quality processing on the first InSAR data and the first GNSS data to obtain second InSAR data and second GNSS data includes: projecting the InSAR pixel coordinates and the GNSS station coordinates onto the same plane coordinate system; determining a first observation time for the first InSAR data and a second observation time for the first GNSS data; if the first observation time and the second observation time are different, determining the average deformation rate during the overlapping period of the first InSAR data and the first GNSS data; performing time interpolation on the first InSAR data or GNSS data according to the average deformation rate to align the first InSAR data and the first GNSS data in time; removing monitoring points with abnormal InSAR data according to a first outlier removal method to obtain the second InSAR data; and removing monitoring points with abnormal GNSS data according to a second outlier removal method to obtain the second GNSS data.

[0066] For example, data alignment includes coordinate alignment and time alignment. Coordinate alignment refers to unifying the coordinates of all data. For instance, projecting the coordinates of all first-source InSAR pixels and the coordinates of all first-source GNSS stations onto the same plane coordinate system. This plane coordinate system can be such as UTM or local engineering coordinates, without a specific requirement for uniqueness. The GNSS elevation datum (i.e., the Z-axis) is consistent with the terrain model (DEM). UTM (Universal Transverse Mercator) is essentially a globally universal map projection system. Its core is to divide the Earth's ellipsoid into fixed longitude zones and convert it into a plane through mathematical projection, solving the problem that distance and area calculations cannot be directly performed on a sphere. It is a fundamental coordinate standard in surveying, navigation, and remote sensing. After projecting the coordinates of multi-source data onto the same plane coordinate system, a fused data is obtained.

[0067] Time alignment refers to aligning the observation times between the first InSAR data and the first GNSS data. If the observation periods of the first InSAR data and the first GNSS data are different, the average deformation rate of the overlapping period is selected by time interpolation to ensure that the data correspond to the "same time window". If the observation periods of the first InSAR data and the first GNSS data are the same, there is no need to perform interpolation.

[0068] Data cleaning refers to removing low-quality data from the first InSAR and first GNSS data. For the first InSAR data, a first outlier removal method can be used. For example, the coherence coefficient and line-of-sight standard deviation of each first InSAR monitoring point are determined, and the first InSAR monitoring points with a coherence coefficient less than a first preset value or a line-of-sight standard deviation greater than a second preset value are deleted to obtain the second InSAR data.

[0069] Meanwhile, the 3σ criterion or box plot method can be used to remove monitoring points with abnormal GNSS data in the first GNSS monitoring points. Spatial isolated points (with no other valid points within a preset radius) can also be marked as high uncertainty points, and their weights can be reduced subsequently to obtain the second GNSS data.

[0070] In one possible embodiment, the plurality of monitoring points includes a plurality of first InSAR monitoring points and a plurality of first GNSS monitoring points; the step of constructing global three-dimensional deformation field data based on the second multi-source data includes: determining K adjacent first GNSS monitoring points for each first InSAR monitoring point; determining a first weight for the K first GNSS monitoring points; determining the target three-dimensional deformation rate and target three-dimensional error of the first InSAR monitoring points based on the first weight, thereby obtaining global three-dimensional deformation field data.

[0071] Specifically, as shown in Table 1, the acquired first InSAR data lacks three-dimensional parameters. After projecting the first InSAR monitoring point and the first GNSS monitoring point onto the same plane coordinate system, it is necessary to interpolate the first InSAR monitoring point to complete its three-dimensional parameters, thereby obtaining complete multi-source fusion data.

[0072] Based on this, the IDW algorithm (Inverse Distance Weighting) can be used to predict the target's three-dimensional deformation rate at i first InSAR monitoring points based on the first GNSS monitoring point, while simultaneously calculating the corresponding target three-dimensional error. For example, for each first InSAR monitoring point, the first weight, target three-dimensional deformation rate (Vx, Vy, Vz), and target three-dimensional error (σ_Vx, σ_Vy, σ_Vz) are calculated from the K nearest first GNSS monitoring points.

[0073] Determine the distance between each first InSAR monitoring point and its K neighboring first GNSS monitoring points. Assuming there are i first InSAR monitoring points, we can obtain i*K distance values. Then, calculate the first weight of each first InSAR monitoring point relative to the K first GNSS monitoring points based on the i*K distance values, that is, obtain i*K first weights, where the sum of the i*K first weights is equal to 1.

[0074] After determining i*K first weights, the K first weights of each first InSAR monitoring point are weighted and calculated with the first three-dimensional deformation rates of the corresponding K first GNSS monitoring points to obtain the target's three-dimensional deformation rate. Then, the target's three-dimensional error is calculated by weighting and calculating the first three-dimensional errors of the K first GNSS monitoring points.

[0075] The following section uses two first GNSS monitoring points and ten first InSAR monitoring points as examples to provide a detailed explanation of how to calculate the first weight, the target's three-dimensional deformation rate, and the target's three-dimensional error.

[0076] 1. Calculate the first weight of the first InSAR monitoring point;

[0077] First, the distances between the 10 first InSAR monitoring points and the 2 first GNSS monitoring points are determined. For each first InSAR monitoring point, two first weights are obtained, denoted as follows: and A total of 20 first weights can be obtained.

[0078] The following formula (2) can be obtained by using the Euclidean distance formula:

[0079]

[0080] in Let be the distance between the i-th first InSAR monitoring point and the j-th first GNSS monitoring point. The x-coordinate of the first InSAR monitoring point The ordinate of the first InSAR monitoring point. The x-coordinate of the first GNSS monitoring point. The ordinate of the first GNSS monitoring point is given.

[0081] Based on formula (2), each first InSAR monitoring point is respectively and the coordinates of the first GNSS monitoring point Substituting into the formula, the first distance from each first InSAR monitoring point to the first first GNSS monitoring point G1 can be calculated. .

[0082] Each first InSAR monitoring point and the coordinates of the second first GNSS monitoring point Substituting into the formula, the second distance from each first InSAR monitoring point to the second first GNSS monitoring point G2 can be calculated. .

[0083] Finally, the first weight is calculated based on the distance. The formula (3) for calculating the first weight is as follows:

[0084]

[0085] in, This is the distance attenuation coefficient, which is usually taken as 2; The first weight of the i-th first InSAR monitoring point relative to the j-th first GNSS monitoring point; The distance attenuation value from the i-th first InSAR monitoring point to the j-th first GNSS monitoring point; It is the sum of the reciprocals of all distance decay values;

[0086] Specifically, each first InSAR monitoring point can ultimately calculate K first weights. Taking 2 first GNSS monitoring points and 10 first InSAR monitoring points as an example, each first InSAR monitoring point can calculate a first distance. Second distance Based on the two distance values, two first weights can be calculated. and The first weight must satisfy: .

[0087] 2. Calculate the target's three-dimensional deformation rate at each first InSAR monitoring point;

[0088] Furthermore, based on the calculated first weight, the target three-dimensional deformation rate of each first InSAR monitoring point is calculated. The calculation formulas for the target three-dimensional deformation rate are as follows: (4-1), (4-2), and (4-3):

[0089] ;

[0090] ;

[0091] ;

[0092] in Let X be the deformation rate in the X direction of the i-th first InSAR monitoring point, which characterizes how fast the point deforms in the east direction (X direction) of the plane; Let be the deformation rate in the Y direction of the i-th InSAR monitoring point, which characterizes how fast the point deforms in the Y direction of the plane; Let be the deformation rate in the Z direction of the i-th InSAR monitoring point, which characterizes how fast the point deforms in the Z direction of the plane.

[0093] Let X be the deformation rate in the X direction of the i-th first GNSS monitoring point, which characterizes how fast the point deforms in the east direction (X direction) of the plane; Let be the deformation rate in the Y direction of the i-th first GNSS monitoring point, which characterizes how fast the point deforms in the eastward (Y direction) plane; Let be the deformation rate in the Z direction of the i-th first GNSS monitoring point, which characterizes how fast the point deforms in the eastward direction (Z direction) of the plane.

[0094] In one optional example, the X direction can be east or west, the Y direction can be northeast or south, and the Z direction is perpendicular to the horizontal plane.

[0095] Taking two first GNSS monitoring points and ten first InSAR monitoring points as examples, formulas (4-1), (4-2), and (4-3) can be transformed as follows:

[0096] ;

[0097] ;

[0098] ;

[0099] Substituting the corresponding data into the formula above, the result for each first InSAR monitoring point can be calculated. , and That is, the target three-dimensional deformation rate.

[0100] 3. Calculate the target three-dimensional error for each first InSAR monitoring point;

[0101] Based on the first 3D error of the first GNSS monitoring point and the calculated first weight, the target 3D error of each first InSAR monitoring point can be calculated by combining formulas (5-1), (5-2), and (5-3). Formulas (5-1), (5-2), and (5-3) are as follows:

[0102] ;

[0103] ;

[0104] ;

[0105] in, Let j be the first three-dimensional error of the first GNSS monitoring point. Let X be the error in the X direction of the i-th first InSAR monitoring point. Let be the error in the Y direction of the i-th first InSAR monitoring point. Let be the error in the Z direction of the i-th first InSAR monitoring point.

[0106] Taking two first GNSS monitoring points and ten first InSAR monitoring points as examples, formulas (5-1), (5-2), and (5-3) can be transformed as follows:

[0107] ;

[0108] ;

[0109] ;

[0110] Substituting the corresponding data into the formula above, the result for each first InSAR monitoring point can be calculated. , and That is, the target's three-dimensional error.

[0111] Specifically, taking the first InSAR monitoring point as an example, the corresponding first weight, target 3D deformation rate, and target 3D error are calculated. Table 1 shows that the coordinates of the first InSAR monitoring point are (X1, Y1) = (100, 200), and the coordinates of the first GNSS monitoring point G1 are... The coordinates of the second GNSS monitoring point G2 Substituting the coordinate values ​​into formula (2), we can obtain:

[0112] ;

[0113] ;

[0114] Then and Substitute them separately You can then obtain: and Then and Substituting into formula (3), we can obtain:

[0115] ;

[0116] ;

[0117] Finally and Substituting the corresponding first three-dimensional deformation rate into formulas (4-1), (4-2), and (4-3), or into formulas (4-4), (4-5), and (4-6), we can obtain:

[0118] ;

[0119] ;

[0120] ;

[0121] and will and By substituting the first three-dimensional error into formulas (5-1), (5-2), and (5-3), or into formulas (5-4), (5-5), and (5-6), we can obtain:

[0122] ;

[0123] ;

[0124] ;

[0125] It is understandable that the calculation methods for the first weight, target 3D deformation rate, and target 3D error of the 2nd to 9th first InSAR monitoring points are the same as those for the 1st first InSAR monitoring point, and will not be elaborated here. The above calculation process can be summarized into a table as shown in Table 2 below:

[0126] Table 2 Global 3D Deformation Field Data

[0127]

[0128] In one possible embodiment, the multi-source data further includes prior engineering structure data and terrain data; the step of constructing a first comprehensive feature vector for each monitoring point based on the second InSAR data, the second GNSS data, and the global three-dimensional deformation field data includes: determining the predicted line-of-sight rate and residual; constructing a second comprehensive feature vector for each first InSAR monitoring point based on the predicted line-of-sight rate, the residual, the global three-dimensional deformation field data, the prior engineering structure data, and the terrain data; and standardizing the second comprehensive feature vector to obtain a first comprehensive feature vector, wherein the first comprehensive feature vector includes the predicted line-of-sight rate, the residual, the target three-dimensional deformation rate, the target three-dimensional error, the terrain data, and the prior engineering structure data.

[0129] In practice, after refining the relevant parameters of each first InSAR monitoring point to obtain the global three-dimensional deformation field data, it is necessary to construct the input parameters required for the subsequent partitioning model based on the global three-dimensional deformation field data, so as to provide data support for the partitioning identification function of the partitioning model.

[0130] Specifically, the predicted line-of-sight rate and residual are first determined. These predicted line-of-sight rate and residual are predicted values, calculated based on the line-of-sight unit vector, the target's three-dimensional deformation rate, and the target's three-dimensional error. The calculation formulas (6) for the predicted line-of-sight rate and (7) for the residual are as follows:

[0131] ;

[0132] ;

[0133] in, It is the observation value of the i-th first InSAR monitoring point, i.e., the LOS in Table 1; while the predicted line-of-sight rate Then you only need to separately , and Substituting the calculated line-of-sight unit vector into formula (6), the predicted line-of-sight rate corresponding to the i-th first InSAR monitoring point can be obtained. .

[0134] Taking the first InSAR monitoring point as an example, substitute Vx1=0.0967, Vy1=0.1967, Vz1=-1.1397 into formula (6) and the line-of-sight unit vector. By performing calculations, the predicted line-of-sight rate corresponding to the first InSAR monitoring point can be obtained:

[0135] =-0.0967*sin30°sin45°+0.1967*sin30°cos45°-1.1397*cos30°≈-0.95;

[0136] The first InSAR monitoring point corresponds to The value is LOS = -1.1 mm / yr. Substituting this value and the predicted line-of-sight rate corresponding to the first InSAR monitoring point into formula (7), the residual of the predicted line-of-sight rate corresponding to the first InSAR monitoring point can be obtained. =-1.1+0.95=-0.15mm / yr.

[0137] It is understandable that the calculation method for the predicted line-of-sight rate of the 2nd to 9th first InSAR monitoring points is the same as that of the 1st first InSAR monitoring point, and will not be elaborated here. The above calculation process can be summarized into a table as shown in Table 3 below:

[0138] Table 3

[0139]

[0140] Thus, based on all currently known parameters, a first comprehensive feature vector is constructed for each first InSAR monitoring point, resulting in a 12-dimensional feature vector set for each first InSAR monitoring point. :

[0141]

[0142] in, For the first Elevation of the first InSAR monitoring point For the first The slope of the first InSAR monitoring point Slope direction, For the first The first InSAR monitoring point to the dam axis (e.g.) Figure 5 The distance shown is [distance]. For the first The distance from the first InSAR monitoring point to critical structures (such as crack zones), Code the engineering structural area (e.g., dam foundation = 1, dam abutment = 2).

[0143] Specifically, the slope aspect of each first InSAR monitoring point is determined by the engineering structure zone it belongs to. The slope aspect of the first InSAR monitoring point belonging to the dam foundation is 0°, the slope aspect of the first InSAR monitoring point belonging to the right abutment is 90°, and the slope aspect of the first InSAR monitoring point belonging to the transition zone (i.e., the transition zone between the abutment and the dam foundation) is 45°. Based on this, it can be known that the first InSAR monitoring points 1-3, 8, and 10 belong to the dam foundation, and therefore these first InSAR monitoring points... =0, =1; the first InSAR monitoring points 4-6 and 9 belong to the right abutment, therefore these first InSAR monitoring points =1, =0; the first InSAR monitoring point 7 belongs to the transition zone, then the first InSAR monitoring point 7's =0.707, =0.707.

[0144] The engineering structural zone of the dam foundation is coded as 1, the engineering structural zone of the right abutment is coded as 2, and the engineering structural zone of the transition zone is coded as 3.

[0145] After constructing the first comprehensive feature vector of each first InSAR monitoring point Next, each first comprehensive feature vector is standardized. A robust standardization method can be used, such as subtracting the median and dividing by the interquartile range (IQR), to avoid interference from extreme values; for distance-type features (… and Logarithmic or truncation processing is performed to suppress the influence of long-tailed distributions. Since this method is existing technology, it will not be elaborated upon here. and Taking the data as an example, we can obtain the following Table 4:

[0146] Table 4

[0147]

[0148] In one possible embodiment, the step of inputting the first integrated feature vector into the partitioning model to obtain the target partition corresponding to each monitoring point includes: determining the second weight corresponding to each first InSAR monitoring point, and normalizing the second weight to obtain the third weight; determining the initial parameters of the partitioning model; inputting the first integrated feature vector into the partitioning model, and iterating the partitioning model according to the first integrated feature vector, the third weight and the initial parameters to obtain the target partition.

[0149] In practical implementation, the partitioning model can adopt the GMM model. Based on the GMM model, a partitioning model specifically for partition identification can be constructed. The second weight is the initial weight of each first InSAR monitoring point in the partitioning model. The second weight is calculated first; it is a raw value obtained directly through a formula and represents a raw comprehensive score of the data reliability and observation uncertainty for each point. However, the calculated second weight is not normalized, and its numerical range is not uniformly limited. It ultimately needs to be normalized to 0.1-1. The InSAR data quality dimension is... (Higher coherence and smaller LOS observation error result in a higher score), the GNSS interpolation uncertainty dimension is... (The smaller the 3D interpolation error, the higher the score.) It is obtained by multiplying the two core dimensions by weight, reflecting the quality advantage of the point. As the original material for the point weight, it intuitively reflects the reliability difference between different points. The larger the second weight, the more reliable the point.

[0150] Specifically, the second InSAR data includes the coherence coefficient and first error for each first InSAR monitoring point; determining the second weight corresponding to each first InSAR monitoring point and normalizing the second weight to obtain the third weight includes: determining the target three-dimensional variance of each first InSAR monitoring point based on the target three-dimensional error of each first InSAR monitoring point; and determining the second weight of each first InSAR monitoring point based on the coherence coefficient, the first error, and the target three-dimensional variance. .

[0151] The formula (9) for calculating the second weight is as follows:

[0152]

[0153] Where β=1, c=0.1, The error of the i-th first InSAR monitoring point (as shown in Table 1 for σ_LOS corresponding to each first InSAR monitoring point), the first variance The error σ_LOS is calculated based on the square of the error in Table 1. The sum of the target's three-dimensional variances is obtained by squared and summed the target's three-dimensional errors from Table 2. The calculation for the first InSAR monitoring point is illustrated below:

[0154] Three-dimensional sum of squared errors:

[0155] InSAR data quality dimensions:

[0156] GNSS interpolation uncertainty dimension:

[0157] Second weight:

[0158] Finally, after normalizing all the second weights, we can obtain the following table:

[0159] Table 5

[0160]

[0161] Furthermore, determining the initial parameters of the partitioning model includes: acquiring prior data of the engineering structure; determining multiple basic partitions based on the structural regions in the prior data of the engineering structure; dividing the multiple first InSAR monitoring points into the multiple basic partitions to obtain multiple preset groups; determining the mean and covariance of the multiple preset groups; and determining the proportion of the first InSAR monitoring points in each preset group to obtain a fourth weight.

[0162] In practical implementation, for the zoning model, the number of zoning (i.e., multiple basic zoning) needs to be determined first before multiple first InSAR monitoring points can be assigned to the corresponding zoning. Assuming the number of zoning is Q, the number of zoning Q is initialized by combining engineering structure priors, preliminary data feature analysis, and anomaly pre-identification, i.e., its initial value is determined.

[0163] The first comprehensive feature vector is calculated based on the data in Table 1, which shows the clearly defined structural engineering zones of the dam body itself: the dam foundation, the right abutment, and the transition zone. This is an inherent structural division of the hydraulic engineering and belongs to the domain's prior knowledge. When initializing the partitioned model, these structural engineering zones can be directly aligned, allowing the model to start training from a point that conforms to engineering logic, avoiding the partitioning caused by random initialization that is out of touch with the actual engineering meaning.

[0164] Specifically, the structural engineering area of ​​the dam foundation is numbered 1, therefore it is divided into partition 1 (dam foundation dominant). Based on partition 1, the data is grouped to obtain preset group 1: containing the first InSAR monitoring points 1-3, 8 and 10, with initial mean μ1 = the feature-weighted mean of this group, covariance Σ1 = the weighted covariance of this group, and the fourth weight of preset partition 1. =5 / 10=0.5.

[0165] The structural engineering area of ​​the right abutment is numbered 2, therefore it is divided into partition 2 (right abutment dominant). Based on partition 2, the data is grouped to obtain preset group 2: including the first InSAR monitoring points 4-6 and 9, the initial mean μ2 = the weighted mean of the features of this group, the covariance Σ2 = the weighted covariance of the features of this group, and the fourth weight of the preset partition 2. =4 / 10=0.4.

[0166] The structural engineering area of ​​the transition zone is numbered 3, therefore it is divided into partition 3. Based on partition 3, the data is grouped to obtain preset group 3 (potential anomaly): containing the first InSAR monitoring point 7 (with the largest residual), the initial mean μ3 = the feature of the first InSAR monitoring point 7, the covariance Σ3 = the diagonal matrix (feature variance + 0.01), and the fourth weight of preset partition 3. =1 / 10=0.1.

[0167] In summary, in this embodiment, the number of partitions Q can be determined as 3 basic partitions. Multiple first InSAR monitoring points are assigned to these 3 basic partitions, resulting in 3 preset groups. Simultaneously, based on the number of first InSAR monitoring points contained in each group, an initial fourth weight is assigned. For example, if preset group 1 contains 5 first InSAR monitoring points, its fourth weight is 0.5; if preset group 1 contains 5 first InSAR monitoring points, its fourth weight is 0.4; if preset group 1 contains 5 first InSAR monitoring points, its fourth weight is 0.1. The fourth weight must satisfy the following conditions: .

[0168] Optionally, if the standard deviation of deformation features within a preset group exceeds a threshold, the centers are further subdivided through local clustering (such as 2-point K-means). The covariance of each preset group is the weighted covariance of the first InSAR monitoring point within each preset group. If the sample size is small, a regularization term εI (ε=0.001) is added to the diagonal matrix (feature variance) to avoid ambiguity.

[0169] Taking the data in Table 4 as an example, after parameter initialization, the initial parameters of the partitioning model are shown in Table 6:

[0170] Table 6

[0171]

[0172] As can be seen, this embodiment initializes the parameters of the partitioning model using prior data and the first comprehensive feature vector, providing data support for the subsequent partitioning recognition function.

[0173] Furthermore, the step of iterating the partitioning model based on the first comprehensive feature vector, the third weight, and the initial parameters to obtain the target partition includes: determining a first probability that each first InSAR monitoring point belongs to one of the multiple basic partitions based on the first comprehensive feature vector, the third weight, the mean, the covariance, and the fourth weight; determining a first label based on the first probability, wherein the first label is used to indicate the basic partition to which the corresponding first InSAR monitoring point belongs; determining the number of effective samples; updating the mean, the covariance, and the fourth weight based on the number of effective samples to obtain a new mean, a new covariance, and a new fourth weight; determining an objective function based on the new mean, the new covariance, and the new fourth weight; repeating the above steps until the rate of change of the objective function is less than a preset value or equal to zero, then stopping the iteration and outputting the current target partition.

[0174] In the specific implementation, the first comprehensive feature vector generated earlier is input into the initialized partition model, and the partition model calculates the first probability that each first InSAR monitoring point belongs to one of the three groups according to formula (10).

[0175]

[0176] in, Let be the multivariate Gaussian likelihood function, representing the eigenvectors. In the mean The probability density under covariance Σ is calculated as follows: , where D is the feature dimension, and the denominator is the sum of probabilities, used for normalization to ensure that the sum of probabilities is 1.

[0177] Because the specific calculation process is very complicated and involves a large number of matrix operations, it is usually done by programming. Here, we take partition 1 as an example to briefly show the calculation idea:

[0178] After the first comprehensive feature vectors of multiple first InSAR monitoring points are input into the partitioning model as input parameters, since the first comprehensive feature vector includes 12-dimensional features, it is necessary to calculate the 12-dimensional covariance value of each first comprehensive feature vector, and then form a covariance diagonal matrix.

[0179] In a partitioning model, the covariance matrix describes the dispersion of sample features across dimensions within a corresponding partition, as well as the linear correlation between these dimensions. During the initialization phase, to simplify calculations and conform to engineering logic (assuming each feature dimension is independent and has no significant linear correlation), this embodiment sets the covariance matrix as a diagonal matrix (all off-diagonal elements are 0), retaining only the variance values ​​of the diagonal elements, thus obtaining the diagonal covariance matrix.

[0180] To calculate the covariance diagonal matrix of partition 1, we first need to determine the sample set of partition 1. Partition 1 includes the first InSAR monitoring points 1, 2, 3, 8, and 10 (5 points in the dam foundation area). We extract the standardized 12-dimensional features of these first InSAR monitoring points. For the d-th dimension feature, the variance of the samples within the partition is calculated using the following formula: , where n=5 (number of samples in partition 1). It is the mean of the d-th feature within partition 1. For the d-th eigenvalue of the i-th sample, the variance values ​​calculated for each dimension are filled into the diagonal matrix in order, and the off-diagonal elements are set to 0, finally obtaining the covariance matrix of partition 1. (12-dimensional diagonal matrix), similarly calculate the covariance matrix of partitions 2 and 3.

[0181] Calculate the 12-dimensional mean for partitions 1, 2, and 3 respectively. , , (Calculate the mean of each dimension of the 12-dimensional features for all first InSAR monitoring points in each partition). Calculate the observation covariance of the first InSAR monitoring point. Calculate the total covariance for each partition. Taking partition 1 as an example, the total covariance is a diagonal matrix, and the variances of each dimension are summed. Similarly, the total covariance of other partitions is calculated, which will not be elaborated here.

[0182] Calculate the Gaussian likelihood of each partition. The likelihood of the three partitions was calculated as follows: , and Then calculate the numerator. Then, the summation is performed to obtain the denominator, followed by normalization to obtain the posterior probability γ_ik for each partition. Finally, the temporary label is determined based on the magnitude of the posterior probability. After all the data has been calculated, the first probability of each first InSAR monitoring point for each partition is compared, and the partition with the highest probability for each first InSAR monitoring point is determined as its target partition. The first label corresponding to the partition is then configured, resulting in the data shown in Table 7.

[0183] Table 7

[0184]

[0185] After calculating the first probability, the current number of valid samples is determined based on the first probability and the second weight. The formula for calculating the number of valid samples (11) is as follows:

[0186] ;

[0187] Then, a new fourth weight is calculated based on the number of valid samples. The formula for calculating the new fourth weight (12) is as follows:

[0188] ;

[0189] Then, based on the number of effective samples, the first probability, the second weight, and the first comprehensive feature vector, a new mean is calculated. The formula for calculating the new mean (13) is as follows:

[0190] ;

[0191] Finally, the new covariance is obtained by updating it using formula (14), where ε = 0.001:

[0192] ;

[0193] Based on the characteristics of the GMM model, it is known that after generating the first probability, it will also generate an objective function. This objective function is the weighted negative log-likelihood function of the GMM model, which is essentially a score of the model's fit to the data. The smaller the value, the better the model parameters fit the actual data, and the better the iterative optimization effect. The calculation formula of the objective function (15) is as follows:

[0194]

[0195] Since the parameters in formula (15) are known, the calculation process will not be described in detail. It can be understood that the convergence judgment is to repeat the E step and M step until "the parameter change norm is <0.5%", "the objective function L decreases by <1% for 5 consecutive rounds" or the maximum number of iterations is reached (which can be set according to the actual situation, generally 100~200 rounds).

[0196] The data obtained from the above calculations can be organized to obtain the contents shown in Table 8:

[0197] Table 8

[0198]

[0199] Based on the parameters in Table 8, the first comprehensive feature vector is iterated until the rate of change of the objective function is less than the preset value or equal to zero. Then the iteration stops and the current target partition is output.

[0200] The data obtained from the iteration process are shown in Table 9. The specific calculation process will not be described in detail here. Table 9 is as follows:

[0201] Table 9

[0202]

[0203] As shown in Table 9, the mean change rate after iteration 3 is <0.5%, and the objective function no longer changes after the third iteration, indicating that it has reached the best fit. Therefore, the corresponding target partition can be output.

[0204] In one possible embodiment, the method further includes: after B iterations, performing a partition rationality check to obtain target iteration data; the rationality check includes splitting check, merging check, outlier handling, and outlier partition handling; generating partitioning results based on the target partitions and the corresponding target iteration data; the partitioning results include the partition label of each first InSAR monitoring point and the probability of belonging to each partition; statistically analyzing the deformation characteristics of each partition to generate a partitioning interpretation report, the partitioning interpretation report including residual variance reduction rate, spatial coherence, structural region consistency, and outlier boundary proportion.

[0205] Specifically, after every B iterations, the rationality of the partitioning is checked. If the covariance principal eigenvalue of the preset partition Q is too large, the residual r_cons exhibits a bimodal distribution, or the BIC (Bayesian Information Criterion, i.e., objective function) decreases after splitting, then the partition is subdivided using a 2-component GMM, and the number of partitions Q increases by 1. If the mean distance between two partitions Q1 and Q2 is less than the threshold and the BIC decreases after merging, then they are merged into one partition, and the number of partitions Q decreases by 1. For example, if the covariance principal eigenvalues ​​of the three partitions are all <0.02 (internal compactness), and the residual distribution does not exhibit a bimodal distribution, then splitting is unnecessary. Another example is that the μ_Vz difference between partitions 1 and 2 (0.783 mm / yr) is >0.5 mm / yr, and the BIC increases after merging (from 75 to 89), then merging is unnecessary.

[0206] In one possible implementation, points that have a likelihood of <0.1 for all partitions and whose measurement errors cannot be explained are marked as outliers and their weight in model updates is reduced; for example, the first InSAR monitoring point 7 has the lowest likelihood for all partitions (residual -1.68 mm / yr) and is retained as partition 3; the first InSAR monitoring point 10 has a weight of 0.1 and is not marked as an outlier separately.

[0207] If outliers form a spatially connected cluster (area > preset threshold), a separate outlier partition is constructed, with its mean being the weighted mean of points within the cluster and its covariance being the weighted covariance within the cluster.

[0208] In one possible implementation, the first probability is only spatially smoothed. For example, E-step correction. When this is done, neighbor weights are introduced (those less than 20m are considered neighbors). =0.5), for example, point 10 (neighboring points 1 and 8 are both in partition 1). The value increased from 0.88 to 0.93, while γ2 decreased to 0.06. and The correction method and The same applies, so I won't go into details here.

[0209] In one possible embodiment, partitioning results are generated based on the target partition and the corresponding target iteration data. The partitioning results include the partition label of each first InSAR monitoring point and the probability of belonging to each partition; the deformation characteristics of each partition are statistically analyzed, and a partitioning interpretation report is generated. The partitioning results are shown in Table 10.

[0210] Table 10

[0211]

[0212] in, and The first The probability that a first InSAR monitoring point belongs to preset partition 1, preset partition 2, or preset partition 3 is shown in the table. For example, monitoring point 1 (i.e., the first InSAR monitoring point) belongs to... and The probabilities are 0.99, 0.01, and 0.00, respectively.

[0213] At the same time, the data table shown in Table 11 is generated:

[0214] Table 11

[0215]

[0216] Among them, zone 1 is a slow settlement zone, zone 2 is a stable zone, and zone 3 is an abnormal zone; These are the core deformation features corresponding to preset partition 1, preset partition 2, and preset partition 3, respectively. These are the residual statistics corresponding to preset partition 1, preset partition 2, and preset partition 3, respectively.

[0217] Specifically, determine the final partition label for each point (take...). The largest Q) and the partition probability vector ( ); and determine the statistical characteristics of each partition: average three-dimensional velocity, residual mean and variance, covariance principal axis (principal deformation direction), spatial range (minimum bounding rectangle), and point percentage; and determine the list of abnormal partitions / abnormal points: abnormal point coordinates, deformation characteristics and spatial location of abnormal partitions;

[0218] Furthermore, the quality of the final partitioned data is evaluated:

[0219] The rate of decrease in residual variance is determined; a higher rate indicates stronger explanatory power for the partition. Specifically, the rate of decrease in residual variance is calculated using formula (16):

[0220]

[0221] Determine the spatial coherence of the data; a higher coherence indicates a more complete partition and less fragmentation. Specifically, the spatial coherence of the data is determined using formula (17):

[0222]

[0223] Determine the consistency between the target partition and the engineering structure area (a higher value indicates a higher logical match between the partition and the engineering structure). Specifically, the consistency between the target partition and the engineering structure area is determined using formula (18):

[0224]

[0225] It can be seen that the technical solution in this application has the following technical effects:

[0226] Upgraded Multi-Source Fusion Mechanism: Breaking through the traditional "simple data stitching" model, a LOS rate prediction model is constructed based on satellite geometric parameters, and the residual is used to predict the rate. The consistency between InSAR and GNSS is quantified, and "range-error dual-factor weighted IDW interpolation" is adopted to achieve accurate extension of sparse GNSS to the global three-dimensional deformation field, thus solving the problem of "mismatch between one-dimensional observation and three-dimensional deformation".

[0227] Feature system dimension expansion: Abandoning the single deformation index, we construct an 11-dimensional feature vector containing "deformation (LOS / 3D rate / residual) + terrain (elevation / slope / aspect) + spatial location (distance to dam axis) + engineering coding", and use "median-IQR" robust standardization to avoid extreme value interference and make the features more in line with actual influencing factors.

[0228] Model robustness design: The first "measurement uncertainty weighted GMM" is proposed, which calculates the weight of points by coherence coefficient and observation error to suppress interference from low-quality data; the addition of dynamic partition adjustment (splitting if the dispersion is too large and merging if the difference is too small) and spatial smoothing constraints solves the problems of "fixed number of partitions not being suitable for complex deformation" and "partition fragmentation".

[0229] Engineering adaptability enhancement: The model is initialized with prior knowledge of the engineering structural zones (number of zones = number of structural zones + potential anomaly zones) to ensure that the starting point aligns with the mechanical logic of the dam body; when outputting the zones, the covariance analysis of the principal deformation direction is combined to achieve the integration of "data-driven" and "engineering cognition" and avoid the disconnect between the results and reality.

[0230] Significantly improved accuracy: After multi-source fusion, the 3D deformation error is reduced from 1.2 mm / yr of single InSAR to 0.4 mm / yr (a reduction of 67%), the residual interpretation rate exceeds 90%, and it can accurately distinguish between real deformation and measurement error.

[0231] Enhanced robustness: The impact of low-quality points on partitions is reduced by 70%, and the consistency of partitions with repeated calculations is improved from 65% to 92%, demonstrating outstanding anti-interference capabilities.

[0232] High practicality: Spatial coherence reaches over 90%, eliminating isolated small zones; structural matching degree exceeds 85%, and zone names (such as "dam foundation stability zone") can directly serve engineering evaluation, reducing decision-making conversion costs.

[0233] Precise decision support: By quantifying the uncertainty of the boundary through posterior probability, the accuracy of anomaly area identification is improved by 50%, enabling rapid location of high-risk areas and providing clear targets for emergency response.

[0234] The above primarily describes the solutions of the embodiments of this application from the perspective of the method execution process. It is understood that, in order to achieve the above functions, mobile electronic devices include corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments provided herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0235] This application embodiment can divide the electronic device into functional units according to the above method example. For example, each function can be divided into a separate functional unit, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this application embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0236] Please see Figure 6 This application also provides a dam deformation zoning identification device, characterized in that it includes:

[0237] An acquisition unit is used to acquire multi-source data, the multi-source data including first InSAR data and first GNSS data;

[0238] A first processing unit is configured to perform data quality processing on the first InSAR data and the first GNSS data to obtain second InSAR data and second GNSS data; the data quality processing includes one or more of data alignment, data cleaning, and assignment.

[0239] A construction unit is configured to construct global three-dimensional deformation field data based on the second InSAR data and the second GNSS data, wherein the global three-dimensional deformation field includes multiple monitoring points; and to construct a first comprehensive feature vector for each monitoring point based on the second InSAR data, the second GNSS data and the global three-dimensional deformation field data.

[0240] The determining unit is used to input the first comprehensive feature vector into the partitioning model to obtain the target partition corresponding to each monitoring point.

[0241] As can be seen, this application first acquires multi-source data, including first InSAR data and first GNSS data; then, it performs data quality processing on the first InSAR data and the first GNSS data to obtain second InSAR data and second GNSS data; the data quality processing includes one or more of data alignment, data cleaning, and assignment; based on the second InSAR data and the second GNSS data, it constructs global three-dimensional deformation field data, which includes multiple monitoring points; based on the second InSAR data, the second GNSS data, and the global three-dimensional deformation field data, it constructs a first comprehensive feature vector for each monitoring point; and it inputs the first comprehensive feature vector into a partitioning model to obtain the target partition corresponding to each monitoring point. In this way, the fusion of multi-source data enables probabilistic output of partitioning of the dam and surrounding areas, providing accurate support for engineering safety assessment and risk identification.

[0242] In one possible embodiment, the plurality of monitoring points include a first InSAR monitoring point and a first GNSS monitoring point. The first InSAR data includes the InSAR pixel coordinates of the first InSAR monitoring point, and the first GNSS data includes the GNSS station coordinates of the first GNSS monitoring point. Regarding the aspect of performing data quality processing on the first InSAR data and the first GNSS data to obtain second InSAR data and second GNSS data, the first processing unit is specifically configured to: project the InSAR pixel coordinates and the GNSS station coordinates onto the same plane coordinate system; determine a first observation time of the first InSAR data and a second observation time of the first GNSS data; if the first observation time and the second observation time are different, determine the average deformation rate of the overlapping period of the first InSAR data and the first GNSS data; perform time interpolation on the first InSAR data or GNSS data according to the average deformation rate to time-align the first InSAR data and the first GNSS data; remove monitoring points with abnormal InSAR data according to a first outlier removal method to obtain the second InSAR data; and remove monitoring points with abnormal GNSS data according to a second outlier removal method to obtain the second GNSS data.

[0243] In one possible embodiment, the plurality of monitoring points includes a plurality of first InSAR monitoring points and a plurality of first GNSS monitoring points; in the aspect of constructing global three-dimensional deformation field data based on the second InSAR data and the second GNSS data, the construction unit is specifically used to: determine K adjacent first GNSS monitoring points for each first InSAR monitoring point; determine a first weight for the K first GNSS monitoring points; and determine the target three-dimensional deformation rate and target three-dimensional error of the first InSAR monitoring points based on the first weights to obtain global three-dimensional deformation field data.

[0244] In one possible embodiment, the multi-source data further includes prior engineering structure data and terrain data; the aspect of constructing a first comprehensive feature vector for each monitoring point based on the second InSAR data, the second GNSS data, and the global three-dimensional deformation field data, wherein the construction unit is specifically used for: determining the predicted line-of-sight rate and residual; constructing a second comprehensive feature vector for each first InSAR monitoring point based on the predicted line-of-sight rate, the residual, the global three-dimensional deformation field data, the prior engineering structure data, and the terrain data; and standardizing the second comprehensive feature vector to obtain a first comprehensive feature vector, wherein the first comprehensive feature vector includes the predicted line-of-sight rate, the residual, the target three-dimensional deformation rate, the target three-dimensional error, the terrain data, and the prior engineering structure data.

[0245] In one possible embodiment, the step of inputting the first comprehensive feature vector into the partitioning model to obtain the target partition corresponding to each monitoring point is specifically configured to: determine the target three-dimensional variance of each first InSAR monitoring point based on the target three-dimensional error of each first InSAR monitoring point; determine the second weight of each first InSAR monitoring point based on the coherence coefficient, the first error, and the target three-dimensional variance of each first InSAR monitoring point; normalize the second weight to obtain a third weight; determine the initial parameters of the partitioning model; input the first comprehensive feature vector into the partitioning model; and iterate the partitioning model based on the first comprehensive feature vector, the third weight, and the initial parameters to obtain the target partition.

[0246] In one possible embodiment, the determining unit is specifically used for: acquiring prior data of the engineering structure; determining multiple basic partitions based on the structural regions in the prior data of the engineering structure; dividing the multiple first InSAR monitoring points into the multiple basic partitions to obtain multiple preset groups; determining the mean and covariance of the multiple preset groups; and determining the proportion of the first InSAR monitoring points in each preset group to obtain a fourth weight.

[0247] In one possible embodiment, the determination unit is specifically used to: determine a first probability that each first InSAR monitoring point belongs to the plurality of basic partitions based on the first comprehensive feature vector, the third weight, and the initial parameters to obtain aspects of the target partition, wherein the determination unit is used to: determine a first probability that each first InSAR monitoring point belongs to the plurality of basic partitions based on the first comprehensive feature vector, the third weight, the mean, the covariance, and the fourth weight; determine a first label based on the first probability, wherein the first label is used to indicate the basic partition to which the corresponding first InSAR monitoring point belongs; determine the number of effective samples; update the mean, the covariance, and the fourth weight based on the number of effective samples to obtain a new mean, a new covariance, and a new fourth weight; determine an objective function based on the new mean, the new covariance, and the new fourth weight; repeat the above steps until the rate of change of the objective function is less than a preset value or equal to zero, then stop the iteration and output the current target partition.

[0248] In one possible embodiment, please refer to Figure 7 The dam deformation zoning identification device further includes:

[0249] The checking unit is used to perform partition rationality checks after B iterations to obtain the target iteration data; the rationality checks include split checks, merge checks, outlier handling, and abnormal partition handling;

[0250] The generation unit is used to generate partitioning results based on the target partition and the corresponding target iteration data. The partitioning results include the partitioning label of each first InSAR monitoring point and the probability of belonging to each partition. The unit also calculates the deformation characteristics of each partition and generates a partitioning interpretation report, which includes the residual variance reduction rate, spatial coherence, structural region consistency, and the proportion of anomalous boundaries.

[0251] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0252] This application also provides an electronic device 10, such as... Figure 8 As shown, it includes at least one processor 11, a display screen 12, and a memory 13, and may also include a communications interface 15 and a bus 14. The processor 11, display screen 12, memory 13, and communications interface 15 can communicate with each other via the bus 14. The display screen 12 is configured to display a preset user guide interface in the initial setup mode. The communications interface 15 can transmit information. The processor 11 can invoke logical instructions in the memory 13 to execute the methods described in the above embodiments.

[0253] Optionally, the electronic device 10 may be a mobile electronic device, an electronic device or other device, and is not limited to a single type.

[0254] Furthermore, the logic instructions in the aforementioned memory 13 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0255] The memory 13, as a computer-readable storage medium, can be configured to store software programs, computer-executable programs, such as program instructions or modules corresponding to the methods in the embodiments of this disclosure. The processor 11 executes functional applications and data processing by running the software programs, instructions, or modules stored in the memory 13, thereby implementing the methods in the above embodiments.

[0256] The memory 13 may include a program storage area and a data storage area. The program storage area may store the operating system and application programs required for at least one function; the data storage area may store data created based on the use of the electronic device 10. Furthermore, the memory 13 may include high-speed random access memory (RAM) and may also include non-volatile memory. For example, various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, may be used, or they may be transient storage media.

[0257] This application also provides a computer storage medium storing a computer program for electronic data interchange, which causes a computer to perform some or all of the steps of any of the methods described in the above method embodiments, wherein the computer includes an electronic device.

[0258] This application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps of any of the methods described in the above method embodiments. The computer program product may be a software installation package, and the computer may include an electronic device.

[0259] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0260] In the several embodiments provided in this application, it should be understood that the disclosed methods, apparatuses, and systems can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for example, the division of units is merely a logical functional division, and other division methods may exist in actual implementation; for example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0261] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0262] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can be physically comprised separately, or two or more units can be integrated into one unit. The integrated unit described above can be implemented in hardware or in the form of hardware plus software functional units.

[0263] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute some steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, volatile memory, or non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM), etc., which are various media capable of storing program code.

[0264] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can easily conceive of variations or substitutions without departing from the spirit and scope of the present invention, and various modifications and alterations can be made, including combinations of the different functions and implementation steps described above, as well as software and hardware implementation methods, all of which are within the protection scope of the present invention.

Claims

1. A method for identifying dam deformation zones, characterized in that, include: Acquire multi-source data, which includes first InSAR data and first GNSS data, wherein the first InSAR data corresponds to multiple first InSAR monitoring points and the first GNSS data corresponds to multiple first GNSS monitoring points; The first InSAR data and the first GNSS data are subjected to data quality processing to obtain second InSAR data and second GNSS data; the data quality processing includes one or more of data alignment, data cleaning and assignment. Constructing global three-dimensional deformation field data based on the second InSAR data and the second GNSS data includes: determining K adjacent first GNSS monitoring points for each first InSAR monitoring point, determining the first weight of the K first GNSS monitoring points, determining the target three-dimensional deformation rate and target three-dimensional error of the first InSAR monitoring point based on the first weight, and obtaining global three-dimensional deformation field data containing multiple monitoring points. A first comprehensive feature vector is constructed for each monitoring point based on the second InSAR data, the second GNSS data, and the global three-dimensional deformation field data. The process of inputting the first comprehensive feature vector into the partitioning model to obtain the target partition corresponding to each monitoring point includes: determining the target three-dimensional variance of each first InSAR monitoring point based on the target three-dimensional error of each first InSAR monitoring point; determining the second weight of each first InSAR monitoring point based on the coherence coefficient, the first error, and the target three-dimensional variance; normalizing the second weight to obtain the third weight; determining the initial parameters of the partitioning model; inputting the first comprehensive feature vector into the partitioning model; iterating the partitioning model based on the first comprehensive feature vector, the third weight, and the initial parameters to obtain the target partition; the partitioning model is based on a Gaussian mixture model, and a Gaussian mixture model specifically designed for partition identification with measurement uncertainty weighting is constructed.

2. The method according to claim 1, characterized in that, The plurality of monitoring points include a first InSAR monitoring point and a first GNSS monitoring point. The first InSAR data includes the InSAR pixel coordinates of the first InSAR monitoring point, and the first GNSS data includes the GNSS station coordinates of the first GNSS monitoring point. The step of performing data quality processing on the first InSAR data and the first GNSS data to obtain second InSAR data and second GNSS data includes: Project the InSAR pixel coordinates and the GNSS station coordinates onto the same plane coordinate system; Determine the first observation time of the first InSAR data and the second observation time of the first GNSS data; If the first observation time is different from the second observation time, then the average deformation rate of the overlapping period of the first InSAR data and the first GNSS data is determined; The first InSAR data or GNSS data is time-interpolated according to the average deformation rate to time-align the first InSAR data with the first GNSS data. The monitoring points with abnormal InSAR data are removed according to the first outlier removal method to obtain the second InSAR data; The second GNSS data is obtained by removing monitoring points with abnormal GNSS data using the second outlier removal method.

3. The method according to claim 2, characterized in that, The multiple monitoring points include multiple first InSAR monitoring points and multiple first GNSS monitoring points; The construction of global three-dimensional deformation field data based on the second InSAR data and the second GNSS data includes: Identify the K adjacent first GNSS monitoring points for each first InSAR monitoring point; Determine the first weights of the K first GNSS monitoring points; Based on the first weight, the target three-dimensional deformation rate and target three-dimensional error of the first InSAR monitoring point are determined, and the global three-dimensional deformation field data are obtained.

4. The method according to claim 3, characterized in that, The multi-source data also includes prior engineering structure data and terrain data; The step of constructing a first comprehensive feature vector for each monitoring point based on the second InSAR data, the second GNSS data, and the global three-dimensional deformation field data includes: Determine the predicted line-of-sight rate and residual; A second comprehensive feature vector is constructed for each first InSAR monitoring point based on the predicted line-of-sight rate, the residual, the global three-dimensional deformation field data, the prior data of the engineering structure, and the terrain data. The second comprehensive feature vector is standardized to obtain the first comprehensive feature vector, which includes the predicted line-of-sight rate, the residual, the target three-dimensional deformation rate, the target three-dimensional error, terrain data, and the prior data of the engineering structure.

5. The method according to claim 1, characterized in that, Determining the initial parameters of the partitioning model includes: Obtain prior data of the engineering structure; Multiple basic zones are determined based on the structural zones in the prior data of the engineering structure. The multiple first InSAR monitoring points are divided into multiple basic partitions to obtain multiple preset groups; Determine the mean and covariance of the multiple preset groups; The proportion of the first InSAR monitoring point in each preset group is determined to obtain the fourth weight.

6. The method according to claim 5, characterized in that, The step of iterating the partitioning model based on the first comprehensive feature vector, the third weight, and the initial parameters to obtain the target partition includes: Based on the first comprehensive feature vector, the third weight, the mean, the covariance, and the fourth weight, the first probability that each first InSAR monitoring point belongs to the plurality of basic partitions is determined; A first label is determined based on the first probability, and the first label is used to indicate the basic partition to which the corresponding first InSAR monitoring point belongs. Determine the number of valid samples; The mean, covariance, and fourth weight are updated based on the number of valid samples to obtain a new mean, a new covariance, and a new fourth weight. The objective function is determined based on the new mean, the new covariance, and the new fourth weight. Repeat the above steps until the rate of change of the objective function is less than the preset value or equal to zero, then stop the iteration and output the current target partition.

7. The method according to any one of claims 1-6, characterized in that, Also includes: After B iterations, a partition rationality check is performed to obtain the target iteration data; the rationality check includes split check, merge check, outlier handling, and abnormal partition handling; Generate partitioning results based on the target partition and the corresponding target iteration data; The partitioning results include the partition label of each first InSAR monitoring point and the probability of belonging to each partition; The deformation characteristics of each partition are statistically analyzed, and a partition interpretation report is generated. The partition interpretation report includes the residual variance reduction rate, spatial coherence, structural region consistency, and the proportion of abnormal boundaries.

8. An electronic device, characterized in that, The method includes a processor, a memory, a communication interface, and one or more programs, said one or more programs being stored in the memory and configured to be executed by the processor, said programs including instructions for performing the steps of the method as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, A computer program for storing electronic data interchange is provided, wherein the computer program causes a computer to execute instructions for the steps of the method as described in any one of claims 1-7.

Images