Dike deformation monitoring method and device and electronic equipment

Abstract

The invention provides a dike deformation monitoring method and device and electronic equipment, and belongs to the technical field of data processing, and the dike deformation monitoring method comprises the steps: obtaining a first three-dimensional deformation observation value; obtaining a first deformation observation value; inputting the first three-dimensional deformation observation value and the first deformation observation value into a space-time fusion model constructed based on Kalman filtering to obtain a three-dimensional deformation field of the dike output by the space-time fusion model; wherein a physical geometric constraint matrix determined based on geometric parameters of the dike is introduced into an observation equation of the space-time fusion model, and the three-dimensional deformation field of the dike is generated under the physical geometric constraint matrix. The three-dimensional deformation observation value of the Beidou satellite and the line-of-sight direction deformation observation value of the synthetic aperture radar are fused, and the rationality of the calculation result is guaranteed by means of the physical geometric constraint matrix, so that the output three-dimensional deformation field better fits the actual deformation rule of the dike, and the precision and reliability of the dike deformation monitoring result are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, apparatus, and electronic device for monitoring the deformation of dikes. Background Technology

[0002] River levees are the foundation of flood control engineering systems and a vital barrier protecting the lives and property of people living along rivers and in riverside areas. The increasing frequency of extreme hydrological events caused by global climate change, coupled with the increasing service life of levee projects, has led to a growing risk of levee subsidence, landslides, and seepage damage. Levee projects are typically characterized by long routes, large spans, and complex geological conditions, and their deformation often exhibits nonlinear, sudden, and localized characteristics. If even minor deformations are not detected in time, they can easily trigger catastrophic consequences such as piping, bank collapses, or even levee breaches.

[0003] In related technical solutions, the deformation of dikes is treated as a purely geometric or statistical problem, which makes the monitoring results of dike deformation deviate from reality and results in poor reliability of the dike deformation monitoring results. Summary of the Invention

[0004] This invention provides a method, device, and electronic device for monitoring the deformation of dikes, in order to solve the problem that related technical solutions treat dike deformation as a purely geometric or statistical problem, and that there are unrealistic sharp abrupt changes or deformation patterns between adjacent monitoring points, which makes the monitoring results of dike deformation deviate from reality and result in poor reliability of the dike deformation monitoring results.

[0005] This invention provides a method for monitoring the deformation of a dike, comprising the following steps: Obtain the first three-dimensional deformation observation value, which is the three-dimensional deformation observation value of the dike determined by the Beidou satellite; Obtain a first deformation observation value, which is the deformation observation value of the embankment in the line of sight direction; The first three-dimensional deformation observation value and the first deformation observation value are input into the spatiotemporal fusion model constructed based on Kalman filtering to obtain the three-dimensional deformation field of the dike output by the spatiotemporal fusion model; The observation equation of the spatiotemporal fusion model incorporates a physical geometric constraint matrix determined based on the geometric parameters of the embankment, and the three-dimensional deformation field of the embankment is generated under the physical geometric constraint matrix.

[0006] The deformation monitoring method for dikes provided by this invention determines the physical geometric constraint matrix using the following steps: Based on the geometric parameters of the embankment, the normal vector of the embankment slope is determined in the station center coordinate system; A pseudo-observation equation is constructed based on the normal vector of the embankment slope. Based on the pseudo-observation equation, the projection of the three-dimensional deformation vector of the embankment slope onto the normal vector of the embankment slope is set to zero, thus obtaining the physical geometric constraint matrix.

[0007] The present invention provides a method for monitoring the deformation of a dike, wherein the geometric parameters of the dike include the slope and aspect of the dike.

[0008] The present invention provides a method for monitoring the deformation of a dike, the method further comprising: Based on the three-dimensional deformation field of the embankment, it is determined that the acceleration of the first observation point down the slope is greater than the preset acceleration, thus determining that there is a first hidden danger event, and outputting the first warning information corresponding to the first hidden danger event. The first observation point is an observation point located on the embankment. Based on the three-dimensional deformation field of the embankment, it is determined that there is continuous vertical settlement at the second observation point, a second hidden danger event is identified, and a second early warning information corresponding to the second hidden danger event is output. The second observation point is located on the embankment, and the second potential incident is that the embankment has foundation consolidation or uneven settlement.

[0009] The method for monitoring the deformation of a dike provided by this invention includes obtaining a first deformation observation value, comprising: Obtain a three-dimensional geometric model of the dike, the three-dimensional geometric model including multiple grid units divided according to the terrain; The geometric distortion index of each grid cell is determined based on the first orbital parameters and the geometric parameters of the grid cell, and the geometric mask is determined based on the geometric distortion index. The first orbital parameters are the orbital parameters of the Earth observation remote sensing satellite when it observes the dike. The Earth observation remote sensing satellite is equipped with synthetic aperture radar. Under the constraints of the geometric mask, permanent and distributed scatterers are identified in the synthetic aperture radar images; Phase unwrapping is performed on the observation points on the permanent scatterer and the observation points on the distributed scatterer to obtain the first deformation observation value.

[0010] The present invention provides a method for monitoring the deformation of a dike, wherein, under the constraint of the geometric mask, the method identifies permanent and distributed scatterers in the synthetic aperture radar image, comprising: Based on the geometric mask, candidate points in the image captured by the synthetic aperture radar are determined; The candidate points are distinguished based on the amplitude deviation index and spatial statistical test to obtain permanent scatterers and distributed scatterers.

[0011] The method for monitoring the deformation of a dike provided by this invention, wherein the identification of permanent scatterers and distributed scatterers in the synthetic aperture radar image under the constraint of the geometric mask, further includes: Identify the homogeneous pixels associated with each monitoring point in the distributed scatterer; The target monitoring point is retained because the number of homogeneous pixels associated with the target monitoring point in the distributed scatterer is greater than a preset number. The target monitoring point is removed if the number of homogeneous pixels associated with the target monitoring point in the distributed scatterer is less than or equal to a preset number. The target monitoring point is a monitoring point in a distributed scatterer.

[0012] The method for monitoring the deformation of a dike provided by this invention, wherein the identification of permanent scatterers and distributed scatterers in the synthetic aperture radar image under the constraint of the geometric mask, further includes: The coherence matrix of each monitoring point is determined based on the homogeneous pixels associated with the monitoring point; The eigenvector corresponding to the largest eigenvalue is determined based on the coherence matrix of the monitoring points; The original noise phase of the monitoring point is replaced with the eigenvector corresponding to the largest eigenvalue.

[0013] This invention provides a deformation monitoring device for dikes, comprising the following modules: The first acquisition module is used to acquire the first three-dimensional deformation observation value, which is the three-dimensional deformation observation value of the dike determined by the Beidou satellite. The second acquisition module is used to acquire the first deformation observation value, which is the deformation observation value of the dike in the line of sight direction; The processing module is used to input the first three-dimensional deformation observation value and the first deformation observation value into the spatiotemporal fusion model constructed based on Kalman filtering, so as to obtain the three-dimensional deformation field of the dike output by the spatiotemporal fusion model; The observation equation of the spatiotemporal fusion model incorporates a physical geometric constraint matrix determined based on the geometric parameters of the embankment, and the three-dimensional deformation field of the embankment is generated under the physical geometric constraint matrix.

[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the deformation monitoring method of the dike as described above.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the deformation monitoring method for a dike as described above.

[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the deformation monitoring method for a dike as described above.

[0017] The method, device, and electronic equipment for monitoring the deformation of dikes provided by this invention integrate three-dimensional deformation observations from BeiDou satellites and line-of-sight deformation observations from synthetic aperture radar. Simultaneously, by leveraging physical geometric constraint matrices to ensure the rationality of the solution results, it not only achieves complementarity in temporal and spatial resolution for dike deformation monitoring but also makes the output three-dimensional deformation field more closely match the actual deformation patterns of the dike. This effectively improves the accuracy and reliability of dike deformation monitoring results, providing precise data support for the timely identification of dike safety hazards. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is one of the flowcharts of the method for monitoring the deformation of dikes provided by the present invention; Figure 2 This is a schematic diagram of the process for obtaining the first deformation observation value provided by the present invention; Figure 3 This is a schematic diagram of the process for identifying permanent and distributed scatterers in synthetic aperture radar images under the constraint of a geometric mask, as provided by the present invention. Figure 4 This is the second flowchart of the method for monitoring the deformation of dikes provided by the present invention; Figure 5 This is a schematic block diagram of the deformation monitoring device for dikes provided by the present invention; Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0021] It should be noted that in the description of this invention, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. The terms "upper," "lower," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Unless otherwise expressly specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly, for example, as a fixed connection, a detachable connection, or an integral connection; a mechanical connection or an electrical connection; a direct connection or an indirect connection through an intermediate medium; or a connection within two elements. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0022] The terms "first," "second," etc., used in this invention are used to distinguish similar objects, not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0023] The following explains the terminology used in this invention: 1. BeiDou satellites, also known as the BeiDou Navigation Satellite System, can provide high-precision three-dimensional coordinates of monitoring points, enabling automated and high-temporal-resolution monitoring. However, its main limitation lies in the high cost of individual point equipment and deployment, resulting in sparse spatial sampling of the monitoring network and an inability to capture the continuous deformation field of the entire dike, both overall and locally.

[0024] 2. Interferometric Synthetic Aperture Radar (InSAR) is a mapping technology that uses a radar system mounted on a satellite or aircraft to acquire complex numerical images and generate surface deformation maps or digital elevation models by utilizing phase differences. The measurement accuracy can reach the centimeter level, and it is mainly used for volcano monitoring, seismic displacement analysis, and engineering stability assessment.

[0025] 3. Radar satellites are a general term for Earth observation remote sensing satellites that carry Synthetic Aperture Radar (SAR).

[0026] 4. Global Navigation Satellite System (GNSS), also known as Global Satellite Navigation System, is a space-based radio navigation and positioning system that can provide users with all-weather three-dimensional coordinates, velocity, and time information at any location on the Earth's surface or in near-Earth space. It is a general term for satellite navigation systems that can achieve global coverage.

[0027] 5. Station-centered coordinate system: also known as the local horizontal coordinate system, is a rectangular coordinate system established with the monitoring point or a certain reference point of the embankment as the origin, with the east (E) as the x-axis, the north (N) as the y-axis, and the vertical upward (U) axis as the z-axis. This coordinate system can intuitively reflect the spatial geometric relationship of the local area of ​​the embankment and is a commonly used coordinate system for calculating the normal vector of the embankment slope.

[0028] 6. Normal vector of embankment slope: refers to the unit vector perpendicular to the embankment slope (including the upstream slope, downstream slope, etc.). Its direction can represent the spatial orientation of the slope and is the core geometric parameter for constructing physical geometric constraints.

[0029] 7. Pseudo-observation equations: These are virtual observation equations constructed based on the physical deformation laws of embankments, and are not directly obtained from measured data. Their function is to provide physical constraints for the solution of the spatiotemporal fusion model, ensuring that the solution results conform to the deformation logic of the actual engineering.

[0030] The following is combined Figures 1-6The present invention describes a method, apparatus, and electronic device for monitoring the deformation of dikes, which aims to solve the problem that related technical solutions treat dike deformation as a purely geometric or statistical problem, and that there are unrealistic sharp abrupt changes or deformation patterns between adjacent monitoring points, causing the monitoring results of dike deformation to deviate from reality and resulting in poor reliability of the dike deformation monitoring results.

[0031] Figure 1 This is one of the flowcharts illustrating the deformation monitoring method for dikes provided by the present invention, such as... Figure 1 As shown, including but not limited to the following steps: Step 101: Obtain the first three-dimensional deformation observation value, which is the three-dimensional deformation observation value of the dike determined by the Beidou satellite.

[0032] Specifically, BeiDou receivers are first deployed along the dike. The deployment locations can be determined based on factors such as the dike's length, geological conditions, and the distribution of historical potential hazards. High-density deployment can be carried out at key locations such as the dike crest and toe, or conventional deployment can be done at preset intervals (e.g., 500 meters to 1 kilometer). The receivers can be high-precision multi-frequency, multi-mode GNSS receivers, capable of continuously receiving positioning signals from BeiDou satellites. Using carrier phase differential technology or precise point positioning technology, the three-dimensional coordinates of each monitoring point at different time points are calculated. Subsequently, the coordinate difference of the same monitoring point at adjacent time points is calculated to obtain the first three-dimensional deformation observation value for that monitoring point.

[0033] For example, if the three-dimensional coordinates of the monitoring point at time T0 are (E0, N0, U0) and the three-dimensional coordinates at time T1 are (E1, N1, U1), then the three-dimensional deformation during this time period is (ΔE=E1-E0, ΔN=N1-N0, ΔU=U1-U0), where ΔE is the eastward deformation, ΔN is the northward deformation, and ΔU is the vertical deformation. For the deployed receivers, real-time transmission of observation data can be achieved through 4G / 5G or BeiDou short message communication networks, ensuring the timeliness of data acquisition.

[0034] Step 102: Obtain the first deformation observation value, which is the deformation observation value of the embankment in the line of sight direction.

[0035] In some embodiments, observational data of a target levee area can be obtained from a publicly available satellite data platform or a customized satellite observation mission using a SAR-equipped Earth observation remote sensing satellite. During the observation process, the satellite emits radar waves toward the levee surface and receives echo signals. By interferometric processing of SAR images of the same area acquired at different times, interferometric phase data of the levee along the satellite's line of sight can be obtained. Combined with information such as the first orbital parameters and radar wavelength, the interferometric phase data is converted into the deformation of the levee along the line of sight, i.e., the first deformation observation value.

[0036] During the data acquisition process, satellites with different resolutions and revisit periods can be selected according to monitoring needs. If it is necessary to monitor minute deformations, high-resolution SAR satellites can be selected first. If it is necessary to achieve high-frequency monitoring, satellites with short revisit periods can be selected.

[0037] Step 103: Input the first three-dimensional deformation observation value and the first deformation observation value into the spatiotemporal fusion model constructed based on Kalman filtering to obtain the three-dimensional deformation field of the embankment output by the spatiotemporal fusion model.

[0038] Among them, the observation equation of the spatiotemporal fusion model introduces a physical geometric constraint matrix determined based on the geometric parameters of the dike, and the three-dimensional deformation field of the dike is generated under the physical geometric constraint matrix.

[0039] Specifically, a spatiotemporal fusion model based on Kalman filtering is first constructed. Kalman filtering mainly includes two core stages: time update and measurement update. In the time update stage, using previous levee deformation data and considering the levee's deformation trend, the three-dimensional deformation state of each monitoring point on the levee at the current moment is predicted. In the measurement update stage, the acquired first three-dimensional deformation observation value and the first deformation observation value are used as inputs to the model to correct the predicted deformation state. The core of this invention is to introduce a physical geometric constraint matrix into the observation equations of the spatiotemporal fusion model. This matrix is ​​determined based on the geometric morphology parameters of the levee, which constrains the model's solution process, ensuring that the three-dimensional deformation field output by the model conforms to the physical structural characteristics of the levee.

[0040] For example, for monitoring points in the embankment slope area, the physical geometric constraint matrix restricts their deformation in the direction perpendicular to the slope, ensuring that their deformation patterns conform to the actual instability mode of the embankment. After inputting the first three-dimensional deformation observation value and the first deformation observation value into this spatiotemporal fusion model, the model outputs a three-dimensional deformation field of the entire embankment area through iterative calculation using Kalman filtering. This three-dimensional deformation field can accurately represent the three-dimensional deformation state of each area of ​​the embankment.

[0041] By integrating three-dimensional deformation observations from BeiDou satellites and line-of-sight deformation observations from synthetic aperture radar, and by leveraging physical geometric constraint matrices to ensure the rationality of the solution results, this approach not only achieves complementarity in temporal and spatial resolution for levee deformation monitoring, but also makes the output three-dimensional deformation field more closely match the actual deformation patterns of the levee. This effectively improves the accuracy and reliability of levee deformation monitoring results, providing precise data support for the timely identification of levee safety hazards.

[0042] Furthermore, the physical geometric constraint matrix, also known as the introduced slope parallel flow constraint, ensures that the component of the deformation vector perpendicular to the slope conforms to the consolidation settlement law. Thus, even with only single-track InSAR data, it is possible to calculate the physically meaningful three-dimensional deformation (eastward, northward, vertical / along the slope direction).

[0043] In some embodiments, the physical geometric constraint matrix is ​​determined using the following steps: Based on the geometric parameters of the levee, the normal vector of the levee slope is determined in the station-centered coordinate system.

[0044] Specifically, for a target slope of a dike, first determine the three-dimensional coordinates of several discrete points on the slope in the station center coordinate system. For example, select three non-collinear points on the slope: P1(E1,N1,U1), P2(E2,N2,U2), and P3(E3,N3,U3). Solve for the slope normal vector through vector cross product operation.

[0045] First, calculate vectors P1P2 = (E2-E1, N2-N1, U2-U1) and P1P3 = (E3-E1, N3-N1, U3-U1). Then, perform a cross product on the two vectors. The result of this cross product is the normal vector of the slope, n = (nE, nN, nU). Finally, normalize this vector to obtain the unit normal vector of the embankment slope, n = [sinβsinα, sinβcosα, cosβ]. T (Where β is the slope and α is the aspect). For embankment slopes with regular shapes, the normal vector in the station center coordinate system can also be directly calculated based on the design parameters or measured parameters of the slope and aspect.

[0046] A pseudo-observation equation is constructed based on the normal vector of the embankment slope.

[0047] Specifically, based on the actual deformation physical laws of embankments, it can be seen that, as geotechnical structures, the instability deformation of embankments mainly manifests as consolidation settlement in the vertical direction and slippage along the slope. Under normal deformation conditions, the displacement component perpendicular to the slope normal is almost zero. Based on this physical law, a pseudo-observation equation is constructed with the unit normal vector of the slope as the core. Let the three-dimensional deformation vector of a monitoring point on the embankment slope be d=[dE,dN,dU] in the station-centered coordinate system. T (where dE is the eastward deformation, dN is the northward deformation, and dU is the vertical deformation), then the pseudo-observation equation can be expressed as Zconstraint=n·[dE,dN,dU] T The equation ≈0 transforms the physical deformation law of the dike into a mathematical constraint relationship, providing a foundation for the subsequent construction of the constraint matrix.

[0048] Based on the pseudo-observation equation, the projection of the three-dimensional deformation vector of the embankment slope onto the normal vector of the embankment slope is set to zero, thus obtaining the physical geometric constraint matrix.

[0049] Transforming the pseudo-observation equation into a matrix form of constraint equations can be written as 0= ·X k +V SPF The matrix form, where X k =[dE,dN,dU] T V is the three-dimensional deformation state vector of the monitoring point. SPF H is the noise term in the constraint equation (which can be set according to the actual error level of the dike deformation). SPF = .

[0050] Subsequently, the constraint noise variance is defined. In the embankment area ( ): Based on the slope sliding model, a smaller setting is used. (like ).

[0051] On the top of the dike / flat area :at this time The constraint equations degenerate into This does not conform to reality (allowing for vertical settlement). Therefore, the system introduces an adaptive judgment: when the slope... When the value is less than the threshold, the constraint is automatically turned off, or it is switched to the "minimum horizontal displacement" constraint (assuming that the flat land mainly experiences vertical settlement and the horizontal displacement is 0).

[0052] Integrate the above model into the extended Kalman filter loop: Time update (prediction): Using Φk, predict the state $X_{k|k-1}$ and covariance P at time $k$. k|k-1 .

[0053] Measurement update (calibration): Constructing a joint observation matrix and joint observation vector : If only InSAR data is available at the current time: ; If BeiDou + InSAR data is available at the current moment: ; Calculate Kalman gain And update the state vector .

[0054] In this embodiment, by introducing The constraint matrix, also known as the physical geometric constraint matrix, even in Missing and only one view In the case of observation matrix The rank can also be restored (from rank deficiency to full rank), thus uniquely solving the three-dimensional deformation. ,in, and It is in InSAR data and joint observation matrix and joint observation vector The corresponding data, and It is the BeiDou data and the joint observation matrix and joint observation vector The corresponding data.

[0055] The solution results can be clearly distinguished: Displacement along the slope: .

[0056] Vertical settlement: .

[0057] To adapt to the observation equation format of the Kalman filter spatiotemporal fusion model, the constraint matrix needs further standardization and adaptation. If the spatiotemporal fusion model incorporates deformation calculations for multiple monitoring points, a block-based physical geometric constraint matrix can be constructed based on the normal vector of the slope to which each monitoring point belongs. This ensures that each monitoring point on the slope is constrained by the corresponding geometric constraints, ultimately forming a physical geometric constraint matrix that can be embedded into the observation equation of the spatiotemporal fusion model.

[0058] The deformation of a levee is limited by its own geotechnical characteristics and stress state, and its deformation direction has a clear physical boundary. Large displacements perpendicular to the slope normal do not conform to the conventional instability mode of levees. This implementation method first solves the slope normal vector, then constructs a pseudo-observation equation, and finally transforms it into a physical geometric constraint matrix. This process converts the physical deformation law of the levee into mathematical constraints, which are then embedded into the observation equation of the Kalman filter. This ensures that when the model integrates BeiDou 3D observations and InSAR line-of-sight observations, it is constrained by physical geometric conditions, avoiding the calculation of deformation results that violate engineering reality.

[0059] By constructing a physical geometric constraint matrix through a standardized process, we can effectively combine the physical deformation law of the dike with the mathematical solution model, and provide a precise geometric constraint basis for the spatiotemporal fusion model. This can significantly improve the physical authenticity and rationality of the three-dimensional deformation field solution results, effectively avoid the problems of deformation abrupt change and logical contradiction that are prone to occur in traditional pure statistical models, and further ensure the reliability of dike deformation monitoring.

[0060] In some embodiments, the geometric parameters of the dike include the slope and aspect of the dike.

[0061] In some embodiments, the method for monitoring the deformation of a dike further includes: Based on the three-dimensional deformation field of the embankment, the acceleration of the first observation point down the slope is determined to be greater than the preset acceleration, thus identifying the existence of the first hidden danger event and outputting the first early warning information corresponding to the first hidden danger event. The first observation point is the observation point located on the embankment.

[0062] Among them, the first observation point refers to the monitoring point located on the slope of the embankment (including the upstream slope and the downstream slope). The deformation state of this type of point is directly related to whether the embankment will experience instability problems such as landslides and creep, and it is the core focus point for embankment safety monitoring.

[0063] Preset acceleration: This refers to an acceleration threshold preset based on the geotechnical properties, historical deformation data, and safety specifications of the embankment project. This threshold can be configured differently according to factors such as the embankment's construction materials, slope, and geological conditions. For example, for a silty clay embankment slope, the preset acceleration can be set to 0.1 mm / d. 2 .

[0064] The first type of hidden danger event refers to the type of safety hazard where monitoring points on the slope of the embankment show accelerated downward displacement along the slope, posing a risk of landslides and slope collapses. This type of hazard is characterized by its suddenness and great harm.

[0065] First warning information: refers to the warning notification generated in response to the first hidden danger event, which includes key information such as the location of the hidden danger, the value of deformation acceleration, and the risk level. It can be output to the outside world through SMS, platform push, sound and light alarms, etc.

[0066] First, deformation data of the observation points is extracted based on the three-dimensional deformation field. Specifically, spatiotemporal deformation sequence data of all monitoring points are extracted from the three-dimensional deformation field of the embankment. For the first observation point, the three-dimensional deformation vector (eastward dE, northward dN, vertical dU) needs to be decomposed into deformation components along the slope direction and perpendicular to the slope direction, based on the slope angle α and slope angle β. Specifically, the deformation along the slope direction can be obtained by coordinate transformation using the slope angle α and slope angle β. d slope =dEsinαsinβ+dNcosαsinβ+dUcosβ; Subsequently, the slope-direction displacement acceleration at the first observation point is calculated based on the deformation sequence. This can be achieved by performing a quadratic difference on the slope-direction deformation over multiple time periods, for example, obtaining the slope-direction deformation d for three consecutive time periods: t0, t1, and t2. slope0 d slope1 d slope2 First, calculate the displacement rate v1 = (d) in adjacent time periods. slope1 -dslope0 ) / (t1-t0), v2=(d slope2 -d slope1 ) / (t2-t1), then calculate the acceleration a=(v2-v1) / (t2-t0).

[0067] The calculated acceleration along the slope at the first observation point is compared with the preset acceleration. If the acceleration value of a certain first observation point is greater than the preset acceleration, and the acceleration state continues for at least 2 monitoring cycles, then the first hidden danger event is determined to exist at that point.

[0068] Based on the three-dimensional deformation field of the levee, it was determined that there was continuous vertical settlement at the second observation point, thus identifying the existence of a second hidden danger event, and outputting the second early warning information corresponding to the second hidden danger event.

[0069] The second observation point refers to the monitoring point located on the top of the embankment. The vertical settlement of the top of the embankment can reflect the degree of consolidation of the embankment foundation and its overall stability.

[0070] The second type of hidden danger event refers to the continuous vertical settlement at the monitoring point on the top of the dike, which indicates a safety hazard due to abnormal foundation consolidation or uneven settlement. Long-term uneven settlement can easily lead to cracking of the dike body and structural damage.

[0071] Second warning information: refers to the warning notice generated in response to the second hidden danger event, which must clearly specify the settlement point, settlement rate, cumulative settlement amount and corresponding risk warning.

[0072] Among them, the vertical deformation dU of the second observation point is directly extracted as the core monitoring data. The vertical settlement of the second observation point in multiple consecutive monitoring periods is statistically analyzed to determine whether it shows a continuous settlement trend.

[0073] For example, a continuous settlement judgment criterion is set. For instance, if the vertical settlement is greater than 0 for three or more consecutive monitoring cycles and the cumulative settlement exceeds 5 mm, or the settlement rate in a single cycle is greater than 0.5 mm / d, then the second observation point is judged to have a second hidden event, that is, the foundation consolidation or uneven settlement of the embankment top.

[0074] The second observation point is located on the embankment, and the second potential incident is the existence of foundation consolidation or uneven settlement of the embankment.

[0075] In this embodiment, different areas of the dike (slope and top) correspond to different instability modes. Accelerated slippage of the slope is a precursor to sudden disasters such as landslides, while continuous vertical settlement of the top reflects long-term deformation problems of the foundation. Based on precise data from the three-dimensional deformation field, this implementation sets differentiated hazard judgment rules for the deformation characteristics of different areas, transforming quantified deformation data into intuitive safety warnings. This achieves closed-loop management from "data monitoring" to "hazard identification" and then to "warning output." Its core is to establish a correspondence between hazards and deformation indicators using the physical laws of dike deformation, ensuring the accuracy and timeliness of warnings.

[0076] By deeply mining and specifically identifying three-dimensional deformation field data, it is possible to quickly identify two core safety hazards: accelerated slippage of the dike slope and continuous settlement of the dike crest. It can also promptly output graded early warning information, which not only fills the gap of traditional monitoring that "only measures but does not judge," but also provides water conservancy management departments with clear guidelines for handling hazards. This effectively improves the proactive prevention and control capabilities of dike safety risks and reduces the probability of catastrophic consequences such as dike breaches and bank collapses caused by the failure to detect hazards in a timely manner.

[0077] In some embodiments, such as Figure 2 As shown, the first deformation observation value is obtained, including: Step 201: Obtain the three-dimensional geometric model of the dike. The three-dimensional geometric model includes multiple grid cells divided according to the terrain.

[0078] Among them, the three-dimensional geometric model refers to a three-dimensional digital model that can accurately represent the overall and local spatial geometric features of the embankment. The model contains multiple grid units divided according to the embankment's landform features. Each grid unit carries key geometric parameters such as elevation, slope, and aspect, which are the basic data carriers for subsequent geometric distortion analysis.

[0079] Grid unit: refers to the smallest geometric unit formed by regularly or irregularly dividing the three-dimensional space of the dike. Its size can be set according to the monitoring accuracy requirements. For example, it can be set to a regular grid of 1m×1m, or an irregular grid with adaptive size can be used according to the complexity of the dike topography. The division of grid units must completely cover all topographic units of the dike, such as the top of the dike, the water-facing slope, the back slope, and the toe of the dike.

[0080] In some embodiments, high-resolution point cloud data of the levee monitoring area can be acquired through methods such as airborne LiDAR scanning and UAV oblique photogrammetry. If airborne LiDAR scanning is used, a professional surveying aircraft equipped with LiDAR equipment can perform multi-strip scanning along the levee's direction to acquire dense three-dimensional point clouds on the levee surface. If UAV oblique photogrammetry is used, the UAV can be controlled to acquire images of the levee from multiple angles, and then point cloud data can be generated using multi-view stereo matching technology. After preprocessing the acquired point cloud data, such as denoising, registration, and stitching, a detailed digital elevation model (DEM) of the levee is constructed. Based on the levee's geomorphic features (levee crest, upstream slope, downstream slope, and levee toe), the DEM is segmented into micro-topographic units, dividing the entire levee into multiple grid units. Simultaneously, the elevation, slope, aspect, and other geometric parameters of each grid unit are calculated, ultimately forming a three-dimensional geometric model of the levee containing the grid units and their corresponding geometric parameters.

[0081] Step 202: Determine the geometric distortion index of each grid cell based on the first orbital parameters and the geometric parameters of the grid cells, and determine the geometric mask based on the geometric distortion index.

[0082] The first orbital parameter is the orbital parameter of the Earth observation remote sensing satellite when it observes the dike. The Earth observation remote sensing satellite is equipped with synthetic aperture radar.

[0083] Among them, the first orbital parameter refers to the orbital parameters of the Earth observation remote sensing satellite carrying synthetic aperture radar when observing the dike area. It mainly includes core parameters such as the satellite incident angle and heading angle. These parameters directly determine the observation angle and coverage of the radar beam and are the key basis for calculating the geometric distortion index.

[0084] For example, the first orbital parameters include the incident angle and the heading angle.

[0085] Geometric distortion index: This refers to a quantitative indicator used to describe the geometric matching relationship between the radar beam and the local topography of the dike. It can intuitively reflect whether there are geometric distortion problems such as perspective shrinkage, top and bottom inversion, and shadows in each grid cell of the SAR image. It is the core criterion for screening effective observation areas.

[0086] Geometric mask: refers to a binary mask layer generated based on the geometric distortion index, used to mark the effective and invalid observation areas in SAR images. The effective areas are marked with "1" and can participate in subsequent scatterer identification and other processing; the invalid areas are marked with "0" and will be removed and will not participate in subsequent calculations.

[0087] Specifically, the first orbital parameters of the Earth observation remote sensing satellite, including the satellite's incident angle and heading angle, are obtained directly from the satellite data header file. Then, for each grid cell in the 3D geometric model, the geometric distortion index (R-Index) of that grid cell is calculated using a geometric projection calculation model, combining its geometric parameters (slope, aspect, elevation) and the satellite's first orbital parameters. The specific calculation process needs to consider factors such as the angle between the radar beam incident direction and the grid cell's slope, and the relative relationship between the slope orientation and the satellite's heading. If the grid cell exhibits perspective contraction, its geometric distortion index will decrease significantly; if there are shadows or inverted top and bottom, the geometric distortion index will approach 0. Subsequently, a geometric distortion index threshold is set, for example, 0.3. Grid cells with a geometric distortion index greater than the threshold are marked as valid observation areas, while those with a geometric distortion index less than or equal to the threshold are marked as invalid observation areas. Based on this marking result, a geometric mask covering the entire levee monitoring area is generated, completing the initial screening of the SAR image observation area.

[0088] Step 203: Under the constraint of the geometric mask, identify permanent scatterers and distributed scatterers in the synthetic aperture radar image.

[0089] Among them, permanent scatterers refer to ground objects on the surface of the dike that have strong scattering characteristics and whose scattering characteristics remain stable over time, such as concrete structures and artificial corner reflectors on the dike, which can provide stable phase observation signals for InSAR monitoring.

[0090] Distributed scatterers: These refer to ground features on the surface of a dike where individual pixels have weak scattering characteristics, but adjacent pixels have similar scattering characteristics. Examples include bare soil on the dike slope and areas with sparse vegetation cover. Effective phase information can be extracted by jointly processing homogeneous pixels.

[0091] Phase unwrapping: refers to the process of restoring the interferometric phase wrapped in the interval (-π,π] into a continuous phase that can reflect the true deformation. It is the core step in InSAR technology to convert the interferometric phase into deformation.

[0092] Step 204: Perform phase unwrapping on the observation points on the permanent scatterer and the observation points on the distributed scatterer to obtain the first deformation observation value.

[0093] For the identified permanent and distributed scatterers, a preliminary extraction of their time-series interferometric phase is performed, followed by three-dimensional phase unwrapping using a minimum-cost flow unwrapping algorithm. During unwrapping, geometric parameters such as slope and aspect from the levee's three-dimensional geometric model can be introduced to construct constraint weights. For example, unwrapping costs are increased for adjacent pixels with large elevation differences to prevent unwrapping errors from propagating across the slope. After phase unwrapping, the unwrapped continuous phase is converted into the levee's deformation along the satellite line-of-sight direction by combining the first orbital parameters and radar wavelength, thus obtaining the first deformation observation value. This observation value accurately reflects the deformation amplitude and distribution characteristics of various areas of the levee along the line-of-sight direction.

[0094] In this embodiment, the complex geometry of the dike can cause geometric distortion in SAR images, thus affecting the effectiveness of InSAR monitoring data. This implementation first constructs a three-dimensional geometric model of the dike using high-precision methods to understand its fine geometric features; then, it calculates the geometric distortion index using first orbital parameters, and removes distorted regions using a geometric mask to ensure the effectiveness of subsequent data processing; finally, it identifies permanent and distributed scatterers to obtain stable phase observation signals, and converts the interferometric phase into line-of-sight deformation through phase unwrapping, thereby providing high-quality first deformation observation values ​​for the spatiotemporal fusion model.

[0095] Based on this, by constructing a three-dimensional geometric morphology model and introducing geometric mask screening, the problem of invalid data caused by the geometric distortion of dikes in InSAR monitoring is fundamentally solved. At the same time, by jointly identifying permanent and distributed scatterers, the monitoring point density in the vegetation-covered area of ​​the dike is significantly improved. The final output of the first deformation observation value has higher accuracy and more complete coverage, providing reliable basic data for subsequent multi-source data fusion, and effectively making up for the shortcomings of traditional InSAR technology in dike monitoring, such as data loss and insufficient accuracy.

[0096] In some embodiments, such as Figure 3 As shown, under the constraint of a geometric mask, permanent and distributed scatterers in synthetic aperture radar images are identified, including: Step 301: Based on the geometric mask, determine candidate points in the images captured by synthetic aperture radar.

[0097] Candidate points refer to pixels in SAR images that are within the effective observation area after being filtered by geometric masks. These pixels have the basic conditions to participate in scatterer identification and are the initial objects for distinguishing between permanent and distributed scatterers.

[0098] Specifically, the generated geometric mask is spatially registered with the time-series single-view complex (SLC) images captured by synthetic aperture radar to ensure that the grid cells of the geometric mask correspond one-to-one with the pixels of the SAR image.

[0099] Then, each pixel in the SAR image is traversed, and candidate points are selected based on the marking results of the geometric mask: if the geometric mask grid cell corresponding to a pixel is marked as an effective observation area (marked as "1", that is, the area has no severe perspective shrinkage, top-bottom inversion or shadows or other geometric distortions), then the pixel is determined as a candidate point; if the corresponding grid cell is marked as an invalid observation area (marked as "0"), then the pixel is directly removed and does not participate in the subsequent scatterer identification process.

[0100] Candidate point selection can be achieved through the mask extraction tool of professional remote sensing data processing software. After selection, a candidate point dataset containing only effective region pixels will be formed, which will serve as the basis for subsequent scatterer classification.

[0101] Specifically, synthetic aperture radar images are presented as a time-series single-view complex (SLC) image set. This indicates that the geometric mask is based on If it means: Traverse every pixel in the image Only when A pixel is marked as a candidate pixel only when it is in a non-shadowed, non-top-bottom inverted effective imaging area.

[0102] Step 302: Based on the amplitude deviation index and spatial statistical test, candidate points are distinguished to obtain permanent scatterers and distributed scatterers.

[0103] Amplitude deviation index: This refers to a quantitative indicator used to measure the stability of the scattering amplitude of SAR image pixels over time. It is calculated as the ratio of the standard deviation of the pixel's time series amplitude to the mean amplitude. The smaller the value, the more stable the scattering characteristics of the pixel.

[0104] Spatial statistical tests refer to the test methods that use statistical algorithms to verify whether the scattering characteristics of adjacent pixels in SAR images are consistent. Commonly used tests include the Kolmogorov-Smirnov (KS) test and the Anderson-Darling (AD) test, which are the core means of identifying distributed scatterers.

[0105] Specifically, for the selected candidate points, the amplitude deviation index (Damp) in their time series is calculated one by one. The specific calculation formula is Damp = μAσA, where σA is the amplitude standard deviation of the candidate point in the time series SAR image, and μA is the amplitude mean of the candidate point in the time series SAR image. A threshold (TPS) for the amplitude deviation index of permanent scatterers is preset in advance. This threshold can be adjusted according to the ground object types in the levee area and the imaging parameters of the SAR image. For example, for C-band SAR images, the threshold TPS can be set to 0.25. If the amplitude deviation index Damp of a certain candidate point is <TPS, it means that the scattering characteristics of this candidate point have strong stability in the time series, and it can be directly determined as a permanent scatterer; if the amplitude deviation index Damp of the candidate point is ≥TPS, it will be classified into the candidate category of distributed scatterers and enter the next spatial statistical test process.

[0106] Exemplarily, calculate the time series amplitude deviation index for each candidate point : where is the standard deviation of the time series amplitude, ' is the mean. If (such as 0.25), it is directly marked as a PS point, that is, a candidate point for distributed scatterers.

[0107] For the candidate points not determined as permanent scatterers, their surrounding homogeneous pixels are identified through spatial statistical test methods, and then distributed scatterers are determined. First, a spatial search window is set for each candidate point to be tested. The window size can be set according to the spatial distribution characteristics of the levee ground objects. Commonly used are square windows of 11×11 pixels or 21×21 pixels. Centered on the candidate point, the time series scattering amplitude data of all adjacent pixels within the search window are extracted. Subsequently, a suitable spatial statistical test algorithm, such as the KS test, is selected to calculate the cumulative distribution function (CDF) difference between the central candidate point and the neighboring pixels. A significance level α (usually taken as 5%) is set. If the CDF difference between the central candidate point and a certain neighboring pixel is less than α, it is determined that the neighboring pixel and the central candidate point are homogeneous pixels, and the two have similar scattering statistical characteristics. If a certain candidate point can find a sufficient number of homogeneous pixels within the search window, it means that the scattering signal in its area can be enhanced through the joint processing of homogeneous pixels. At this time, the candidate point and the corresponding homogeneous pixel group are determined as distributed scatterers, and the final distinction between the two types of scatterers is completed.

[0108] In this embodiment, the core characteristic of permanent scatterers is the strong stability of their scattering properties over time, which can be quickly screened using the amplitude deviation index. Distributed scatterers, on the other hand, have weaker and less stable scattering signals from individual pixels, but adjacent pixels exhibit statistically consistent scattering characteristics, allowing for group identification through spatial statistical testing. This implementation first uses a geometric mask to eliminate invalid observation pixels, and then employs a two-step method of "initial screening using the amplitude deviation index + detailed subdivision using spatial statistical testing" to accurately distinguish between permanent and distributed scatterers. The principle behind this approach is to combine the temporal stability and spatial statistical consistency characteristics of scatterers to design targeted identification logic.

[0109] By employing standardized candidate point screening and scatterer differentiation processes, the accuracy of scatterer identification results is ensured while significantly improving identification efficiency. Furthermore, by jointly identifying permanent and distributed scatterers, the shortcomings of single permanent scatterer identification in monitoring points in vegetation-covered embankment areas are effectively compensated for. This provides sufficient and high-quality observation points for subsequent phase unwrapping and acquisition of first deformation observations, further enhancing the integrity and reliability of InSAR monitoring data.

[0110] In some embodiments, identifying permanent and distributed scatterers in synthetic aperture radar images, under the constraint of a geometric mask, further includes: Identify the homogeneous pixels associated with each monitoring point in the distributed scatterer.

[0111] Among them, homogeneous pixels refer to SAR image pixels in the neighborhood that have the same scattering statistical characteristics as the target monitoring point in the distributed scatterer. These pixels have a high degree of consistency in scattering amplitude, phase and other characteristics in the time series, and can be determined by spatial statistical tests.

[0112] Specifically, based on the initial identification of distributed scatterers, for each monitoring point within a distributed scatterer, the time-series scattering data of all neighboring pixels within the pre-defined spatial search window (e.g., 11×11 pixels, 21×21 pixels) are retrieved. A selected spatial statistical test algorithm is used to calculate the difference in the Cumulative Distribution Function (CDF) between the target monitoring point and each neighboring pixel. A significance level is set (usually 5%). If the CDF difference between a neighboring pixel and the target monitoring point is less than this significance level, the neighboring pixel is determined to be a homogeneous pixel of the target monitoring point; if the difference is greater than or equal to the significance level, it is determined to be a non-homogeneous pixel and not included in the association scope. After completing the test for all neighboring pixels, the set of homogeneous pixels associated with each distributed scatterer monitoring point is summarized, forming a "monitoring point - homogeneous pixel" correspondence table.

[0113] For example, calculate the center pixel (That is, the target monitoring point in this invention) and neighboring pixels The difference in the cumulative distribution function. If the difference is less than the significance level. (e.g., 5%), then determine for Homogeneous pixels.

[0114] The target monitoring point is retained if the number of homogeneous pixels associated with the target monitoring point in the distributed scatterer is greater than the preset number.

[0115] If the number of homogeneous pixels associated with a target monitoring point in a distributed scatterer is less than or equal to a preset number, the target monitoring point is removed.

[0116] Among them, the target monitoring point is the monitoring point in the distributed scatterer.

[0117] Specifically, a preset threshold for the number of homogeneous pixels is set. This threshold should balance monitoring accuracy and data integrity. For example, in densely vegetated embankment areas, the preset number can be appropriately reduced (e.g., 15) to retain more effective monitoring points; in hardened embankment top areas, the preset number can be increased (e.g., 25) to ensure the reliability of monitoring points. All target monitoring points in the distributed scattering system are traversed, and the number of homogeneous pixels associated with each point is counted. If the number of homogeneous pixels associated with the target monitoring point is greater than the preset number, it means that the scattering signal in the area where the monitoring point is located can be effectively enhanced through the joint processing of homogeneous pixels, and has a stable phase observation capability. Therefore, the target monitoring point is retained and included in the subsequent phase optimization and phase unwrapping processing scope. If the number of homogeneous pixels associated with a target monitoring point is less than or equal to the preset number, it indicates that the scattered signal strength of the monitoring point is insufficient and the stability is poor, and it cannot provide reliable phase data for InSAR monitoring. Therefore, the target monitoring point is removed and will not participate in the subsequent data processing.

[0118] For example, the number N of SHPs for each DS candidate point (i.e., target monitoring point) is counted. SHP If N SHP If Tnum (e.g., 20) is selected, the point is retained as a valid DS point; otherwise, it is discarded. This step can effectively identify mixed scattering signals of bare soil or sparse vegetation on vegetated embankment slopes.

[0119] After the screening is completed, the final effective distributed scatterer monitoring point dataset is formed. This dataset, together with the permanent scatterer monitoring point data, will serve as the basis for subsequent phase unwrapping.

[0120] The monitoring capability of distributed scatterers depends on the group scattering characteristics of their region. The scattering signal from a single monitoring point often suffers from high noise and poor stability. However, a sufficient number of homogeneous pixels can cancel out some of the noise through signal superposition, thereby improving the signal-to-noise ratio of the overall observed signal. This implementation method, by first identifying homogeneous pixels and then filtering monitoring points according to a quantity threshold, essentially performs a secondary verification of the effectiveness of distributed scatterer observations. Its principle is to utilize the spatial statistical consistency characteristics of scatterers to eliminate monitoring points with unstable signals, ensuring the reliability of subsequent InSAR data processing.

[0121] By setting a threshold for the number of homogeneous pixels, the monitoring points of the distributed scatterer are precisely screened. This process not only eliminates invalid monitoring points with weak signals and insufficient stability, thus avoiding interference with subsequent phase unwrapping and deformation observation calculations, but also retains monitoring points with effective observation capabilities, ensuring the monitoring point density in the levee monitoring area (especially in vegetation-covered areas). At the same time, this screening process further improves the quality of the distributed scatterer data, laying a solid foundation for obtaining high-precision first deformation observations and effectively reducing the error of InSAR monitoring data.

[0122] In some embodiments, such as Figure 4 As shown, under the constraint of a geometric mask, the identification of permanent and distributed scatterers in synthetic aperture radar images also includes: Step 401: Determine the coherence matrix of each monitoring point based on the homogeneous pixels associated with the monitoring point.

[0123] Among them, the coherence matrix refers to the covariance matrix constructed based on the complex observation vectors of the monitoring points of the distributed scatterer and their associated homogeneous pixels. It is the core matrix that characterizes the correlation and stability of the scattering signals of the monitoring points and their homogeneous pixels, and can be used for subsequent phase optimization calculations.

[0124] Specifically, after selecting target monitoring points, for each retained distributed scatterer target monitoring point, the time-series normalized complex observation vectors of all associated homogeneous pixels are retrieved. Let N be the number of homogeneous pixels associated with a given target monitoring point, and let yk∈CN×1 be the normalized complex observation vector of the k-th homogeneous pixel (where N is the time-series length of the SAR image, and C represents the complex domain). Then, the formula for calculating the coherence matrix C of this target monitoring point is: ; in It is the first Normalized complex observation vectors of SHP This indicates the conjugate transpose.

[0125] This formula can aggregate the scattering signals of multiple homogeneous pixels. The coherence matrix can comprehensively reflect the scattering phase correlation characteristics of the target monitoring point and its homogeneous pixel group, providing a data basis for subsequent phase optimization.

[0126] Step 402: Determine the eigenvector corresponding to the largest eigenvalue based on the coherence matrix of the monitoring points.

[0127] Among them, the eigenvector corresponding to the largest eigenvalue refers to the eigenvector corresponding to the largest eigenvalue after eigenvalue decomposition of the coherence matrix. This vector contains the optimal phase information of the monitoring point and its homogeneous pixel group, which can effectively reduce noise interference in the original phase.

[0128] Specifically, the coherence matrix C obtained in the above steps is subjected to eigenvalue decomposition, that is, solving for the eigenvalues ​​λi and corresponding eigenvectors vi (i=1,2,...,N) that satisfy Cvi=λivi. All the decomposed eigenvalues ​​are sorted in descending order, the largest eigenvalue λmax is selected, and its corresponding eigenvector vmax is extracted. In the statistical characteristics of the scattered signal, the eigenvector corresponding to the largest eigenvalue represents the phase component with the strongest signal energy and highest correlation in a homogeneous pixel group, effectively removing the noise component from the original phase and preserving the true deformation phase information.

[0129] Step 403: Replace the original noise phase of the monitoring point with the eigenvector corresponding to the largest eigenvalue.

[0130] Among them, the original noise phase refers to the unoptimized interferometric phase extracted directly from SAR images by distributed scatterer monitoring points. This type of phase usually contains a lot of random noise and system noise, and cannot be directly used for accurate deformation inversion.

[0131] Specifically, firstly, the phase component of the eigenvector vmax corresponding to the largest eigenvalue is extracted to obtain the optimized phase vector. (in, (This refers to the optimized phase corresponding to the nth time).

[0132] The optimized phase vector The maximum solution for the following objective function is: in, The optimal phase vector is the core objective to be solved, and its dimension is consistent with the number of time phases in the SAR time series image (if there are N time phases of imagery, then φopt=[φ1,φ2,...,φN]). T), representing the true deformation phase of the distributed scatterer monitoring point at each time phase, can effectively eliminate noise in the original phase; φ is the candidate phase vector, which is all possible phase combinations traversed during the optimization process, and its value range is a complex phase vector wrapped in the interval (-π, π]. H : Represents the conjugate transpose of the candidate phase vector φ. Since the phase vector is a complex vector, the conjugate transpose is a standard operation for calculating the dot product of vectors in the complex field, used to ensure the real nature of the optimization objective and |C|. -1 |C| is the matrix obtained by taking the modulo of each element of the coherence matrix C. -1 The inverse of the modulus matrix, This represents the Hadamard product (element-by-element multiplication).

[0133] Solving for the coherence matrix using eigenvalue decomposition. The eigenvector corresponding to the largest eigenvalue is the optimal phase vector. The phase portion.

[0134] The original noise phase vector of the distributed scatterer target monitoring point was then replaced with the optimized phase vector. This completes the phase optimization processing for distributed scatterer monitoring points. For permanent scatterer monitoring points, due to their stable scattering characteristics and low phase noise, their original phase can be directly retained without the need for this phase replacement step.

[0135] In this study, the original phase of a single monitoring point of a distributed scatterer is affected by factors such as vegetation cover and radar signal interference, resulting in a large amount of noise. However, its associated homogeneous pixels possess similar scattering statistical characteristics. By constructing a coherence matrix, the effective phase information of the group signal can be aggregated. Eigenvalue decomposition can extract the phase component with the strongest energy from the coherence matrix. This component corresponds to the true deformed phase of the homogeneous pixel group. Replacing the original noisy phase with this component essentially utilizes spatial statistical correlation to achieve noise filtering and phase enhancement, ensuring the accuracy of subsequent phase unwrapping.

[0136] Phase optimization of distributed scatterer monitoring points was achieved through coherence matrix construction and eigenvalue decomposition, effectively filtering out noise interference in the original phase and significantly improving the phase signal-to-noise ratio in weak scattering areas such as embankment vegetation cover. At the same time, the phase-replaced data has higher stability and accuracy, providing a high-quality phase data source for subsequent phase unwrapping and acquisition of the first deformation observation value, further reducing the error of InSAR monitoring and ensuring the accuracy and reliability of the final deformation monitoring results.

[0137] In some embodiments, phase unwrapping is the process of untangling the phase... The observed phase of the interval is restored to a continuous true phase. Due to significant elevation changes (from top to toe) and steep slopes in the dike, conventional unwrapping is prone to integer jump errors across the slope. This invention employs a three-dimensional minimum cost flow unwrapping algorithm that takes into account geometric constraints. Specifically, it includes the following steps: 1. Construct a three-dimensional unwrapped network.

[0138] Specifically, a three-dimensional grid is constructed that includes a "spatial dimension" (image plane x, y) and a "temporal dimension" (image sequence t).

[0139] 2. Define geometric constraint weights.

[0140] Specifically, in related technical solutions, the untangling path always tends to pass through the arc segment with the minimum "cost". This invention introduces the geometric distortion index (R-Index) and the terrain slope gradient (i.e., the physical geometric constraint matrix) into the construction of the cost function as suppression factors.

[0141] For connected pixels and The arc segment, its cost Defined as: in: To connect pixels and The cost of the arc segment, It is a pixel The coherence between them.

[0142] in, : Geometric distortion penalty term. Utilizing Specifically: like For smaller values ​​(close to perspective shrinkage or shadow), the cost increases exponentially, forcing the unwrapping path to bypass areas of severe geometric distortion, where... It is a pixel and pixels Corresponding to , A very small positive number (such as 10⁻⁶) is called a "smoothing factor".

[0143] in, : Terrain gradient penalty term.

[0144] ; in, ΔH is the terrain gradient penalty factor connecting pixel i and pixel j. i,j H represents the elevation difference between pixel i and pixel j.threshold This is the preset elevation difference threshold.

[0145] If the elevation difference between adjacent pixels (provided by the LiDAR DEM prior) is too large (such as crossing a steep embankment), the cost is increased to prevent unwrapping errors from propagating in a direction perpendicular to the embankment's direction (i.e., the direction across the embankment slope).

[0146] The flow in the network is calculated using a minimum cost flow solver, and the integer multiples of 2π that need to be added or subtracted are determined to obtain the final untangled phase.

[0147] Table 1 shows the comparative advantages of the dike deformation monitoring method proposed in this invention with related technical solutions.

[0148] Table 1 Comparison Dimensions Related technologies This invention Basic principles Statistical optimization treats monitoring points as independent samples. Geometric-mathematical coupled optimization, taking into account the overall structural integrity of the dike. Data integration methods Sequential correction or weighted average based on prior accuracy Synchronization and joint solution of BeiDou / GNSS, InSAR and geometric models under a unified framework Physical authenticity This cannot be guaranteed and may result in outcomes that violate structural continuity. By using strong constraints based on geometric equations, the geometric continuity and smoothness of the results are ensured. 3D Deformation Solution Relying on simplified projection assumptions results in significant accuracy loss. Rigorous calculation of dense, high-precision three-dimensional deformation fields Utilization of prior knowledge Unutilized geometric information of the dike Deeply integrating the design form of dikes as prior knowledge Quantification of uncertainty Lack of rigorous overall accuracy assessment Provides rigorous and accurate estimation of displacement components at each point based on deformation decomposition theory. The proposed method for monitoring levee deformation introduces a high-precision LiDAR DEM and R-Index masking method. This invention addresses the geometric distortion problem of InSAR on complex levee terrain from a physical perspective, avoiding false deformation signals caused by overlay or shadows, and significantly improving the reliability of monitoring results at key parts of the levee slope. Utilizing the inherent geometry of the levee (slope and aspect) as strong constraints, a quasi-three-dimensional decomposition of single-track InSAR data is achieved through a Kalman filter algorithm. This accurately distinguishes between levee crest settlement (vertical) and levee slope creep (aspect-wise), providing precise data for engineering management.

[0149] Furthermore, it integrates the high spatial coverage of InSAR with the high temporal resolution of BeiDou. Through the time-series prediction function of Kalman filtering, even during satellite revisit intervals, it can use BeiDou data to drive models to deduce the deformation state of the entire field, achieving integrated "air-ground" continuous monitoring.

[0150] It should be noted that the deformation monitoring device for dikes provided by the present invention can execute the deformation monitoring method for dikes in any of the above embodiments during specific operation, which will not be elaborated in this embodiment.

[0151] like Figure 5 As shown, the deformation monitoring device for dikes provided by the present invention includes: The first acquisition module 501 is used to acquire the first three-dimensional deformation observation value, which is the three-dimensional deformation observation value of the dike determined by the Beidou satellite. The second acquisition module 502 is used to acquire the first deformation observation value, which is the deformation observation value of the embankment in the line of sight direction; The processing module 503 is used to input the first three-dimensional deformation observation value and the first deformation observation value into the spatiotemporal fusion model constructed based on Kalman filtering, so as to obtain the three-dimensional deformation field of the embankment output by the spatiotemporal fusion model. Among them, the observation equation of the spatiotemporal fusion model introduces a physical geometric constraint matrix determined based on the geometric parameters of the dike, and the three-dimensional deformation field of the dike is generated under the physical geometric constraint matrix.

[0152] In this embodiment, by fusing the three-dimensional deformation observations from BeiDou satellites and the line-of-sight deformation observations from synthetic aperture radar, and by using a physical geometric constraint matrix to ensure the rationality of the solution results, not only is the temporal and spatial resolution of levee deformation monitoring achieved, but the output three-dimensional deformation field is also made to better fit the actual deformation law of the levee. This effectively improves the accuracy and reliability of levee deformation monitoring results and provides accurate data support for the timely identification of levee safety hazards.

[0153] Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 6 As shown, the electronic device may include a processor 610, a communication interface 620, a memory 630, and a communication bus 640. The processor 610, communication interface 620, and memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute a method for monitoring the deformation of a levee. This method includes: acquiring a first three-dimensional deformation observation value, which is a three-dimensional deformation observation value of the levee determined using BeiDou satellites; acquiring the first deformation observation value, which is a deformation observation value of the levee in the line-of-sight direction; inputting the first three-dimensional deformation observation value and the first deformation observation value into a spatiotemporal fusion model constructed based on Kalman filtering to obtain a three-dimensional deformation field of the levee output by the spatiotemporal fusion model; wherein, the observation equation of the spatiotemporal fusion model incorporates a physical geometric constraint matrix determined based on the geometric parameters of the levee, and the three-dimensional deformation field of the levee is generated under the physical geometric constraint matrix.

[0154] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes 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.

[0155] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by the computer, the computer can execute the deformation monitoring method for the dike provided in the above embodiments. The method includes: acquiring a first three-dimensional deformation observation value, which is a three-dimensional deformation observation value of the dike determined by the Beidou satellite; acquiring the first deformation observation value, which is a deformation observation value of the dike in the line-of-sight direction; inputting the first three-dimensional deformation observation value and the first deformation observation value into a spatiotemporal fusion model constructed based on Kalman filtering to obtain a three-dimensional deformation field of the dike output by the spatiotemporal fusion model; wherein, the observation equation of the spatiotemporal fusion model introduces a physical geometric constraint matrix determined based on the geometric parameters of the dike, and the three-dimensional deformation field of the dike is generated under the physical geometric constraint matrix.

[0156] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program is implemented to perform the deformation monitoring method for the dike provided in the above embodiments. The method includes: acquiring a first three-dimensional deformation observation value, wherein the first three-dimensional deformation observation value is a three-dimensional deformation observation value of the dike determined using Beidou satellites; acquiring the first deformation observation value, wherein the first deformation observation value is a deformation observation value of the dike in the line-of-sight direction; inputting the first three-dimensional deformation observation value and the first deformation observation value into a spatiotemporal fusion model constructed based on Kalman filtering to obtain a three-dimensional deformation field of the dike output by the spatiotemporal fusion model; wherein the observation equation of the spatiotemporal fusion model introduces a physical geometric constraint matrix determined based on the geometric parameters of the dike, and the three-dimensional deformation field of the dike is generated under the physical geometric constraint matrix.

[0157] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and 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 modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0158] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0159] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring the deformation of a dike, characterized in that, include: Obtain the first three-dimensional deformation observation value, which is the three-dimensional deformation observation value of the dike determined by the Beidou satellite; Obtain a first deformation observation value, which is the deformation observation value of the embankment in the line of sight direction; The first three-dimensional deformation observation value and the first deformation observation value are input into the spatiotemporal fusion model constructed based on Kalman filtering to obtain the three-dimensional deformation field of the dike output by the spatiotemporal fusion model; The observation equation of the spatiotemporal fusion model incorporates a physical geometric constraint matrix determined based on the geometric parameters of the embankment, and the three-dimensional deformation field of the embankment is generated under the physical geometric constraint matrix.

2. The method for monitoring the deformation of a dike according to claim 1, characterized in that, The physical geometric constraint matrix is ​​determined using the following steps: Based on the geometric parameters of the embankment, the normal vector of the embankment slope is determined in the station center coordinate system; A pseudo-observation equation is constructed based on the normal vector of the embankment slope. Based on the pseudo-observation equation, the projection of the three-dimensional deformation vector of the embankment slope onto the normal vector of the embankment slope is set to zero, thus obtaining the physical geometric constraint matrix.

3. The method for monitoring the deformation of a dike according to claim 2, characterized in that, The geometric parameters of the embankment include its slope and aspect.

4. The method for monitoring the deformation of a dike according to claim 1, characterized in that, The method for monitoring the deformation of the dike also includes: Based on the three-dimensional deformation field of the embankment, it is determined that the acceleration of the first observation point down the slope is greater than the preset acceleration, thus determining that there is a first hidden danger event, and outputting the first warning information corresponding to the first hidden danger event. The first observation point is an observation point located on the embankment. Based on the three-dimensional deformation field of the embankment, it is determined that there is continuous vertical settlement at the second observation point, a second hidden danger event is identified, and a second early warning information corresponding to the second hidden danger event is output. The second observation point is located on the embankment, and the second potential incident is that the embankment has foundation consolidation or uneven settlement.

5. The method for monitoring the deformation of a dike according to any one of claims 1 to 4, characterized in that, The process of obtaining the first deformation observation value includes: Obtain a three-dimensional geometric model of the dike, the three-dimensional geometric model including multiple grid units divided according to the terrain; The geometric distortion index of each grid cell is determined based on the first orbital parameters and the geometric parameters of the grid cell, and a geometric mask is determined based on the geometric distortion index. The first orbital parameters are the orbital parameters of the Earth observation remote sensing satellite when it observes the dike. The Earth observation remote sensing satellite is equipped with a synthetic aperture radar. Under the constraints of the geometric mask, permanent and distributed scatterers are identified in the synthetic aperture radar images; Phase unwrapping is performed on the observation points on the permanent scatterer and the observation points on the distributed scatterer to obtain the first deformation observation value.

6. The method for monitoring the deformation of a dike according to claim 5, characterized in that, The process of identifying permanent and distributed scatterers in the synthetic aperture radar image under the constraint of the geometric mask includes: Based on the geometric mask, candidate points in the image captured by the synthetic aperture radar are determined; The candidate points are distinguished based on the amplitude deviation index and spatial statistical test to obtain permanent scatterers and distributed scatterers.

7. The method for monitoring the deformation of a dike according to claim 6, characterized in that, The step of identifying permanent and distributed scatterers in the synthetic aperture radar image under the constraint of the geometric mask further includes: Identify the homogeneous pixels associated with each monitoring point in the distributed scatterer; The target monitoring point is retained because the number of homogeneous pixels associated with the target monitoring point in the distributed scatterer is greater than a preset number. The target monitoring point is removed if the number of homogeneous pixels associated with the target monitoring point in the distributed scatterer is less than or equal to a preset number. The target monitoring point is a monitoring point in a distributed scatterer.

8. The method for monitoring the deformation of a dike according to claim 7, characterized in that, The step of identifying permanent and distributed scatterers in the synthetic aperture radar image under the constraint of the geometric mask further includes: The coherence matrix of each monitoring point is determined based on the homogeneous pixels associated with the monitoring point; The eigenvector corresponding to the largest eigenvalue is determined based on the coherence matrix of the monitoring points; The original noise phase of the monitoring point is replaced with the eigenvector corresponding to the largest eigenvalue.

9. A deformation monitoring device for a dike, characterized in that, include: The first acquisition module is used to acquire the first three-dimensional deformation observation value, which is the three-dimensional deformation observation value of the dike determined by the Beidou satellite. The second acquisition module is used to acquire the first deformation observation value, which is the deformation observation value of the dike in the line of sight direction; The processing module is used to input the first three-dimensional deformation observation value and the first deformation observation value into the spatiotemporal fusion model constructed based on Kalman filtering, so as to obtain the three-dimensional deformation field of the dike output by the spatiotemporal fusion model; The observation equation of the spatiotemporal fusion model incorporates a physical geometric constraint matrix determined based on the geometric parameters of the embankment, and the three-dimensional deformation field of the embankment is generated under the physical geometric constraint matrix.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the deformation monitoring method for the dike as described in any one of claims 1 to 8.

Images