A road deformation risk area identification method, device, equipment and storage medium

Abstract

The application relates to a road area deformation risk area identification method and device, equipment and a storage medium, and relates to the technical field of highway geological disaster remote sensing monitoring and risk assessment. The method comprises the following steps: obtaining a plurality of deformation monitoring points and deformation information thereof by solving based on time sequence SAR images and DEM data of a target road area range, projecting the deformation information to a slope surface direction to obtain slope surface deformation information; constructing a multi-dimensional feature vector based on the slope surface deformation information, and calculating a comprehensive anomaly index of each deformation monitoring point; determining an adaptive threshold according to the distribution characteristics of the comprehensive anomaly index, and determining a deformation monitoring point with a comprehensive anomaly index greater than the adaptive threshold as a deformation anomaly point; and performing density clustering on the deformation anomaly points, and determining each clustering cluster obtained through the clustering as a deformation risk area. The method solves the problem that a fixed threshold screening method in the prior art cannot adapt to noise characteristics of different regions, a large number of noise points are misjudged as anomaly points, or real hidden danger points are missed, and deformation risks cannot be accurately identified.

Description

Technical Field

[0001] This invention relates to the field of remote sensing monitoring and risk assessment technology for highway geological disasters, specifically to a method, device, equipment, and storage medium for identifying road deformation risk zones. Background Technology

[0002] With the rapid development of transportation infrastructure, the highway network is expanding and extending into mountainous areas with complex geological conditions. Geological disasters such as landslides and ground subsidence seriously threaten operational safety. Therefore, utilizing remote sensing technology for early identification and risk assessment of deformation hazards over large road areas is of great significance for disaster warning and maintenance decision-making. Synthetic Aperture Radar Interferometry (InSAR) technology has become the main monitoring method in this field due to its advantages of large-scale, high-precision, and all-weather monitoring.

[0003] In existing technologies, road deformation monitoring methods based on InSAR technology typically use time-series SAR images to calculate deformation rates, and then use a fixed deformation rate threshold to screen out anomalies. Subsequently, a clustering algorithm based on spatial distance is used to delineate the risk zone, and then the risk level is assessed based on deformation indicators.

[0004] However, when facing a large and complex road environment, the existing fixed threshold screening method is difficult to adapt to the noise characteristics of different areas because the road area covers a variety of landforms and the background noise level varies greatly. This results in a large number of noise points being misjudged as abnormal points or real hidden danger points being missed, and there is a problem that it is difficult to accurately identify deformation risks. Summary of the Invention

[0005] This application provides a method, apparatus, device, and storage medium for identifying road deformation risk zones. It can solve the problem in the prior art that when facing a large-scale and complex road environment, due to the large-scale road area covering various landform types and huge differences in background noise levels, the existing fixed threshold screening method is difficult to adapt to the noise characteristics of different areas, resulting in a large number of noise points being misjudged as abnormal points or real hidden danger points being missed, and thus making it difficult to accurately identify deformation risks.

[0006] In a first aspect, embodiments of this application provide a method for identifying road deformation risk zones, comprising: Based on the time-series SAR images and DEM data of the target road area, multiple deformation monitoring points and their deformation information are calculated, and the deformation information is projected onto the slope direction to obtain the slope deformation information. Based on the slope deformation information, a multi-dimensional feature vector is constructed, and the comprehensive anomaly index of each deformation monitoring point is calculated. An adaptive threshold is determined based on the distribution characteristics of the comprehensive anomaly index, and deformation monitoring points whose comprehensive anomaly index is greater than the adaptive threshold are identified as deformation anomaly points. Density clustering is performed on the deformation anomaly points, and each cluster obtained from the clustering is identified as a deformation risk area.

[0007] In one implementation, the time-series SAR imagery and DEM data based on the target road area are used to calculate multiple deformation monitoring points and their deformation information, including: Acquire ascending orbit time-series SAR images and descending orbit time-series SAR images covering the target road area; The ascending orbit time-series SAR images and descending orbit time-series SAR images are registered and interferometrically processed to generate interferogram sequences. The terrain phase was removed using the DEM data, and deformation monitoring points were extracted using permanent scatterer (PS) and distributed scatterer (DS) technologies. The deformation monitoring points are jointly solved based on the time series analysis model to obtain the deformation rate and time series deformation information of each deformation monitoring point.

[0008] In one embodiment, projecting the deformation information onto the slope direction to obtain slope deformation information includes: Based on the DEM data, calculate the slope and aspect of each deformation monitoring point. The deformation rate and time-series deformation information are decomposed into vertical deformation components and east-west deformation components. A geometric projection model of the radar line of sight and the slope direction is constructed. Using the slope, aspect, vertical deformation component and east-west deformation component, the slope deformation information of each deformation monitoring point along the slope direction is calculated.

[0009] In one implementation, the joint calculation of the deformation monitoring points based on the time-series analysis model to obtain the deformation rate and time-series deformation information of each deformation monitoring point includes: We constructed an interferogram network based on spatiotemporal baseline thresholds, and the observation equations relating interferometric phase to deformation parameters, topographic errors, and atmospheric phase. The observation equations are solved using the least squares method or singular value decomposition method to separate the linear deformation components and the nonlinear deformation components. The deformation rate is obtained from the linear deformation component, and the time-series deformation information is obtained from the nonlinear deformation component.

[0010] In one implementation, the step of constructing a multi-dimensional feature vector based on the slope deformation information and calculating the comprehensive anomaly index for each deformation monitoring point includes: Extract the deformation rate, deformation acceleration, cumulative deformation, and data coherence from the slope deformation information of each deformation monitoring point; The deformation rate, deformation acceleration, cumulative deformation, and data coherence are combined to form the multidimensional feature vector. The features of each dimension in the multidimensional feature vector are standardized to eliminate the influence of units; Based on the weight coefficients of each dimension of features, the weighted sum of the standardized features of each dimension is calculated to obtain the comprehensive anomaly index.

[0011] In one implementation, determining an adaptive threshold based on the distribution characteristics of the comprehensive anomaly index, and identifying deformation monitoring points where the comprehensive anomaly index is greater than the adaptive threshold as deformation anomaly points, includes: The comprehensive anomaly index of all deformation monitoring points within the target road area is statistically analyzed, and a frequency distribution histogram of the comprehensive anomaly index is constructed. Identify the trough positions between the background noise distribution and the abnormal signal distribution in the frequency distribution histogram, and determine the comprehensive anomaly index value corresponding to the trough position as the adaptive threshold. The comprehensive anomaly index of all deformation monitoring points is compared with the adaptive threshold, and the deformation monitoring points that are greater than the adaptive threshold are identified as deformation anomaly points.

[0012] In one implementation, the step of performing density clustering on the deformation anomaly points and identifying each cluster as a deformation risk region includes: Calculate the spatial Euclidean distance and eigenvector distance between any two deformation anomaly points; Set the spatial neighborhood radius and feature similarity threshold; When the spatial Euclidean distance between two deformation anomaly points is less than the spatial neighborhood radius and the feature vector distance is less than the feature similarity threshold, it is determined that the density of the two deformation anomaly points is achievable. Based on density reachability, the deformation anomaly points are divided into several independent clusters, and each cluster is identified as a deformation risk zone.

[0013] Secondly, embodiments of this application also provide a road deformation risk zone identification device, which includes: The deformation information calculation module is used to calculate multiple deformation monitoring points and their deformation information based on time-series SAR images and DEM data of the target road area, and project the deformation information onto the slope direction to obtain slope deformation information. The index calculation module is used to construct a multi-dimensional feature vector based on the slope deformation information and calculate the comprehensive anomaly index of each deformation monitoring point. An anomaly point determination module is used to determine an adaptive threshold based on the distribution characteristics of the comprehensive anomaly index, and to determine deformation monitoring points whose comprehensive anomaly index is greater than the adaptive threshold as deformation anomaly points. The risk zone determination module is used to perform density clustering on the deformation anomaly points and determine each cluster obtained by clustering as a deformation risk zone.

[0014] Thirdly, this application also provides a road deformation risk zone identification device, which includes a processor, a memory, and a road deformation risk zone identification program stored in the memory and executable by the processor. When the road deformation risk zone identification program is executed by the processor, it implements the steps of the road deformation risk zone identification method described above.

[0015] Fourthly, this application embodiment also provides a storage medium storing a road deformation risk zone identification program, wherein when the road deformation risk zone identification program is executed by a processor, it implements the steps of the road deformation risk zone identification method described above.

[0016] The beneficial effects of the technical solutions provided in this application include: By calculating deformation information from time-series SAR imagery and DEM data within the target road area and projecting it onto the slope direction, this method overcomes the limitation of traditional InSAR technology, which only provides one-dimensional deformation along the radar line of sight. It can more realistically reflect the actual sliding state of the slope along the slope direction, improving the physical interpretability and accuracy of deformation monitoring results. By constructing multi-dimensional feature vectors to calculate a comprehensive anomaly index and determining an adaptive threshold based on its distribution characteristics, it can effectively adapt to the complex and varied background noise levels across a large road area, significantly reducing the risk of false alarms and missed alarms caused by a single fixed threshold. Furthermore, density clustering of deformation anomaly points enables precise and automatic delineation from discrete monitoring points to continuous risk zones, closely matching the true boundaries of irregular deformation areas. This provides a reliable spatial object basis and technical support for scientific early warning and maintenance decisions regarding highway geological disasters. This invention addresses the problem that existing technologies struggle to accurately identify deformation risks when facing large-scale, complex road environments. These environments often involve diverse terrain types and varying background noise levels, making it difficult for existing fixed-threshold screening methods to adapt to the noise characteristics of different areas. This results in numerous noise points being misjudged as anomalies or real potential hazards being missed. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the first embodiment of a method for identifying road deformation risk zones according to the present invention.

[0019] Figure 2 This is a schematic diagram of the hardware structure of the road deformation risk zone identification device involved in the embodiments of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0021] Firstly, this application provides a method for identifying road deformation risk zones.

[0022] In one embodiment, reference is made to Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the road deformation risk zone identification method of this application. Figure 1 As shown, the method for identifying road deformation risk zones includes: S1: Based on the time-series SAR images and DEM data of the target road area, multiple deformation monitoring points and their deformation information are calculated, and the deformation information is projected onto the slope direction to obtain the slope deformation information; S2: Construct a multi-dimensional feature vector based on slope deformation information and calculate the comprehensive anomaly index for each deformation monitoring point; S3: Determine the adaptive threshold based on the distribution characteristics of the comprehensive anomaly index, and identify deformation monitoring points with a comprehensive anomaly index greater than the adaptive threshold as deformation anomaly points; S4: Perform density clustering on the deformation anomaly points, and identify each cluster as a deformation risk area.

[0023] By calculating deformation information from time-series SAR imagery and DEM data within the target road area and projecting it onto the slope direction, this method overcomes the limitation of traditional InSAR technology, which only provides one-dimensional deformation along the radar line of sight. It can more realistically reflect the actual sliding state of the slope along the slope direction, improving the physical interpretability and accuracy of deformation monitoring results. By constructing multi-dimensional feature vectors to calculate a comprehensive anomaly index and determining an adaptive threshold based on its distribution characteristics, it can effectively adapt to the complex and varied background noise levels across a large road area, significantly reducing the risk of false alarms and missed alarms caused by a single fixed threshold. Furthermore, density clustering of deformation anomaly points enables precise and automatic delineation from discrete monitoring points to continuous risk zones, closely matching the true boundaries of irregular deformation areas. This provides a reliable spatial object basis and technical support for scientific early warning and maintenance decisions regarding highway geological disasters.

[0024] In this example, after identifying each cluster obtained from clustering as a deformation risk region, the process also includes: Based on the acquired deformation risk areas, multidimensional deformation features and terrain features of each deformation risk area are extracted; Based on the multidimensional deformation characteristics and topographic features of each deformation risk zone, a comprehensive risk assessment is conducted for each deformation risk zone, and the risk classification results are output.

[0025] Specifically, multi-dimensional feature extraction of deformation risk areas: For each polygon in the deformation risk area, extract deformation intensity, spatial features, terrain features, and data reliability features respectively; if the... Deformation risk area The number of monitoring points is The extracted features of each type are as follows: Deformation intensity characteristics Deformation rate : Using the deformation rate along the slope direction, the maximum absolute value of the deformation within the polygon is taken. Compared with the average Weighted composite value This reflects the severity of the current deformation. The higher the deformation rate, the more violent the slope movement and the greater the probability of instability.

[0026] Cumulative deformation : This represents the cumulative deformation average of all monitoring points within the polygon; Deformation acceleration The maximum acceleration within the polygon is taken to reflect the dynamic evolution trend of the deformation process. Spatial features Polygon area : Get the area range of the polygon; the larger the area, the wider the potential influence range. Deformation gradient The maximum value of the average gradient of the longitudinal and transverse sections inside the polygon is taken, which reflects the rate of change of deformation space in the most unfavorable direction. Polygon clustering neighborhood consistency This reflects the spatial clustering strength of outliers within a cluster during the clustering process. It is obtained by aggregating the proportion of outliers in the neighborhood, as shown below: ,in, For point The percentage of outliers within the neighborhood. The closer the value is to 1, the higher the spatial clustering of anomalies within the deformation risk zone, and the more reliable the clustering results.

[0027] Topographic features terrain slope : Take the average terrain slope within the polygon; the steeper the slope, the higher the probability of slope instability. Deformation-aspect consistency coefficient This reflects the degree of agreement between the deformation direction and the slope aspect. The calculation formula is as follows: ,in, For the first The deformation direction angle of each monitoring point (with true north as 0° and clockwise as positive). For the first The slope angle of a pixel point The closer the value is to 1, the more consistent the slope direction within the polygon, which is more conducive to determining that they are the same sliding body.

[0028] Location sensitivity index The degree of threat to highway safety is quantified based on the distance relationship between the deformation zone and highway facilities. ,in, Deformation risk zone polygon The shortest distance to the highway alignment. For safe distance threshold, The attenuation coefficient is... It is a natural constant; the index takes a maximum value of 1.0 within the safe distance range. The greater the distance, the lower the threat level, reflecting that the closer the deformation risk zone is to the highway, the greater the threat to highway safety.

[0029] Data reliability characteristics coherence The average coherence of each deformation point within the polygon is taken to reflect the overall reliability of the InSAR monitoring data in the deformation risk area. The higher the value, the more reliable the monitoring results. Time series morphology types within polygons The value is determined by the mode of the time series pattern at each point, and the calculation formula is as follows: ,in, For the first The time series displacement function of points. The time series morphology classification function determines the morphology type based on the sign of the second derivative of the displacement-time curve: if the second derivative is greater than 0 for more than 3 consecutive windows in the time series, it is classified as an accelerating type; if there is a step change (the displacement increment between adjacent periods exceeds 3 times the absolute deviation of the median), it is classified as a step type; if the absolute value of the second derivative is consistently less than the threshold, it is classified as a linear type; if there is a significant annual cycle fluctuation, it is classified as a seasonal fluctuation type; if the second derivative is consistently less than 0 and the cumulative displacement tends to stabilize, it is classified as a decelerating and stabilizing type. The mode function is used to select the most frequently occurring shape type within the polygon as the overall shape type of the deformation risk zone. The risk scores for each shape type are as follows: Accelerating type 1.0, Step type 0.8, Linear type 0.6, Seasonal fluctuation type 0.4, and Decelerating and stabilizing type 0.2.

[0030] Multi-indicator weighted risk scoring model: The weighted summation model is used to calculate the comprehensive risk score for each deformation risk zone. ,in, For the first Comprehensive risk score of each deformation risk zone polygon; For the first The weights of each indicator satisfy ; For the first The first deformation risk zone The normalized values ​​of each indicator.

[0031] The normalization method depends on the type of indicator. For indicators that are more dangerous the larger they are (deformation rate, cumulative deformation, deformation acceleration, deformation gradient, area, neighborhood consistency, slope, aspect consistency, location sensitivity), the normalization method is as follows: ,in, For the first The first deformation risk zone The original values ​​of each indicator The corresponding maximum value; This is a minimum function that truncates the normalized result to 1.0 to avoid extreme values ​​having an excessive impact on the score.

[0032] For coherence indicators, since they reflect the reliability of monitoring data rather than the safety level of the slope itself, the normalization method differs from that of direct hazard indicators such as deformation intensity. Higher coherence indicates more reliable monitoring results, and the risk score should be given a higher confidence level; lower coherence indicates insufficient data quality, and the risk score should be appropriately lowered to avoid false alarms. The normalization method is as follows: ,in, For the first The average coherence of each deformation risk region The corresponding maximum value is ; when the coherence is close to the upper limit of the reference value, the normalized risk contribution value approaches 0 and does not impose an additional burden on the comprehensive risk score; the lower the coherence, the larger the normalized risk contribution value, indicating that the risk classification result of the deformation risk area should be appropriately lowered in confidence due to insufficient data reliability.

[0033] Deformation risk classification: based on comprehensive risk score The deformation risk zone is divided into four risk levels:

[0034] Based on the above, the risk levels of each deformation risk zone can be classified. The characteristics and recommended measures for each risk level are as follows: Level I (High Risk): The deformation zone shows a clear accelerating trend or a large cumulative deformation, and has high sensitivity to topographic conditions and location, with high consistency with neighboring areas and good data reliability. Priority should be given to on-site verification and emergency response. It is recommended to complete the on-site investigation within 7 working days and initiate emergency monitoring as needed.

[0035] Level II (Medium Risk): Persistent deformation exists in the deformation zone, but the rate and acceleration have not yet reached dangerous levels, and the deformation signal is reliable. Regular monitoring and on-site investigation are required. It is recommended to complete the on-site verification within one month and include it in the regular monitoring plan.

[0036] Level III (low-risk) deformation zones exhibit minor deformation at a low rate or with the deformation tending to stabilize, indicating a low risk. These zones can be included in routine inspections, with deformation patterns checked quarterly or semi-annually.

[0037] Level IV (Negligible Risk): Deformation signals are weak or data reliability is low; no obvious signs of slope instability have been found. No specific action is required at this time, but continued monitoring in subsequent monitoring cycles is recommended.

[0038] Furthermore, in one embodiment, the temporal SAR imagery and DEM data based on the target road area are used to calculate multiple deformation monitoring points and their deformation information, including: Acquire rising orbit time-series SAR images and falling orbit time-series SAR images covering the target road area; Registration and interferometry were performed on ascending-orbit and descending-orbit time-series SAR images respectively to generate interferogram sequences. Topographic phases were removed using DEM data, and deformation monitoring points were extracted using permanent scatterer (PS) and distributed scatterer (DS) technologies. The deformation monitoring points are jointly solved based on the time series analysis model to obtain the deformation rate and time series deformation information of each deformation monitoring point.

[0039] In this embodiment, by acquiring ascending and descending time-series SAR images covering the target road area and performing joint calculations, the geometric distortion limitations of single-orbit radar line-of-sight observations are overcome. This allows for the inversion of vertical and east-west deformation components that more closely approximate the actual surface movement, significantly improving the accuracy and reliability of deformation calculations. Simultaneously, the use of permanent scatterer (PS) technology and distributed scatterer (DS) technology to extract deformation monitoring points effectively solves the problem of sparse monitoring points in natural slope areas using traditional methods, greatly improving the spatial coverage and point density of road area deformation monitoring. Furthermore, based on the time-series analysis model, deformation rate and time-series deformation information are obtained, not only revealing the current state of deformation but also preserving the historical evolution of deformation, providing rich and high-quality data support for subsequent construction of multi-dimensional feature vectors and analysis of deformation trends.

[0040] Furthermore, in one embodiment, the step of projecting the deformation information onto the slope direction to obtain slope deformation information includes: Based on DEM data, calculate the slope and aspect of each deformation monitoring point. The deformation rate and time-series deformation information are decomposed into vertical deformation components and east-west deformation components; A geometric projection model of the radar line of sight and the slope direction is constructed. Using the slope, aspect, vertical deformation components and east-west deformation components, the slope deformation information of each deformation monitoring point along the slope direction is calculated.

[0041] In this embodiment, by projecting deformation information onto the slope direction, the limitation of traditional InSAR technology, which can only provide deformation along the radar line of sight but cannot directly reflect the actual sliding direction of the slope, is overcome. This makes the deformation data more consistent with the physical and mechanical mechanism of slope instability and sliding along the slope. By constructing a geometric projection model using slope, aspect, and vertical and east-west deformation components, the projection error of deformation components caused by satellite orbit geometry is eliminated, significantly improving the physical interpretability of deformation monitoring results for slope stability. In particular, for slopes with a large angle between the direction and the radar line of sight, this method effectively improves the sensitivity of deformation monitoring, ensuring that subsequent risk assessment is based on more realistic and critical slope deformation data, thereby improving the accuracy of risk identification.

[0042] Furthermore, in one embodiment, the joint calculation of deformation monitoring points based on the time-series analysis model to obtain the deformation rate and time-series deformation information of each deformation monitoring point includes: We constructed an interferogram network based on spatiotemporal baseline thresholds, and the observation equations relating interferometric phase to deformation parameters, topographic errors, and atmospheric phase. The observation equations are solved using the least squares method or singular value decomposition method to separate linear and nonlinear deformation components; The deformation rate is obtained from the linear deformation component, and the time-series deformation information is obtained from the nonlinear deformation component.

[0043] In this embodiment, by constructing an interferogram network based on spatiotemporal baseline thresholds and an observation equation between the interferometric phase and multiple error sources, deformation parameters, topographic errors, and atmospheric phases can be effectively constrained, improving the reliability and anti-interference capability of parameter calculation. By using the least squares method or singular value decomposition method to solve the observation equations, linear and nonlinear deformation components can be robustly separated. This not only yields the deformation rate reflecting the long-term stable trend but also retains the time-series deformation information reflecting the dynamic evolution of deformation. This precise separation of linear and nonlinear components provides the necessary data foundation for the subsequent construction of multidimensional feature vectors including deformation acceleration, enhancing the temporal resolution and accuracy of deformation monitoring results, thereby improving the sensitivity and accuracy of risk identification.

[0044] Specifically, during PS and DS point extraction, a combination of amplitude deviation index method and phase stability analysis is used to identify highly coherent PS points in the road area for both time-series ascending and descending SAR images. The FaSHPS algorithm is then used for homogeneous pixel identification and coherence estimation, and DS points are selected based on the number of homogeneous pixels and coherence. The extracted PS and DS points are then fused to construct a dense deformation monitoring point cloud covering the road area. During ascending and descending deformation extraction, time-series InSAR methods such as PS-InSAR are used to perform spatiotemporal network joint calculation and deformation information inversion processing on the identified PS and DS points, respectively, to obtain deformation information of the road area under ascending and descending orbits. For east-west and vertical deformation calculation, the LOS deformation rates obtained from ascending and descending orbits are used... , These are respectively represented as east-west displacements ( ) and vertical displacement ( The linear combination of , ignoring the north-south component (due to the satellite's near-polar orbit characteristics making it extremely insensitive to north-south motion), is expressed as follows: , where the projection coefficient , Angle of incidence from the satellite and flight azimuth Decide: After substituting the angle parameters into the ascending and descending orbits, the east-west horizontal deformation rate and vertical deformation rate at each monitoring point can be directly calculated. When obtaining the deformation rate along the slope direction, since the main direction of motion for highway slope instability is usually consistent with the slope direction (direction of maximum gradient), by using the obtained vertical and east-west deformation rates, combined with the slope and aspect calculated from the DEM topographic data, and projecting the deformation rate onto the slope direction, the stability state of the slope can be reflected more directly. Let the slope at a certain point be... The slope direction is Then the unit vector in the slope direction is Where the x-axis points east, the y-axis points north, and the z-axis points towards the zenith (upwards). If the three-dimensional deformation rate vector of a monitoring point is... , Since the deformation is in the north-south direction, it can be approximated as negligible. Therefore, the deformation rate in the slope direction is: When the conditions for joint calculation of ascending and descending orbits are lacking, an approximate estimate can be made using the deformation rate in the LOS direction of a single orbit and the satellite incident angle: Through the aforementioned projection, the deformation rate along the slope direction at each deformation monitoring point is obtained. This parameter directly reflects the deformation intensity of the slope along the potential sliding direction and is a key input for subsequent deformation risk zone identification and risk classification. During multidimensional deformation feature extraction, based on the acquired original time-series displacement, vertical / east-west deformation rate, and slope-direction deformation rate, multidimensional deformation features are extracted from both temporal and spatial dimensions to construct a multidimensional feature system for road area deformation risk identification. Specifically, this includes: cumulative deformation... The total displacement during the observation period, extracted from the original time series displacement, reflects the overall magnitude of slope deformation, specifically: ,in, This represents the displacement increment between two consecutive observation periods. represents the observation period. To suppress the influence of time series noise, a robust linear fitting method can be used to estimate the cumulative deformation: For deformation sequences exhibiting significant seasonal fluctuations (such as slopes affected by freeze-thaw cycles and seasonal rainfall variations), the periodic term should be removed before calculating the cumulative trend to eliminate the interference of periodic environmental factors on the cumulative deformation estimation. The decomposition model is as follows: ,in, for The original cumulative deformation at time; The net deformation rate after removing seasonal fluctuations represents the long-term trend movement of the slope. For seasonal amplitude, For the initial phase, It is an annual cycle. This represents random noise.

[0045] Deformation acceleration: Based on the original time-series displacement, a sliding window piecewise fitting method is used to calculate the deformation acceleration, reflecting whether the deformation process is accelerating and deteriorating, serving as a dynamic and forward-looking indicator for risk assessment. The deformation time series is divided into continuous sliding windows (window length...). sliding step size To obtain the local deformation rate of a window, a linear fit is performed on the displacement sequence within each window. ,in, For the first A sliding window, The average time within the window represents the deformation acceleration, and the rate change between adjacent windows is the deformation acceleration. The sign of deformation acceleration has a clear physical meaning: when At that time, the deformation is accelerating; when When, the deformation remains uniform; when At that time, the deformation is decelerating. Deformation acceleration, as a dynamic indicator for risk assessment, is incorporated into the subsequent risk grading indicator system.

[0046] Slope deformation rate: reflects the deformation intensity of the slope along the potential sliding direction, which is obtained directly in the previous step.

[0047] Deformation gradient: Based on the deformation rate along the slope direction, the longitudinal gradient and the cross-sectional gradient perpendicular to the road direction are calculated respectively. This reflects the rate of change of deformation over a short distance and characterizes the degree of uneven deformation. Assume the road alignment starts from the centerline point... Defined as follows: the rate of deformation of the slope at each point is... The adjacent spacing is Then the deformation gradient along the route direction is: For discrete PS / DS point cloud data, the inverse distance weighting method is used to interpolate the point cloud deformation rate to the highway centerline point: The cross-sectional deformation gradient perpendicular to the route direction is: ,in, The sampling interval for the cross section. and These represent the deformation rates on both sides of the midline point along the normal direction.

[0048] Furthermore, in one embodiment, the construction of a multi-dimensional feature vector based on slope deformation information and the calculation of the comprehensive anomaly index for each deformation monitoring point include: Extract the deformation rate, deformation acceleration, cumulative deformation, and data coherence from the slope deformation information of each deformation monitoring point; The deformation rate, deformation acceleration, cumulative deformation, and data coherence are combined to form a multidimensional feature vector; Standardize the features of each dimension in the multidimensional feature vector to eliminate the influence of dimensions; Based on the weight coefficients of each dimension of features, the weighted sum of the standardized features is calculated to obtain the comprehensive anomaly index.

[0049] In this embodiment, a multi-dimensional feature vector is constructed based on slope deformation information, including deformation rate, deformation acceleration, cumulative deformation, and data coherence. This overcomes the limitations of existing technologies that rely solely on a single deformation rate indicator for judgment. It can simultaneously characterize the current intensity, evolution trend, cumulative effect, and data reliability of deformation. In particular, the introduction of deformation acceleration features can effectively capture potential hidden dangers that develop at low speeds but accelerate. By standardizing the features of each dimension, the differences between different physical dimensions are effectively eliminated, ensuring a balanced contribution of each feature in the evaluation and preventing a single dominant feature from obscuring other key information. A comprehensive anomaly index is calculated based on weighting coefficients, fusing multi-dimensional complex information into a single quantifiable indicator. This not only improves the comprehensiveness and accuracy of anomaly point identification but also effectively eliminates low-quality noise points, providing a scientific and reliable basis for subsequent adaptive threshold screening.

[0050] Further, in one embodiment, the step of determining an adaptive threshold based on the distribution characteristics of the comprehensive anomaly index, and identifying deformation monitoring points with a comprehensive anomaly index greater than the adaptive threshold as deformation anomaly points, includes: The comprehensive anomaly index of all deformation monitoring points within the target road area is statistically analyzed, and a frequency distribution histogram of the comprehensive anomaly index is constructed. Identify the trough locations between the background noise distribution and the abnormal signal distribution in the frequency distribution histogram, and determine the comprehensive anomaly index value corresponding to the trough location as the adaptive threshold. The comprehensive anomaly index of all deformation monitoring points is compared with an adaptive threshold, and deformation monitoring points that exceed the adaptive threshold are identified as deformation anomaly points.

[0051] In this embodiment, an adaptive threshold is determined by constructing a comprehensive anomaly index frequency distribution histogram and identifying the trough positions between background noise and anomalous signal distributions. This overcomes the shortcomings of traditional fixed threshold methods, which cannot adapt to the large differences in background noise levels across a wide road area. This method can automatically distinguish between background noise and real anomalous signals based on the statistical distribution characteristics of the data itself, significantly reducing the risk of false detection and false negative detection caused by improper threshold setting. Especially in areas with complex geological backgrounds, it can effectively eliminate low-quality noise points while retaining real low-speed deformation hazard points, improving the robustness and accuracy of deformation anomaly point screening, and providing high-quality data input for subsequent risk area clustering.

[0052] Furthermore, in one embodiment, the density clustering of deformation anomaly points, and the determination of each cluster as a deformation risk region, includes: Calculate the spatial Euclidean distance and eigenvector distance between any two deformation anomaly points; Set the spatial neighborhood radius and feature similarity threshold; When the spatial Euclidean distance between two deformation anomaly points is less than the spatial neighborhood radius, and the feature vector distance is less than the feature similarity threshold, it is determined that the density of the two deformation anomaly points can reach the target value. Based on density reachability, deformation anomalies are divided into several independent clusters, and each cluster is identified as a deformation risk zone.

[0053] In this embodiment, by applying dual-constraint density clustering—spatial neighborhood constraints and deformation feature similarity constraints—to deformation anomaly points, the shortcomings of traditional clustering algorithms that rely solely on spatial distance while ignoring the consistency of deformation patterns are overcome. This effectively avoids the erroneous merging of spatially adjacent but significantly different noise points with real potential hazards. By setting spatial neighborhood radius and feature similarity threshold to determine density accessibility, clusters with independent deformation features can be automatically identified and divided, achieving a precise transformation from discrete anomaly points to continuous risk areas. This approach closely matches the true boundaries of irregular deformation regions, significantly improving the accuracy and reliability of risk area delineation and providing a spatially accurate and clearly defined object foundation for subsequent risk assessment.

[0054] Specifically, adaptive pre-screening of deformation anomalies: Traditional InSAR deformation clustering methods rely solely on absolute rate values ​​for threshold truncation, ignoring data quality, dynamic trends, and local background differences of deformation points. This can easily lead to missed detections of low-speed acceleration hazards and false alarms from noisy points. This scheme integrates deformation rate, temporal acceleration, coherence, and local spatial outlier levels to achieve highly reliable anomaly identification. The specific steps are as follows: Multidimensional anomaly component calculation: Calculate the slope deformation rate anomaly component for each deformation monitoring point. Acceleration anomaly components Coherence confidence component Spatial outlier components Abnormal components.

[0055] Anomaly component of slope deformation rate This reflects the degree to which the slope rate at that point deviates from the overall background level; it is calculated using a method based on the median absolute deviation. ,in, For the first The slope deformation rate at each deformation point This represents the median slope deformation rate at all deformation points in the study area. The median absolute deviation of the slope deformation rate sequence is a constant. This is a scaling factor for converting MAD into a consistent estimator of standard deviation.

[0056] Acceleration Anomaly Components It reflects the accelerating trend of deformation and is a more forward-looking indicator than the rate. It is calculated using a method based on the absolute deviation of the median. ,in, For the first Deformation acceleration at each deformation point This represents the median of the deformation accelerations at all deformation points in the study area. The median absolute deviation of the deformation acceleration sequence is a constant. This is a scaling factor for converting MAD into a consistent estimator of standard deviation.

[0057] Coherence confidence component This reflects the quality of InSAR data. The lower the coherence, the less reliable the data, and the lower the confidence level of the anomaly index should be. Define the data quality penalty term as follows: ,in, For the first The average coherence of each deformation point ranges from [0, 1] (0 represents complete incoherence, and 1 represents complete coherence). The larger the value of this component, the less reliable the data is; the smaller the value, the more reliable the data is.

[0058] Spatial outlier components This reflects the deformation difference of the point relative to its local neighborhood; using the concept of local spatial autocorrelation, the standardized deviation of the rate at this point from the average rate of the neighborhood is calculated. ,in, For the first The slope deformation rate at each deformation point For the first Centered on a point, spatial radius The average rate of all neighboring points within the range. This is the standard deviation component of the velocity values ​​within the corresponding neighborhood; this component can effectively identify isolated high-speed points and relatively stable points in the overall slip zone.

[0059] Multidimensional Comprehensive Anomaly Index Fusion: The four components mentioned above are robustly weighted and fused to construct a comprehensive anomaly index. The details are as follows: ,in, , , Let be the weighting coefficients of each component, satisfying... , This is a data quality penalty coefficient, which is applied when relevance is low ( (larger), denominator increases, The value is suppressed, thereby reducing the false detection rate of low-quality data; For the first The multidimensional comprehensive anomaly index of points.

[0060] Adaptive threshold determination and iterative truncation: An iterative truncation strategy is used to determine the threshold. Filtering threshold; initial calculation of all points The sequence is used to estimate its background distribution parameters using MAD: ,in, for the median of the sequence; for The absolute deviation of the median of the sequence; For all monitoring points in the study area A set of values. The initial screening criteria are: ,in, As the initial confidence factor, the system is re-estimated after successively removing the currently selected outliers. and The process continues until two consecutive estimated rates of change are less than a set threshold, at which point convergence occurs. The final adaptive threshold is determined by the converged statistic. ,in, and After iterative convergence The median and absolute deviation of the median of the sequence; According to The optimal confidence factor for adaptive adjustment of distribution skewness increases when the distribution is severely right-skewed. To reduce the false positive rate, and conversely, to increase the sensitivity; This is the final adaptive filtering threshold.

[0061] Pre-screening results output: will meet the requirements The points are marked as candidate points for deformation anomalies, and then proceed to the next step of slope constraint and cluster analysis.

[0062] Topographic slope constraints and morphological optimization: The topographic slope of each monitoring point was calculated using a DEM (Digital Elevation Model), and false detection points in gently sloping areas (non-slope areas) were eliminated. A multi-scale slope calculation method was adopted, taking into account both local details and regional trends. Details are as follows: Multi-scale slope calculation: Based on the standard 3×3 neighborhood Horn algorithm, a 5×5 neighborhood slope is added as a supplement to suppress the interference of local terrain noise on slope estimation. 3×3 neighborhood slope. The calculation method is as follows: , , ,in, The elevation change rate in the east-west direction of the 3×3 neighborhood. The elevation change rate is 3×3 in the north-south direction. For the 3×3 neighborhood window, the first Line 1 DEM elevation values ​​of the columns of pixels, Spatial resolution of the DEM raster; 5×5 neighborhood slope The slope is calculated in a similar way, and the final slope is the weighted average of the two values: ,in, The final terrain slope after multi-scale slope fusion; The weighting coefficient for the 3×3 neighborhood slope is adaptively determined based on the DEM data quality, and appropriately reduced when the data noise is high. To enhance the smoothing effect of the 5×5 neighborhood.

[0063] Dual slope screening: Applying slope constraints to each deformation anomaly point , For the first The final terrain slope at each monitoring point This is the lower limit for terrain slope, used to eliminate false positives in flat areas. It can be set... ; This is the upper limit of the terrain slope, used to exclude areas of rocky cliffs that are too steep to form continuous landslides. It can be set... .

[0064] Morphological filtering and spatial continuity enhancement: After slope screening, some isolated outliers may be retained due to residual terrain noise. A secondary filtering is then performed by introducing spatial continuity constraints; for each candidate point, its spatial radius is calculated. Candidate point density within : ,in, For the first The spatial density of each candidate point represents the degree of aggregation of candidate points in its local neighborhood. For the point Center, radius The total number of candidate points within the circular neighborhood. Retaining the points that satisfy... , The density threshold is used to constrain spatial continuity. Points below this threshold are considered isolated noise points and are removed, thereby enhancing the spatial continuity of deformation anomalies.

[0065] Density Clustering Based on Multi-Source Feature Fusion and Dual Constraints: This invention employs a density clustering algorithm based on multi-dimensional feature vector similarity and spatial neighborhood constraints to ensure consistency in clustering results across deformation features and terrain conditions. The specific steps are as follows: Multidimensional feature vector construction: A multidimensional feature vector is constructed for each deformation anomaly point that passes pre-screening and slope constraints, comprehensively representing its deformation intensity, dynamic evolution, and terrain conditions. ,in, For the first Multidimensional feature vectors of each deformation point; For the first The slope deformation rate at each point; For the first Cumulative deformation at each point; For the first Deformation acceleration at each point; For the first The terrain slope at each point; For the first The average coherence of each point. This feature vector characterizes the comprehensive features of each deformation point from five dimensions: deformation rate, cumulative displacement, dynamic evolution, terrain conditions, and data quality.

[0066] Adaptive feature normalization: Due to the different dimensions of each feature, robust Z-Score normalization is adopted, using the median and MAD instead of the mean and standard deviation to reduce the interference of outliers on the normalization parameters. ,in, For the first The point, the first Robust Z-Score normalized values ​​for each feature; For the first The point, the first The original values ​​of each feature; For all monitoring points The set of values ​​for the i-th feature; This is a median function.

[0067] A composite distance metric that integrates spatial and feature distances: A clustering distance metric is defined that combines spatial distance and feature distance, as follows: , ,in, For the first Point and the Composite cluster distance between points; The normalized spatial Euclidean distance; The Mahalanobis distance between eigenvectors before normalization; and The weighting coefficients for spatial distance and feature distance (satisfying) Adaptively adjust according to the application scenario: increase when the deformation area is relatively dispersed. When the characteristic differences are significant, it increases . , The first Point, number The multidimensional feature vector of a point; Let covariance be the eigenvectors of all monitoring points. It is its inverse matrix.

[0068] Adaptive Neighborhood Radius and Minimum Number of Points Setting: Traditional DBSCAN algorithms require manual setting of the neighborhood radius and minimum number of points. This invention employs a method based on the spatial distribution of deformed point clouds to adaptively determine these parameters. First, the parameters for each point are calculated relative to its neighboring point. Distance of neighbors Take all points Sequences, using their distribution characteristics to determine neighborhood radius : ,in, This represents the total number of deformation monitoring points in the study area. For adaptive nearest neighbor number; This is the adaptive adjustment coefficient for the neighborhood radius; This is a function of the absolute deviation of the median. Minimum number of points. Determined based on the average number of points within the neighborhood radius: ,in, For the point Center, radius The total number of points in the neighborhood of . The minimum number of points is the adaptive scaling factor. ∈(0,1]; This is the floor function; It is a function for maximizing the value, ensuring that the minimum number of points is at least 3.

[0069] Cluster core point determination based on dual constraints: for each deformation anomaly point Using the KDTree data structure, search for all deformation points within a spatial radius R, and statistically analyze the deformation points in the neighborhood. Number of outliers pre-screened and percentage : , ,in, For the point Center, radius Spatial neighborhood; For any point in the neighborhood The comprehensive abnormality index, The adaptive filtering threshold determined earlier; For point The total number of points in the neighborhood.

[0070] point To be identified as a cluster core point, it must simultaneously meet both spatial density constraints and feature similarity constraints in order to pass cluster verification and participate in subsequent cluster expansion and merging.

[0071] Among them, the spatial density constraint is that the number of outliers in the neighborhood must reach a minimum cardinality threshold. Furthermore, the percentage of outliers must reach the minimum threshold. : and Feature similarity constraint: The Mahalanobis distance between the feature vectors of each point in the neighborhood and the center point must be less than the feature similarity threshold. : ,in, for Point and Neighboring Points The characteristic Mahalanobis distance between them This is a maximum value function that takes the maximum value of the characteristic distances between all points in the neighborhood and the center point.

[0072] Priority-based cluster expansion and merging: Starting from the cluster core, merging based on cluster distance... The clusters are expanded sequentially from smallest to largest, grouping neighboring points with a distance less than a threshold into the same cluster, until no new points can be added. This process is repeated until all core points are assigned to clusters. For clusters with spatial overlap, they are merged based on a priority defined by the maximum comprehensive anomaly index within the cluster. , ,in, For the first The merging of individual clusters takes priority, and they are sorted by the largest comprehensive anomaly index within the cluster, with higher values ​​indicating higher priority. For the first A cluster, The total number of clusters; The final cluster after priority merging; For set union operation, For set difference operations, clusters with spatial overlap are processed in descending order of priority. If the overlap area between a high-priority cluster and a low-priority cluster exceeds 50% of the area of ​​the low-priority cluster, the two clusters are merged. Otherwise, the complete boundary of the high-priority cluster is preserved, and the overlapping part is deducted from the low-priority cluster.

[0073] Clustering quality assessment: The silhouette coefficient and Davies-Bouldin index are introduced to assess the quality of the clustering results. When the quality index is below the threshold, the spatial neighborhood radius R and the minimum number of points are readjusted. Feature similarity threshold Re-cluster the data with parameters such as [parameter name] until the optimal clustering or the maximum number of iterations is reached.

[0074] Intelligent generation of deformation risk zone boundaries: For the clustered and merged point set, adaptive buffer generation, improved Alpha Shape boundary extraction, morphological smoothing, Douglas-Peucker simplification, and multi-dimensional attribute statistics are performed in sequence to generate the final deformation risk zone polygon.

[0075] Adaptive buffer generation: For each deformation point in a cluster, a circular buffer is generated according to a preset buffer radius, transforming the discrete point set into a continuous planar region. The buffer radius is adaptively determined based on the deformation point density and spatial distribution. ,in, For adaptive buffer radius; The average nearest neighbor distance of all points within the cluster; The standard deviation of the nearest neighbor distance within the cluster; The weighting coefficient for the average nearest neighbor distance. This is the weighting coefficient for the standard deviation of the nearest neighbor distance.

[0076] Improved Alpha Shape boundary extraction: The Alpha Shape algorithm is used to extract the concave hull boundary of the cluster point set; unlike the convex hull, Alpha Shape can adapt to irregularly shaped deformation regions and accurately characterize the true boundary of the deformation risk area. ,in, For the first The boundary of a polygonal deformation risk zone; This is the Alpha Shape algorithm function; For the first The point set of each cluster (as input to the Alpha Shape algorithm); For the first The Alpha Shape parameter of each cluster controls the degree of boundary concavity. This invention employs an adaptive Alpha parameter selection strategy, expressed as follows: , For the first The average nearest neighbor distance of points within each cluster; This is the adjustment coefficient (usually taken as 0.8~2.0), when... When the value is large, the boundary tends to form a convex hull; when When the value is smaller, the boundary more closely matches the true shape of the point set.

[0077] Morphological boundary smoothing: The original boundaries extracted from the Alpha Shape are smoothed using morphological opening and closing operations to eliminate jagged boundaries.

[0078] Douglas-Peucker boundary simplification: The smoothed boundary polygon is simplified using the Douglas-Peucker algorithm, which removes redundant nodes while preserving the overall shape of the boundary.

[0079] Secondly, this application also provides a road deformation risk zone identification device, which includes: The deformation information calculation module is used to calculate multiple deformation monitoring points and their deformation information based on time-series SAR images and DEM data of the target road area, and project the deformation information onto the slope direction to obtain the slope deformation information. The index calculation module is used to construct a multi-dimensional feature vector based on slope deformation information and calculate the comprehensive anomaly index for each deformation monitoring point. The anomaly point determination module is used to determine an adaptive threshold based on the distribution characteristics of the comprehensive anomaly index, and to identify deformation monitoring points whose comprehensive anomaly index is greater than the adaptive threshold as deformation anomaly points. The risk zone determination module is used to perform density clustering on deformation anomaly points and determine each cluster obtained from the clustering as a deformation risk zone.

[0080] The functions of each module in the above-mentioned road deformation risk zone identification device correspond to the steps in the above-mentioned road deformation risk zone identification method embodiment, and their functions and implementation processes will not be described in detail here.

[0081] The deformation information calculation module, index calculation module, anomaly point determination module, and risk area determination module can be devices with data processing capabilities, such as personal computers (PCs), laptops, and servers.

[0082] like Figure 2 As shown, in a third aspect, this application also provides a road deformation risk zone identification device, which includes a processor, a memory, and a road deformation risk zone identification program stored in the memory and executable by the processor, wherein when the road deformation risk zone identification program is executed by the processor, it implements the steps of the road deformation risk zone identification method described above.

[0083] In this embodiment of the application, a road deformation risk zone identification device may further include a communication interface and a communication bus.

[0084] The communication bus can be of any type and is used to interconnect the processor, memory, and communication interface.

[0085] The communication interface includes input / output (I / O) interfaces, physical interfaces, and logical interfaces used for interconnecting internal components of the road deformation risk zone identification device, as well as interfaces used for interconnecting the road deformation risk zone identification device with other devices (such as other computing devices or user equipment). Physical interfaces can be Ethernet interfaces, fiber optic interfaces, ATM interfaces, etc.; user equipment can be displays, keyboards, etc.

[0086] Memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0087] The processor can be a general-purpose processor, which can call the road deformation risk zone identification program stored in the memory and execute the road deformation risk zone identification method provided in the embodiments of this application. For example, the general-purpose processor can be a central processing unit (CPU). The method executed when the road deformation risk zone identification program is called can be referred to the various embodiments of the road deformation risk zone identification method of this application, and will not be repeated here.

[0088] Fourthly, embodiments of this application also provide a storage medium.

[0089] The storage medium stores a road deformation risk zone identification program, wherein when the road deformation risk zone identification program is executed by the processor, the steps of the road deformation risk zone identification method described above are implemented.

[0090] The method implemented when the road deformation risk zone identification procedure is executed can be referred to in various embodiments of the road deformation risk zone identification method of this application, and will not be repeated here.

[0091] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0092] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.

[0093] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.

[0094] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0095] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0096] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, 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 is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.

[0097] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for identifying road deformation risk zones, characterized in that, include: Based on the time-series SAR images and DEM data of the target road area, multiple deformation monitoring points and their deformation information are calculated, and the deformation information is projected onto the slope direction to obtain the slope deformation information. Based on the slope deformation information, a multi-dimensional feature vector is constructed, and the comprehensive anomaly index of each deformation monitoring point is calculated. An adaptive threshold is determined based on the distribution characteristics of the comprehensive anomaly index, and deformation monitoring points whose comprehensive anomaly index is greater than the adaptive threshold are identified as deformation anomaly points. Density clustering is performed on the deformation anomaly points, and each cluster obtained from the clustering is identified as a deformation risk area.

2. The method for identifying road deformation risk zones as described in claim 1, characterized in that, The aforementioned time-series SAR imagery and DEM data based on the target road area are used to calculate multiple deformation monitoring points and their deformation information, including: Acquire ascending orbit time-series SAR images and descending orbit time-series SAR images covering the target road area; The ascending orbit time-series SAR images and descending orbit time-series SAR images are registered and interferometrically processed to generate interferogram sequences. The terrain phase was removed using the DEM data, and deformation monitoring points were extracted using permanent scatterer (PS) and distributed scatterer (DS) technologies. The deformation monitoring points are jointly solved based on the time series analysis model to obtain the deformation rate and time series deformation information of each deformation monitoring point.

3. The method for identifying road deformation risk zones as described in claim 2, characterized in that, The process of projecting the deformation information onto the slope direction to obtain slope deformation information includes: Based on the DEM data, calculate the slope and aspect of each deformation monitoring point. The deformation rate and time-series deformation information are decomposed into vertical deformation components and east-west deformation components. A geometric projection model of the radar line of sight and the slope direction is constructed. Using the slope, aspect, vertical deformation component and east-west deformation component, the slope deformation information of each deformation monitoring point along the slope direction is calculated.

4. The method for identifying road deformation risk zones as described in claim 2, characterized in that, The aforementioned joint solution of the deformation monitoring points based on the time-series analysis model yields the deformation rate and time-series deformation information for each deformation monitoring point, including: We constructed an interferogram network based on spatiotemporal baseline thresholds, and the observation equations relating interferometric phase to deformation parameters, topographic errors, and atmospheric phase. The observation equations are solved using the least squares method or singular value decomposition method to separate the linear deformation components and the nonlinear deformation components. The deformation rate is obtained from the linear deformation component, and the time-series deformation information is obtained from the nonlinear deformation component.

5. The method for identifying road deformation risk zones as described in claim 1, characterized in that, The construction of a multi-dimensional feature vector based on the slope deformation information and the calculation of the comprehensive anomaly index for each deformation monitoring point include: Extract the deformation rate, deformation acceleration, cumulative deformation, and data coherence from the slope deformation information of each deformation monitoring point; The deformation rate, deformation acceleration, cumulative deformation, and data coherence are combined to form the multidimensional feature vector. The features of each dimension in the multidimensional feature vector are standardized to eliminate the influence of units; Based on the weight coefficients of each dimension of features, the weighted sum of the standardized features of each dimension is calculated to obtain the comprehensive anomaly index.

6. The method for identifying road deformation risk zones as described in claim 1, characterized in that, The step of determining an adaptive threshold based on the distribution characteristics of the comprehensive anomaly index, and identifying deformation monitoring points whose comprehensive anomaly index is greater than the adaptive threshold as deformation anomaly points, includes: The comprehensive anomaly index of all deformation monitoring points within the target road area is statistically analyzed, and a frequency distribution histogram of the comprehensive anomaly index is constructed. Identify the trough positions between the background noise distribution and the abnormal signal distribution in the frequency distribution histogram, and determine the comprehensive anomaly index value corresponding to the trough position as the adaptive threshold. The comprehensive anomaly index of all deformation monitoring points is compared with the adaptive threshold, and the deformation monitoring points that are greater than the adaptive threshold are identified as deformation anomaly points.

7. The method for identifying road deformation risk zones as described in claim 1, characterized in that, The aforementioned density clustering of the deformation anomaly points, and the determination of each cluster as a deformation risk region, includes: Calculate the spatial Euclidean distance and eigenvector distance between any two deformation anomaly points; Set the spatial neighborhood radius and feature similarity threshold; When the spatial Euclidean distance between two deformation anomaly points is less than the spatial neighborhood radius and the feature vector distance is less than the feature similarity threshold, it is determined that the density of the two deformation anomaly points is achievable. Based on density reachability, the deformation anomaly points are divided into several independent clusters, and each cluster is identified as a deformation risk zone.

8. A road deformation risk zone identification device, characterized in that, include: The deformation information calculation module is used to calculate multiple deformation monitoring points and their deformation information based on time-series SAR images and DEM data of the target road area, and project the deformation information onto the slope direction to obtain slope deformation information. The index calculation module is used to construct a multi-dimensional feature vector based on the slope deformation information and calculate the comprehensive anomaly index of each deformation monitoring point. An anomaly point determination module is used to determine an adaptive threshold based on the distribution characteristics of the comprehensive anomaly index, and to determine deformation monitoring points whose comprehensive anomaly index is greater than the adaptive threshold as deformation anomaly points. The risk zone determination module is used to perform density clustering on the deformation anomaly points and determine each cluster obtained by clustering as a deformation risk zone.

9. A road deformation risk zone identification device, characterized in that, The road deformation risk zone identification device includes a processor, a memory, and a road deformation risk zone identification program stored in the memory and executable by the processor, wherein when the road deformation risk zone identification program is executed by the processor, it implements the steps of the road deformation risk zone identification method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores a road deformation risk zone identification program, wherein when the road deformation risk zone identification program is executed by the processor, it implements the steps of the road deformation risk zone identification method as described in any one of claims 1 to 7.

Images