Urban building deformation monitoring leak detection point identification method and system
By constructing a parameterized three-dimensional confidence space that takes into account factors such as radar line-of-sight direction and building height, the problem of distinguishing between PS points on the building itself and non-target scattering points in PS-InSAR false negative rate assessment is solved, improving the accuracy and automated assessment capability of false negative rate assessment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUNAN UNIV
- Filing Date
- 2026-02-05
- Publication Date
- 2026-05-01
AI Technical Summary
Existing PS-InSAR false negative rate assessment methods fail to effectively distinguish between PS points on the building itself and surrounding non-target scattering points in complex urban environments, especially in areas with dense high-rise buildings, resulting in insufficient accuracy of false negative rate assessment results.
A parameterized three-dimensional confidence space conforming to the physical mechanism of SAR side-view imaging is constructed, taking into account factors such as radar line of sight direction, incident angle, building height and occlusion effect. The spatial attribution relationship between PS points and buildings is improved by judging the visibility of building facades and the existence of PS points in the three-dimensional confidence space.
It significantly improves the reliability of missed detection rate assessment, enables scientific and reasonable determination of PS point attribution in complex urban environments, and supports automated assessment of building safety monitoring in smart cities.
Smart Images
Figure CN121955918A_ABST
Abstract
Description
A method and system for identifying missed detection points in urban building deformation monitoring Technical Field
[0001] This invention belongs to the field of urban building health monitoring technology, specifically relating to a method and system for identifying missed detection points in urban building deformation monitoring. Background Technology
[0002] Persistent Scatterer Interferometric Synthetic Aperture Radar (PS-InSAR) is an important technology for monitoring land deformation. It can acquire land and building deformation information with millimeter-level accuracy under long-term time series conditions and has been widely used in urban land subsidence monitoring, infrastructure safety assessment and building health diagnosis.
[0003] Level 1 (LOD1) building models are a commonly used form of 3D city model representation. They typically simplify buildings into regular geometric shapes (such as cuboids) with uniform height information, offering advantages such as simple structure, small data size, and ease of storage and processing. With the development of data acquisition methods such as aerial imagery, remote sensing imagery, and laser point clouds, generating LOD1 level building white models using automated methods such as deep learning has become an important technical approach for 3D city modeling.
[0004] Accurate registration of PS-InSAR deformation monitoring points with LOD1 level building models can realize the spatial correspondence between PS points and specific buildings, thereby quickly determining the building unit to which the deformation information belongs. This can be further used to calculate the missed detection rate and coverage rate of PS-InSAR monitoring results at the city scale, which is of great significance for improving the accuracy of urban safety monitoring and the reliability of building deformation analysis.
[0005] Most existing PS-InSAR false negative rate assessment methods are based on the concept of a two-dimensional planar buffer zone for buildings. This involves extending a certain distance outward from the building's planar area and assessing the building's detection integrity by counting the number of PS points within the buffer zone. However, these methods do not fully consider the spatial location and line-of-sight characteristics of radar imaging. Simply using an isotropic planar buffer lacks physical meaning and easily introduces PS deformation points that are not part of the building itself, leading to an overestimation of PS point coverage and insufficient accuracy in the assessment results.
[0006] Building upon this foundation, some existing studies attempt to incorporate building outline information to constrain or correct the spatial distribution of PS points, thereby improving the spatial correspondence between PS points and buildings. However, these methods are generally based on two-dimensional geometric assumptions, primarily focusing on the positional correction of PS points in the planar direction. They do not adequately consider the mechanism of the true projected position under synthetic aperture radar side-looking imaging conditions, which is formed by the combined effects of imaging line-of-sight direction, elevation differences, and occlusion effects. In complex urban environments, especially in areas with dense high-rise buildings, relying solely on two-dimensional building outlines for matching or buffering is still insufficient to effectively distinguish between PS points on the building itself and surrounding non-target scattering points. The accuracy and stability of the missed detection rate assessment results need further improvement. Summary of the Invention
[0007] This invention provides a method and system for identifying missed detection points in urban building deformation monitoring. It abandons the isotropic assumption of the traditional two-dimensional planar buffer and constructs a parameterized three-dimensional confidence space that conforms to the physical mechanism of SAR side-view imaging. It fully considers multiple factors such as radar line of sight direction, incident angle, building height and occlusion effect, effectively avoids misjudgment of non-target scattering points, makes the spatial attribution relationship between PS points and buildings more accurate, significantly improves the reliability of missed detection rate assessment, and thus effectively solves at least one of the technical problems involved in the background art.
[0008] To solve the above-mentioned technical problems, the present invention is implemented as follows: a method for identifying missed detection points in urban building deformation monitoring, comprising the following steps: Step S1, acquiring white model data of buildings in the target area and PS point data of PS-InSAR deformation monitoring; Step S2, registering the PS point data with the white model data of buildings, so that the PS points are attached to the building surface; Step S3, projecting the PS points and the building onto the same two-dimensional plane, determining whether there are PS points within the planar projection outline area of the building. If so, the building is determined to be successfully monitored, the determination ends, and the PS points within the outline area are marked and do not participate in the subsequent identification process; otherwise, the subsequent steps are executed; Step S4, determining the visibility of each building facade by the angle between the building facade and the radar line of sight, using the visible facade of the building as a reference surface, constructing a parameterized three-dimensional confidence space to describe the spatial uncertainty between the PS points and the building, determining whether there are PS points within the three-dimensional confidence space. If so, the building is determined to be successfully monitored, the determination ends; otherwise, the building is determined to have missed detection.
[0009] As a preferred improvement, the architectural white model data is at LOD1 level.
[0010] As a preferred improvement, the architectural white model data includes the building's outline and height, which are predicted by deep learning technology using remote sensing imagery.
[0011] As a preferred improvement, the prediction process for building outline and height data specifically includes the following steps: Step S11, acquiring high-resolution remote sensing images of the target area; Step S12, extracting building outlines and predicting height information from the remote sensing images based on deep learning prediction methods; Step S13, fusing the two-dimensional building outlines with the three-dimensional height information to ensure that the building polygons predicted by deep learning are strictly aligned in space with the predicted height raster map. For each building polygon, the median of all height raster pixel values covered within it is taken as its corresponding building height.
[0012] As a preferred improvement, the process for determining the visibility of a building facade includes the following steps: Step S411, calculating the unit vector of the radar line of sight direction based on the incident angle and heading angle of the SAR satellite; Step S412, calculating the normal vector of each facade based on the white model data of the building; Step S413, calculating the angle between the facade normal vector and the radar line of sight direction unit vector. If the angle is less than 90°, the facade is determined to be visible under the SAR satellite; otherwise, the facade is determined to be invisible under the SAR satellite.
[0013] As a preferred improvement, the determination of whether a PS point exists in the three-dimensional confidence space is specifically performed as follows: Step S421, determine whether the projection depth of the PS point along the radar line of sight to the reference surface satisfies the projection depth constraint. If not, the PS point is determined not to fall within the three-dimensional confidence space; otherwise, proceed to the next step. Step S422, determine whether the orthographic projection of the projection depth of the PS point along the radar line of sight to the reference surface to the reference surface satisfies the in-plane lateral expansion constraint. If yes, the PS point is determined to fall within the three-dimensional confidence space; otherwise, the PS point is determined not to fall within the three-dimensional confidence space. Step S423, traverse all PS points. If a PS point falls within the three-dimensional confidence space, it indicates that a PS point exists within the three-dimensional confidence space.
[0014] As a preferred improvement, the projection depth constraint is expressed as: In the formula, This represents the projection depth of point PS onto the reference surface along the radar line of sight. Indicates the maximum confidence depth; where: In the formula, Indicates the position of point PS; This indicates the position of the orthographic projection of point PS onto the reference plane; Represents the unit vector of the radar line-of-sight direction; In the formula, For radar resolution, For the elevation error of the building model, For correction factor, This is an empirical correction factor. The building complexity is represented by a value of [0, 1], which is used to correct the signal multipath effect in complex urban contexts through an exponential decay term.
[0015] As a preferred improvement, the in-plane lateral expansion constraint is expressed as: In the formula, Indicates projection depth Orthographic projection onto the reference plane; Indicates the angle of incidence of the SAR satellite; This is the radar azimuth resolution angle.
[0016] As a preferred improvement, when the same PS point falls within the three-dimensional confidence space of multiple buildings simultaneously, a weighted response discrimination mechanism based on building height and line-of-sight distance is introduced to define the PS point as the [missing information - likely a specific building or structure]. Height-weighted response value of a building for: In the formula, For the first The height of the building The maximum height of all buildings within the target area; For PS, click to the first The projected distance of the building reference plane. These are non-negative weighting coefficients. and These factors are used to balance the relative influence of building height and radar line-of-sight geometric constraints on PS point assignment; the PS point is then assigned to the height-weighted response value. The largest single building.
[0017] A system for performing the above-described method for identifying missed detection points in urban building deformation monitoring includes: a data acquisition module for acquiring white model data of buildings in the target area and PS-InSAR deformation monitoring PS point data; a registration module for registering the PS point data with the white model data of buildings, so that the PS points are attached to the building surface; a contour attribution determination module for projecting the PS points and the building onto the same two-dimensional plane, determining whether there are PS points within the planar projection contour area of the building. If so, the building is determined to be successfully monitored, the determination ends, and the PS points within the contour area are marked and do not participate in the subsequent identification process; otherwise, the subsequent steps are executed; and a confidence space determination module for determining the visibility of each building facade by the angle between the building facade and the radar line of sight. Using the visible facade of the building as a reference surface, a parameterized three-dimensional confidence space is constructed to describe the uncertainty of spatial attribution between the PS points and the building. The system determines whether there are PS points within the three-dimensional confidence space. If so, the building is determined to be successfully monitored, the determination ends; otherwise, the building is determined to have missed detection.
[0018] Compared with the prior art, the present invention has the following beneficial effects: (1) It abandons the isotropic assumption of the traditional two-dimensional planar buffer and constructs a parameterized three-dimensional confidence space that conforms to the physical mechanism of SAR side-view imaging on the basis of strong contour determination. It fully considers multiple factors such as radar line of sight direction, incident angle, building height and occlusion effect, effectively avoids misjudgment of non-target scattering points, makes the spatial attribution relationship between PS points and buildings more accurate, and significantly improves the reliability of the missed detection rate assessment; (2) By introducing a weighted response discrimination mechanism based on building height and line of sight distance, it effectively solves the problem of PS point attribution ambiguity in dense high-rise building areas, making the PS point attribution determination in complex urban environments more scientific and reasonable; (3) From building white model data acquisition, PS point registration, three-dimensional confidence space construction to missed detection rate statistics, the entire process does not require manual intervention and can be directly embedded into the existing PS-InSAR monitoring platform to realize automated assessment of missed detection rate of large-scale building groups at the city level, meet the actual needs of smart city building safety monitoring, and provide reliable data support for subsequent building health monitoring and risk assessment. Attached Figure Description
[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Among them: Figure 1 is a schematic diagram of the target area provided in Embodiment 1; Figure 2 is a schematic diagram 1 of the positional relationship between the white model of the building and the PS point before registration in Embodiment 1; Figure 3 is a schematic diagram 2 of the positional relationship between the white model of the building and the PS point before registration in Embodiment 1; Figure 4 is a side view of the positional relationship between the white model of the building and the PS point after registration in Embodiment 1; Figure 5 is a top view of the positional relationship between the white model of the building and the PS point after registration in Embodiment 1; Figure 6 is a schematic diagram of the visible facade of Embodiment 1; Figure 7 is a schematic diagram of the three-dimensional confidence space of Embodiment 1; Figure 8 is a partial enlarged view of the three-dimensional confidence space shown in Figure 7; Figure 9 is a schematic diagram of buildings with and without PS points in the urban building complex provided in Embodiment 1; Figure 10 is a schematic diagram of the relationship between buildings with and without PS points and PS points in the urban building complex provided in Embodiment 1. Detailed Implementation
[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0021] As shown in Figures 1-10, this embodiment provides a method for matching building white models with time-series InSAR deformation points, including the following steps: Step S1, acquire building white model data and PS point data of PS-InSAR deformation monitoring of the target area.
[0022] The building white model data is LOD1 level building white model data, including the building outline and height. The building outline and height data are predicted by deep learning technology through remote sensing imagery, specifically including the following steps: Step S11, acquire high-resolution remote sensing imagery of the target area.
[0023] Step S12: Extract building outlines and predict height information from remote sensing images based on deep learning prediction methods.
[0024] For building outline extraction, the model learns to identify areas belonging to buildings from remote sensing image pixels and outputs a binary polygon vector outline; for height prediction, the model learns to predict the relative or absolute height of the surface where each pixel is located from single or stereo images and generates a height raster map aligned with the image.
[0025] Step S13: The two-dimensional building outline is fused with the three-dimensional height information to ensure that the building polygons predicted by deep learning are strictly aligned in space with the predicted height raster map. For each building polygon, the median of all height raster pixel values covered inside it is taken as its corresponding building height.
[0026] Compared to the mean, the median is less sensitive to outliers (such as extremely high or low noise caused by prediction errors, trees, billboards, and other objects not on the top of the building) and better represents the general height of the building body.
[0027] The obtained architectural white models also need to be preprocessed to extract the geometric parameters of each building, including the building center coordinates ( , , Building height In addition, the coordinates of the vertices of the building's base outline should be obtained. Furthermore, the acquired white model data and PS point data should be calibrated to a single coordinate system, for example, by simultaneously switching to a coordinate system in meters, to facilitate subsequent calculations.
[0028] Step S2: Register the PS point data with the building white model data so that the PS points are attached to the building surface.
[0029] The registration of PS point data and building white model data adopts conventional techniques in this field. For example, the ICP method (Iterative Closest Point Method) disclosed in prior art 1 (MATCHING OF PERSISTENT SCATTERERS TO BUILDINGS, Alexander Schunert et al.) is used to register the PS point set with the building.
[0030] Step S3: Project the PS point and the building onto the same two-dimensional plane, and determine whether there is a PS point in the planar projection outline area of the building. If so, the building is successfully monitored, the determination ends, and the PS point in the outline area is marked and does not participate in the subsequent recognition process; otherwise, the subsequent steps are executed.
[0031] The presence of a PS point within the building's outline area indicates that the building has been properly "illuminated" by the radar, and there are no missed detections. If no PS point is present within the building's outline area, further judgment is required to determine whether the building has been successfully detected. Furthermore, there is a unique correspondence between each PS point within the outline area and a successfully detected building; they cannot belong to other buildings. Therefore, marking PS points within the outline area and excluding them from subsequent identification processes reduces data processing and improves identification efficiency.
[0032] Step S4: Determine the visibility of each building facade by the angle between the building facade and the radar line of sight. Using the visible facade of the building as a reference surface, construct a parameterized three-dimensional confidence space to describe the spatial uncertainty between the PS point and the building. Determine whether there is a PS point in the three-dimensional confidence space. If so, the building is successfully monitored and the determination ends; otherwise, the building is missed.
[0033] PS-InSAR is a detection method based on radar technology. Radar waves propagate in a straight line, and only surfaces that can be directly illuminated by the beam (where the angle between the normal direction and the radar line of sight is less than 90°) can generate strong echoes. Surfaces facing away from the radar (where the angle between the normal direction and the radar line of sight is greater than or equal to 90°) cannot receive direct energy. Based on this principle, the visibility of building facades can be screened to determine the facades facing the satellite.
[0034] The process of determining the visibility of a building facade specifically includes the following steps: Step S411, based on the incident angle of the SAR satellite... and heading angle Calculate the unit vector of the radar line-of-sight direction.
[0035] Among them, heading angle The angle between the flight path and true north, and the unit vector of the radar line-of-sight direction. Represented as: Step S412: Calculate the normal vector of each facade based on the architectural white model data of the high-rise building.
[0036] For a LOD1 building model, each building facade can be represented by a planar equation: In the formula, , , Represents the three-dimensional spatial coordinates of any point on the plan of the building facade; , , , Let represent the coefficients of the plane equation, and let be vectors. Let be the normal vector of the building's facade; then the normal vector of the building's facade... Represented as: Step S413: Calculate the angle between the facade normal vector and the unit vector of the radar line of sight. If the angle is less than 90°, the facade is determined to be visible under SAR satellite; otherwise, the facade is determined to be invisible under SAR satellite.
[0037] because and Both are unit vectors, therefore the included angle It can be represented as: when: ,Right now When the angle is acute, it indicates that the facade faces the direction of the satellite's line of sight, and the facade is visible under SAR satellite observation conditions; conversely, when the angle is acute, it indicates that the facade faces away from the direction of the satellite's line of sight, and the facade is not visible under SAR satellite observation conditions.
[0038] In step S3, the determination method based on the PS point and the building outline is a theoretically strong determination method. In practical applications, due to the influence of multiple factors such as radar line-of-sight direction, incident angle, building height, and measurement errors caused by occlusion effects, it is very easy to miss cases. For example, there may be cases where the PS point is located outside the building outline but close to it. To solve this problem, this invention introduces the concept of a three-dimensional confidence space. Taking the visible facade of the building as the reference surface, a three-dimensional trapezoidal space is extended outward along the radar line-of-sight direction. As long as the PS point falls within the trapezoidal space, it can be determined that the PS point belongs to the target building. This method can effectively avoid measurement errors caused by various factors and reduce the occurrence of missed cases. The structure of the three-dimensional confidence space is shown in Figures 7 and 8.
[0039] The boundary of the three-dimensional confidence space is defined by projection depth constraints and in-plane lateral expansion constraints. The determination of whether a PS point exists in the three-dimensional confidence space is made in the following way: Step S421, determine whether the projection depth of the PS point along the radar line of sight to the reference surface satisfies the projection depth constraints. If not, it is determined that the PS point does not fall into the three-dimensional confidence space; otherwise, the subsequent steps are executed.
[0040] According to the incident angle of the SAR satellite and heading angle Calculate the unit vector of the radar line-of-sight direction. Based on this, the maximum confidence depth along the radar line of sight is defined. , is represented as: In the formula, For radar resolution, For the elevation error of the building model, For correction factor, This is an empirical correction factor. Let the building complexity be [0, 1], used to correct for signal multipath effects in complex urban environments through an exponential decay term; then, the projection depth constraint is expressed as: In the formula, This represents the projection depth of point PS onto the reference surface along the radar line of sight, where: In the formula, Indicates the position of point PS; This indicates the position of the orthographic projection of point PS onto the reference plane.
[0041] Step S422: Determine whether the projection depth of point PS along the radar line of sight onto the reference surface and its orthographic projection onto the reference surface satisfy the in-plane lateral expansion constraint. If yes, determine that point PS falls within the three-dimensional confidence space; otherwise, determine that point PS does not fall within the three-dimensional confidence space. The in-plane lateral expansion constraint is expressed as: In the formula, This indicates that the PS point is on the facade. The in-plane distance between the orthogonal projection points; The azimuth resolution angle is used to describe the angular spread characteristics of the radar beam during propagation; in step S423, all PS points are traversed. If a PS point falls into the three-dimensional confidence space, it indicates that a PS point exists in the three-dimensional confidence space.
[0042] When a point PS simultaneously satisfies the aforementioned depth constraint along the line of sight and the lateral expansion constraint in the plane, it is determined that the point PS is located inside the parametric three-dimensional trapezoidal confidence space generated outward from the geometric plane of the exterior facade.
[0043] Furthermore, when the same PS point When a point falls within the 3D confidence space of multiple buildings, to eliminate ambiguity in attribution caused by similar building heights or unclear front-to-back occlusion relationships, a weighted response discrimination mechanism based on building height and line-of-sight distance is introduced; the PS point is defined as the first... Height-weighted response value of a building for: In the formula, For the first The height of the building This represents the maximum height of all buildings within the target area, used for height normalization. For PS, click to the first The projected distance of the building reference plane. These are non-negative weighting coefficients. and These factors are used to balance the relative influence of building height and radar line-of-sight geometric constraints on PS point attribution. In areas with significant differences in building height and complex building occupancy relationships, the influence can be appropriately increased. The value of [value] is adjusted to enhance the weight of building height in PS point attribution determination; in areas with large radar incidence angles or densely packed buildings along the line of sight, the value can be appropriately increased. The value of is chosen to enhance the influence of line-of-sight distance constraints in the discrimination process. In a preferred embodiment, the weighting coefficient is... The value range is 0.4-0.7. The value ranges from 0.3 to 0.6, and satisfies... When a PS point corresponds to multiple buildings, the PS point is assigned to the height-weighted response value. The largest single building.
[0044] Based on the above judgment process, the final evaluation process for building missed detection identification is as follows: when a PS point exists within the outline of a building, the building is considered successfully monitored; when no PS point exists within the outline of a building, but a PS point exists in the three-dimensional confidence space, the building is considered successfully monitored; when no PS point exists within the outline of a building and no PS point exists in the three-dimensional confidence space, the building is considered missed; by statistically analyzing the number of missed detections of individual buildings within the target area and the total number of buildings, the distribution of the building missed detection rate at the city scale is obtained.
[0045] This embodiment also provides a system for the above-described method for identifying missed detection points in urban building deformation monitoring, comprising: a data acquisition module for acquiring building white model data and PS-InSAR deformation monitoring PS point data of the target area; a registration module for registering the PS point data with the building white model data, so that the PS points are attached to the building surface; and a contour attribution determination module for projecting the PS points and the building onto the same two-dimensional plane, determining whether there are PS points within the planar projection contour area of the building. If so, the building is determined to have been successfully monitored, the determination ends, and the PS points within the contour area are marked. If the building's facade is not visible, it will not participate in the subsequent identification process; otherwise, the subsequent steps will be executed. The confidence space determination module determines the visibility of each building facade by the angle between the building facade and the radar line of sight. Using the visible facade of the building as a reference surface, a parameterized three-dimensional confidence space is constructed to describe the spatial uncertainty between the PS point and the building. It determines whether there is a PS point in the three-dimensional confidence space. If so, the building is determined to be successfully monitored, and the determination ends; otherwise, the building is determined to have missed detection. Example 1: This example uses the urban building deformation monitoring missed detection point identification method provided by the present invention to perform missed detection in a certain target area. Point identification was performed. Figure 1 shows the LOD1 level building white model data for the target area, containing 1783 buildings. Twenty 3m×3m (ascending orbit) CSK strip images of the area were acquired from September 2022 to May 2024. The SAR images were processed, and the deformation time series of stable scattering points within the area was extracted, resulting in 52853 PS points. The satellite incident angle was 24.31°, and the heading angle was 11.24°. The building white models and PS point results were uniformly converted to the CGCS2000 national geodetic coordinate system. Using the Gauss-Kruger 3-degree zone projection, the initial white model and the positional relationship of the PS points are shown in Figures 2 and 3. The PS points are moved as a whole to register with the building, so that the PS points are attached to the building surface. The matching results are shown in Figures 4 and 5. In the plane coordinate system, the spatial inclusion relationship of the PS points is determined based on the building's bottom contour. The first attribution determination between the PS points and the building unit is performed. It is determined that 16286 points are inside the building, and there are 36570 PS points outside the contour. The process then enters the 3D spatial attribution determination process. Based on the incident angle of the SAR satellite... and heading angle Calculate the unit vector in the radar line-of-sight direction: Calculate the building facade normal vector to determine the facade's visibility, as shown in Figure 6. For the PS points marked as outside the outline in step S3, construct a parametric three-dimensional confidence space by combining the building's visible facade, SAR imaging geometry, and radar resolution information. Determine the spatial uncertainty between the PS points and the building. The CSK satellite's radar resolution is 3m, therefore... The model's elevation error is on the order of meters. Empirical correction coefficient Architectural complexity =0.5, then Therefore, the model extends [8.95, 1.78, 20.21] m along the line of sight. For the height-weighted response value... Since the target area is a densely populated urban area with many tall buildings, we choose... , The final result was that 29,505 PS points were assigned in the confidence space, while 7,065 PS points were still not assigned. Based on the combined results of strong assignment within the outline and the results of three-dimensional confidence space outside the outline, the monitoring integrity of individual buildings was evaluated. 1,494 buildings were successfully monitored, and 289 buildings were missed. The missed detection rate in urban monitoring was 16.21%. The distribution of the missed detection rate of buildings at the urban scale is shown in Figures 9 and 10.
[0046] The embodiments of the present invention have been described above, but the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit of the present invention, and all of these forms are within the protection scope of the present invention.
Claims
1. A method for identifying missed detection points in urban building deformation monitoring, characterized in that, The process includes the following steps: Step S1, acquiring building white model data and PS point data from PS-InSAR deformation monitoring of the target area; Step S2, registering the PS point data with the building white model data to attach the PS points to the building surface; Step S3, projecting the PS points and the building onto the same two-dimensional plane, determining whether there are PS points within the planar projection outline area of the building. If so, the building is successfully monitored, the determination ends, and the PS points within the outline area are marked and do not participate in the subsequent identification process; otherwise, the subsequent steps are executed; Step S4, determining the visibility of each building facade by the angle between the building facade and the radar line of sight. Using the visible facade of the building as a reference surface, constructing a parameterized three-dimensional confidence space to describe the spatial uncertainty between the PS points and the building, determining whether there are PS points within the three-dimensional confidence space. If so, the building is successfully monitored, the determination ends. Conversely, if the building is not inspected, it is determined that a missed inspection has occurred.
2. The method for identifying missed detection points in urban building deformation monitoring according to claim 1, characterized in that, The architectural white model data is at LOD1 level.
3. The method for identifying missed detection points in urban building deformation monitoring according to claim 1, characterized in that, Building white model data includes the building's outline and height, which are predicted using deep learning technology through remote sensing imagery.
4. The method for identifying missed detection points in urban building deformation monitoring according to claim 3, characterized in that, The prediction process for building outline and height data specifically includes the following steps: Step S11, acquiring high-resolution remote sensing images of the target area; Step S12, extracting building outlines and predicting height information from the remote sensing images based on deep learning prediction methods; Step S13, fusing the two-dimensional building outlines with the three-dimensional height information to ensure that the building polygons predicted by deep learning are strictly aligned in space with the predicted height raster map. For each building polygon, the median of all height raster pixel values covered within it is taken as its corresponding building height.
5. The method for identifying missed detection points in urban building deformation monitoring according to claim 1, characterized in that, The process for determining the visibility of a building facade includes the following steps: Step S411, calculate the unit vector of the radar line of sight direction based on the incident angle and heading angle of the SAR satellite; Step S412, calculate the normal vector of each facade based on the white model data of the building; Step S413, calculate the angle between the facade normal vector and the radar line of sight direction unit vector. If the angle is less than 90°, the facade is determined to be visible under the SAR satellite; otherwise, the facade is determined to be invisible under the SAR satellite.
6. The method for identifying missed detection points in urban building deformation monitoring according to claim 1, characterized in that, The determination of whether a PS point exists in the three-dimensional confidence space is made in the following way: Step S421, determine whether the projection depth of the PS point along the radar line of sight to the reference surface meets the projection depth constraint. If not, it is determined that the PS point does not fall into the three-dimensional confidence space. Otherwise, proceed with the following steps; Step S422: Determine whether the projection depth of the PS point along the radar line of sight onto the reference surface and its orthographic projection onto the reference surface satisfy the in-plane lateral expansion constraint. If yes, determine that the PS point falls within the three-dimensional confidence space; otherwise, determine that the PS point does not fall within the three-dimensional confidence space; Step S423: Traverse all PS points. If a PS point falls within the three-dimensional confidence space, it indicates that a PS point exists within the three-dimensional confidence space.
7. The method for identifying missed detection points in urban building deformation monitoring according to claim 6, characterized in that, The projection depth constraint is expressed as: In the formula, This represents the projection depth of point PS onto the reference surface along the radar line of sight. Indicates the maximum confidence depth; in: In the formula, Indicates the position of point PS; This indicates the position of the orthographic projection of point PS onto the reference plane; Represents the unit vector of the radar line-of-sight direction; In the formula, For radar resolution, For the elevation error of the building model, For correction factor, This is an empirical correction factor. The building complexity is represented by a value of [0, 1], which is used to correct the signal multipath effect in complex urban contexts through an exponential decay term.
8. The method for identifying missed detection points in urban building deformation monitoring according to claim 7, characterized in that, The in-plane lateral expansion constraint is represented as: In the formula, Indicates projection depth Orthographic projection onto the reference plane; Indicates the angle of incidence of the SAR satellite; This is the radar azimuth resolution angle.
9. The method for identifying missed detection points in urban building deformation monitoring according to claim 6, characterized in that, When the same PS point falls within the 3D confidence space of multiple buildings simultaneously, a weighted response discrimination mechanism based on building height and line-of-sight distance is introduced to define the relationship between the PS point and the first building. Height-weighted response value of a building for: In the formula, For the first The height of the building The maximum height of all buildings within the target area; For PS, click to the first The projected distance of the building reference plane. These are non-negative weighting coefficients. and These factors are used to balance the relative influence of building height and radar line-of-sight geometric constraints on PS point assignment; the PS point is then assigned to the height-weighted response value. The largest single building.
10. A system for implementing the method for identifying missed detection points in urban building deformation monitoring according to any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire building white model data and PS point data for PS-InSAR deformation monitoring of the target area; The registration module registers the PS point data with the building white model data, attaching the PS points to the building surface. The outline attribution determination module projects the PS points and the building onto the same two-dimensional plane, determining whether there are PS points within the building's planar projection outline area. If so, the building is successfully monitored, the determination ends, and the PS points within the outline area are marked and do not participate in subsequent recognition processes; otherwise, subsequent steps are executed. The confidence space determination module determines the visibility of each building facade by the angle between the building facade and the radar line of sight. Using the visible facade of the building as a reference surface, a parametric three-dimensional confidence space is constructed to describe the uncertainty of spatial attribution between PS points and the building. The module determines whether there are PS points within the three-dimensional confidence space. If so, the building is successfully monitored, and the determination ends. Conversely, if the building is not inspected, it is determined that a missed inspection has occurred.