A building health status detection system

Abstract

The present invention discloses a building health status detection system, which relates to the field of photogrammetry and image processing technology. The system first obtains the four-dimensional spatiotemporal coordinate data of multiple monitoring points on the target building based on satellite image analysis, and constructs a four-dimensional spatiotemporal network model containing spatial topology and temporal evolution information based on this data. Then, the system maps the nodes of the network model to a predefined standard reference network associated with the local gravity direction. Then, the system uses a graph self-attention network to analyze the dynamic correlation within the four-dimensional network model, and at the same time calculates the relative relationship changes of the monitoring points relative to the predefined standard reference network through geometric analysis. The deformation interpretation module distinguishes and determines the internal relative deformation, overall displacement and absolute posture change (including settlement and tilt) of the building based on the above two types of analysis results and with reference to the surrounding environment information. The status assessment module integrates this information to generate a health status report with risk level.

Description

Technical Field

[0001] The present invention relates to the technical field of photogrammetry and image processing, and in particular to a building health status detection system. Background Art

[0002] Buildings, especially large infrastructure such as bridges, dams, high-rise buildings, and tunnels, are affected by various factors during their use, including loads, environmental erosion, material aging, and geological activities. These factors may cause deformation, displacement, cracking, and other defects. Therefore, regular and accurate health status monitoring of buildings is crucial. It is a key link in ensuring their safe operation, formulating scientific maintenance strategies, and extending their service life.

[0003] Traditional building health monitoring methods include manual visual inspection, deployment of contact sensors (such as strain gauges, displacement gauges, and inclinometers), geodetic surveying (such as leveling, GPS / GNSS, and total stations), and the recently developed three-dimensional laser scanning. However, these methods often have several limitations: manual inspection is highly subjective, inefficient, and difficult to detect early minor damage; contact sensors require physical deployment, which is costly, fragile, and can only monitor a limited number of points; while geodetic surveying offers high accuracy, it is usually point-based monitoring and labor-intensive, making monitoring large areas or inaccessible areas costly and inefficient; and while three-dimensional laser scanning can acquire high-precision geometric models, it usually requires ground operations, is subject to line of sight obstructions, and is costly and complex for periodic monitoring of large structures or areas. Therefore, the development of automated, non-contact, large-scale, and efficient building health status monitoring technologies has important practical significance and application value.

[0004] In order to achieve more efficient structural deformation monitoring, some existing technologies have been explored, such as:

[0005] Chinese invention patent CN113819902B discloses a single-axis gyroscope-based three-dimensional measurement method for a travel trajectory, with IPC classification number G01C. This method improves the Z coordinate accuracy by measuring the three-dimensional coordinates of each point on the travel route in a local geographic coordinate system and compensating for gyroscope errors (including the influence of the Earth's rotation). It is used for deformation detection of structures such as bridges and dams. This method has its advantages in improving the measurement accuracy of specific sensors and facilitates local high-precision trajectory measurement.

[0006] Chinese invention patent CN115908424B discloses a building health monitoring method based on 3D laser scanning, with IPC classification number G06T. This method plans scanning sites, obtains point cloud data, generates 3D models, and combines image comparison to identify building anomalies, conduct health assessments, and issue early warnings. This method utilizes the advantages of 3D laser scanning technology to obtain high-precision geometric models and is committed to the information management of the entire life cycle of buildings.

[0007] Although the above existing technologies have made certain progress in specific aspects, they still have certain limitations, especially in achieving large-scale, long-term, all-round, and automated monitoring. Specifically, the sensors physically move along a specific trajectory, making it difficult to achieve rapid, non-contact monitoring of buildings in a wide area or in areas that are inconvenient for people to reach. Ground-based three-dimensional laser scanning requires on-site operations and is easily blocked by ground vision. In addition, the cost is high and the efficiency is low for periodic monitoring of large areas or numerous buildings. The ability to reveal complex relative deformation patterns within the structure (such as torsion and shear strain distribution) is limited, and there is a lack of a mechanism for directly analyzing the changes in the building's posture relative to the absolute gravity reference (such as changes in pitch angle and roll angle). Summary of the Invention

[0008] The purpose of the present invention is to overcome the deficiencies of the prior art and to provide a building health status detection system to solve the above-mentioned problems.

[0009] The purpose of the present invention is achieved through the following technical solutions: a building health status detection system, including a four-dimensional spatiotemporal network construction module, which is used to construct a four-dimensional spatiotemporal network model of a target building based on four-dimensional spatiotemporal coordinate data, wherein the model includes nodes representing the spatiotemporal positions of monitoring points, spatial edges connecting nodes in the same timestamp, and time edges connecting nodes of the same monitoring point at different timestamps; a standard network mapping module, which is used to map the nodes in the four-dimensional spatiotemporal network model to reference points of a predefined standard reference network associated with the local gravity direction to establish a mapping relationship; a first analysis module, which is configured to use a graph self-attention network to process the four-dimensional spatiotemporal network model, and is used to output attention weights representing the internal spatiotemporal dynamic correlation between nodes and dynamic embedding vectors encoding node states and neighborhood information; a second analysis module, which is configured to process the mapping relationship and output a time change index representing the spatiotemporal relative geometric relationship of the monitoring point relative to the reference point of the standard reference network through geometric analysis; a deformation interpretation module, which is used to perform the following steps: step S1, based on dynamic The time change pattern of the state embedding vector or the attention weight is compared with the pattern of the surrounding reference buildings to filter out the common mode effect, and the internal relative deformation information of the target building is determined; step S2, based on the four-dimensional space-time coordinate data, the rigid body transformation is fitted or the center of mass is tracked, and the overall displacement information of the target building is determined by comparing the motion with the surrounding reference buildings; step S3, based on the time change index, the absolute attitude change information of the target building relative to the standard reference network, including the absolute vertical settlement rate, pitch angle and roll angle, is determined by the conversion model; a state assessment module is used to integrate the internal relative deformation information, the overall displacement information and the absolute attitude change information, and generate a health status assessment report including a risk level assessment according to a preset threshold or change trend analysis; a verification module is used to compare the health status assessment report with the ground measured benchmark data to generate an accuracy assessment index; and a calibration module is used to generate a calibration feedback instruction based on the accuracy assessment index to adjust the parameters or model of at least one other module of the system through iterative feedback.

[0010] When constructing spatial edges, the four-dimensional spatiotemporal network building module connects structurally adjacent monitoring point nodes based on the roof geometric topology of the target building, which is known or extracted from the image.

[0011] The standard network mapping module is configured to map each monitoring point node in the four-dimensional space-time network model to three reference points with predetermined relative geometric positions in the standard reference network.

[0012] The attention weights and dynamic embedding vectors output by the first analysis module include time-varying inter-node attention weight values and / or time-varying node embedding vector coordinates.

[0013] The time-varying index of the spatiotemporal relative geometric relationship output by the second analysis module includes the time-varying rate or cumulative variation of the relative distance or relative angle defined by the monitoring point node and its mapped standard reference network reference point set.

[0014] When determining internal relative deformation information, the deformation interpretation module compares the time evolution pattern of the target building's dynamic embedding vector or attention weight with the corresponding time evolution pattern obtained from surrounding stable reference buildings to identify and quantify specific internal deformations indicating stress concentration or uneven settlement.

[0015] When determining the absolute attitude change information, the deformation interpretation module uses the time change index of the relative geometric relationship in space and time through a preset conversion model to calculate the absolute vertical settlement rate expressed in millimeters / year or similar units and the tilt angle change value including the pitch angle and roll angle expressed in degrees or radians.

[0016] The ground-based measured reference data used by the verification module is selected from at least one of: the three-dimensional coordinate time series of a global navigation satellite system (GNSS) receiver deployed on the target building, the elevation change data of precise leveling measurement, the angle change data recorded by the inclinometer, or the multi-period point cloud data obtained by terrestrial three-dimensional laser scanning (TLS).

[0017] The calibration feedback instructions generated by the calibration module are used to adjust at least one of the following: the model structure or hyperparameters of the graph self-attention network in the first analysis module, the definition parameters of the standard reference network, the mapping rules in the standard network mapping module, or the classification threshold used to distinguish different deformation levels in the deformation interpretation module.

[0018] The system also includes a data interface and preprocessing module, which is used to receive the initial monitoring point detection results generated by analyzing a series of high-resolution optical satellite images (wherein the satellite images contain stereo observation information) acquired over time and covering the target building, and perform coordinate solution, time series matching and noise filtering, and output four-dimensional space-time coordinate data to the four-dimensional space-time network construction module.

[0019] The beneficial effects of the present invention are:

[0020] 1. By constructing a four-dimensional spatiotemporal network model and combining it with analytical methods such as graph self-attention networks, this system can deeply explore the complex spatiotemporal dynamic correlations between building monitoring points, revealing internal relative deformation patterns and early subtle changes that are difficult to detect based on displacement alone. At the same time, by introducing a standard reference network associated with gravity and specialized geometric analysis, it can realize the solution of the building's absolute posture (settlement, tilt, including pitch and roll angles). Combined with the analysis of overall displacement, the system can provide multi-dimensional and comprehensive deformation information, including internal relative deformation, overall displacement, and absolute posture changes.

[0021] 2. This system adopts a multi-strategy deformation interpretation method. By comparing and analyzing the dynamic patterns or overall motion of the target building with the surrounding stable reference buildings, it can effectively filter out regional common-mode noise (such as atmospheric delay residuals, seasonal influences, etc.) and clearly distinguish the building's specific internal deformation and overall displacement relative to the environment. Combined with absolute posture analysis based on a standard reference network, it achieves effective decoupling and identification of deformation information with different physical sources and forms.

[0022] 3. The system has built-in verification and calibration modules, forming a closed-loop feedback mechanism. By comparing the system output with ground-based benchmark data (such as high-precision measurement results from GNSS, levels, inclinometers, etc.), the accuracy of the system's monitoring results can be objectively evaluated. More importantly, the calibration module can automatically or semi-automatically generate feedback instructions based on the evaluation results, and iteratively adjust and optimize key parameters, model structures, or processing rules within the system, thereby continuously improving the system's monitoring accuracy and long-term operational reliability.

[0023] 4. This system not only provides raw deformation data but also integrates and analyzes the interpreted multi-dimensional deformation information through the status assessment module. Combining preset thresholds and change trend judgments, it automatically generates a health status report including risk level assessment. This provides managers with an intuitive and quantitative decision-making basis, helps to achieve condition-based predictive maintenance, and improves the efficiency and safety of building asset management.

[0024] 5. This invention primarily relies on high-resolution satellite imagery as a data source, enabling non-contact remote monitoring of building health. This technology is particularly suitable for scenarios where ground personnel have difficulty accessing or where deploying traditional sensors is costly (e.g., high-rise buildings, large bridges, and buildings in hazardous areas). The system's automated processing flow (from data preprocessing and analysis to evaluation) reduces manual intervention and has the potential to be applied to large-scale, periodic building surveys and long-term health monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a system architecture diagram of the present invention;

[0026] Figure 2 This is a timing diagram of the present invention. DETAILED DESCRIPTION

[0027] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0028] It is to be noted that the directions of "left", "right", "up", "down", "front", "back", "inside" and "outside" in the following schemes are all relative directions and are not listed here one by one.

[0029] Example 1:

[0030] like Figure 1 and Figure 2 As shown, this embodiment describes in detail the basic construction and core analysis front-end part of a building health status detection system. This embodiment aims to illustrate how the system starts from raw data input, constructs a structured four-dimensional space-time network model, establishes a mapping relationship with a standard reference network, and uses advanced analysis modules to generate key intermediate indicators required for subsequent deformation interpretation. The implementation method of this embodiment mainly corresponds to the functions of the four-dimensional space-time network construction module, the standard network mapping module, the first analysis module, the second analysis module, and the data interface and preprocessing module.

[0031] In this embodiment, the building health status detection system implements its basic construction and core analysis functions in the following manner:

[0032] The data interface and preprocessing module is equipped with a data interface for receiving a series of high-resolution optical satellite images covering the target building and its surrounding area, acquired over time and provided by commercial satellite providers (such as Maxar and Airbus). These images have a spatial resolution better than 0.5 meters and contain stereo observation information (for example, stereo pairs from the same or different orbits) to support the calculation of three-dimensional coordinates. The received images first undergo rigorous preprocessing, including radiometric calibration, atmospheric correction, orthorectification, and high-precision time-series image registration. Computer vision techniques, such as automatic building outline recognition algorithms combined with convolutional neural networks (such as U-Net or Mask R-CNN), are used to determine the scope of the target building. Based on this, deep learning corner detection algorithms such as SuperPoint, or combined with traditional methods such as Harris and FAST for sub-pixel positioning, are used to identify clear and stable corner points such as building roofs to generate initial monitoring point detection results. For time-series images, robust feature matching and tracking algorithms (such as photometric matching based on Siamese networks combined with geometrically constrained graph optimization tracking, or combined with Kalman filter prediction) are used to continuously track the corresponding positions of these monitoring points in different images.

[0033] Utilizing the stereo observation information of satellite images, the three-dimensional geodetic coordinates (longitude, latitude, elevation) or projection coordinates (x, y, z) of the successfully tracked monitoring points at each time point are calculated through photogrammetry principles (such as spatial resection, forward intersection, and bundle adjustment). The necessary noise filtering and outlier removal are performed on the solution results. After the above processing, the data interface and preprocessing module finally output reliable four-dimensional space-time coordinate data containing the three-dimensional coordinates and time information of each monitoring point at each time stamp.

[0034] The four-dimensional spatiotemporal network construction module receives the four-dimensional spatiotemporal coordinate data from the data interface and preprocessing module. For each target building, at each timestamp tk, the corresponding monitoring point three-dimensional coordinates (xi, yi, zi) and timestamp tk are encapsulated into a network node vi(tk).

[0035] Within the same timestamp tk, based on the known roof geometric topology information of the target building (which can be imported from design drawings, BIM models, or inferred by analyzing the spatial distribution pattern of monitoring points), an undirected spatial edge eij(tk) is established between the monitoring point nodes that represent physical proximity in the structure (such as along the same ridge line or eaves line). For the same physical monitoring point i, a directed time edge vi(tk)→vi(tk+1) is established between consecutive timestamps tk and tk+1 to represent the temporal evolution relationship of the point.

[0036] The constructed nodes, spatial edges, temporal edges and their attributes (such as coordinates, timestamps, and edge types) are organized into a graph data structure (for example, represented by an adjacency list) and can be stored in a graph database (such as Neo4j) or an in-memory data structure.

[0037] A structured four-dimensional spatiotemporal network model is generated for each target building, which encodes the building's spatial form and its dynamic changes over time.

[0038] The standard network mapping module receives a four-dimensional spatiotemporal network model, the geographic location of target buildings, and predefined standard reference network construction rules and mapping rules.

[0039] According to the geographical location of the building, the local vertical direction of gravity is determined using a high-precision geoid model (such as EGM2008) or a local gravity field model. On this basis, a standard reference network associated with the building is instantiated. The standard reference network can be designed as a virtual coordinate system or a set of points, whose Z axis is strictly aligned with the opposite direction of local gravity (vertically upward), and the X and Y axes are defined according to the geographic direction (such as due north and due east). The reference points have predefined and stable relative coordinates in the standard reference network.

[0040] Using specific geometric mapping rules, each monitoring point node vi(tk) in the four-dimensional space-time network model is mapped to three reference points with predetermined relative geometric positions in its associated standard reference network. For example, these three points can be a projection point on the origin of the standard reference network, and two other reference points located on the horizontal and vertical planes of the standard reference network that can reflect the relative posture information of the node. The mapping relationship between the monitoring point node and its corresponding standard reference network reference point is established and stored. This relationship can be attached to the model as a node attribute or stored as independent associated data.

[0041] The first analysis module receives a four-dimensional spatiotemporal network model and adopts a spatiotemporal graph neural network model, specifically a graph self-attention network (GAT) or its spatiotemporal extension (such as ST-GAT). The network takes a four-dimensional spatiotemporal network model (including node coordinate features, spatial edges, and temporal edges) as input. Through network-level calculations, the model can learn the complex spatiotemporal dependencies between nodes. The self-attention mechanism enables the model to dynamically assign different importance (attention weights) to different neighboring nodes (in space or time), capturing the synergistic / differential patterns of force transmission or deformation within the structure. At the same time, the model generates a dynamic embedding vector for each node vi(tk), which is a highly condensed representation of the node's own information and its spatiotemporal neighborhood structure and dynamics.

[0042] Output internal spatiotemporal dynamic correlation indicators, including the time-varying inter-node attention weight values and the time-varying dynamic embedding vector coordinates.

[0043] The second analysis module receives the mapping relationship and the four-dimensional spatiotemporal coordinate data of each monitoring point node. For each monitoring point node vi(tk) and its standard reference network reference point set associated with the mapping relationship, this module performs geometric calculations.

[0044] Calculate and track the relative distance (e.g., the distance to each reference point) and relative angle (e.g., the angle formed by the node and the two reference points, or the angle of the node relative to the standard reference network coordinate axis) defined by the monitoring point node vi(tk) relative to its corresponding three standard reference network reference points, calculate the rate of change or cumulative change of these geometric parameters over time, and output the time change index of the spatiotemporal relative geometric relationship, specifically in the form of time series data of relative distance / angle, change rate or cumulative change.

[0045] Working process

[0046] The workflow begins with the data interface and preprocessing module, which first receives a series of high-resolution optical satellite images covering the target building acquired in chronological order. These images preferably include stereo observation capabilities. Through the image processing and computer vision algorithms integrated within the data interface and preprocessing module (including target recognition, feature point detection, cross-temporal tracking, three-dimensional coordinate solution based on stereo pairs, and necessary noise filtering and optimization), the raw image data is converted into a set of reliable and time-aligned four-dimensional spatiotemporal coordinate data. This data is the basis for all subsequent analysis. It assigns three-dimensional spatial coordinates (x, y, z) to each stably tracked target building monitoring point in a continuous time series (t1, t2, ..., tn).

[0047] Subsequently, the four-dimensional space-time network construction module receives this time series coordinate data. It not only stores these points, but also gives structural meaning to these discrete space-time coordinate points. For each timestamp tk, the module constructs a spatial graph based on the geometric topology of the building (such as the connection relationship between the roof edges), using nodes to represent monitoring points and spatial edges to represent the structural connections between them. More importantly, it connects the nodes of the same physical monitoring point at different timestamps tk and tk+1 by establishing time edges. In this way, the four-dimensional space-time network construction module transforms (constructs) the input point cloud time series data into a dynamic four-dimensional space-time network model that can reflect the spatial form of the building and its evolution over time.

[0048] On this basis, the standard network mapping module starts working. It receives the constructed four-dimensional space-time network model and, based on the geographic location information of the target building and predefined rules, instantiates a standard reference network aligned with the local gravity field. The core task is to perform mapping: each dynamic node in the four-dimensional space-time network model is established with a set of static reference points in its associated standard reference network according to specific geometric rules (for example, the 1-to-3 mapping used in this embodiment). Establish a clear correspondence, that is, a mapping relationship. This step provides a key reference benchmark for the subsequent analysis of absolute posture changes.

[0049] Next, the system launches two core analysis engines in parallel, each performing in-depth mining on the previously generated data:

[0050] The first analysis module focuses on the four-dimensional spatiotemporal network model itself. It leverages the powerful capabilities of the graph self-attention network to deeply analyze the internal structure encoded in the four-dimensional spatiotemporal network model and its dynamic evolution. By learning the spatiotemporal dependencies between nodes (attention weights) and generating dynamic embedding vectors reflecting the node states, it aims to capture the complex patterns and correlations of relative motion between various parts of the building. Its output mainly reflects the internal relative dynamic characteristics of the building.

[0051] The second analysis module uses mapping relationships and the coordinate data of monitoring points. It does not focus on the relationship between points inside the building, but focuses on the relative geometric relationship (such as distance and angle) between each monitoring point and its corresponding standard reference network reference point. By calculating and tracking the changes of these geometric parameters over time, the module aims to quantify the movement of the monitoring point relative to the external absolute reference system (gravity direction). The time change index it outputs mainly reflects the absolute posture change trend of the building.

[0052] The attention weights and dynamic embedding vectors output by the first analysis module, and the temporal variation index output by the second analysis module, together constitute the core outputs of this embodiment (system infrastructure construction and core analysis phase). The dynamic embedding vectors and temporal variation index reveal the behavioral characteristics of the building from two different dimensions: "internal relative dynamics" and "external absolute reference." They are structured and stored and passed to the subsequent deformation interpretation module as key input for interpreting three specific deformation types: internal relative deformation, overall displacement, and absolute posture change.

[0053] Through the above detailed workflow, this embodiment gradually transforms the original satellite remote sensing observation data into structured analysis-ready data containing deep features and multi-dimensional reference information, laying a solid foundation for the subsequent implementation and comprehensive building health status assessment.

[0054] Use Cases

[0055] Consider long-term health monitoring of a super-high-rise building located in the city center. Due to the complex surrounding environment, the deployment of ground-based monitoring methods is difficult and costly. The technical solution of this embodiment can play an important role: the data interface and preprocessing module regularly (e.g., quarterly) obtains high-resolution stereo satellite images of the area covered by the building. Through an automated processing process, dozens of corner points on the roof and its ancillary structures are identified and tracked, generating four-dimensional spatiotemporal coordinate data of these corner points, with a time span of up to several years.

[0056] The four-dimensional space-time network construction module constructs a four-dimensional space-time network model of the super high-rise building based on the four-dimensional space-time coordinate data and the basic structural information of the building. The standard network mapping module establishes an associated standard reference network for the building and maps the network nodes to reference points to establish a mapping relationship.

[0057] The first analysis module analyzes the four-dimensional space-time network model, and the output dynamic embedding vector reveals subtle differences in the response patterns of different parts of the building (such as the core tube and outer frame) under wind loads or temperature changes. The second analysis module analyzes the mapping relationship and outputs time-varying indicators that directly reflect the slight swing of the building top relative to the direction of gravity or the overall settlement trend.

[0058] The structured four-dimensional space-time network model, dynamic embedding vectors, and time-varying indicators generated by this embodiment provide high-quality input and basis for subsequent interpretation of possible differential settlement, overall tilt, torsional deformation, etc. of the super high-rise building.

[0059] Through the four-dimensional space-time network construction module, discrete monitoring point data are converted into a four-dimensional space-time network model with topological structure and temporal relationship, which greatly enhances the analyzability of the data and facilitates the subsequent application of models such as graph neural networks.

[0060] The standard reference network associated with gravity and its mapping relationship established through the standard network mapping module introduces a stable absolute reference system for the system, making it possible to subsequently distinguish between relative deformation and absolute posture changes (such as settlement and tilt).

[0061] The first analysis module uses the graph self-attention mechanism to automatically learn and extract deep features (attention weights and dynamic embedding vectors) that reflect the complex dynamic responses inside the building, which has the potential to detect early, subtle or nonlinear deformation patterns.

[0062] The second analysis module directly quantifies the changes in the geometric relationship of the monitoring points relative to the stable reference system (time-varying indicators), providing a direct basis for the subsequent calculation of absolute settlement and tilt angle.

[0063] The structured model and two types of core analysis indicators dynamically embedded vectors and time-varying indicators generated in this embodiment lay a solid data and feature foundation for achieving refined, multi-strategy deformation interpretation and state assessment in subsequent embodiments.

[0064] Combined with data interfaces and pre-processing modules, the entire process is based on satellite remote sensing data and has the potential for non-contact, large-scale, and periodic monitoring. It also has a high degree of automation and is suitable for monitoring scenarios where human intervention is difficult or inconvenient.

[0065] Example 2:

[0066] like Figure 1 and Figure 2 As shown, this embodiment describes how the building health status detection system performs core deformation interpretation and status assessment based on the aforementioned embodiment one (system construction and core analysis basis). This embodiment focuses on converting the intermediate analysis results (internal correlation indicators and posture change-related indicators) generated by embodiment one into building deformation information with clear physical meaning, and ultimately forming a health status report including risk assessment. The implementation method of this embodiment mainly corresponds to the functions of the deformation interpretation module and the status assessment module, and also covers the logic of determining the overall displacement information.

[0067] In this embodiment, the system's deformation interpretation and state assessment functions are implemented through the following modules and their internal logic:

[0068] The input of the deformation interpretation module is: the attention weight and dynamic embedding vector from the first analysis module, the time change index of the spatiotemporal relative geometric relationship from the second analysis module, and the original four-dimensional spatiotemporal coordinate data from the data interface and preprocessing module (or its cache).

[0069] Surrounding environment reference information obtained from the system or input externally (for example, analysis results obtained by applying the process of embodiment 1 to adjacent buildings identified as stable).

[0070] The algorithms within the deformation interpretation module analyze the time series of the input dynamic embedding vectors or attention weights. For example, time series clustering or anomaly detection algorithms can be used to identify monitoring points or groups of points where the embedding vector trajectory deviates significantly (relative to its own historical trajectory or relative to other nodes), or to analyze abnormal and continuous changes in the connection strength in the attention weight graph.

[0071] In order to distinguish real internal deformation from regional common-mode effects (such as large-scale ground subsidence, seasonal thermal expansion and contraction, satellite orbit or atmospheric error residuals), the system compares the above-mentioned change pattern of the target building with the corresponding pattern extracted from the surrounding environmental reference information (i.e., stable reference building), for example, calculating the correlation or difference between the change patterns of the target building and the reference building.

[0072] Only those statistically significant relative change patterns that are significantly different from those of the reference building are identified as specific internal deformations indicating internal stress concentration, uneven settlement, loose component connections, or local damage. The internal relative deformation information is quantified as relative displacement rate between point pairs, strain index, or marked as abnormal areas.

[0073] The deformation interpretation module uses the input four-dimensional space-time coordinate data to calculate the geometric center of mass of the point cloud composed of all monitoring points of the target building for each timestamp tk, or fits a rigid body transformation model (including translation and rotation) through the least squares method.

[0074] Extract the translation vector of the target building at time tk relative to time tk−1.

[0075] At the same time, the average translation vector of the reference building is calculated using the surrounding environment reference information (specifically the four-dimensional space-time coordinate data of the reference building).

[0076] The translation vector of the target building is subtracted from the average translation vector of the reference building to obtain the specific overall displacement of the target building relative to its local stable environment, forming the overall displacement information, which is usually expressed as a velocity vector (such as the velocity in the horizontal X and Y directions) and the cumulative displacement trajectory.

[0077] The deformation interpretation module receives the temporal change index of the spatiotemporal relative geometric relationship from the second analysis module, which is usually a time series of parameters such as the distance and angle of the monitoring point relative to its standard reference network reference point.

[0078] A pre-established transformation model (e.g., based on trigonometric functions, coordinate transformations, or small calibration models) is applied based on a standard reference network structure and mapping rules, which indices the relative geometric changes of the input into absolute pose parameters with well-defined physical units and directions.

[0079] The output absolute attitude change information specifically includes: the absolute vertical settlement rate of the building as a whole or a specific part relative to the direction of gravity (for example, in millimeters per year), and the tilt angle change value that describes the change of its tilt state over time. This value is further decomposed into the pitch angle (Pitch) and roll angle (Roll) commonly used in engineering (for example, in degrees or arc seconds).

[0080] The deformation interpretation module ultimately outputs three independent but potentially related sets of deformation information: internal relative deformation information, global displacement information, and absolute posture change information.

[0081] The state assessment module receives three types of deformation information from the deformation interpretation module: internal relative deformation information, overall displacement information, and absolute posture change information. It may also include auxiliary information such as the building type, design specifications, historical monitoring data, and preset deformation control thresholds.

[0082] First, the three inputs were cross-validated to check for logical contradictions, for example, whether a significant tilt was accompanied by a corresponding internal differential sedimentation pattern.

[0083] Compare the calculated deformations (such as settlement rate, tilt angle, and maximum relative displacement) with the allowable thresholds in the design code or industry standard corresponding to this type of building.

[0084] Analyze the time series of deformation parameters (such as settlement rate and rate of change of tilt angle) to determine whether the deformation is converging, remaining stable, or showing an accelerating trend.

[0085] Based on the results of the above integrated analysis (deformation magnitude, rate, trend, distribution, and relationship with thresholds), the status assessment module uses preset assessment rules (which may be based on expert systems or machine learning models) to grade the current health status of the building, for example, dividing it into risk level assessments of "normal", "concern", "warning", "dangerous", etc.

[0086] The relative deformation information within the quantified deformation indicators, overall displacement information and absolute posture change information, time series charts, possible deformation visualization diagrams (such as cloud maps, vector maps), threshold comparison results and final risk level assessment are integrated to generate a comprehensive health status assessment report.

[0087] Working process

[0088] The deformation interpretation module, as the core processing unit of this stage, is activated first. It receives the key outputs from the two parallel analysis modules in Example 1: the attention weights and dynamic embedding vectors generated by the first analysis module, which reflect the internal dynamic correlation, and the time change index of the spatiotemporal relative geometric relationship, which reflects the geometric change relative to the absolute reference frame, generated by the second analysis module. At the same time, in order to perform comprehensive interpretation, the deformation interpretation module also calls the original four-dimensional spatiotemporal coordinate data (for overall motion analysis) and the surrounding environment reference information obtained by system processing (for filtering common mode effects).

[0089] After receiving the required input, the deformation interpretation module activates its three unique interpretation strategies in parallel or in a predetermined logical sequence, each independently analyzing different dimensions of data:

[0090] Through time series analysis, pattern recognition or anomaly detection algorithms, the temporal evolution characteristics of attention weights and dynamic embedding vectors are deeply explored. After identifying significant changes in the correlation between nodes within the target building or abnormal drift of node status, the strategy effectively filters out regional common mode influences by strictly comparing them with the corresponding patterns of stable buildings in the surrounding environment reference information. Ultimately, the strategy outputs quantitative internal relative deformation information, indicating the possible stress concentration or uneven deformation areas and degree inside the building.

[0091] Calculate the center of mass position of the target building monitoring point cloud at each time step or obtain the overall translation vector by fitting the rigid body transformation model. Then, compare this motion vector with the average motion vector calculated from the surrounding environment reference information (reference building coordinate data) by the same method (vector subtraction). In this way, the overall displacement information of the target building relative to its local stable environment (such as horizontal velocity, cumulative displacement trajectory) is separated and output.

[0092] By applying a preset conversion model based on the geometry and mapping relationship of the standard reference network, the time-varying indicators of the geometric relationship of the monitoring point relative to the reference point of the standard reference network (such as relative distance and angle change rate) are converted into absolute attitude parameters used in engineering practice. The final output is absolute attitude change information including the absolute vertical settlement rate and the decomposed pitch angle and roll angle time series.

[0093] After completing the calculations for the three strategies, the deformation interpretation module integrates the relative deformation information, overall displacement information, and absolute posture change information within their respective output results. This integrated structured data set containing multi-dimensional deformation information is passed to the state assessment module.

[0094] The state assessment module receives the relative deformation information, overall displacement information and absolute attitude change information within the complete deformation information set from the module. It performs the following operations: performs consistency checks and comprehensive analysis of the information, for example, verifies whether the observed tilt is consistent with the internal deformation pattern, quantitatively compares key deformation parameters (such as maximum settlement rate, maximum tilt angle, maximum relative displacement, etc.) with the safety thresholds or performance standards preset in the system for this type of building, analyzes the time series data of each deformation parameter, and determines whether the deformation tends to be stable, continues to develop, or shows an accelerated deterioration trend. Based on the results of the above quantitative comparison and trend analysis, the built-in assessment rules or models are applied to make an objective risk level assessment of the current health status of the building (for example, "normal", "concern", "warning", "dangerous").

[0095] All key quantitative deformation indicators, time series graphs, risk level assessment results and possible recommendations are automatically compiled into a structured and information-rich health status assessment report.

[0096] The status assessment module outputs a final health status assessment report. This report is not only the final deliverable of this embodiment (deformation interpretation and status assessment phase), which can be reviewed and used by users for decision-making, but also serves as a key input to be passed to the verification module in embodiment 3 for external verification of system performance.

[0097] This embodiment effectively transforms the intermediate analysis data generated by the first embodiment into a comprehensive, in-depth and actionable understanding and assessment of the building's health status.

[0098] By implementing three different interpretation strategies, the system can comprehensively diagnose the health of a building from multiple dimensions, including internal relative motion, overall motion relative to the environment, and posture changes relative to an absolute reference frame, thus avoiding the one-sidedness that may be caused by a single indicator.

[0099] The status assessment module not only presents data, but also provides a clear risk level assessment through integrated analysis, threshold comparison and trend judgment, providing direct and intelligent support for managers' maintenance decisions (such as whether inspection is needed, when to inspect, and inspection focus).

[0100] By comparing with the surrounding reference information, common mode noise is effectively suppressed, the relative accuracy and reliability of internal deformation and overall displacement monitoring are improved, and by introducing a standard reference network and conversion, the quantitative monitoring capability of absolute settlement and tilt is provided.

[0101] It can automatically generate health status assessment reports containing rich information, unify assessment standards and output formats, and improve the efficiency and standardization of monitoring work.

[0102] Example 3:

[0103] like Figure 1 and Figure 2 As shown, this embodiment describes in detail the crucial subsequent steps of the building health status detection system after completing deformation interpretation and status assessment (as described in Example 2): verification and adaptive calibration of system output results. This embodiment focuses on ensuring the accuracy and reliability of system monitoring results and establishing a feedback mechanism to achieve continuous optimization and performance improvement of the system. The implementation method of this embodiment mainly corresponds to the functions of the verification module and the calibration module.

[0104] In this embodiment, the verification and calibration functions of the system are implemented through the following modules and their internal logic:

[0105] The verification module receives the health status assessment report from the status assessment module, which contains quantitative deformation information (such as key indicators of internal relative deformation, overall displacement rate, settlement rate and tilt angle of absolute posture change, etc.), and independently acquired, high-precision ground-based benchmark data.

[0106] The verification module is equipped with an interface for accessing or importing ground-based benchmark data. These data come from a variety of sources, including: three-dimensional coordinate time series obtained by continuous or periodic observations of Global Navigation Satellite System (GNSS) receivers deployed at key locations on the target building; elevation change data obtained by traditional precision leveling; angle change time series recorded by inclinometers installed on the structure; or displacement information of homonymous points extracted from multi-period high-density point cloud data obtained by terrestrial three-dimensional laser scanning (TLS).

[0107] The verification module implements a spatiotemporal alignment algorithm that matches the deformation of the monitoring point (or the specific location inferred from it) calculated by the system in the health status assessment report with the measurement results at the same geographical location and the same time period (or the closest time point) in the ground-based benchmark data. This may involve operations such as coordinate system conversion, time interpolation, and point association (for example, associating the roof corner points monitored by satellite with GNSS points deployed on the ground).

[0108] For successfully matched data pairs, the difference (i.e., error) between the output value of this system and the ground measured true value is calculated. For example, the difference in sedimentation rate, the difference in tilt angle, the difference in displacement vector of a specific point, etc.

[0109] Based on the calculated errors, a series of accuracy evaluation indicators are statistically generated. These indicators may include: root mean square error (RMSE), mean absolute error (MAE), bias, standard deviation of the error, and the correlation coefficient between the two, etc., to comprehensively and quantitatively evaluate the performance of the system in monitoring various deformation parameters.

[0110] Output a set of quantitative accuracy evaluation indicators to reflect the accuracy level and error characteristics of the system's current monitoring results. The calibration module receives the accuracy evaluation indicators from the verification module and also refers to the system's current configuration parameters and historical calibration records.

[0111] The calibration module receives and analyzes accuracy assessment indicators. For example, if the RMSE is large, it analyzes whether the main cause is random errors or systematic deviations, whether the deviations are positive or negative, and whether the errors are related to a specific time period, a specific area, or a specific deformation type. Through this analysis, it attempts to infer which upstream processing link may be the main source of the error (for example, satellite data preprocessing, three-dimensional coordinate solution accuracy, standard reference network definition, applicability of the analysis model [3,4], deviation of the interpretation algorithm, etc.).

[0112] Based on the results of the error analysis and diagnosis, the calibration module applies a preset calibration strategy (which can be a rule-based expert system or a machine learning-based adaptive optimization algorithm) to generate specific calibration feedback instructions. The goal of these instructions is to adjust the parameters or models within the system in order to reduce the identified errors in subsequent processing. The objects that can be adjusted include (but are not limited to):

[0113] The model structure (such as the number of layers, node feature dimensions) or hyperparameters (such as learning rate, regularization coefficient) of the graph self-attention network in the first analysis module; the definition parameters of the standard reference network (such as the origin position, fine-tuning of the coordinate axis orientation); the mapping rules in the standard network mapping module (such as adjusting the mapping algorithm or parameters); the classification threshold used to distinguish different deformation levels (risk levels) in the deformation interpretation module; it is recommended to adjust certain parameters in the data interface and preprocessing module (such as filtering strength).

[0114] The generated calibration feedback instructions are passed back to the corresponding modules or parameter configuration files in the system. After receiving the instructions, the system can automatically update the relevant parameters or prompt the operator to make manual adjustments. These adjustments will take effect when the next monitoring task is performed or the model is retrained, thereby realizing iterative feedback optimization of the system and outputting calibration feedback instructions pointing to specific modules or parameters.

[0115] Working process

[0116] The workflow of this embodiment constitutes a closed loop for quality assurance and performance improvement of the entire monitoring system. After the health status assessment report is generated in Example 2, the verification module is activated. It actively obtains or receives relevant ground-based benchmark data. The module's core task is to perform strict spatiotemporal alignment and compare the key deformation quantification results in the system report with the benchmark data one by one. It calculates a series of accuracy assessment indicators that reflect the current system performance. These objective indicators are then fed into the calibration module, which conducts in-depth analysis of these indicators to diagnose the source of errors and determine the optimization direction. Based on the diagnostic results, the calibration module generates specific and actionable calibration feedback instructions, clearly indicating which modules (such as to) and which parameters or models in the system need to be adjusted. These instructions are fed back to the system through internal mechanisms, allowing the system to apply these adjustments in subsequent operations, achieving self-correction and continuous improvement of performance. This "assessment-diagnosis-feedback-adjustment" process forms a complete iterative optimization cycle.

[0117] The verification module provides objective and quantitative accuracy assessment indicators for the system output results by comparing them with recognized high-precision ground measurement methods, allowing users to understand the reliability level of the monitoring results. The iterative feedback mechanism established by the calibration module enables the system to self-adjust and optimize according to the actual verification results, which is expected to gradually improve the monitoring accuracy and adapt to changes in different monitoring objects and environmental conditions. Through continuous verification and calibration cycles, it can timely discover and correct possible systematic deviations or model drifts in the system, thereby significantly improving the stability and reliability of the system in the long-term operation. The clear verification links and transparent (or traceable) calibration process enhance users' trust in the results provided by this automated and intelligent monitoring system. The calibration module's diagnostic analysis of error sources can also provide valuable feedback information for the system's continued research and development and algorithm improvement, guiding subsequent optimization priorities.

[0118] The above description is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein and should not be regarded as excluding other embodiments. Instead, it can be used in various other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or technology or knowledge in related fields. The changes and modifications made by those skilled in the art do not depart from the spirit and scope of the present invention and should be within the scope of protection of the claims attached to the present invention.

Claims

1. A building health status detection system, characterized in that: include: A four-dimensional space-time network construction module is used to construct a four-dimensional space-time network model of the target building based on the four-dimensional space-time coordinate data. The model contains nodes representing the space-time positions of monitoring points, spatial edges connecting nodes within the same timestamp, and temporal edges connecting nodes of the same monitoring point at different timestamps. a standard network mapping module, configured to map nodes in the four-dimensional space-time network model to reference points of a predefined standard reference network associated with a local gravity direction, and establish a mapping relationship; a first analysis module configured to process the four-dimensional spatiotemporal network model using a graph self-attention network, and output attention weights representing the internal spatiotemporal dynamic correlations between nodes and dynamic embedding vectors encoding node states and neighborhood information; a second analysis module configured to process the mapping relationship and output, through geometric analysis, a time-varying index representing the spatiotemporal relative geometric relationship of the monitoring point relative to the reference point of the standard reference network; a deformation interpretation module configured to perform the following steps: step S1, based on the time-varying pattern of the dynamic embedding vector or attention weight, determine the internal relative deformation information of the target building by comparing it with the pattern of surrounding reference buildings to filter out common-mode effects; step S2, based on the four-dimensional spatiotemporal coordinate data, fit the rigid body transformation or track the center of mass, and determine the overall displacement information of the target building by comparing it with the motion of surrounding reference buildings; step S3, based on the time-varying index, determine the absolute attitude change information of the target building relative to the standard reference network, including the absolute vertical settlement rate, pitch angle, and roll angle, through a conversion model; a state assessment module configured to integrate the internal relative deformation information, overall displacement information, and absolute attitude change information, and generate a health status assessment report including a risk level assessment based on a preset threshold or change trend analysis; a verification module configured to compare the health status assessment report with the ground-based measured benchmark data to generate an accuracy assessment index; and a calibration module for generating a calibration feedback instruction based on the accuracy evaluation index, so as to adjust the parameters or model of at least one other module of the system through iterative feedback.

2. A building health status detection system according to claim 1, characterized in that: When constructing the spatial edge, the four-dimensional spatiotemporal network construction module connects structurally adjacent monitoring point nodes based on the roof geometric topology of the target building that is known or extracted from the image.

3. A building health status detection system according to claim 1, characterized in that: The standard network mapping module is configured to map each monitoring point node in the four-dimensional spatiotemporal network model to three reference points with predetermined relative geometric positions in the standard reference network.

4. A building health status detection system according to claim 1, characterized in that: The attention weight and dynamic embedding vector output by the first analysis module include the inter-node attention weight values that change over time and / or the node embedding vector coordinates that change over time.

5. A building health status detection system according to claim 1, characterized in that: The time-varying index of the spatiotemporal relative geometric relationship output by the second analysis module includes the time-varying rate or the cumulative changing amount of the relative distance or relative angle defined by the monitoring point node and its mapped standard reference network reference point set.

6. A building health status detection system according to claim 1, characterized in that: When determining the internal relative deformation information, the deformation interpretation module compares the time evolution pattern of the dynamic embedding vector or attention weight of the target building with the corresponding time evolution pattern obtained from the surrounding stable reference buildings to identify and quantify specific internal deformations indicating stress concentration or uneven settlement.

7. A building health status detection system according to claim 1, characterized in that: When determining the absolute attitude change information, the deformation interpretation module uses the time change index of the space-time relative geometric relationship through a preset conversion model to calculate the absolute vertical settlement rate and the tilt angle change value including the pitch angle and roll angle expressed in degrees or radians.

8. The building health status detection system according to claim 1, characterized in that: The ground measured reference data used by the verification module is selected from: the three-dimensional coordinate time series of the global navigation satellite system GNSS receiver deployed on the target building, the elevation change data of the precision leveling measurement, the angle change data recorded by the inclinometer, or at least one of the multi-period point cloud data obtained by the ground three-dimensional laser scanning TLS.

9. The building health status detection system according to claim 1, characterized in that: The calibration feedback instructions generated by the calibration module are used to adjust at least one of the following: the model structure or hyperparameters of the graph self-attention network in the first analysis module, the definition parameters of the standard reference network, the mapping rules in the standard network mapping module, or the classification threshold used to distinguish different deformation levels in the deformation interpretation module.

10. A building health status detection system according to claim 1, characterized in that: The system also includes a data interface and preprocessing module, which is used to receive initial monitoring point detection results generated by analyzing a series of high-resolution optical satellite images covering the target building acquired over time, and perform coordinate solution, time series matching and noise filtering, and output the four-dimensional space-time coordinate data to the four-dimensional space-time network construction module.

Images