Bridge damage detection quantification method and device based on air-ground data fusion

Abstract

The application discloses a bridge damage detection quantification method and device based on air-ground data fusion, and the method comprises the following steps: acquiring three-dimensional point cloud data of a bridge collected in multiple time periods, and constructing a time sequence point cloud model of the bridge in each time period; based on the time sequence point cloud model, the geometric curvature features of each sampling point are calculated, and the geometric curvature features are enhanced to obtain enhanced curvature features; the geometric curvature features in multiple time periods are subjected to time sequence difference analysis to obtain structural damage features, and a suspected damage area is determined according to the structural damage features; the real structural damage area is identified according to the enhanced curvature features of the suspected damage area; and the damage grade of the real structural damage area is determined according to the corresponding structural damage features of the real structural damage area, so that high-precision damage identification, multi-bridge type adaptive matching and accurate damage grade determination in a complex environment are realized.

Description

Technical Field

[0001] This application relates to the field of bridge health monitoring, specifically to a method and device for quantitative detection of bridge damage based on air-to-ground data fusion. Background Technology

[0002] As a critical node in transportation networks, the structural health of bridges directly impacts the safe operation of public transportation. During long-term service, bridges are susceptible to structural damage such as cracks, spalling, and deformation due to various factors including vehicle overloading, temperature stress, structural aging, and environmental erosion. Therefore, timely and accurate identification of potential bridge damage is crucial for preventing catastrophic accidents and optimizing maintenance decisions.

[0003] Currently, bridge damage detection methods mainly include traditional manual inspection, contact sensor monitoring, and non-contact detection based on three-dimensional laser scanning. However, with the expansion of transportation networks and the diversification of bridge structural forms, existing detection technologies have gradually revealed many limitations in practical engineering applications, making it difficult to meet the needs for high-precision, full-coverage, and automated detection.

[0004] In terms of data acquisition, traditional manual inspection and contact-based detection methods are inefficient, often requiring traffic closures or scaffolding, which are not only costly but also pose safety hazards, making it difficult to achieve rapid detection of long-distance bridge clusters. Although 3D laser scanning technology has been applied to bridge inspection, existing technologies mostly rely on a single ground scanning platform, which is limited by line-of-sight obstructions and has scanning blind spots in areas such as the inner side of bridge piers, the bottom of box girders, and the top of bridge towers. This makes it difficult to obtain complete 3D data of the entire bridge area, leading to the risk of missed damage identification.

[0005] In terms of data processing and damage identification, bridge structures are subjected to the coupled influence of multiple environmental factors such as temperature changes, vehicle loads, and wind-induced vibrations during service. The resulting non-structural deformations often overlap with the geometric features of actual structural damage. Existing technologies struggle to achieve high signal-to-noise ratio damage feature extraction under complex environmental interference, leading to a high rate of false damage assessment. Furthermore, existing methods often rely on single-dimensional geometric indicators for discrimination, which are insufficiently robust to point cloud noise or registration errors, easily resulting in false damage determinations. Simultaneously, existing discrimination criteria lack adaptability to the differentiated characteristics of bridges with different structural forms and struggle to achieve accurate quantitative assessments from geometric appearance to mechanical performance degradation, limiting their engineering applicability in large-scale bridge cluster inspections.

[0006] In summary, existing detection technologies cannot meet the engineering requirements of high-precision, full-coverage, low-false-judgment, early warning, automation, and real-time damage detection for bridge clusters. There is an urgent need for a quantitative method for full-area, accurate, and automated damage detection that is suitable for bridge clusters. Summary of the Invention

[0007] This application provides a bridge damage detection and quantification method and device based on air-ground data fusion, which can solve the technical problems existing in the prior art, such as incomplete bridge detection coverage, high damage misjudgment rate due to environmental interference, poor adaptability to different bridge types, and inability to accurately quantify and assess damage.

[0008] In a first aspect, embodiments of this application provide a bridge damage detection quantification method based on air-to-ground data fusion, including: Acquire 3D point cloud data of the bridge through collaborative collection of aerial and ground equipment at multiple time periods, and construct a time-series point cloud model of the entire bridge area at various time periods; Based on the temporal point cloud model, the geometric curvature features of each sampling point are calculated, and the geometric curvature features are enhanced based on the local curvature entropy to obtain enhanced curvature features; Temporal difference analysis is performed on the geometric curvature features over multiple time periods to obtain structural damage features, and suspected damage areas are determined based on the structural damage features. Based on the enhanced curvature characteristics of the suspected damaged areas, the actual structural damaged areas are identified. Based on the structural damage characteristics corresponding to the actual structural damage area, and combined with adaptive discrimination parameters that match the structural type information of the bridge, the damage level of the actual structural damage area is determined.

[0009] Secondly, embodiments of this application provide a bridge damage detection and quantification device based on air-to-ground data fusion, characterized in that it includes: The module is used to acquire 3D point cloud data of the bridge collected collaboratively by aerial and ground equipment at multiple time periods, and to construct a time-series point cloud model of the entire bridge at various time periods. The calculation module is used to calculate the geometric curvature features of each sampling point based on the time-series point cloud model, and to enhance the geometric curvature features based on the local curvature entropy to obtain enhanced curvature features. The analysis module is used to perform temporal difference analysis on the geometric curvature features over multiple time periods to obtain structural damage features, and to determine suspected damage areas based on the structural damage features. The identification module is used to identify the actual structural damage area based on the enhanced curvature characteristics of the suspected damage area; The determination module is used to determine the damage level of the actual structural damage area based on the structural damage characteristics corresponding to the actual structural damage area and in combination with adaptive discrimination parameters that match the structural type information of the bridge.

[0010] The bridge damage detection quantification method and apparatus based on air-to-ground data fusion provided in this application acquires three-dimensional point cloud data of the bridge collected collaboratively by aerial and ground equipment over multiple time periods, constructing a temporal point cloud model of the entire bridge area for each time period; based on the temporal point cloud model, the geometric curvature features of each sampling point are calculated, and the geometric curvature features are enhanced based on local curvature entropy to obtain enhanced curvature features; temporal difference analysis is performed on the geometric curvature features of multiple time periods to obtain structural damage features, and suspected damage areas are determined based on the structural damage features; based on the enhanced curvature features of the suspected damage areas... The method identifies the actual structural damage area based on the structural damage characteristics corresponding to the actual structural damage area. Combined with adaptive discrimination parameters that match the structural type information of the bridge, the damage level of the actual structural damage area is determined. This solves the technical problems in related technologies, such as incomplete coverage due to scanning blind spots in bridge detection, high false positive rates due to temperature and load interference from single geometric features, poor versatility due to fixed thresholds not adapting to different bridge types, and lack of graded and quantitative basis for damage assessment. It achieves full-area coverage without blind spots in bridges, high accuracy in damage identification under complex environments, adaptive matching for multiple bridge types, and accurate determination of damage levels. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating an embodiment of the bridge damage detection and quantification method based on air-to-ground data fusion according to this application. Figure 2 This is a schematic diagram of the functional modules of an embodiment of the bridge damage detection and quantification device based on air-to-ground data fusion according to this application. Detailed Implementation

[0012] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0013] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0014] In one aspect, embodiments of this application provide a method for quantifying bridge damage detection based on air-to-ground data fusion.

[0015] In one embodiment, reference is made to Figure 1 , Figure 1This is a flowchart illustrating the first embodiment of the bridge damage detection quantification method based on air-to-ground data fusion proposed in this application. Figure 1 As shown, the bridge damage detection quantification method based on air-to-ground data fusion includes: Step S1: Acquire 3D point cloud data of the bridge collected collaboratively by aerial and ground equipment at multiple time periods, and construct a time-series point cloud model of the entire bridge area at each time period.

[0016] In one embodiment, the airborne equipment can be an airborne lidar, and the ground equipment can be a ground-based 3D laser scanner. The airborne lidar and the ground-based 3D laser scanner are used together to collect a full-area 3D point cloud of the bridge, with data acquired at 6 to 12-month intervals. , , Three-phase time-series point cloud data.

[0017] The 3D point cloud data collected collaboratively by aerial and ground equipment is converted to a unified geographic coordinate system, such as the WGS-84 coordinate system, to eliminate spatial reference errors caused by inconsistencies in coordinate systems between different devices. Point cloud data collected by aerial and ground equipment within the same time period are spatially aligned to stitch together point clouds from multiple measurement stations. This allows the aerial perspective to supplement the top blind spots of the ground scan, and the ground perspective to supplement the bottom blind spots of the aerial scan. Subsequently, filtering algorithms are used to remove noise, redundant points, and irrelevant objects from the point cloud, and temporal synchronization and spatial fusion are performed to obtain a temporal point cloud model.

[0018] For example, in this embodiment, the multi-site point cloud stitching adopts a two-stage registration strategy of first coarse registration and then fine registration to achieve high-precision spatial alignment of air-to-ground multi-source point clouds.

[0019] In the coarse registration stage, corresponding feature points are manually selected. These are feature points that share the same physical location and can be clearly identified in both the airborne and ground point clouds, such as bridge corners, sign edges, and pier apexes. Typically, at least three pairs of corresponding feature points are selected to provide sufficient spatial constraints. Based on these corresponding feature points, a least-squares rigid transformation is used to complete the coarse registration between the aerial and ground point cloud data. The transformation formula is:

[0020] in, This is a rotation matrix used to describe the rotation angle of the point cloud in three-dimensional space. This is a translation vector used to describe the displacement of the point cloud in three-dimensional space. The original point, i.e., the point cloud coordinates before registration. These are the coarse registration points, i.e., the coordinates of the point cloud after rotation and translation transformations. The least-squares rigid transformation finds the optimal rotation matrix by minimizing the sum of squared coordinate errors between all corresponding feature points. Translation vector This allows the airborne point cloud to be roughly aligned with the ground point cloud in space, providing a good initial position for subsequent fine registration.

[0021] Then, the fine registration stage begins, employing the Iterative Closest Point (ICP) algorithm for iterative optimization. The objective function for fine registration is:

[0022] in, Let i be the i-th sampling point in the source point cloud, i.e., a point in the airborne or ground point cloud after coarse registration. In the target point cloud The corresponding nearest point, i.e., with The point with the smallest spatial distance. and Let the rotation matrix and translation vector be the ones to be optimized. The number of point pairs participating in the registration. The objective function is used to find the optimal... and This minimizes the sum of squared Euclidean distances between all corresponding point pairs in the transformed source point cloud and the target point cloud, thereby achieving sub-millimeter-level alignment between the air and ground points. This ensures high-precision spatial overlap of multi-source point clouds and provides a reliable data foundation for subsequent curvature calculations and temporal difference analysis.

[0023] Furthermore, in this embodiment, noise removal employs a combination of statistical filtering and radius filtering. Statistical filtering is based on the 3σ criterion, calculating the average distance from each point to its neighboring points. If the difference between this distance and the global mean satisfies... If it is, then it is determined to be an outlier and should be removed. For the first The average distance from a point to its neighboring points The neighborhood mean The standard deviation is used to remove sparse discrete noise. Radius filtering sets the search radius r, counts the number of neighborhood points N within that radius, and if the number of neighborhood points N < Then it is determined to be an isolated point and removed, where The minimum neighbor number threshold is set to remove residual noise points with excessively low local density.

[0024] After denoising, time-series synchronization is performed to ensure consistency of time bases for multiple data periods, facilitating subsequent time-series differential analysis. Redundant points are removed to balance the density of overlapping areas and avoid data weight bias. Spatial fusion is then performed to merge multi-source data, ultimately constructing a clean time-series point cloud model with no scanning blind spots and low noise.

[0025] This embodiment uses aerial equipment (such as airborne LiDAR) and ground equipment (such as ground-based 3D laser scanners) to collaboratively collect data. The aerial perspective is used to supplement the top blind spots of the ground scan, and the ground perspective is used to supplement the bottom blind spots of the aerial scan. Combined with registration and temporal fusion techniques, a clean temporal point cloud model with no scanning blind spots and low noise is constructed, providing a complete and high-quality data foundation for subsequent damage identification.

[0026] Step S2: Based on the time-series point cloud model, calculate the geometric curvature features of each sampling point, and enhance the geometric curvature features based on the local curvature entropy to obtain enhanced curvature features.

[0027] It is worth noting that in this embodiment, the geometric unevenness of the bridge surface is quantified by calculating Gaussian curvature, and an adaptive curvature entropy enhancement algorithm is introduced to amplify the entropy value of regions with weak curvature changes, thereby strengthening the characteristics of minor damage and solving the problem of missing detection of minor damage.

[0028] In a specific implementation, a local coordinate system is first established for each sampling point in the time-series point cloud model to facilitate surface fitting within a local area. Neighboring points are selected centered on the current sampling point, and a quadratic surface fitting model is constructed based on the coordinates of the current sampling point and its neighbors. The model equation is as follows:

[0029] in, , , , , , These are the surface fitting parameters for the quadratic surface fitting model. , , These are the three-dimensional coordinate values ​​of the point cloud's local coordinate system.

[0030] To constrain the surface shape, improve fitting reliability, and avoid unreasonable distortion of the fitted surface due to noise, this embodiment uses a least-squares objective function based on normal vector constraints to solve for the surface fitting parameters of the quadratic surface fitting model. The objective function is:

[0031] in, The number of neighboring points, For neighborhood points The measured elevation value, The elevation value of the fitted surface is calculated based on the quadratic surface fitting model. These are the normal vector constraint weight coefficients. The normal vector of the fitted surface. The original normal vector of the sampling point. Euclidean distance calculation; After solving for the surface fitting parameters, substitute these parameters into the constrained Gaussian curvature formula derived from the quadratic surface fitting model to calculate the geometric curvature characteristics of the sampling points. The formula is as follows:

[0032] in, Sampling points Geometric curvature characteristics; , Sampling points The local coordinates.

[0033] Furthermore, the geometric curvature features are enhanced based on local curvature entropy to obtain enhanced curvature features, specifically including: Based on the geometric curvature characteristics of neighboring points within the neighborhood of each sampling point, the neighborhood curvature probability density of the corresponding sampling point is determined. Specifically, this can be achieved by statistically analyzing the geometric curvature characteristics of all neighboring points within the neighborhood, constructing a curvature value distribution histogram, and normalizing the frequency of each partition in the histogram into a probability value. For the m-th neighboring point within the neighborhood, the probability value of the partition to which its geometric curvature characteristics belong is used as the neighborhood curvature probability density. This probability density reflects the degree of concentration of curvature values ​​within the neighborhood.

[0034] Calculate the local curvature entropy of the corresponding sampling point based on the neighborhood curvature probability density:

[0035] in, Sampling points The local curvature entropy, The number of neighboring points, For the first in the neighborhood Geometric curvature characteristics of a neighborhood point Let ε be the neighborhood curvature probability density, where ε is a very small positive number used to prevent the logarithmic operation from being zero. The larger the local curvature entropy, the more disordered the curvature distribution in that region, and the higher the probability of damage.

[0036] Based on the ratio of local curvature entropy to global maximum curvature entropy, and combined with preset enhancement adjustment coefficients, the adaptive enhancement weights for the corresponding sampling points are calculated:

[0037] in, Sampling points Adaptive augmentation weights, and To enhance the adjustment coefficient, The global maximum curvature entropy refers to the maximum local curvature entropy of all sampling points in the temporal point cloud model for the current time period. (Enhancement adjustment coefficient) and These are preset empirical coefficients, which can be determined through experimental calibration. Enhancement gain used to control the curvature characteristics of high-entropy regions. Used to set the base weights for low-entropy regions.

[0038] The geometric curvature features of the corresponding sampling points are weighted according to the adaptive enhancement weights to obtain the corresponding enhanced curvature features:

[0039] in, Sampling points Enhanced curvature features, Sampling points The geometric curvature characteristics.

[0040] By introducing the curvature entropy adaptive enhancement algorithm, the geometric curvature features are weighted and amplified by the degree of disorder (entropy value) of the local curvature distribution. This effectively enhances the damage features in areas with weak curvature changes, solves the problem that weak damage signals are easily submerged by noise in traditional methods, leading to missed detection, and achieves accurate capture of early minor damage to bridges.

[0041] Step S3: Perform temporal difference analysis on the geometric curvature features of multiple time periods to obtain structural damage features, and determine the suspected damage area based on the structural damage features.

[0042] In one embodiment, step S3 specifically includes: acquiring the geometric curvature features of three time periods, denoted as the first period geometric curvature features. Geometric curvature characteristics of the second period and the geometric curvature characteristics of the third period This corresponds to the t1, t2, and t3 data in the time-series point cloud model.

[0043] For each sampling point, based on the geometric curvature characteristics of the first period and the second periodic geometric curvature feature The first-order time difference was calculated. This is used to extract short-term curvature fluctuations; Based on the geometric curvature characteristics of the second period and the third periodic geometric curvature feature Calculate the second-order time difference It is used to capture long-term curvature trends.

[0044] Because non-structural deformations caused by temperature, temporary vehicle loads, and wind-induced vibrations are short-term, random, and reversible, while permanent structural damage is cumulative, a weighted summation of the first-order and second-order time-series differences is used to obtain a comprehensive time-series difference result. This comprehensive time-series difference result is then used as the structural damage characteristic, as shown in the formula:

[0045] in, Indicates structural damage characteristics, This represents the first-order temporal difference. This represents the second-order temporal difference. and Represents adaptive time-series weights, and satisfies + =1. Adaptive time-series weights can be calibrated experimentally or preset based on data fluctuation characteristics to balance short-term volatility sensitivity and long-term trend stability.

[0046] Furthermore, the structural damage features of each sampling point are compared with a preset damage discrimination threshold E, where the damage discrimination threshold E is an adaptive discrimination threshold. When When the value is greater than E, it is determined to be a permanent structural damage and retained. Sampling points whose structural damage characteristics are greater than the damage discrimination threshold are identified as suspected damage points. When the value is ≤E, it is determined to be a non-structural deformation and is removed, thereby effectively filtering out environmental interference.

[0047] Sampling points whose structural damage features are greater than the damage discrimination threshold are identified as suspected damage points. One or more adjacent suspected damage points constitute the suspected damage region, realizing the transformation from discrete points to continuous damage regions and providing target objects for subsequent multidimensional verification.

[0048] By constructing a three-phase time-series difference model, and utilizing the difference between the cumulative nature of structural damage and the stochastic reversibility of non-structural deformations (such as temperature changes, vehicle loads, and wind-induced vibrations), combined with adaptive weight screening, the interference caused by environmental factors can be accurately separated and eliminated, while retaining the true characteristics of permanent structural damage and significantly reducing the false alarm rate.

[0049] Step S4: Identify the actual structural damage area based on the enhanced curvature characteristics of the suspected damage area.

[0050] In one embodiment, based on the enhanced curvature characteristics of each suspected damage point within the suspected damage area, and the average enhanced curvature characteristics of neighboring points within the neighborhood of the suspected damage point, the curvature abrupt change characteristics of each suspected damage point are calculated:

[0051] in, Suspected damage point The curvature abrupt change feature, Suspected damage point Enhanced curvature features, The mean of the neighborhood enhanced curvature feature, The larger the value, the more significant the geometric abrupt change, indicating a higher probability of damage at that point.

[0052] Texture and modal features are extracted from each suspected damage point within the suspected damage area to obtain the non-geometric physical features. The texture damage features are obtained by calculating the variance of the point cloud surface normal vector and the texture dispersion to obtain texture feature values. The larger the value, the more obvious the damage such as surface cracks, spalling, or exposed reinforcement; modal features are constructed by fusing the bridge's natural frequency offset and mode curvature deviation to construct modal feature values. This reflects the variation in the overall mechanical properties of the structure.

[0053] Based on curvature abrupt change features, texture damage features, and modal features, the damage discrimination index of each suspected damage point is calculated:

[0054] in, Suspected damage point Damage discrimination index, Features of texture damage Modal features, This is a feature representing the global maximum curvature abrupt change. This represents the largest texture damage feature globally. It is the global maximum modal feature. , , These are adaptive weights that match the structural type information of the bridge, and + + =1. Adaptive weights in this embodiment. , , Automatic matching can be achieved by querying a preset bridge type feature database. For example, curvature features can be assigned higher weights to beam bridges, while modal features can be assigned higher weights to cable-stayed bridges, thus enabling adaptive discrimination of different bridge types. The global maximum eigenvalue refers to the maximum value of the features corresponding to all suspected damage points, which is used to normalize features in each dimension and eliminate dimensional differences.

[0055] If the damage discrimination index of the suspected damage point is greater than the preset false damage filtering threshold... ,Right now ≥ If the suspected damage point is determined to be a real damage point, a double verification is performed to filter out false damage signals caused by registration errors or local noise; if the damage discrimination index is less than the false damage filtering threshold, i.e. < If so, then it will be removed.

[0056] Ultimately, the actual structural damage region is composed of one or more adjacent actual damage points.

[0057] By fusing geometric curvature abrupt change features, surface texture damage features, and structural modal features, a three-dimensional damage discrimination model is constructed to double-verify suspected damage. A single feature anomaly is not considered damage; only when multiple features are supported is it confirmed as genuine damage. This effectively filters out false damage signals caused by registration errors and local noise, ensuring the accuracy of the identification results. This represents a leap from single geometric feature judgment to multi-dimensional fusion judgment, significantly reducing the false positive rate.

[0058] Step S5: Based on the structural damage characteristics corresponding to the actual structural damage area, and combined with adaptive discrimination parameters that match the structural type information of the bridge, determine the damage level of the actual structural damage area.

[0059] In one embodiment, step S5 specifically includes the following steps: Obtain the structural damage features corresponding to the actual structural damage region. Based on the structural type information of the bridge, a corresponding adaptive discrimination threshold is matched from a preset bridge type feature database.

[0060] Specifically, by calling the built-in bridge type feature database, the system can automatically identify bridge types, such as beam bridges, arch bridges, or cable-stayed bridges, and match the corresponding structural feature parameters to construct an adaptive threshold formula for multiple bridge types. ,in, This is a general benchmark threshold. For multi-bridge adaptive discrimination threshold, This refers to the bridge type adaptation coefficient, with different coefficient values ​​corresponding to different bridge types. For example, beam bridges, arch bridges, and cable-stayed bridges each correspond to different empirical coefficients. is a correction coefficient for structural features such as bridge span and material. Through this formula, adaptive adjustment of the threshold under different bridge types and different structural parameters can be realized, so as to avoid misjudgment caused by a fixed threshold. The bridge type feature database stores historical detection data of various bridges, and the mapping relationship between structural parameters and threshold coefficients, which is used to support the above matching process.

[0061] Then the ratio of the structural damage feature to the adaptive discrimination threshold is calculated to obtain a damage grading index, and the formula is: , where is the damage grading index, is the structural damage feature.

[0062] Finally, according to the value range where the damage grading index lies, the damage level of the real structural damage area is determined.

[0063] Exemplarily, in this embodiment, the detection area is divided into five levels: no damage, early damage, potential damage, moderate damage, and severe damage. If the damage grading index is equal to zero (i.e. =0), it is determined that the real structural damage area has no damage; if the damage grading index is greater than zero and less than or equal to the first grading threshold 0.3 (i.e. 0<DI≤0.3), it is determined that the real structural damage area has early damage; if the damage grading index is greater than the first grading threshold and less than or equal to the second grading threshold 0.6 (i.e. 0.3< ≤0.6), it is determined that the real structural damage area has potential damage; if the damage grading index is greater than the second grading threshold and less than or equal to the third grading threshold 1.0 (i.e. 0.6< ≤1.0), it is determined that the real structural damage area has moderate damage; if the damage grading index is greater than the third grading threshold (i.e. >1.0), it is determined that the real structural damage area has severe damage.

[0064] Through the preset bridge type feature database, the adaptive discrimination threshold is automatically matched according to the structural type information of the bridge (such as girder bridge, arch bridge, cable-stayed bridge) and structural parameters (span, material), which overcomes the defect that the fixed threshold in the prior art cannot adapt to the structural characteristics of different bridge types, realizes unified quantitative evaluation of different bridge damage standards, and is suitable for automatic inspection of large-scale bridge clusters.

[0065] In one embodiment, the method further comprises: inputting the structural damage feature and the damage area area corresponding to the real structural damage area into a bridge finite element model, establishing a linkage relationship between the measured curvature and the simulated mechanical response, and converting the structural damage feature into a finite element simulation deformation value based on the curvature-displacement mapping relationship:

[0066] in, The deformation value is the value obtained from the finite element simulation. The structural damage features, The area of ​​the damaged region. The curvature deformation mapping coefficient is related to the bridge material properties and is used to convert the geometric curvature dimension into the mechanical displacement dimension, thereby achieving the coupling of geometric features and mechanical parameters.

[0067] Based on the coordinate differences of the same sampling point in the time-series point cloud model across multiple time periods, the measured deformation value Ureal is calculated. The measured deformation value is derived from direct measurement of the point cloud data, reflecting the actual displacement changes of the bridge structure and serving as the true value benchmark for model calibration.

[0068] Based on the error between the finite element simulation deformation value and the measured deformation value, calculate the comprehensive damage parameters:

[0069] in, The damage comprehensive parameters are as follows: These are the inversion weighting coefficients. This refers to the actual structural damage area. The measured deformation value is used. This step accumulates the simulation and measurement errors within the damaged area through integration calculations, thereby quantifying the degree of damage. The comprehensive damage parameters include damage depth, damage range, and expansion trend.

[0070] The model parameters of the bridge finite element model are iteratively calibrated according to a pre-built model correction coefficient until the error between the simulated deformation value and the measured deformation value meets a preset convergence condition. The calibration process includes dynamically adjusting the finite element stiffness, boundary conditions, and material properties using the model correction coefficient γ to reduce the simulation and measurement errors, as shown in the following formula:

[0071] in, These are the correction coefficients for the model. γ represents the error between the finite element simulation deformation value and the measured deformation value. When γ approaches 1 or the error is less than the preset threshold, the comprehensive damage parameters obtained from the inversion based on the calibrated model are the accurate quantification results, thereby realizing the calibration and update of the bridge finite element model.

[0072] By linking Gaussian curvature with the finite element model for calibration, and using measured deformation values ​​to iteratively correct the finite element model parameters, mechanical indicators such as damage depth, damage range, and expansion trend are obtained through inversion. This upgrades damage assessment from traditional qualitative description to quantitative quantification, providing scientific and accurate data support for bridge maintenance and reinforcement decisions.

[0073] Secondly, embodiments of this application also provide a bridge damage detection and quantification device based on air-to-ground data fusion.

[0074] In one embodiment, reference is made to Figure 2 , Figure 2 This is a functional module diagram of an embodiment of the bridge damage detection and quantification device based on air-to-ground data fusion according to this application. Figure 2 As shown, the bridge damage detection and quantification device based on air-to-ground data fusion includes: The module is used to acquire 3D point cloud data of the bridge collected collaboratively by aerial and ground equipment at multiple time periods, and to construct a time-series point cloud model of the entire bridge at various time periods. The calculation module is used to calculate the geometric curvature features of each sampling point based on the time-series point cloud model, and to enhance the geometric curvature features based on the local curvature entropy to obtain enhanced curvature features. The analysis module is used to perform temporal difference analysis on the geometric curvature features over multiple time periods to obtain structural damage features, and to determine suspected damage areas based on the structural damage features. The identification module is used to identify the actual structural damage area based on the enhanced curvature characteristics of the suspected damage area; The determination module is used to determine the damage level of the actual structural damage area based on the structural damage characteristics corresponding to the actual structural damage area and in combination with adaptive discrimination parameters that match the structural type information of the bridge.

[0075] Furthermore, in one embodiment, the computing module is also used for: A local coordinate system is established for each sampling point in the time-series point cloud model; A quadratic surface fitting model is constructed based on the coordinates of each sampling point and the neighboring points within the sampling point's neighborhood. Based on the least squares objective function constrained by the normal vector, the surface fitting parameters of the quadratic surface fitting model are solved:

[0076] in, The number of neighboring points, For neighborhood points The measured elevation value, The elevation value of the fitted surface is calculated based on the quadratic surface fitting model. These are the normal vector constraint weight coefficients. The normal vector of the fitted surface. The original normal vector of the sampling point. Euclidean distance calculation; Substituting the surface fitting parameters into the constrained Gaussian curvature formula derived from the quadratic surface fitting model, the geometric curvature characteristics of the sampling points are calculated.

[0077] Furthermore, in one embodiment, the computing module is also used for: Based on the geometric curvature characteristics of the neighborhood points within the neighborhood of each sampling point, the neighborhood curvature probability density of the corresponding sampling point is determined. Calculate the local curvature entropy of the corresponding sampling point based on the neighborhood curvature probability density:

[0078] in, Sampling points The local curvature entropy, The number of neighboring points, For the first in the neighborhood Geometric curvature characteristics of a neighborhood point Let ε be the neighborhood curvature probability density, where ε is a positive number; Based on the ratio of the local curvature entropy to the global maximum curvature entropy, and in conjunction with a preset enhancement adjustment coefficient, the adaptive enhancement weight for the corresponding sampling point is calculated:

[0079] in, Sampling points Adaptive augmentation weights, and To enhance the adjustment coefficient, The global maximum curvature entropy; The geometric curvature features of the corresponding sampling points are weighted according to the adaptive enhancement weights to obtain the corresponding enhanced curvature features:

[0080] in, Sampling points Enhanced curvature features, Sampling points The geometric curvature characteristics.

[0081] Furthermore, in one embodiment, the analysis module is also used for: Obtain the geometric curvature features of three time periods, which are denoted as the first period geometric curvature feature, the second period geometric curvature feature, and the third period geometric curvature feature, respectively. For each sampling point, the first-order temporal difference is calculated based on the geometric curvature characteristics of the first period and the geometric curvature characteristics of the second period; The second-order temporal difference is calculated based on the geometric curvature characteristics of the second period and the geometric curvature characteristics of the third period. The first-order temporal difference and the second-order temporal difference are weighted and summed to obtain a comprehensive temporal difference result, and the comprehensive temporal difference result is used as the structural damage feature. The structural damage characteristics of each sampling point are compared with the preset damage discrimination threshold; Sampling points whose structural damage features are greater than the damage discrimination threshold are identified as suspected damage points, and one or more adjacent suspected damage points constitute the suspected damage region.

[0082] Furthermore, in one embodiment, the identification module is also used for: Based on the enhanced curvature characteristics of each suspected damage point within the suspected damage area, and the mean of the enhanced curvature characteristics of neighboring points within the neighborhood of the suspected damage point, the curvature abrupt change characteristics of each suspected damage point are calculated. Extract the texture damage features and modal features of each suspected damage point within the suspected damage area to obtain the non-geometric physical features; Based on the curvature abrupt change feature, the texture damage feature, and the modal feature, calculate the damage discrimination index for each suspected damage point:

[0083] in, The damage discrimination index is the damage discrimination index for suspected damage point i. It is characterized by abrupt changes in curvature. Features of texture damage Modal features, This is a feature representing the global maximum curvature abrupt change. This represents the largest texture damage feature globally. It is the global maximum modal feature. , , These are adaptive weights that match the structural type information of the bridge, and + + =1; If the damage discrimination index of the suspected damage point is greater than the preset false damage filtering threshold, then the suspected damage point is determined to be a real damage point, and the real structural damage region is composed of one or more adjacent real damage points.

[0084] Furthermore, in one embodiment, the determining module is further configured to: Obtain the structural damage features corresponding to the actual structural damage region; Based on the structural type information of the bridge, a corresponding adaptive discrimination threshold is matched from a preset bridge type feature database; The ratio of the structural damage feature to the adaptive discrimination threshold is calculated to obtain the damage grading index; The damage level of the actual structural damage area is determined based on the numerical range of the damage grading index.

[0085] Furthermore, in one embodiment, the determining module is further configured to: If the damage grading index is equal to zero, then the actual structural damage area is determined to be undamaged. If the damage grading index is greater than zero and less than or equal to the first grading threshold, then it is determined that the actual structural damage area has early damage. If the damage grading index is greater than the first grading threshold and less than or equal to the second grading threshold, then it is determined that the actual structural damage area has potential damage. If the damage grading index is greater than the second grading threshold and less than or equal to the third grading threshold, then it is determined that the actual structural damage area has moderate damage. If the damage grading index is greater than the third grading threshold, then it is determined that the actual structural damage area has severe damage.

[0086] Furthermore, in one embodiment, the device further includes a calibration module for: The structural damage features and the area of ​​the damaged region corresponding to the actual structural damage area are input into the bridge finite element model, and the structural damage features are converted into finite element simulation deformation values ​​based on the curvature-displacement mapping relationship.

[0087] in, The deformation value is the value obtained from the finite element simulation. The structural damage features, The area of ​​the damaged region. These are the mapping coefficients; Based on the coordinate differences of the same sampling point in the time-series point cloud model across multiple time periods, the measured deformation value is calculated. Based on the error between the finite element simulation deformation value and the measured deformation value, calculate the comprehensive damage parameters:

[0088] in, The damage comprehensive parameters are as follows: These are the inversion weighting coefficients. This refers to the actual structural damage area. The measured deformation value; The model parameters of the bridge finite element model are iteratively calibrated according to pre-built model correction coefficients until the error between the finite element simulation deformation value and the measured deformation value meets the preset convergence condition.

[0089] in, These are the correction coefficients for the model. This represents the error between the finite element simulation deformation value and the measured deformation value.

[0090] Furthermore, in one embodiment, the building module is also used for: The three-dimensional point cloud data collected collaboratively by the aerial and ground equipment is converted to a unified geographic coordinate system; Spatially align point cloud data collected by aerial equipment with point cloud data collected by ground equipment within the same time period; A filtering algorithm is used to remove noise from the point cloud, and temporal synchronization and spatial fusion are performed to obtain the temporal point cloud model.

[0091] The functions of each module in the bridge damage detection and quantification device based on air-to-ground data fusion correspond to the steps in the embodiment of the bridge damage detection and quantification method based on air-to-ground data fusion. Their functions and implementation processes will not be described in detail here.

[0092] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0093] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.

[0094] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.

[0095] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0096] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0097] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.

[0098] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A bridge damage detection and quantification method based on air-to-ground data fusion, characterized in that, include: Acquire 3D point cloud data of the bridge through collaborative collection of aerial and ground equipment at multiple time periods, and construct a time-series point cloud model of the entire bridge area at various time periods; Based on the temporal point cloud model, the geometric curvature features of each sampling point are calculated, and the geometric curvature features are enhanced based on the local curvature entropy to obtain enhanced curvature features; Temporal difference analysis is performed on the geometric curvature features over multiple time periods to obtain structural damage features, and suspected damage areas are determined based on the structural damage features. Based on the enhanced curvature characteristics of the suspected damaged areas, the actual structural damaged areas are identified. Based on the structural damage characteristics corresponding to the actual structural damage area, and combined with adaptive discrimination parameters that match the structural type information of the bridge, the damage level of the actual structural damage area is determined.

2. The bridge damage detection and quantification method based on air-to-ground data fusion as described in claim 1, characterized in that, Based on the aforementioned temporal point cloud model, the geometric curvature features of each sampling point are calculated, including: A local coordinate system is established for each sampling point in the time-series point cloud model; A quadratic surface fitting model is constructed based on the coordinates of each sampling point and the neighboring points within the sampling point's neighborhood. Based on the least squares objective function constrained by the normal vector, the surface fitting parameters of the quadratic surface fitting model are solved: in, The number of neighboring points, For neighborhood points The measured elevation value, The elevation value of the fitted surface is calculated based on the quadratic surface fitting model. These are the normal vector constraint weight coefficients. The normal vector of the fitted surface. The original normal vector of the sampling point. This is for Euclidean distance calculation; Substituting the surface fitting parameters into the constrained Gaussian curvature formula derived from the quadratic surface fitting model, the geometric curvature characteristics of the sampling points are calculated.

3. The bridge damage detection and quantification method based on air-to-ground data fusion as described in claim 1, characterized in that, The geometric curvature features are enhanced based on local curvature entropy to obtain enhanced curvature features, including: Based on the geometric curvature characteristics of the neighborhood points within the neighborhood of each sampling point, the neighborhood curvature probability density of the corresponding sampling point is determined. Calculate the local curvature entropy of the corresponding sampling point based on the neighborhood curvature probability density: in, Sampling points The local curvature entropy, The number of neighboring points, For the first in the neighborhood Geometric curvature characteristics of a neighborhood point Let ε be the neighborhood curvature probability density, where ε is a positive number; Based on the ratio of the local curvature entropy to the global maximum curvature entropy, and in conjunction with a preset enhancement adjustment coefficient, the adaptive enhancement weight for the corresponding sampling point is calculated: in, Sampling points Adaptive augmentation weights, and To enhance the adjustment coefficient, The global maximum curvature entropy; The geometric curvature features of the corresponding sampling points are weighted according to the adaptive enhancement weights to obtain the corresponding enhanced curvature features: in, Sampling points Enhanced curvature features, Sampling points The geometric curvature characteristics.

4. The bridge damage detection and quantification method based on air-to-ground data fusion as described in claim 1, characterized in that, Temporal difference analysis is performed on the geometric curvature features over multiple time periods to obtain structural damage features, and suspected damage areas are determined based on these structural damage features, including: Obtain the geometric curvature features of three time periods, which are denoted as the first period geometric curvature feature, the second period geometric curvature feature, and the third period geometric curvature feature, respectively. For each sampling point, the first-order temporal difference is calculated based on the geometric curvature characteristics of the first period and the geometric curvature characteristics of the second period; The second-order temporal difference is calculated based on the geometric curvature characteristics of the second period and the geometric curvature characteristics of the third period. The first-order temporal difference and the second-order temporal difference are weighted and summed to obtain a comprehensive temporal difference result, and the comprehensive temporal difference result is used as the structural damage feature. The structural damage characteristics of each sampling point are compared with the preset damage discrimination threshold; Sampling points whose structural damage features are greater than the damage discrimination threshold are identified as suspected damage points, and one or more adjacent suspected damage points constitute the suspected damage region.

5. The bridge damage detection and quantification method based on air-to-ground data fusion as described in claim 4, characterized in that, Based on the enhanced curvature characteristics of the suspected damaged area, the actual structural damaged area is identified, including: Based on the enhanced curvature characteristics of each suspected damage point within the suspected damage area, and the mean of the enhanced curvature characteristics of neighboring points within the neighborhood of the suspected damage point, the curvature abrupt change characteristics of each suspected damage point are calculated. Extract the texture damage features and modal features of each suspected damage point within the suspected damage area to obtain non-geometric physical features; Based on the curvature abrupt change feature, the texture damage feature, and the modal feature, calculate the damage discrimination index for each suspected damage point: in, The damage discrimination index is the damage discrimination index for suspected damage point i. It is characterized by abrupt changes in curvature. Features of texture damage Modal features, This is a feature representing the global maximum curvature abrupt change. This represents the largest texture damage feature globally. It is the global maximum modal feature. , , These are adaptive weights that match the structural type information of the bridge, and + + =1; If the damage discrimination index of the suspected damage point is greater than the preset false damage filtering threshold, then the suspected damage point is determined to be a real damage point, and the real structural damage region is composed of one or more adjacent real damage points.

6. The bridge damage detection and quantification method based on air-to-ground data fusion as described in claim 1, characterized in that, Based on the structural damage characteristics corresponding to the actual structural damage area, and combined with adaptive discrimination parameters that match the structural type information of the bridge, the damage level of the actual structural damage area is determined, including: Obtain the structural damage features corresponding to the actual structural damage region; Based on the structural type information of the bridge, a corresponding adaptive discrimination threshold is matched from a preset bridge type feature database; The ratio of the structural damage feature to the adaptive discrimination threshold is calculated to obtain the damage grading index; The damage level of the actual structural damage area is determined based on the numerical range of the damage grading index.

7. The bridge damage detection and quantification method based on air-to-ground data fusion as described in claim 6, characterized in that, Based on the numerical range of the damage grading index, the damage level of the actual structural damage region is determined, including: If the damage grading index is equal to zero, then the actual structural damage area is determined to be undamaged. If the damage grading index is greater than zero and less than or equal to the first grading threshold, then it is determined that the actual structural damage area has early damage. If the damage grading index is greater than the first grading threshold and less than or equal to the second grading threshold, then it is determined that the actual structural damage area has potential damage. If the damage grading index is greater than the second grading threshold and less than or equal to the third grading threshold, then it is determined that the actual structural damage area has moderate damage. If the damage grading index is greater than the third grading threshold, then it is determined that the actual structural damage area has severe damage.

8. The bridge damage detection quantification method based on air-to-ground data fusion as described in claim 1, characterized in that, The method also includes: The structural damage features and the area of ​​the damaged region corresponding to the actual structural damage area are input into the bridge finite element model, and the structural damage features are converted into finite element simulation deformation values ​​based on the curvature-displacement mapping relationship. in, The deformation value is the value obtained from the finite element simulation. The structural damage features, The area of ​​the damaged region. These are the mapping coefficients; Based on the coordinate differences of the same sampling point in the time-series point cloud model across multiple time periods, the measured deformation value is calculated. Based on the error between the finite element simulation deformation value and the measured deformation value, calculate the comprehensive damage parameters: in, The damage comprehensive parameters are as follows: These are the inversion weighting coefficients. This refers to the actual structural damage area. The measured deformation value; The model parameters of the bridge finite element model are iteratively calibrated according to pre-built model correction coefficients until the error between the finite element simulation deformation value and the measured deformation value meets the preset convergence condition. in, These are the correction coefficients for the model. This represents the error between the finite element simulation deformation value and the measured deformation value.

9. The bridge damage detection and quantification method based on air-to-ground data fusion as described in claim 1, characterized in that, Acquire 3D point cloud data of the bridge collected collaboratively by aerial and ground equipment over multiple time periods, and construct a temporal point cloud model of the entire bridge area for each time period, including: The three-dimensional point cloud data collected collaboratively by the aerial and ground equipment is converted to a unified geographic coordinate system; Spatially align point cloud data collected by aerial equipment with point cloud data collected by ground equipment within the same time period; A filtering algorithm is used to remove noise from the point cloud, and temporal synchronization and spatial fusion are performed to obtain the temporal point cloud model.

10. A bridge damage detection and quantification device based on air-to-ground data fusion, characterized in that, include: The module is used to acquire 3D point cloud data of the bridge collected collaboratively by aerial and ground equipment at multiple time periods, and to construct a time-series point cloud model of the entire bridge at various time periods. The calculation module is used to calculate the geometric curvature features of each sampling point based on the time-series point cloud model, and to enhance the geometric curvature features based on the local curvature entropy to obtain enhanced curvature features. The analysis module is used to perform temporal difference analysis on the geometric curvature features over multiple time periods to obtain structural damage features, and to determine suspected damage areas based on the structural damage features. The identification module is used to identify the actual structural damage area based on the enhanced curvature characteristics of the suspected damage area; The determination module is used to determine the damage level of the actual structural damage area based on the structural damage characteristics corresponding to the actual structural damage area and in combination with adaptive discrimination parameters that match the structural type information of the bridge.

Images