Bridge full-field deformation monitoring method based on bidirectional high-dimensional vector Euclidean distance

Abstract

The present application relates to the technical field of bridge construction, and particularly relates to a bridge full-field deformation monitoring method based on bidirectional high-dimensional vector Euclidean distance, and specifically comprises the following steps: S1, establishment of a structure monitoring database; S2, matching of a topological model; S3, calculation of a displacement distribution map, in the present application, image sensors are used to collect image sequences of a structure monitoring cycle, a topological model of a target structure is recovered by processing and analyzing digital images, a plurality of virtual displacement monitoring points of a structure monitoring surface are formed, a real spatial scale relationship of the topological model of the structure of each monitoring cycle and a matching relationship of the virtual displacement monitoring points are recovered by using homonymous points and a spatial scale conversion relationship, a displacement dynamic distribution of the structure monitoring area is obtained by calculating the offset of the virtual displacement monitoring points of each monitoring cycle, and economic, convenient, safe and reliable deformation monitoring of the structure is realized.

Description

Technical Field

[0001] This invention relates to the field of bridge construction technology, specifically to a method for monitoring the full-field deformation of bridges based on bidirectional high-dimensional vector Euclidean distance. Background Technology

[0002] To accurately understand the safety status of bridge structures during construction and operation, and to ensure the safety of people's lives and property, it is necessary to monitor the deformation of the structure under load. Currently, the monitoring methods commonly used in engineering are mainly of two types. The first is manual inspection using equipment such as total stations and theodolites. This method has a low overall cost, but it is too reliant on manual labor, has low efficiency, and the data is relatively simple and difficult to accurately reflect the safety status of the structure. The second is a long-term health monitoring system established by installing various sensors on the bridge structure. This method provides more intuitive monitoring data and can basically meet the needs of real-time display on the monitoring platform. However, such systems are usually expensive and cannot be widely promoted. The sensors are generally embedded parts, and maintenance personnel cannot calibrate them regularly according to maintenance requirements, resulting in a gradual increase in noise data and a gradual decrease in reliability over time. Moreover, the lifespan is shorter than the bridge's design life, and the cost of later maintenance and replacement is high. Therefore, this invention proposes a bridge full-field deformation monitoring method based on bidirectional high-dimensional vector Euclidean distance. Summary of the Invention

[0003] In view of the above-mentioned shortcomings of the existing technology, the first objective of the present invention is to provide a method for monitoring the full-field deformation of bridges based on bidirectional high-dimensional vector Euclidean distance, thereby solving the problems in the background technology.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for monitoring the full-field deformation of bridges based on bidirectional high-dimensional vector Euclidean distance includes the following steps:

[0006] S1. Establishment of a structural monitoring database;

[0007] S2, Registration of the topology model;

[0008] S3. Calculation of displacement distribution diagram.

[0009] Preferably, the specific steps in step S1 are as follows:

[0010] 3. The specific steps in step S1 are as follows:

[0011] S1.1 Define the initial state of the structure. When the structure has no external load or the load influence is negligible, the structural state at this time is defined as the project baseline state. The working condition of the structure at this time is the baseline working condition T0. Using mobile devices such as drones, the bridge structure is continuously photographed by mobile image sensors on the mobile devices to establish a set of structural sequence images under working condition T0.

[0012] S1.2, Flight distance parameters between the camera and the structure plane: The camera plane pixel value is α × β, f is the camera focal length, d is the distance between the camera and the structure plane, the camera imaging sensor size is μ × φ, and ψ is the highest structure size represented by a single pixel. That is, the distance between the camera and the structure should satisfy the following formula:

[0013] S1.3 Camera flight speed parameter requirements: In monitoring (inspection) projects, the minimum overlap between two adjacent images is set to ε, the camera shutter interval is t, and the moving speed v of the UAV and other equipment along the structure's shooting surface should satisfy the following formula:

[0014] S1.4. The project baseline state can be obtained from steps S1.1, S1.2 and S1.3, that is, under the initial working condition T0, the structural series has a highly overlapping digital image set {image0, image1, ...}.

[0015] S1.5. Use camera intrinsic parameters to correct all digital images under the current working condition, and restore the topological model M0 of the structural scene under the initial working condition T0 through 3D image reconstruction.

[0016] S1.6. Segment the structural topology model under the current working condition and remove irrelevant data from the monitoring (inspection) scenario;

[0017] S1.7, Define the project test condition T i According to the requirements of the monitoring (inspection) task, a series of structural working conditions {T1, T2, ...} are set, and the load environment of the structure under each set working condition is marked. Steps S1.1-S1.6 are repeated to reconstruct the corresponding structural topology model {M1, M2, ...} under each test working condition, forming a structural monitoring database.

[0018] The above technical solution utilizes mobile devices such as drones, and employs image acquisition devices mounted on these devices to continuously capture images of the bridge structure, establishing a set of structural sequence images under T0 working conditions. This flexible data acquisition method can improve the efficiency and comprehensiveness of data acquisition.

[0019] Preferably, step S2 specifically comprises:

[0020] Step S2 specifically involves:

[0021] S2.1, Synchronous project datum T0 and arbitrary test condition T under the same world coordinate system i The following topological model;

[0022] S2.2, Set the project baseline T0 and test condition T i The relatively stable region in the lower structural scene serves as the project registration reference, marked as T0 and T1. i The topology model is registered with a reference region. Using the reference region in the T0 model as a reference, T0 and T... i Distribution of 3D feature points within the reference region of the topological model:

[0023] S2.3. Based on the perspective relationship between the image and the 3D reality, create T0 and T1 respectively. i The indices of the three-dimensional feature points within the central reference region and the projected image are associated as follows:

[0024]

[0025] S2.4. Based on the perspective relationship between the three-dimensional feature points and the projected feature points on the imaging plane, calculate T0 and T1. i The pixel coordinates of each 3D feature point in the reference region on the projected image are indexed and associated, as follows:

[0026]

[0027] S2.5 Calculate T0 and T respectively. i The SIFT description vectors of the projection points of each 3D feature point in the reference region across all associated images are used to obtain the mean intrinsic SIFT description vector of each 3D feature point.

[0028] S2.6 Calculate T0 and T i The Euclidean distance between the mean intrinsic SIFT descriptor vectors of 3D feature points in the reference region, if there exists a mean intrinsic descriptor vector of point A in T0 that is equal to T... i The distance between the mean eigenvectors of midpoint A' and T is... i The shortest distance between the mean intrinsic descriptor vectors of all feature points in T0 is also the shortest distance between A' and the mean intrinsic descriptor vectors of all feature points in T0, which is why it is determined to be... For pairs of points with the same name, we get T0 and T. i Set of pairs of points with the same name

[0029] S2.7 Calculate the sets respectively Internal project baseline T0, test condition T i The average coordinates of the three-dimensional feature points along the x, y, and z axes are used to obtain the working conditions T0 and T. i Centroid coordinates of the marked region Using the topology model under T0 as a reference, move T i Working condition topology model, the direction of movement is The distance moved is To P T0 P Ti Overlapping as P T (x0,y0,z0);

[0030] S2.8, Check T0 and T after the movement. i Are the orientations of the topological models basically consistent? If not, maintain the centroid coordinates P. T Without changing T, adjust i The topological model has [x', y', z']. T =R·[x, y, z] T R is the rotation matrix, [x, y, z] T Let [x', y', z'] be the initial coordinates of the point in space. T To adjust the coordinates of the spatial points until T0, T i The orientation of the topological models remains largely consistent.

[0031] S2.9 Optimize the orientation adjustment results of step S2.8, using the topological model under T0 as a reference, centroid P T Optimize T as the center of rotation i The rotation matrix R parameters of the working condition topology model, at this time, for each monitoring (inspection) working condition T0, T i Centroid P of the marked region T The vectors to the points with the same name in set M are: The sum of the angles between the registration point and the centroid ξ is the total angle limit; complete T0 and T... i Orientation calibration of the topology model;

[0032] S2.10, Check T0 and T after movement. i Are the dimensions of the topological models basically consistent? If not, maintain the centroid coordinates P. T Without changing P T Adjust T around the center i The topology model and adjustment formulas are based on the following: λ is the scaling factor;

[0033] S2.11. Optimize the orientation adjustment results of step S2.10, using the topology model under T0 as a reference, and optimize T. i The scaling factor λ of the working condition topology model is used to calculate the distance between registration point pairs until... As the limit of the sum of distances, T0 and T i Topology model registration is complete.

[0034] The above technical solution is adopted because digital images store the relative size relationships of structures. When reconstructing the topological model of a structure using images, there is a lack of absolute size constraints, which leads to certain differences in the spatial orientation and scale of the topological model reconstructed from images under different monitoring conditions. Therefore, the above technical solution is adopted to process and analyze digital images to extract the feature distribution of the target structure and form several virtual displacement monitoring points on the structural surface. The detection surface constructed by several virtual displacement monitoring points will make the data detection more accurate.

[0035] Preferably, step S3 specifically comprises:

[0036] S3.1, Use step S2 to compare the project baseline T0 with the test condition T. i The topology model is accurately matched to restore the true scale relationship of the structural mechanics topology model under each monitoring (inspection) condition;

[0037] S3.2, Segmentation Project Baseline T0 and Test Condition T i Topological model structural elements are extracted, and topological model data of the deformation analysis area of ​​the main monitoring structure are extracted.

[0038] S3.3. Based on the mapping relationship between the image and the 3D reality, create T0 and T1 respectively. i Each 3D feature point in the deformation analysis region is associated with an index in the related image, as follows:

[0039]

[0040]

[0041] S3.4. Based on the mapping relationship between spatial points and projection points on the imaging plane, calculate T0 and T1. i The projection coordinates of each three-dimensional feature point within the deformation area of ​​the main structure onto the associated image are used to establish an index association.

[0042]

[0043]

[0044] S3.5 Calculate T0 and T respectively. iThe SIFT description vectors of the projection points of each 3D feature point in the deformation region across all associated images are obtained by calculating the mean intrinsic SIFT description vectors of each 3D feature point in the deformation region as follows:

[0045] S3.6 Calculate T0 and T i The Euclidean distance between the mean intrinsic SIFT descriptor vectors of three-dimensional feature points within the deformation region, if there exists a mean intrinsic descriptor vector of point A in T0 that is equal to T... i The distance between the mean eigenvectors of midpoint A' and T is... i The shortest distance between the mean intrinsic descriptor vectors of all feature points in T0 is also the shortest distance between A' and the mean intrinsic descriptor vectors of all feature points in T0, which is why it is determined to be... For virtual displacement sensor point pairs, obtain T0 and T i set of point pairs

[0046] S3.7 Calculate the ground offset vector σ of the virtual sensor points in set N and perform noise reduction processing.

[0047] S3.8 Calculate the true displacement vector of each virtual sensor: v = ξXσ, ξ is the deformation coefficient, v is the displacement vector, and obtain the displacement vector distribution (including magnitude and direction) of the entire field of the deformation region.

[0048] The above technical solution utilizes the corresponding points and spatial scale transformation relationship to restore the structural topology model and feature point matching relationship for each monitoring cycle, enabling the full-field displacement calculation of the structure's texture under natural conditions. This can greatly reduce the amount of on-site construction work and the difficulty of the project for engineers.

[0049] Preferably, in step S2.10, the model size is adjusted until the sizes of the T0 and T0 topology models are basically consistent.

[0050] Beneficial effects

[0051] Compared with known public technologies, the technical solution provided by this invention has the following beneficial effects:

[0052] In this invention, a sequence of images of the structure is acquired by tilt-shifting, thereby realizing a full-field deformation acquisition technology for the structure. Currently, the existing methods in the world can only achieve deformation monitoring of fixed-axis images, and can only monitor deflection deformation through the overlap of contour lines. Displacement vectors can only be obtained by installing sensors at a small number of points or by DIC (Displacement Injection) methods. Secondly, the registration method of the topology model differs from the point cloud model acquired by the 3D laser scanner. Although the topology model restored from the image is also a point cloud model, the point cloud model obtained by the 3D laser scanner is of absolute size, and the coordinates of each point obtained by the 3D laser scanner are absolute coordinates with consistent dimensions. However, the topology model obtained from the image restoration is a relative size model, and it has a linear relationship with the actual size of the structure. It needs to be multiplied by a scaling factor to obtain the actual size of the structure, so this data processing is relatively complicated. The topology model registration method in this invention ensures the accuracy of data processing and the acquisition of full-field displacement. Currently, to obtain the full-field displacement of the structure, only DIC technology can be used. This technology has significant drawbacks. One is that it requires speckle spraying, and the other is that it is fixed-axis, and the monitoring area is relatively narrow. It is generally used for deformation monitoring of mechanical parts in factories, and it is not suitable for large structures such as civil engineering. In this invention, the topology model and feature point matching relationship of the structure in each monitoring period are restored by using corresponding points and spatial scale transformation relationships. This allows the full-field displacement calculation of the structure under natural conditions to be performed, which can greatly reduce the amount of on-site construction work and the difficulty of the project. Attached Figure Description

[0053] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0055] The present invention will be further described below with reference to embodiments.

[0056] Example 1

[0057] like Figure 1 As shown, a method for monitoring the full-field deformation of a bridge based on bidirectional high-dimensional vector Euclidean distance specifically includes the following steps:

[0058] S1. Establishment of a structural monitoring database;

[0059] S2, Registration of the topology model;

[0060] S3. Calculation of displacement distribution diagram.

[0061] The specific steps in step S1 are as follows:

[0062] S1.1 Define the initial state of the structure. When the structure has no external load or the load influence is negligible, the structural state at this time is defined as the project baseline state. The working condition of the structure at this time is the baseline working condition T0. Using mobile devices such as drones, the bridge structure is continuously photographed by mobile image sensors on the mobile devices to establish a set of structural sequence images under working condition T0.

[0063] S1.2, Flight distance parameters between the camera and the structure plane: The camera plane pixel value is α × β, f is the camera focal length, d is the distance between the camera and the structure plane, the camera imaging sensor size is μ × φ, and ψ is the highest structure size represented by a single pixel. That is, the distance between the camera and the structure should satisfy the following formula:

[0064] S1.3 Camera flight speed parameter requirements: In monitoring (inspection) projects, the minimum overlap between two adjacent images is set to ε, the camera shutter interval is t, and the moving speed v of the UAV and other equipment along the structure's shooting surface should satisfy the following formula:

[0065] S1.4. The project baseline state can be obtained from steps S1.1, S1.2 and S1.3, that is, under the initial working condition T0, the structural series has a highly overlapping digital image set {image0, image1, ...}.

[0066] S1.5. Use camera intrinsic parameters to correct all digital images under the current working condition, and restore the topological model M0 of the structural scene under the initial working condition T0 through 3D image reconstruction.

[0067] S1.6. Segment the structural topology model under the current working condition and remove irrelevant data from the monitoring (inspection) scenario;

[0068] S1.7, Define the project test condition T i According to the monitoring (inspection) task requirements, a series of structural working conditions {T1, T2, ...} are set, and the load environment of the structure under each set working condition is marked. Steps S1.1-S1.6 are repeated to reconstruct the corresponding structural topology model {M1, M2, ...} under each test working condition, forming a structural monitoring database. Using mobile devices such as drones, the bridge structure is continuously photographed by mobile imaging devices, and a set of structural sequence images under working condition T0 is established. This flexible data acquisition method can improve the efficiency and comprehensiveness of data acquisition.

[0069] Specifically, step S2 is as follows:

[0070] S2.1, Synchronous project datum T0 and arbitrary test condition T under the same world coordinate system i The following topological model;

[0071] S2.2, Set the project baseline T0 and test condition T i The relatively stable region in the lower structural scene serves as the project registration reference, marked as T0 and T1. i The topology model is registered with a reference region. Using the reference region in the T0 model as a reference, T0 and T... i Distribution of 3D feature points within the reference region of the topological model:

[0072] S2.3. Based on the perspective relationship between the image and the 3D reality, create T0 and T1 respectively. i The indices of the three-dimensional feature points within the central reference region and the projected image are associated as follows:

[0073]

[0074] S2.4. Based on the perspective relationship between the three-dimensional feature points and the projected feature points on the imaging plane, calculate T0 and T1. i The pixel coordinates of each 3D feature point in the reference region on the projected image are indexed and associated, as follows:

[0075]

[0076] S2.5 Calculate T0 and T respectively. i The SIFT description vectors of the projection points of each 3D feature point in the reference region across all associated images are used to obtain the mean intrinsic SIFT description vector of each 3D feature point.

[0077] S2.6 Calculate T0 and T i The Euclidean distance between the mean intrinsic SIFT descriptor vectors of 3D feature points in the reference region, if there exists a mean intrinsic descriptor vector of point A in T0 that is equal to T... i The distance between the mean eigenvectors of midpoint A' and T is... i The shortest distance between the mean intrinsic descriptor vectors of all feature points in T0 is also the shortest distance between A' and the mean intrinsic descriptor vectors of all feature points in T0, which is why it is determined to be... For pairs of points with the same name, we get T0 and T. i Set of pairs of points with the same name

[0078] S2.7 Calculate the sets respectively Internal project baseline T0, test condition T i The average coordinates of the three-dimensional feature points along the x, y, and z axes are used to obtain the working conditions T0 and T. i Centroid coordinates of the marked region Using the topology model under T0 as a reference, move T i Working condition topology model, the direction of movement is The distance moved is To P T0 P Ti Overlapping as P T (x0,y0,z0);

[0079] S2.8, Check T0 and T after the movement. i Are the orientations of the topological models basically consistent? If not, maintain the centroid coordinates P. T Without changing T, adjust i The topological model has [x', y', z']. T =R·[x, y, z] T R is the rotation matrix, [x, y, z] T Let [x', y', z'] be the initial coordinates of the point in space. T To adjust the coordinates of the spatial points until T0, T i The orientation of the topological models remains largely consistent.

[0080] S2.9 Optimize the orientation adjustment results of step S2.8, using the topological model under T0 as a reference, centroid P T Optimize T as the center of rotation i The rotation matrix R parameters of the working condition topology model, at this time, for each monitoring (inspection) working condition T0, T i Centroid P of the marked region T The vectors to the points with the same name in set M are: The sum of the angles between the registration point and the centroid ξ is the total angle limit; complete T0 and T... i Orientation calibration of the topology model;

[0081] S2.10, Check T0 and T after movement. i Are the dimensions of the topological models basically consistent? If not, maintain the centroid coordinates P. T Without changing P T Adjust T around the center i The topology model and adjustment formulas are based on the following: λ is the scaling factor;

[0082] S2.11. Optimize the orientation adjustment results of step S2.10, using the topology model under T0 as a reference, and optimize T. i The scaling factor λ of the working condition topology model is used to calculate the distance between registration point pairs until... As the limit of the sum of distances, T0 and T i Once the topology model registration is complete, the digital image stores the relative size relationships of the structure. However, when reconstructing the topology model from the image, the lack of absolute size constraints leads to certain differences in spatial orientation and scale between the reconstructed topology models from images under different monitoring conditions. Therefore, the above-mentioned technical solution is adopted to process and analyze the digital image to extract the feature distribution of the target structure and form several virtual displacement monitoring points on the structural surface. The detection surface constructed by these virtual displacement monitoring points will make the data detection more accurate.

[0083] Specifically, step S3 is as follows:

[0084] S3.1. Use step S2 to accurately match the topology model of the project baseline T0 with the test condition Ti, and restore the true scale relationship of the structural mechanics topology model under each monitoring (inspection) condition.

[0085] S3.2, Segmentation Project Baseline T0 and Test Condition T i Topological model structural elements are extracted, and topological model data of the deformation analysis area of ​​the main monitoring structure are extracted.

[0086] S3.3. Based on the mapping relationship between the image and the 3D reality, create T0 and T1 respectively. i Each 3D feature point in the deformation analysis region is associated with an index in the related image, as follows:

[0087]

[0088]

[0089] S3.4. Based on the mapping relationship between spatial points and projection points on the imaging plane, calculate T0 and T1. i The projection coordinates of each three-dimensional feature point within the deformation area of ​​the main structure onto the associated image are used to establish an index association.

[0090]

[0091]

[0092] S3.5 Calculate T0 and T respectively. iThe SIFT description vectors of the projection points of each 3D feature point in the deformation region across all associated images are obtained by calculating the mean intrinsic SIFT description vectors of each 3D feature point in the deformation region as follows:

[0093] S3.6 Calculate T0 and T i The Euclidean distance between the mean intrinsic SIFT descriptor vectors of three-dimensional feature points within the deformation region, if there exists a mean intrinsic descriptor vector of point A in T0 that is equal to T... i The distance between the mean eigenvectors of midpoint A' and T is... i The shortest distance between the mean intrinsic descriptor vectors of all feature points in T0 is also the shortest distance between A' and the mean intrinsic descriptor vectors of all feature points in T0, which is why it is determined to be... For virtual displacement sensor point pairs, obtain T0 and T i set of point pairs

[0094] S3.7 Calculate the ground offset vector σ of the virtual sensor points in set N and perform noise reduction processing.

[0095] S3.8 Calculate the true displacement vector of each virtual sensor: v = ξXσ, ξ is the deformation coefficient, v is the displacement vector, and obtain the displacement vector distribution (including magnitude and direction) of the entire field of the deformation region. By using the corresponding points and spatial scale transformation relationship, the structural topology model and feature point matching relationship of each monitoring cycle are restored, so that the texture of the structure itself under natural conditions can be used to calculate the displacement of the entire field of the structure, which can greatly reduce the amount of on-site construction work and the difficulty of the work for engineers.

[0096] In step S1.3, the minimum overlap ε between two adjacent images is greater than 80%.

[0097] In step S2.10, the model size is adjusted until the sizes of the T0 and T0 topology models are basically consistent.

[0098] Comparative Example 1

[0099] This comparative example is largely the same as the method of the provided embodiment 1, the main difference being that: no mobile device such as a drone is used in step S1;

[0100] Comparative Example 2

[0101] This comparative example is largely the same as the method of the provided Example 1, the main difference being that no algorithm is used in step S2.

[0102] Comparative Example 3

[0103] This comparative example is largely the same as the method in the provided Example 1, the main difference being that no algorithm is used in step S3.

[0104] Performance testing

[0105] According to the "Technical Specification for Highway Bridge Construction JTG F50", the cost and accuracy of a bridge full-field deformation monitoring method based on bidirectional high-dimensional vector Euclidean distance provided in Example 1 and Comparative Examples 1-3 are compared:

[0106] cost Accuracy Example 1 10％ 99.9％ Comparative Example 1 45％ 87％ Comparative Example 2 10％ 32％ Comparative Example 3 10％ 27％

[0107] In step S1, mobile devices such as drones are used to capture continuous images of the bridge structure using image acquisition devices mounted on the mobile devices, thus establishing a set of structural sequence images under the T0 working condition. This flexible data acquisition method can improve the efficiency and comprehensiveness of data acquisition.

[0108] In step S2, the algorithm uses digital images to store the relative size relationships of structures. When reconstructing the topological model of a structure using images, there is a lack of absolute size constraints, which leads to certain differences in spatial orientation and scale of the topological models reconstructed from images under different monitoring conditions. Therefore, the above technical solution is adopted to process and analyze digital images to extract the feature distribution of the target structure and form several virtual displacement monitoring points on the structural surface. The detection surface constructed by several virtual displacement monitoring points will make the data detection more accurate.

[0109] In step S3, the algorithm uses corresponding points and spatial scale transformation relationships to restore the structural topology model and feature point matching relationships for each monitoring period, enabling the full-field displacement calculation of the structure's texture under natural conditions. This can greatly reduce the amount of on-site construction work and the difficulty of the project for engineers.

[0110] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring the full-field deformation of a bridge based on bidirectional high-dimensional vector Euclidean distance, characterized in that, Specifically, the following steps are included: S1. Establishment of a structural monitoring database; S2. Registration of the topology model: Step S2 specifically involves: S2.1, Synchronize T0 working condition and any test working condition T under the same world coordinate system i Lower topology model; S2.2, Set T0 working condition and test working condition T i The relatively stable region in the lower structural scene serves as the project registration reference, marked as T0 and T1. i The topology model is registered with a reference region. Using the reference region in the T0 model as a reference, T0 and T... i Distribution of 3D feature points within the reference region of the topological model: S2.

3. Based on the perspective relationship between the image and the 3D reality, create T0 and T1 respectively. i The indices of the three-dimensional feature points within the central reference region and the projected image are associated as follows: S2.4, according to the perspective relationship between the three-dimensional feature points and the projection feature points on the imaging plane, calculate T0, T i The pixel coordinates of each three-dimensional feature point in the reference region on the projection image are indexed and associated, specifically as follows: S2.5, calculate T0, T i The SIFT description vector of the projection point of each three-dimensional feature point in the reference region in all related images is obtained, and the mean intrinsic SIFT description vector of each three-dimensional feature point is calculated as follows: S2.6 Calculate T0 and T i The Euclidean distance between the mean intrinsic SIFT descriptor vectors of 3D feature points in the reference region, if there exists a mean intrinsic descriptor vector of point A in T0 that is equal to T... i The distance between the mean eigenvectors of midpoint A' and T is... i The shortest distance between the mean intrinsic descriptor vectors of all feature points in T0 is also the shortest distance between A' and the mean intrinsic descriptor vectors of all feature points in T0, which is why it is determined to be... For pairs of points with the same name, we get T0 and T. i Set of pairs of points with the same name S2.7 Calculate the sets respectively Internal T0 working condition, test working condition T i The average coordinates of the three-dimensional feature points along the x, y, and z axes are used to obtain the working conditions T0 and T. i Centroid coordinates of the marked region Using the topology model under T0 as a reference, move T i Working condition topology model, the direction of movement is The distance moved is To P T0 P Ti Overlapping as P T (x0,y0,z0); S2.8, check T0, T after moving i topological model orientation is basically consistent, if not consistent, keep the centroid coordinates P T unchanged, adjust T i topological model, [x', y', z'] T = R·[x, y, z] T , R is a rotation matrix, [x, y, z] T is the initial coordinates of the space point, [x', y', z'] T is the adjusted coordinates of the space point, until T0, T i topological model orientation is basically consistent; S2.9 Optimize the orientation adjustment results of step S2.8, using the topological model under T0 as a reference, centroid P T Optimize T as the center of rotation i The rotation matrix R parameter of the topology model under the working condition is given, where the basic working condition is T0, T i Centroid P of the marked region T The vectors to the points with the same name in set M are: The sum of the angles between the registration point and the centroid ξ is the total angle limit; complete T0 and T... i Orientation calibration of the topology model; S2.10, Check T0 and T after the movement. i Are the dimensions of the topological models basically consistent? If not, maintain the centroid coordinates P. T Without changing P T Adjust T around the center i The topology model and adjustment formulas are based on the following: λ is the scaling factor, and in step S2.10, the model size is adjusted until the sizes of T0 and T0 topology models are basically consistent; S2.

11. Optimize the orientation adjustment results of step S2.10, using the topology model under T0 as a reference, and optimize T. i The scaling factor λ of the working condition topology model is used to calculate the distance between registration point pairs until... As the limit of the sum of distances, T0 and T i Topology model registration is complete; S3. Calculation of displacement distribution diagram: Step S3 specifically involves: S3.1, Using step S2, perform T0 condition and test condition T i The topology model is accurately matched to restore the true scale relationship of the structural mechanics topology model under each monitoring (inspection) condition; S3.2, Segmenting T0 working condition and test condition T i Topological model structural elements are extracted, and topological model data of the deformation analysis area of ​​the main monitoring structure are extracted. S3.

3. Based on the mapping relationship between the image and the 3D reality, create T0 and T1 respectively. i Each 3D feature point in the deformation analysis region is associated with an index in the related image, as follows: S3.

4. Based on the mapping relationship between spatial points and projection points on the imaging plane, calculate T0 and T1. i The projection coordinates of each three-dimensional feature point within the deformation area of ​​the main structure onto the associated image are used to establish an index association. S3.5 Calculate T0 and T respectively. i The SIFT description vectors of the projection points of each 3D feature point in the deformation region across all associated images are obtained by calculating the mean intrinsic SIFT description vectors of each 3D feature point in the deformation region as follows: S3.6 Calculate T0 and T i The Euclidean distance between the mean intrinsic SIFT descriptor vectors of three-dimensional feature points within the deformation region, if there exists a mean intrinsic descriptor vector of point A in T0 that is equal to T... i The distance between the mean eigenvectors of midpoint A' and T is... i The shortest distance between the mean intrinsic descriptor vectors of all feature points in T0 is also the shortest distance between A' and the mean intrinsic descriptor vectors of all feature points in T0, which is why it is determined to be... For virtual displacement sensor point pairs, obtain T0 and T i set of point pairs S3.7 Calculate the ground offset vector σ of the virtual sensor points in set N and perform noise reduction processing. S3.8 Calculate the true displacement vector of each virtual sensor: v = ξXσ, ξ is the deformation coefficient, v is the displacement vector, and obtain the displacement vector distribution of the entire field in the deformation region.

2. The method for monitoring full-field deformation of a bridge based on bidirectional high-dimensional vector Euclidean distance according to claim 1, characterized in that: The specific steps in step S1 are as follows: S1.1 Define the initial state of the structure. When the structure has no external load or the load influence is negligible, the structural state at this time is defined as the project baseline state. The working condition of the structure at this time is the baseline working condition T0. Using a drone mobile device, the image sensor device on the mobile device is used to take continuous mobile images of the bridge structure and establish a set of structural sequence images under working condition T0. S1.2, Flight distance parameters between the camera and the structure plane: The camera plane pixel value is α × β, f is the camera focal length, d is the distance between the camera and the structure plane, the camera imaging sensor size is μ × φ, and ψ is the highest structure size represented by a single pixel. That is, the distance between the camera and the structure should satisfy the following formula: S1.3 Camera flight speed parameter requirements: In the monitoring (inspection) project, the minimum overlap between two adjacent images is set to ε, the camera shutter interval is t, and the speed v of the UAV equipment moving along the structure's shooting surface should satisfy the following formula: S1.

4. The project baseline state can be obtained from steps S1.1, S1.2 and S1.3, that is, under the initial working condition T0, the structural series has a highly overlapping digital image set {image0, image1, ...}. S1.

5. Use camera intrinsic parameters to correct all digital images under the current working condition, and restore the topological model M0 of the structural scene under the initial working condition T0 through 3D image reconstruction. S1.

6. Segment the structural topology model under the current working condition and remove irrelevant data from the monitoring (inspection) scenario; S1.7, Define the project test condition T i According to the requirements of the monitoring (inspection) task, a series of structural working conditions {T1, T2, ...} are set, and the load environment of the structure under each set working condition is marked. Steps S1.1-S1.6 are repeated to reconstruct the corresponding structural topology models M1, M2, ...} under each test working condition, forming a structural monitoring database.

Images