A method for identifying the service line shape of bridges based on multiple three-dimensional laser scans

Abstract

The application discloses a bridge service linear identification method based on multiple three-dimensional laser scanning, comprising the following steps: defining and acquiring the bridge girder linear; acquiring the three-dimensional point cloud data of the bridge under the condition of non-interrupted traffic in different service periods; converting the three-dimensional point cloud data of the bridge obtained through multiple scanning under the same coordinate system; obtaining the bridge service linear with statistical significance and meeting the preset confidence condition through regression prediction; comparing the bridge service linear in different periods to identify the spatial form change of the bridge girder, evaluate the bridge structure state in the bridge service period, and track the spatial form change of the bridge in the bridge service period. The application realizes the accurate identification of the bridge service linear, solves the problem of insufficient spatial information acquisition in the traditional measurement method, and realizes the bridge service linear identification with high confidence in view of the complex vibration noise problem of the point cloud array caused by the bridge vibration, thereby improving the accuracy of the bridge service state control.

Description

Technical Field

[0001] This invention belongs to the field of bridge engineering and bridge service life appearance inspection technology, specifically involving a bridge service life alignment recognition method based on multiple three-dimensional laser scanning. Background Technology

[0002] Bridge service alignment is a crucial evaluation indicator for the visual inspection of bridges during their service life. It reflects the bridge's mechanical response to different load effects under its current condition. Bridges with alignments significantly deviating from the original design alignment may already be in poor operational condition. Therefore, identifying the alignment of the main girder is a key aspect of bridge visual inspection. Conventional inspection methods such as total stations, levels, and GPS are inefficient and can only acquire a limited number of single-point information points, failing to provide a complete understanding of the bridge's spatial information. Three-dimensional point cloud data is a massive collection of points containing complete three-dimensional coordinate information of the object's surface. However, during actual operation, the main girder of a bridge is subjected to complex vibrations under loads, and this vibration effect intensifies with the increase of the bridge span. This results in complex vibration noise in the acquired point cloud data. How to accurately identify the bridge service alignment reflecting the bridge's true operational condition from noisy point cloud data is a pressing issue that needs to be addressed. Summary of the Invention

[0003] Technical problem solved: This invention proposes a bridge service line shape identification method based on multiple three-dimensional laser scanning, which is easy to program and significantly improves operational feasibility and calculation accuracy compared with traditional detection methods.

[0004] Technical solution:

[0005] A method for identifying the service alignment of bridges based on multiple three-dimensional laser scans, the method comprising the following steps:

[0006] S1, the bridge service alignment is defined as the equivalent alignment that reflects the statistical characteristics of the main beam vibration, including vehicle / wind-induced vibration factors, at any service period of the bridge; the bridge service alignment is specifically represented by the bridge main beam alignment obtained by multiple three-dimensional laser scans and statistical methods.

[0007] S2. Using a three-dimensional laser scanning device, the bridge structure is scanned multiple times during different service periods to obtain three-dimensional point cloud data of the bridge under non-interrupted traffic conditions during different service periods.

[0008] S3. Select a fixed coordinate origin position, use the coordinate transformation baseline to adjust the initial coordinate system of the point cloud to the target coordinate system, and make the two horizontal coordinate axes in the target coordinate system parallel to the longitudinal or transverse direction of the bridge. Convert the 3D point cloud data of the bridge obtained from multiple scans in step S2 into the same coordinate system. Wherein, if it is a planar straight beam, the longitudinal direction of the bridge is the direction of the intersection line between the vertical plane of the main beam and the horizontal plane of the scanner, and the transverse direction of the bridge is the direction perpendicular to the longitudinal direction of the bridge. If it is a planar curved beam, the transverse direction of the bridge is the transverse width line of the pier, abutment or support, and the longitudinal direction of the bridge is the direction perpendicular to the transverse direction of the bridge.

[0009] S4. Extract the three-dimensional coordinate data of all points along the span direction, at the center of the web thickness, or at the intersection of the web and the flange / bottom plate in the three-dimensional point cloud data of the bridge obtained from each scan. Then, fuse the data extracted from multiple measurements using the Gaussian process algorithm and use regression prediction to obtain the bridge service alignment that is statistically significant and meets the preset confidence conditions.

[0010] S5 identifies changes in the spatial morphology of the main beam of a bridge by comparing the alignment of bridges in different service periods, assesses the structural condition of the bridge during its service period, and tracks changes in the spatial morphology of the bridge during its service period.

[0011] Furthermore, in step S1, the bridge service alignment based on multiple three-dimensional laser scans is represented as a set of three-dimensional spatial coordinate points; the bridge service vertical alignment and bridge service horizontal alignment are obtained by projecting the bridge evaluation index onto different planes, wherein the bridge service alignment projected onto the longitudinal vertical plane of the bridge is the bridge service vertical alignment, and the bridge service horizontal alignment projected onto the horizontal plane of the bridge is the bridge service horizontal alignment.

[0012] Furthermore, in step S2, the process of acquiring bridge 3D point cloud data under uninterrupted traffic conditions at different service periods includes the following sub-steps:

[0013] S21, set up one or more measuring stations at different locations, scan the target bridge from the measuring station, and obtain complete point cloud data of the target bridge; wherein, when there are multiple measuring stations, the complete point cloud data of the target bridge is obtained by stitching together the point cloud data scanned at all measuring stations.

[0014] S22, for different service periods, the target bridge is continuously scanned n times at the same station to obtain n sets of three-dimensional point cloud data of the same bridge structure. The temperature difference between any two scans is less than the preset temperature difference threshold. The scanning time period and temperature conditions corresponding to the bridge point cloud data of different service periods are consistent. n is greater than or equal to 5.

[0015] Furthermore, the preset temperature difference threshold is 3 degrees Celsius.

[0016] Furthermore, when the temperature difference between two scans during different service periods exceeds a preset temperature difference threshold, the bridge service alignment is corrected by combining the actual temperatures during the two scans.

[0017] Further, in step S3, selecting a fixed coordinate origin position, using coordinate transformation baseline, adjusting the initial coordinate system of the point cloud to the target coordinate system, and making the two horizontal coordinate axes in the target coordinate system parallel to the longitudinal or transverse direction of the bridge, includes the following sub-steps:

[0018] S31, Select the coordinate origin position. The coordinate origin of bridge point cloud data in different service periods should be consistent, and the coordinate origin position should be on an object in a long-term stable or nearly stable state.

[0019] S32, extract the coordinate transformation baseline from the acquired bridge point cloud data. This coordinate transformation baseline represents the longitudinal or transverse information of the bridge structure. For straight planar bridges, the longitudinal edge line of the main beam is directly extracted as the coordinate transformation baseline. For curved planar bridges, for rectangular piers, any straight line is directly extracted from the plane of the pier surface as the coordinate transformation baseline. For cylindrical piers, the intersection of the vertical plane containing the pier's transverse width line and the main beam is used as the coordinate transformation baseline. The horizontal coordinates of N points on the coordinate transformation baseline are: {(x1, y1), (x2, y2)…(x…y1)} i y i )…(x N y N )}, i∈{1,2...N};

[0020] The slope k of the baseline is calculated using the least squares method based on the coordinates of N points on the baseline after coordinate transformation.

[0021]

[0022] The angle α between the baseline and the current plane x-axis is calculated based on the slope k. x :

[0023] α x =arctan(k)

[0024] In the formula, α x ∈[0°, 90°];

[0025] S33, Based on the calculated angle α between the coordinate transformation baseline and the current plane coordinate axis, calculate the horizontal rotation matrix T required to transform the point cloud from the initial coordinate system to the target coordinate system. R Using the horizontal rotation matrix T R Perform spatial transformation on the horizontal plane on the overall bridge point cloud.

[0026] Further, in step S33, if the extracted baseline represents the longitudinal information of the bridge structure, including the longitudinal edge line of the main beam and the straight line on the side surface of the pier, and the baseline is rotated clockwise to a direction parallel to the current plane x-axis, then T R for:

[0027]

[0028] If the baseline is rotated counterclockwise to a direction parallel to the current plane's x-axis, then T R for:

[0029]

[0030] If the extracted baseline represents the transverse information of the bridge structure, including the straight line on the front surface of the pier, the intersection of the vertical plane containing the transverse width line of the pier and the main beam, and if the baseline is rotated clockwise to a direction parallel to the current plane's x-axis, then T R for:

[0031]

[0032] If the baseline is rotated counterclockwise to a direction parallel to the current plane's x-axis, then T R for:

[0033]

[0034] Further, in step S4, the three-dimensional coordinate data of all points along the span direction, at the center of the web thickness, or at the intersection of the web and the flange / bottom plate in the three-dimensional point cloud data of the bridge obtained from each scan are extracted from the target span main beam point cloud. The data extracted from multiple measurements are then fused using the Gaussian process algorithm, and the process of regression prediction to obtain the bridge service alignment with statistical significance and satisfying the preset confidence conditions includes the following steps:

[0035] S41. In the multi-source point cloud data obtained from n scans, extract the three-dimensional coordinates of all points in the full span direction at the same lateral position of the target span main beam point cloud in each station's data and fuse them to form a bridge point cloud training dataset.

[0036] S42, calculate the set of predicted vertical coordinates of the bridge.

[0037]

[0038] In the formula, j∈{1,2...N}, and N is the total number of prediction points in the bridge service alignment prediction set;

[0039] S43, calculate the set of predicted transverse coordinates of the bridge.

[0040]

[0041] S44, based on the predicted vertical and horizontal coordinates of the bridge, the service alignment of the bridge for a specific service period is obtained:

[0042]

[0043] Furthermore, in step S42, the set of predicted vertical coordinates of the bridge is calculated. The process includes the following sub-steps:

[0044] S421, For bridge vertical coordinate prediction, let the bridge point cloud training dataset G be... test for:

[0045]

[0046] Wherein, P1, P1...P n This represents the set of three-dimensional coordinates of all points along the entire span at the same lateral position in the point cloud of the target main beam. n is the number of scans, m is the number of points obtained in each scan's point cloud data, and x... nm P represents n The vertical coordinate of the m-th point, z nm P represents n The vertical coordinate of the m-th point; the number of points is controlled by setting a fixed longitudinal interval d, the value of d is related to the span of the main beam of the target span, and the value of d is not greater than 1 / 100 of the span;

[0047] According to the Gaussian process principle, the single longitudinal coordinate x of a bridge in the bridge point cloud training dataset... i The corresponding vertical coordinate random variable It follows a one-dimensional Gaussian distribution:

[0048]

[0049] In the formula, i∈{1,2...M}, and M is the total number of samples in the bridge point cloud training dataset. express The mean, σ 2 express The variance;

[0050] All vertical coordinate random variables in the bridge point cloud training dataset follow a joint Gaussian distribution:

[0051]

[0052] In the formula, μ represents the mean function of the joint Gaussian distribution of the vertical coordinate random variables in the bridge point cloud training dataset, and K = K(X, X) represents the expression of the Gaussian kernel function k and the independent variables X = {x1, x2, ..., x}. i The covariance matrix formed by}

[0053] S422, Set up a bridge service alignment prediction set. The longitudinal coordinate of a single prediction point in the bridge service alignment prediction set is x. j* The interval between the longitudinal coordinates of the predicted points should be as small as possible and no greater than 1 / 200 of the main girder span to maximize the richness of coordinate information of the bridge's service alignment; then the random variable of the vertical coordinate of a single predicted point is... Based on the Gaussian process assumption, and They belong to the same joint Gaussian distribution, that is:

[0054]

[0055] Among them, K * =K(X) * ,X),K ** =K(X) * X * ), X * ={x 1* x 2* ...x j*};

[0056] S423, Considering the observed values ​​of random variables under objective conditions, assume that the noise ε in the bridge point cloud training dataset follows an independent and identically distributed Gaussian distribution. Denote the variance of ε; transform the joint distribution in step S422 into:

[0057]

[0058] Where I is the identity matrix;

[0059] S424, Based on Bayesian estimation, the expression for inferring the bridge service alignment prediction points is:

[0060]

[0061] Based on the principle of Gaussian process function space perspective, in actual calculation process, it is assumed that the mean function in the joint Gaussian distribution is 0;

[0062] The Gaussian kernel function k in the joint Gaussian distribution is determined to be of the following form: squared exponential covariance function.

[0063]

[0064] In the formula, x and x′ represent any two independent variables in a Gaussian distribution;

[0065] Substituting the Gaussian kernel function k from the joint Gaussian distribution into the bridge point cloud training dataset and the bridge service line prediction point data, we get:

[0066]

[0067] Where σ f l and l are undetermined hyperparameters in the SE covariance function;

[0068] S425, Establish the negative logarithmic marginal likelihood function to solve for the hyperparameter θ(σ). f ,1,σ n ):

[0069]

[0070] in,

[0071] For the SE covariance function, the partial derivative of its log-marginal likelihood function with respect to each hyperparameter is expressed as follows:

[0072]

[0073]

[0074]

[0075] Where tr(*) represents finding the trace of the matrix;

[0076] Using the negative log-marginal likelihood function as the minimization objective function and the hyperparameters as optimization variables, we solve for the optimal values ​​of each hyperparameter.

[0077] S426, based on the expression and solution of the bridge service alignment prediction points, the optimal hyperparameters are obtained, and the predicted mean of the prediction points is calculated. and prediction variance for:

[0078]

[0079]

[0080] Predict the mean σ is the predicted vertical coordinate value of the predicted point location in the bridge's service alignment. * Prediction variance σ * Used to quantify the reliability of prediction results;

[0081] S427, calculates a set of predicted vertical service alignment coordinates for the bridge with high confidence. in Indicates the vertical predicted coordinates:

[0082]

[0083] Furthermore, in step S5, by comparing the bridge service alignment corresponding to other lateral positions of the main beam in different periods, the morphological changes of the main beam's vertical deflection and lateral offset are identified, and the spatial morphological changes, including torsion and distortion, are analyzed based on the bridge service alignment of the main beam at different lateral positions.

[0084] Beneficial effects:

[0085] The bridge service alignment identification method based on multiple three-dimensional laser scans of this invention defines the bridge service alignment precisely. By utilizing multi-source point cloud data obtained from multiple three-dimensional laser scans, and through coordinate system adjustment based on coordinate transformation baseline and regression prediction based on Gaussian process algorithm, the accurate identification of the bridge service alignment is achieved. This solves the problem of insufficient spatial information acquisition in traditional measurement methods. Furthermore, it addresses the challenge of complex vibration noise in point cloud arrays caused by bridge vibrations, achieving high-confidence bridge service alignment identification and improving the accuracy of bridge service status control. Attached Figure Description

[0086] Figure 1 This is a flowchart of a bridge service alignment identification method based on multiple three-dimensional laser scans according to an embodiment of the present invention;

[0087] Figure 2 This is a schematic diagram of the coordinate transformation baseline selection and coordinate system adjustment;

[0088] Figure 3 This is a schematic diagram illustrating the extraction of training datasets from multi-source bridge point clouds;

[0089] Figure 4 This is a schematic diagram of the regression results of the Gaussian process algorithm. Detailed Implementation

[0090] The following embodiments are provided to enable those skilled in the art to more fully understand the present invention, but do not limit the invention in any way.

[0091] like Figure 1 As shown in the figure, this embodiment proposes a method for identifying the service alignment of bridges based on multiple three-dimensional laser scans, which mainly includes the following steps:

[0092] (1) Define the equivalent alignment of a bridge during a specific service period as the bridge service space alignment that reflects the statistical characteristics of the main beam vibration (vehicle / wind-induced vibration). The bridge main beam alignment obtained by multiple laser scans and statistical methods is the bridge service space alignment based on multiple three-dimensional laser scans.

[0093] (2) Using a three-dimensional laser scanning device, the bridge structure was scanned multiple times during different service periods to obtain three-dimensional point cloud data of the bridge under non-interrupted traffic conditions during different service periods.

[0094] (3) To facilitate the coordinate positioning of bridge point cloud data, a fixed coordinate origin position is first selected. Then, the initial coordinate system of the point cloud is adjusted to the target coordinate system using the coordinate transformation baseline, and the two horizontal coordinate axes in the target coordinate system are made parallel to the longitudinal or transverse direction of the bridge. If it is a straight planar beam, the longitudinal direction of the bridge can be determined as the direction of the intersection line between the vertical plane of the main beam and the horizontal plane of the scanner, and the transverse direction of the bridge is the direction perpendicular to the longitudinal direction of the bridge. If it is a curved planar beam, the transverse direction of the bridge can be determined as the transverse width line of the pier, abutment or support, and the longitudinal direction of the bridge is the direction perpendicular to the transverse direction of the bridge.

[0095] (4) After the coordinate system is transformed, the three-dimensional point cloud models obtained from multiple scans are all in the same coordinate system. The three-dimensional coordinate data of all points along the span direction of the target span main beam point cloud in each scan data are extracted at the center position of the web thickness or the intersection position of the web and the flange plate / bottom plate. The data extracted from multiple measurements are fused using the Gaussian process (GP) algorithm, and regression prediction is used to obtain the bridge service alignment with statistical significance and high confidence.

[0096] (5) By comparing the spatial alignment of bridges in different service periods, the changes in the spatial shape of the main beam of the bridge can be accurately identified, thereby enabling a precise assessment of the service status of the bridge.

[0097] In this embodiment, the basic conditions that need to be met are:

[0098] (1) The acquired point cloud data should be as complete as possible in the longitudinal direction of the bridge to enable the identification of the complete service alignment of the bridge. For bridges with large spans that cannot be completely scanned by the scanner, the missing point cloud data should be supplemented by setting up reflective targets or by using other equipment with the required accuracy.

[0099] (2) Point cloud data from different service periods should be in the same coordinate system. First, select a fixed coordinate origin position, and then use the coordinate transformation baseline to adjust the initial coordinate system of the point cloud to the target coordinate system, so that the two horizontal coordinate axes in the target coordinate system are parallel to the longitudinal or transverse direction of the bridge.

[0100] In this embodiment, step (1) specifically includes the following definitions and steps:

[0101] Step 1.1: Wind / vehicle-induced vibration makes it difficult to obtain stable and continuous spatial morphological coordinates of the target bridge using existing measurement methods, making it impossible to assess the bridge's service status and track changes in spatial morphology through multi-period measurement comparisons. The defined bridge service alignment is a stable and continuous spatial alignment of the bridge's main girder based on statistical concepts, reflecting the characteristics of wind / vehicle-induced vibration effects, and having an equivalent substitution effect, thus ensuring comparability between multi-period measurement data of the bridge's spatial morphology.

[0102] Step 1.2: The bridge service alignment, based on multiple 3D laser scans, is represented as a set of 3D spatial coordinate points. According to bridge evaluation indicators, the bridge service vertical alignment and bridge service lateral alignment can be obtained by projecting onto different planes. The bridge service alignment projected onto the longitudinal vertical plane is the bridge service vertical alignment, and the bridge service lateral alignment projected onto the horizontal plane is the bridge service lateral alignment.

[0103] In this embodiment, step (2) specifically includes the following steps:

[0104] Step 2.1: Select a 3D laser scanning device with sufficient accuracy and range to meet the observation requirements. Determine a station location with good field of view and minimal external interference. Adjust the scanning parameters to ensure the instrument can completely capture the target bridge. When a single station cannot achieve a complete measurement, multiple stations should be set up as needed, and the complete target bridge point cloud data should be obtained by stitching together the data from these multiple stations. To achieve accurate stitching, the point cloud data from adjacent stations should have at least 50% overlap.

[0105] Step 2.2: During the scanning process, the ambient temperature variation at all stations should not exceed 3°C within the total scanning time to avoid interference with the identification of the service alignment and the comparison of service status caused by bridge deformation due to temperature differences.

[0106] Step 2.3: According to the adjusted scanning parameters, perform n consecutive 3D laser scans on the target at the same station (n≥5 to ensure sufficient data volume, and the number of measurements should be as high as possible while meeting the conditions in 2.2), to obtain n sets of 3D point cloud data of the same bridge structure. For cases with multiple stations at different locations, each station should perform n 3D laser scans, and the point cloud data from different stations should be stitched together n times to ultimately obtain n sets of 3D point cloud data of the same bridge structure.

[0107] If the temperature fluctuates significantly on the day of the scan, or if the bridge structure is too large, resulting in a long scan time per session, and the surveyors are unable to complete at least 5 scans at each station under the constraints outlined in section 2.2, the scan time should be changed to ensure that each station has sufficient scans within the same day. If the above conditions do not permit, a supplementary scan can be performed the following day after the initial scan. The supplementary scan time and temperature range should be limited to a similar time and temperature range as the initial scan.

[0108] Step 2.4: When acquiring bridge point cloud data for different service periods, it is important to control the acquisition under similar temperature conditions as much as possible. Assume that the bridge service space lineage obtained from multiple 3D laser scans at an average temperature t1 is denoted as 11, and the bridge service space lineage obtained from multiple 3D laser scans at an average temperature t2 is denoted as 12 (average temperature is defined as the average of the scan start temperature, the temperature at the end of each hour after the scan begins, and the scan end temperature; if the total scan time is less than one hour, the average of the scan start temperature and the scan end temperature is used). If the difference between t1 and t2 is small (≤3℃), comparing 11 and 12 can directly reflect the difference in the bridge's service status under non-temperature effects (dead load, live load, etc.) between the two scan periods. If the difference between t1 and t2 is large (>3℃), comparing 11 and 12 also includes the influence of temperature on the bridge's service status. Considering that the impact of temperature on bridge morphological changes is difficult to quantify accurately across all dimensions of the bridge, to achieve bridge service status assessment and spatial morphological change tracking, three-dimensional laser scanning of the bridge's service alignment at different service stages should ideally be conducted within the same season or at the same time of day when temperature differences are minimal. If these conditions cannot be met due to engineering requirements, the final service alignment comparison should incorporate the impact of temperature differences on the bridge's service alignment through reasonable calculations.

[0109] In this embodiment, step (3) specifically includes the following steps:

[0110] Step 3.1: The origin of the coordinate system should be selected on an object that is in a long-term stable or nearly stable state. The origin of the coordinate system for bridge point cloud data from different service periods should remain consistent.

[0111] Step 3.2, refer to Figure 2 In the acquired bridge point cloud data, a coordinate transformation baseline is extracted (for multi-station cases, multi-station point cloud stitching needs to be completed first). This baseline should represent the longitudinal or transverse information of the bridge structure. For straight planar bridges, the longitudinal edge line of the main beam can be directly extracted as the coordinate transformation baseline; for curved planar bridges, for rectangular cross-section piers, any straight line can be directly extracted from the plane of the pier surface as the coordinate transformation baseline, and for cylindrical piers, the intersection line of the vertical plane containing the transverse width line of the pier and the main beam can be used as the coordinate transformation baseline. The horizontal coordinates of each point on the coordinate transformation baseline are: {(x1, y1), (x2, y2)…(x…y1)} i y i )}, i∈{1,2…N}.

[0112] The slope k of the baseline is calculated using the least squares method based on the coordinates of N points on the baseline:

[0113]

[0114] The angle α between the baseline and the current plane x-axis is calculated based on the slope k. x (α x ∈[0°, 90°]):

[0115] α x =arctan(k).

[0116] Step 3.3: Based on the calculated angle α between the baseline and the current plane coordinate axis, calculate the horizontal rotation matrix T required to transform the point cloud from the initial coordinate system to the target coordinate system. R This matrix is ​​used to perform spatial transformation of the overall bridge point cloud in the horizontal plane.

[0117] The specific implementation process will vary depending on the baseline selection. If the extracted baseline represents the longitudinal information of the bridge structure, such as the longitudinal edge line of the main beam, the straight line of the pier side surface, etc., and the baseline is rotated clockwise to a direction parallel to the current plane x-axis, then T R for:

[0118]

[0119] If the baseline is rotated counterclockwise to a direction parallel to the current plane's x-axis, then T R for:

[0120]

[0121] If the captured baseline represents the transverse information of the bridge structure, such as the straight line on the front surface of the pier, the intersection of the vertical plane containing the transverse width line of the pier and the main beam, and the baseline is rotated clockwise to a direction parallel to the current plane's x-axis, then T R for:

[0122]

[0123] If the baseline is rotated counterclockwise to a direction parallel to the current plane's x-axis, then T R for:

[0124]

[0125] In this embodiment, step (4) specifically includes the following steps:

[0126] Step 4.1, refer to Figure 3In the multi-source point cloud data obtained from n scans, the 3D coordinates of all points along the entire span (center of the web thickness direction or intersection of the web and flange / bottom plate) at the same lateral position in the target span main girder point cloud of each station are extracted and fused to form a bridge point cloud training dataset. The bridge vertical coordinate prediction and bridge lateral coordinate prediction should be calculated separately. Taking the bridge vertical coordinate prediction as an example, the bridge point cloud training dataset G... test It only contains longitudinal and vertical coordinates.

[0127]

[0128] Wherein, P1, P1…P n This represents the set of three-dimensional coordinates of all points along the entire span at the same transverse position in the point cloud of the target main beam. n is the number of scans, m is the number of points obtained in each scan's point cloud data, and the number of points can be controlled by setting a fixed longitudinal interval d. nm P represents n The vertical coordinate of the m-th point, z nm P represents n The vertical coordinate of the m-th point. The value of d can be appropriately selected based on the span of the main beam of the target span (not greater than 1 / 100 of the span).

[0129] Step 4.2, refer to Figure 4 In this embodiment, multiple scans of point cloud data are directly input into the algorithm. Based on the Gaussian process principle, the x-coordinate of a single longitudinal bridge in the bridge point cloud training dataset... i (i∈{1,2…M}, where M is the total number of samples in the bridge point cloud training dataset) corresponding to the vertical coordinate random variable It follows a one-dimensional Gaussian distribution:

[0130]

[0131] In the formula, i∈{1,2…M}, and M is the total number of samples in the bridge point cloud training dataset. express The mean, σ 2 express The variance; and all vertical coordinate random variables in the bridge point cloud training dataset follow a joint Gaussian distribution:

[0132]

[0133] Where μ represents the mean function of the joint Gaussian distribution of the vertical coordinate random variables in the bridge point cloud training dataset. K = K(X, X) represents the sum of the Gaussian kernel function k and the independent variables X = {x1, x2, ..., x}. i The covariance matrix formed by}

[0134] Establish a bridge service spatial alignment prediction set, where the longitudinal bridge coordinate of a single prediction point in the prediction set is x. j* (j∈{1, 2…N}, where N is the total number of prediction points for the bridge service spatial alignment prediction set). The interval between the longitudinal coordinates of the prediction points should be as small as possible (no more than 1 / 200 of the main girder span) to ensure the richness of coordinate information for the bridge service spatial alignment. Then, the random variable for the vertical coordinate of a single prediction point is... Based on the Gaussian process assumption, Know They belong to the same joint Gaussian distribution, that is:

[0135]

[0136] Among them, K * =K(X) * ,X),K ** =K(X) * X * ), X * ={x 1* x 2* ...x j*}

[0137] Considering the observed values ​​of random variables under objective conditions, i.e., noise always exists in the bridge point cloud training dataset, and assuming that the noise follows an independent and identically distributed Gaussian distribution, Then the joint distribution mentioned above is:

[0138]

[0139] Where I is the identity matrix. Let ε represent the variance.

[0140] Based on Bayesian estimation, the expression for predicting the spatial alignment of bridge service points can be derived as follows:

[0141]

[0142] Based on the principle of Gaussian process function space perspective, in actual calculations, it can be assumed that the mean function in the joint Gaussian distribution is 0.

[0143] The Gaussian kernel function k in the joint Gaussian distribution is determined to be in the form of the following squared exponential (SE) covariance function:

[0144]

[0145] Where x and x′ represent any two independent variables in a Gaussian distribution.

[0146] After substituting the bridge point cloud training dataset and the bridge service space alignment prediction point data, the specific results are as follows:

[0147]

[0148] Where σ f 1 is a hyperparameter to be determined in the SE covariance function.

[0149] Secondly, the negative logarithmic marginal likelihood function is established to solve for the hyperparameter θ(σ). f ,l,σ n ):

[0150]

[0151] in,

[0152] For the SE covariance function, the partial derivative of its log-marginal likelihood function with respect to each hyperparameter is expressed as follows:

[0153]

[0154]

[0155]

[0156] Where tr(*) represents finding the trace of a matrix.

[0157] Using the negative logarithmic marginal likelihood function as the minimization objective function and the hyperparameters as optimization variables, the optimal values ​​of each hyperparameter are solved using the above formula.

[0158] Finally, based on the expression and solution of the bridge service spatial alignment prediction points, the optimal hyperparameters are obtained, and the predicted mean of the prediction points can be obtained. and prediction variance for:

[0159]

[0160]

[0161] The predicted mean When used as the predicted vertical coordinate of a point location in the bridge's service space alignment, it can usually guarantee a small regression error; σ * Prediction variance can be used to quantify the reliability of prediction results.

[0162] Reference Figure 4 Finally, a set of predicted vertical coordinates of the bridge with high confidence was obtained. As shown in the following formula:

[0163]

[0164] In the formula, j∈{1,2…N}, and N is the number of prediction points in the spatial linearity prediction set for bridge service. This indicates the vertical predicted coordinates.

[0165] The scatter points in the figure represent the input training set data, the curve is the bridge service curve obtained by Gaussian regression, and the gray area represents the 95% confidence interval.

[0166] Step 4.3 uses the same regression prediction method as step 4.2 for the lateral service alignment of the bridge, ultimately yielding a set of predicted lateral coordinates for the bridge with high confidence. As shown in the following formula:

[0167]

[0168] Step 4.4: Based on the predicted vertical and horizontal coordinates of the bridge, the service alignment of the bridge for a specific service period is obtained, i.e., the service alignment of the bridge based on multiple three-dimensional laser scans, as shown in the following formula:

[0169]

[0170] In this embodiment, step (5) specifically includes the following steps:

[0171] Depending on the requirements, in addition to the center position of the web thickness or the intersection position of the web with the flange plate / bottom plate mentioned above, the bridge service alignment corresponding to other transverse positions of the main beam can also be calculated; based on the comparison of bridge service alignments at different stages, the analysis of key morphological changes such as vertical deflection and transverse offset of the main beam can be realized; and the analysis of complex spatial morphological changes such as torsion and distortion can also be realized based on the bridge service alignments at different transverse positions of the main beam.

[0172] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1.A bridge service line identification method based on multiple three-dimensional laser scanning, characterized in that, The bridge service line shape identification method comprises the following steps: S1, defining the bridge service line shape as an equivalent line shape reflecting the statistical characteristics of the main girder vibration of the bridge in any service period, including vehicle / wind-induced vibration factors; the bridge service line shape is specifically a bridge main girder line shape obtained by multiple three-dimensional laser scanning and statistical methods; S2, using a three-dimensional laser scanning device, performing multiple three-dimensional laser scanning on the bridge structure at different service periods to obtain bridge three-dimensional point cloud data under non-interrupted traffic at different service periods; S3, selecting a fixed coordinate origin position, adjusting the point cloud initial coordinate system to the target coordinate system by using the coordinate conversion baseline, and making the two horizontal plane coordinate axes in the target coordinate system parallel to the bridge longitudinal direction or transverse direction, and converting the bridge three-dimensional point cloud data obtained by multiple scanning in step S2 in the same coordinate system; wherein, if it is a plane straight girder, the bridge longitudinal direction is the intersection line direction of the main girder vertical plane and the scanner horizontal plane, and the bridge transverse direction is perpendicular to the bridge longitudinal direction; if it is a plane curved girder, the bridge transverse direction is the transverse bridge width line direction of the pier, abutment or support, and the bridge longitudinal direction is perpendicular to the bridge transverse direction; S4, extracting the three-dimensional coordinate data of all points of the target span main girder point cloud along the span direction, at the center position of the web thickness or at the intersection position of the web and the flange plate / bottom plate from the bridge three-dimensional point cloud data obtained by each scanning, and fusing the extracted data by multiple measurements by using a Gaussian process algorithm to obtain a bridge service line shape with statistical significance and satisfying the preset reliability condition through regression prediction; S5, identifying the spatial form change of the bridge main girder by comparing the bridge service line shapes at different periods, evaluating the bridge structure state at the bridge service period, and tracking the spatial form change of the bridge at the bridge service period. 2.The bridge service linear identification method based on multiple three-dimensional laser scanning according to claim 1, wherein, In step S1, the bridge service line shape based on multiple three-dimensional laser scanning is a set of three-dimensional space coordinate points; according to the bridge evaluation index, the bridge service line shape is projected to different planes to obtain the bridge service vertical line shape and the bridge service transverse line shape, wherein the bridge service line shape projected to the bridge longitudinal vertical plane is the bridge service vertical line shape, and projected to the bridge horizontal plane is the bridge service transverse line shape. 3.The bridge service linear identification method based on multiple three-dimensional laser scanning according to claim 1, wherein, In step S2, the process of obtaining bridge three-dimensional point cloud data under non-interrupted traffic at different service periods comprises the following sub-steps: S21, setting one or more measuring stations at different positions, scanning the target bridge from the measuring station to obtain the point cloud data of the complete target bridge; wherein when the number of measuring stations is multiple, the point cloud data scanned at all measuring stations is spliced to obtain the point cloud data of the complete target bridge; S22, for different service periods, continuously scanning the target bridge n times by three-dimensional laser scanning at the same measuring station to obtain n groups of three-dimensional point cloud data of the same bridge structure, and the air temperature difference between any two scans is less than a preset air temperature difference threshold; the bridge point cloud data at different service periods correspond to the same scanning period and air temperature condition; n is greater than or equal to 5. 4.The bridge service linear identification method based on multiple three-dimensional laser scanning according to claim 3, wherein, The preset air temperature difference threshold is 3 degrees Celsius. 5.The bridge service linear identification method based on multiple three-dimensional laser scanning according to claim 3, wherein, When the temperature difference between two scans in different service periods is greater than the preset temperature difference threshold, the bridge service linear is corrected in combination with the actual temperature at the time of the two scans. 6.The bridge service linear identification method based on multiple three-dimensional laser scanning according to claim 1, wherein, In step S3, a fixed coordinate origin position is selected, the initial coordinate system of the point cloud is adjusted to the target coordinate system by using the coordinate conversion baseline, and the two horizontal coordinate axes in the target coordinate system are parallel to the longitudinal or transverse direction of the bridge. The process includes the following sub-steps: S31, select the coordinate origin position, the coordinate origin positions of the bridge point cloud data in different service periods are consistent, and the coordinate origin position is on an object in a long-term stable state; S32, intercept a coordinate conversion baseline in the acquired bridge point cloud data, the coordinate conversion baseline representing longitudinal or transverse information of the bridge structure; wherein, for a planar straight bridge, directly extracting a main beam longitudinal edge line as the coordinate conversion baseline; for a planar curved bridge, directly intercepting an arbitrary straight line on a plane of a rectangular cross-section form pier on a pier surface as the coordinate conversion baseline, and using an intersection line of a vertical plane where a bridge pier transverse width line is located and the main beam as the coordinate conversion baseline; horizontal plane coordinates of N points on the coordinate conversion baseline are: {(x1, y1), (x2, y2)…(xN, yN)}, i∈{1, 2…N}; S33, determining a coordinate conversion matrix of the coordinate conversion baseline according to the horizontal plane coordinates of the N points on the coordinate conversion baseline; wherein, the coordinate conversion matrix is determined by using a least square method; S34, converting the coordinate system of the bridge point cloud data to the coordinate system of the coordinate conversion baseline according to the coordinate conversion matrix; wherein, the coordinate system of the bridge point cloud data is converted to the coordinate system of the coordinate conversion baseline, and the coordinate system of the coordinate conversion baseline is the same as the coordinate system of the coordinate conversion baseline. i ; and i ; and N ; and N ​ The slope k of the baseline is calculated by using the least square method according to the coordinates of the N points on the baseline: According to the slope k, the angle a of the baseline and the current plane x coordinate axis is calculated x : α x = arctan(k) wherein α x ∈ [0°, 90°]; S33, according to the angle a of the coordinate conversion baseline and the current plane coordinate axis calculated, calculating the horizontal rotation matrix T required for the point cloud to transform from the initial coordinate system to the target coordinate system R , using the horizontal rotation matrix T R to perform spatial transformation of the whole bridge point cloud in the horizontal plane. 7.The bridge service linear identification method based on multiple three-dimensional laser scanning according to claim 6, characterized in that, In step S33, if the intercepted baseline represents the longitudinal information of the bridge structure including the longitudinal edge line of the main beam and the straight line of the pier side surface, and the baseline is rotated to the direction parallel to the current plane x coordinate axis in the clockwise direction, T R is: If the baseline rotation to a direction parallel to the current plane x-coordinate axis is a counter-clockwise rotation, then T R is: If the intercepted baseline represents the transverse information of the bridge structure including the straight line of the vertical surface of the pier, the intersection line of the vertical plane where the width line of the pier transverse to the bridge is located and the main beam, and the baseline is rotated to the direction parallel to the current plane x coordinate axis clockwise, then T R is: If the baseline rotation to a direction parallel to the current plane x-coordinate axis is a counter-clockwise rotation, then T R is: 8.The bridge service linear identification method based on multiple three-dimensional laser scanning according to claim 1, wherein, In step S4, the three-dimensional coordinate data of all points of the target span girder point cloud in the bridge three-dimensional point cloud data obtained by each scanning along the span direction, at the center position of the web thickness, or at the intersection position of the web and the flange plate / bottom plate are extracted, and the extracted data obtained by multiple measurements are fused by using the Gaussian process algorithm to obtain the bridge service linear with statistical significance and satisfying the preset reliability condition. The process includes the following steps: S41, in the multi-source point cloud data obtained by n times of scanning, the three-dimensional coordinates of all points of the target span girder point cloud at the same transverse position in each station data are extracted and fused to form a bridge point cloud training data set; S42, calculate the bridge vertical predicted coordinate set In the formula, j ∈ {1, 2…N}, N is the number of the bridge service linear prediction set points, represents the vertical prediction coordinate of the jth prediction point; S43, calculate the bridge transverse predicted coordinate set S44, the bridge service linear in a specific service period is obtained according to the vertical and transverse predicted coordinate values of the bridge: 9.The bridge service linear identification method based on multiple three-dimensional laser scanning according to claim 8, characterized in that, In step S42, the bridge vertical predicted coordinate set is calculated The process includes the following sub-steps: S421, for bridge vertical coordinate prediction, set the bridge point cloud training dataset G test is: P1, P1...P n represents the three-dimensional coordinate set of all points in the full span direction of the same transverse position of the target cross-girder point cloud; n is the number of scans, m is the number of points taken in the point cloud data of each scan, x nm represents the longitudinal coordinate of the mth point coordinate in P n represents the vertical coordinate of the mth point coordinate in P nm represents the vertical coordinate of the mth point coordinate in P n controls the number of points by setting a fixed point longitudinal interval d, the value of d is related to the span of the target cross-girder, and the value of d is not greater than 1 / 100 of the span. According to the Gaussian process principle, a single longitudinal bridge coordinate x i The corresponding vertical coordinate random variable obeys a one-dimensional Gaussian distribution: In the formula, i is in {1, 2,..., M}, M is the total sample quantity of the bridge point cloud training data set, denotes the mean of 2 denotes the variance of ;​ All vertical coordinate random variables in the bridge point cloud training data set follow a joint Gaussian distribution: In the formula, μ represents a mean function of a joint Gaussian distribution of vertical coordinate random variables in a bridge point cloud training data set, K = K(X, X) represents a covariance matrix composed of a Gaussian kernel function k and an independent variable X = {x1, x2…x i} S422, set the bridge service linear prediction set, the longitudinal bridge coordinate of a single prediction point in the bridge service linear prediction set is x j* The interval of the longitudinal bridge coordinate of the prediction point should be as small as possible and not more than 1 / 200 of the main beam span to maximize the coordinate information richness of the bridge service linear; the vertical coordinate random variable of a single prediction point is According to the Gaussian process assumption, and belong to the same joint Gaussian distribution, that is: where K * = K(X * ,X), K ** = K(X * ,X * ), X * = {x 1* ,x 2* …x j*}; S423, considering the observation value of the random variable under the objective condition, assuming that the noise ε in the bridge point cloud training data set satisfies the independent and identically distributed Gaussian distribution, representing the variance of ε; the joint distribution in step S422 is converted into: Where I is the unit matrix; S424, according to Bayesian estimation, the expression of the bridge service linear prediction point is inferred as: Based on the principle of Gaussian process function space perspective, it is assumed that the mean function in the joint Gaussian distribution is 0 in the actual calculation process; The form of the Gaussian kernel function k in the joint Gaussian distribution is determined as the following square exponential covariance function: where x and x ′ denotes the argument in any two Gaussian distributions; After substituting the Gaussian kernel function k in the joint Gaussian distribution into the bridge point cloud training data set and the bridge service linear prediction point data, we have: where σ f and l are hyperparameters to be determined in the SE covariance function; S425, establish negative log marginal likelihood function to solve hyperparameters θ(σ f , l, σ n ): wherein For the SE covariance function, the partial derivative expression of the logarithmic marginal likelihood function with respect to each hyperparameter is: Where tr(*) represents the trace of the matrix; Taking the negative logarithmic marginal likelihood function as the minimization objective function and taking the hyperparameters as the optimization variables, the optimal values of the hyperparameters are solved. S426, according to the expression of the bridge service linear prediction point and the optimal super parameter obtained by solving, the predicted mean value of the prediction point is calculated and the prediction variance is: the predicted mean value as the predicted value of the vertical coordinate at the position of the prediction point in the bridge service linear; σ * the prediction variance σ * for quantifying the reliability of the prediction result; S427, calculate the bridge vertical service linear prediction coordinate set with high confidence wherein denotes the vertical prediction coordinate: 10.The bridge service linear identification method based on multiple three-dimensional laser scanning according to claim 1, wherein, In step S5, the shape changes of the vertical deflection and transverse offset of the main girder are identified by comparing the bridge service linear of the other transverse positions of the main girder in different periods, and the spatial shape changes including torsion and distortion are analyzed based on the bridge service linear of different transverse positions of the main girder.

Images