Suspension bridge main cable strand height difference real-time cloud measurement method and system

Abstract

The application discloses a kind of suspension bridge main cable strand height difference real-time cloud measurement method and system, it is related to suspension bridge construction measurement field, comprising: the measurement data of reference target and monitoring target measured by obtaining measuring device measurement, and each target positioning data is obtained by solving;Temperature data of installed strand and strand to be installed are obtained;Positioning data and temperature data are transmitted to cloud processing module;Cloud processing module, according to the positioning data of each reference target, the actual linear type of installed strand is fitted, according to the positioning data of each monitoring target, the actual linear type of strand to be installed is fitted, and is corrected in combination with temperature data, obtains the relative strand height difference of installed strand and strand to be installed.The application realizes the automation, intelligentization and high-precision of suspension bridge main cable strand height difference measurement, significantly improves construction efficiency and linear control precision.

Description

Technical Field

[0001] This invention relates to the field of suspension bridge construction surveying technology, and more specifically to a method and system for real-time cloud measurement of the height difference of the main cable strands of a suspension bridge. Background Technology

[0002] The erection of the main cable strands for suspension bridges is a crucial step in the construction of long-span suspension bridges. The strand height difference and alignment control directly affect the quality of the main cable formation, the structural stress safety, and the service life of the bridge. Accurate measurement of the strand height difference is a core prerequisite for ensuring efficient strand erection and is of great significance for improving construction precision and reducing safety risks.

[0003] However, in the current process of erecting main cable strands, the measurement methods mainly rely on manual total stations and levels, which have problems such as low measurement efficiency, poor real-time performance, and delayed data feedback. Furthermore, the length of the strands is affected by factors such as manufacturing temperature and batch, and the traditional marking method is not accurate enough, resulting in large errors in the control of the strand length. The reference strands and the strands to be erected are in a non-uniform temperature field, and the alignment is distorted due to the constraints of sunlight and strand groups. Adjustment can only be carried out during the low temperature window in the early morning, resulting in a narrow working window. At the same time, it relies on manual observation and qualitative judgment, lacks automated measurement and intelligent adjustment devices, and has not achieved cloud-based fusion processing of positioning, temperature, and alignment data, making it difficult to achieve millimeter-level quantitative control and all-weather accurate erection requirements.

[0004] Therefore, developing a real-time cloud measurement system and method for the height difference of the main cable strands of a suspension bridge that is highly real-time, accurate, resistant to temperature interference, and has cloud-based intelligent calculation capabilities is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method and system for real-time cloud measurement of the height difference of the main cable strands of a suspension bridge, in order to solve the above problems.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: This invention discloses a method for real-time cloud measurement of the height difference of the main cable strands of a suspension bridge, comprising: Acquire the measurement data of the reference target and monitoring target measured by the measuring device, and calculate the positioning data of each target; Obtain temperature data for installed and uninstalled cable strands; The location data and temperature data are transmitted to the cloud processing module; The cloud processing module fits the actual line shape of the installed cable strands based on the positioning data of each benchmark target, fits the actual line shape of the cable strands to be installed based on the positioning data of each monitoring target, and makes corrections based on the temperature data to obtain the relative height difference between the installed cable strands and the cable strands to be installed.

[0007] Furthermore, the acquisition of measurement data from the reference target and monitoring target measured by the measuring device specifically includes: A reference target is set on the installed cable strand, and a monitoring target is set on the cable strand to be installed; both the reference target and the monitoring target are microprism retroreflective targets, and the target body is a flexible magnetic structure that is bent to fit the curved surface of the cable strand and is fixed to the outside of the cable strand by magnetic attraction. Several laser ranging and point cloud acquisition devices are deployed around the suspension bridge. The laser ranging and point cloud acquisition devices emit lasers to each of the reference targets and each of the monitoring targets, and collect the laser reflection point cloud data corresponding to each target in real time. The three-dimensional coordinates of the laser ranging and point cloud acquisition devices themselves, the three-dimensional coordinates of the laser reflection point cloud of each target, and the reflected light intensity data are used as measurement data.

[0008] Furthermore, the calculation yields target positioning data, specifically including: Step 1: Calculate edge parameters for the collected laser reflection point cloud data and determine the edge parameter threshold. Perform edge detection on the point cloud data based on the edge parameter threshold to extract the target region point cloud. Step 2: Based on the reflected light intensity data of the point cloud, perform light intensity filtering, denoising and normalization on the point cloud of the target area after edge detection, and remove points with abnormal light intensity and invalid point clouds. Step 3: Perform spatial interpolation to complete the target region point cloud after removing outliers and invalid point clouds; Step 4: Based on the interpolated and completed 3D coordinates of the target area point cloud, calculate and determine the direction vector from the laser ranging and point cloud acquisition equipment to the corresponding target; Step 5: Construct an input matrix using the interpolated target area point cloud 3D coordinates and reflected light intensity. Input the input matrix and the direction vector into the pre-trained positioning and detection model, and output the single positioning data of the target relative to the laser ranging and point cloud acquisition device. Step 6: Summarize the single positioning data of the same target under all laser ranging and point cloud acquisition devices, and obtain the final positioning data of the target by fitting the data using the least squares method.

[0009] Furthermore, the edge parameter calculation includes: Based on the three-dimensional coordinates of the laser reflection point cloud and the reflected light intensity, a two-dimensional reflection intensity map is generated by projection. , , These are the coordinates of the two-dimensional reflectance intensity map; edge parameters are calculated pixel-by-pixel based on the improved Zernike moments, specifically: First, the two-dimensional reflectance intensity map of the target area. Step Tzernik Defined as: ; in, The radial distance is in polar coordinates. This represents the distance from the calculation point to the center of the unit circle; The polar coordinate azimuth angle represents the angle between the line connecting the calculation point and the center of the unit circle and the x-axis. Zernike kernel function The conjugate of complex numbers, , For Zernike radial polynomials, It is a natural constant. The imaginary unit; Then, taking advantage of the rotational invariance of Zernike moments, we obtain lower-order moments. , Calculate edge parameters: ; in, For background reflection intensity, The edge step intensity; The determination of the edge parameter threshold includes: Based on the differences in reflection intensity between the target, the cable strands, and the background, as well as geometric constraints, adaptive edge detection conditions are constructed: ;in, The threshold for edge step intensity. , The maximum reflection intensity of the point cloud in the target area; This is the reflection intensity range coefficient; To determine the background reflection intensity threshold, a binary search method is used for iterative optimization, specifically: First, calculate the average reflectance of the point cloud in the target area. Set the search range , This is a value used to adjust the search range. Then, the point clouds of the two arc-shaped boundary contours inside and outside the target are extracted, and the least squares method is used to perform ring curve fitting to construct the curve fitting loss function: ;in and These are the fitting residuals for the inner arc-shaped profile and the fitting residuals for the outer arc-shaped profile, respectively. Binary search within the interval, with loss function An intensity value less than the preset residual threshold is determined to be the background reflection intensity threshold. .

[0010] Furthermore, the localization detection model includes an input layer, an encoding layer, a feature alignment layer, and a decoding regression layer; The input layer receives and standardizes the input data, constructs a 4-channel matrix from the interpolated target point cloud 3D coordinates and reflected light intensity, and inputs it into the encoding layer. The coding layer consists of three levels of sampling, grouping, and convolution. The first level performs FPS farthest point sampling, local neighborhood grouping, and extracts coarse geometric features of the target on the point cloud. The second level fuses spatial coordinates and reflected light intensity features through an MLP multilayer perceptron. The third level performs downsampling and aggregates global features to output a global feature vector of the target. The feature alignment layer aligns the direction vector with the global feature vector in spatial pose to eliminate viewpoint differences and obtain the aligned feature vector. The decoding regression layer consists of a flattening layer and three fully connected layers. The flattening layer flattens the aligned feature vectors in one dimension. Then, the three fully connected layers perform nonlinear mapping and position fitting on the flattened features to obtain the final three-dimensional coordinates.

[0011] Furthermore, the acquisition of temperature data of the installed and uninstalled cable strands specifically includes: using a drone equipped with an infrared thermal imager to conduct a flight path inspection along the extension direction of the installed and uninstalled cable strands, continuously imaging the entire length of the cable strands with the infrared thermal imager, collecting infrared temperature information of each measuring point on the surface of the cable strands in real time, and obtaining a temperature data sequence of each point distributed along the length of the cable strands.

[0012] Furthermore, the actual line shapes of the installed strands and the strands to be installed are obtained by fitting the data in the following way: The final positioning data of all reference targets on the installed cable strands are uniformly transformed to the overall coordinate system of the suspension bridge construction. Abnormal positioning points are eliminated and smoothing preprocessing is performed. Based on the catenary theory, a basic mathematical model of the installed cable strand alignment is constructed. The three-dimensional coordinates of the preprocessed reference targets are used as constraint points. The least squares method is used for nonlinear fitting to obtain the actual catenary alignment equation of the installed cable strands. The final positioning data of all monitoring targets on the cable strand to be installed are uniformly transformed to the overall coordinate system of the suspension bridge construction. Abnormal positioning points are eliminated and smoothing preprocessing is performed. Based on the catenary theory, a basic mathematical model of the cable strand to be installed is constructed. The three-dimensional coordinates of the preprocessed monitoring targets are used as constraint points. The least squares method is used for nonlinear fitting to obtain the actual catenary line shape equation of the cable strand to be installed.

[0013] Furthermore, the step of correcting the temperature data to obtain the relative height difference between the installed strands and the strands to be installed specifically includes: Calculate the length of the strand to be installed based on the actual cable profile. The specific formula is as follows: ;in, For the length of the strand, The weight per unit length of the strand. For horizontal tension of the cable, It is a hyperbolic sine function. The span of the bridge; Based on the actual line shape of the cable strands to be installed, generate a coordinate data sequence corresponding to the temperature data sequence; The temperature data sequences of the installed and uninstalled cable strands, the coordinate data sequence of the uninstalled cable strands, and the length of the uninstalled cable strands, the actual span of the bridge, and the self-weight per unit length of the cable strands are input into the correction model to obtain the corrected alignment of the uninstalled cable strands. Along the length of the cable strand, at preset intervals, extract the height difference between the corresponding coordinate points on the equation of the actual alignment of the installed cable strand and the corrected alignment of the cable strand to be installed, and use the height difference sequence as the relative height difference between the installed cable strand and the cable strand to be installed.

[0014] Furthermore, the correction model is a temporal correction model based on an improved Transformer, which consists of a sequence input layer, an improved Transformer encoder, a static feature embedding layer, a feature fusion layer, and a regression output layer. The sequence input layer receives the temperature data sequence and the coordinate data sequence, performs data standardization, and outputs a spatial feature sequence of fixed length and fixed dimensions. The improved Transformer encoder is composed of multiple layers of lightweight coding units stacked together. It performs global correlation extraction on spatial feature sequences and outputs global spatial features. The static feature embedding layer converts the length of the cable strand to be installed, the actual span of the bridge, and the self-weight per unit length of the cable strand into a static embedding vector through standardization and linear mapping. The feature fusion layer concatenates the global spatial features with the static embedding vector through channels to obtain fused features. The regression output layer consists of multiple fully connected networks, and the formula parameters of the corrected line type of the cable strand to be installed are obtained based on the fusion feature mapping.

[0015] This invention also discloses a real-time cloud measurement system for the height difference of the main cable strands of a suspension bridge, comprising: Positioning module: acquires measurement data of the reference target and monitoring target measured by the measuring device, and calculates the positioning data of each target; Temperature detection module: acquires temperature data of installed and uninstalled cable strands; Transmission module: transmits the positioning data and temperature data to the cloud processing module; Cloud processing module: Fits the actual line shape of the installed cable strands based on the positioning data of each benchmark target, fits the actual line shape of the cable strands to be installed based on the positioning data of each monitoring target, and corrects it in combination with the temperature data to obtain the relative height difference between the installed cable strands and the cable strands to be installed.

[0016] As can be seen from the above technical solution, compared with the prior art, the present invention provides a method and system for real-time cloud measurement of the height difference of the main cable strands of a suspension bridge, which has the following beneficial effects: This invention employs a combination of laser point cloud and microprism retroreflective targets to achieve automated acquisition and high-precision calculation of target positioning data, significantly improving measurement efficiency and real-time performance, and solving the problems of low efficiency and data lag in traditional manual measurements. It utilizes UAV infrared inspection to collect cable strand temperature data throughout the process, and combines this with an improved Transformer time-series correction model to eliminate the interference of temperature field inhomogeneity on the alignment, enabling all-weather cable adjustment. A cloud-based processing module completes the fusion calculation and alignment fitting correction of positioning, temperature, and alignment data, achieving millimeter-level quantitative control of cable strand height differences, effectively improving the main cable forming quality and structural stress safety. The flexible magnetic target is reusable and fits tightly, reducing construction costs and installation difficulty. This invention achieves automation, intelligence, and high precision in measuring the height difference of suspension bridge main cable strands, significantly improving construction efficiency and alignment control accuracy. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the overall process of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] This invention discloses a method for real-time cloud measurement of the height difference of the main cable strands of a suspension bridge, such as... Figure 1 As shown, it includes: Acquire the measurement data of the reference target and monitoring target measured by the measuring device, and calculate the positioning data of each target; Obtain temperature data for installed and uninstalled cable strands; The location data and temperature data are transmitted to the cloud processing module; The cloud-based processing module fits the actual alignment of the installed cable strands based on the positioning data of each benchmark target, and fits the actual alignment of the cable strands to be installed based on the positioning data of each monitoring target, and makes corrections based on temperature data to obtain the relative height difference between the installed and the cable strands to be installed. Finally, the relative height difference and the designed height difference guide the precise adjustment of the cable strands to be installed.

[0021] In one specific embodiment, acquiring measurement data of the reference target and the monitoring target measured by the measuring device specifically includes: A reference target is set on the installed cable strand, and a monitoring target is set on the cable strand to be installed. Both the reference target and the monitoring target are microprism retroreflective targets. The target body is a flexible magnetic structure (a layer of flexible magnetic layer is covered with a layer of microprism retroreflective layer, which facilitates the target to be recycled and reused). It is bent to fit the curved surface of the cable strand and is fixed to the outside of the cable strand by magnetic attraction. Several laser ranging and point cloud acquisition devices are deployed around the suspension bridge. Lasers are emitted to each benchmark target and each monitoring target through the laser ranging and point cloud acquisition devices. Based on the retroreflection characteristics of the microprism, the laser returns along the original path. The laser reflection point cloud data corresponding to each target is collected in real time. The three-dimensional coordinates of the laser ranging and point cloud acquisition devices themselves, the three-dimensional coordinates of the laser reflection point cloud of each target, and the reflected light intensity data are used as measurement data.

[0022] In a specific embodiment, the target positioning data is calculated, specifically including: Step 1: Calculate edge parameters for the collected laser reflection point cloud data, and determine the edge parameter threshold based on the differences in point cloud features between the target, cable, and background. Perform edge detection on the point cloud data based on the edge parameter threshold to extract the point cloud of the target region. Step 2: Based on the reflected light intensity data of the point cloud, perform light intensity filtering, denoising and normalization on the point cloud of the target area after edge detection, and remove points with abnormal light intensity and invalid point clouds. Step 3: Spatial interpolation is performed on the target region point cloud after removing abnormal and invalid point clouds. Spatial interpolation is performed by filling the missing point cloud with an interpolation algorithm based on the spatial distribution characteristics of the target region point cloud, so that the target region point cloud is restored to a complete and continuous distribution. Step 4: Based on the interpolated and completed 3D coordinates of the target area point cloud, calculate and determine the direction vector from the laser ranging and point cloud acquisition device to the corresponding target; the direction vector is a spatial vector with the coordinates of the laser ranging and point cloud acquisition device as the starting point and the coordinates of the centroid of the target area point cloud as the ending point. Step 5: Construct an input matrix using the interpolated target area point cloud 3D coordinates and reflected light intensity. Input the input matrix and direction vector into the pre-trained positioning and detection model, and output the single positioning data of the target relative to the laser ranging and point cloud acquisition device. Step 6: Summarize the single positioning data of the same target under all laser ranging and point cloud acquisition devices, and obtain the final positioning data of the target by fitting the data using the least squares method.

[0023] In one specific embodiment, edge parameter calculation includes: Based on the three-dimensional coordinates of the laser reflection point cloud and the reflected light intensity, a two-dimensional reflection intensity map is generated by projection. , , These are the coordinates of the two-dimensional reflectance intensity map, which corresponds to a unit circle. An interpolation algorithm is used to fill in missing regions. Edge parameters are calculated pixel-by-pixel based on improved Zernike moments, specifically: First, the two-dimensional reflectance intensity map of the target area. Step Tzernik Defined as: ; in, The radial distance is in polar coordinates. This represents the distance from the calculation point to the center of the unit circle; The polar coordinate azimuth angle represents the angle between the line connecting the calculation point and the center of the unit circle and the x-axis. Zernike kernel function The conjugate of complex numbers, , For Zernike radial polynomials, It is a natural constant. The imaginary unit; Then, taking advantage of the rotational invariance of Zernike moments, we obtain lower-order moments. , Calculate edge parameters: ; in, For background reflection intensity, The edge step intensity; Determining the threshold values ​​for edge parameters includes: Based on the differences in reflection intensity between the target, the cable strands, and the background, as well as geometric constraints, adaptive edge detection conditions are constructed: ;in, The threshold for edge step intensity. , The maximum reflection intensity of the point cloud in the target area; This is the reflection intensity range coefficient, with a value ranging from 15 to 30; To determine the background reflection intensity threshold, a binary search method is used for iterative optimization, specifically: First, calculate the average reflectance of the point cloud in the target area. Set the search range , This is a value used to adjust the search range; it is usually taken as... ; Then, the point clouds of the two arc-shaped boundary contours inside and outside the target are extracted, and the least squares method is used to perform ring curve fitting to construct the curve fitting loss function: ;in and These are the fitting residuals for the inner arc contour and the outer arc contour, respectively. The smaller the fitting residual, the more regular the edge contour extracted under the current threshold. Binary search within the interval, with loss function An intensity value less than the preset residual threshold is determined to be the optimal background reflection intensity threshold. .

[0024] In one specific embodiment, the localization detection model includes an input layer, an encoding layer, a feature alignment layer, and a decoding regression layer; The input layer receives and standardizes the input data, constructs a 4-channel matrix from the interpolated target point cloud 3D coordinates and reflected light intensity, and inputs it into the encoding layer. The encoding layer consists of three levels of sampling, grouping, and convolution. The first level performs FPS farthest point sampling, local neighborhood grouping, and extracts coarse geometric features of the target on the point cloud. The second level uses an MLP multilayer perceptron to fuse spatial coordinates and reflected light intensity features to enhance edge and centroid features. The third level performs downsampling and aggregates global features to output a global feature vector of the target. Specifically, the coding layer consists of three cascaded SA sub-modules, each of which adopts a unified sampling, grouping, and point cloud convolution structure, as follows: The first-level SA submodule is used for coarse feature extraction, and is executed sequentially. (1) FPS farthest point sampling: Select core sampling points from the input point cloud in a geometrically uniform manner to reduce point cloud redundancy and retain target contour features; (2) K-nearest neighbor local neighborhood grouping: Taking each sampling point as the center, search for the K points with the closest Euclidean distance to form a local neighborhood, thus forming a local receptive field; (3) Local point cloud convolution: Convolution calculation is performed on each local neighborhood using shared weight MLP to extract coarse geometric features such as target geometric position, curvature, and normal.

[0025] The second-level SA submodule is used for feature fusion.

[0026] (1) FPS downsampling: The farthest point of the first-level output feature points is sampled again to further compress the data volume; (2) Expanding neighborhood grouping: Using a larger radius neighborhood search to capture more global target structure information; (3) Joint feature convolution: The three-dimensional coordinate features and reflected light intensity features are concatenated and then input into the MLP convolution to output geometric and light intensity fusion features, which enhance the target edge and center features.

[0027] The third-level SA submodule is used for global feature aggregation.

[0028] (1) Minimize sampling: retain only a small number of key feature points and focus on the centroid region of the target; (2) Global neighborhood aggregation: Construct a global neighborhood with all feature points as a whole to realize global information association; (3) Max pooling and feature integration: Max pooling is performed on local convolutional features to extract the most representative high-dimensional semantic features and output the target global feature vector.

[0029] The feature alignment layer aligns the orientation vector with the global feature vector in spatial pose, eliminating viewpoint differences and obtaining aligned feature vectors. Specifically, the feature alignment layer normalizes the spatial pose of point cloud features and orientation vectors, eliminating positioning errors caused by different angles. First, the orientation vector is normalized to obtain the target's observation pose reference relative to the acquisition device. Then, a spatial rotation transformation matrix is ​​constructed using the orientation vector as a rotation reference. Next, the global feature vector output by the feature encoding layer is spatially rotated using the transformation matrix. Finally, the rotated global feature vector is concatenated and fused with the orientation vector to complete the pose alignment.

[0030] The decoding regression layer consists of a flattening layer and three fully connected layers. The flattening layer flattens the aligned high-dimensional feature vectors into a one-dimensional shape, transforming them into a continuous one-dimensional feature sequence that fits the input of the fully connected layers. Then, the three fully connected layers perform nonlinear mapping and position fitting on the flattened features to obtain the final three-dimensional coordinates. Specifically, the first fully connected layer enhances the representation of features related to the target centroid; the second fully connected layer compresses redundant features and strengthens key localization information; and the third fully connected layer maps the features to the three-dimensional spatial coordinate dimension.

[0031] The overall processing flow is as follows: the input matrix and direction vector are fed into the model; geometric and light intensity joint features are extracted through the feature encoding layer; the viewpoint deviation is eliminated through the feature alignment layer; and the three-dimensional coordinates of the target center are output through the decoding regression layer to obtain single positioning data.

[0032] In one specific embodiment, acquiring temperature data of installed and uninstalled cable strands includes: using a drone equipped with an infrared thermal imager to conduct a flight path inspection along the extension direction of the installed and uninstalled cable strands; continuously imaging the entire length of the cable strands with the infrared thermal imager; collecting infrared temperature information of each measuring point on the surface of the cable strands in real time; and obtaining a temperature data sequence of each point distributed along the length of the cable strands after temperature calculation and spatial registration.

[0033] In a specific embodiment, the actual line shapes of the installed strands and the strands to be installed are obtained by fitting the following method: The final positioning data of all reference targets on the installed cable strands are uniformly transformed to the overall coordinate system of the suspension bridge construction. Abnormal positioning points are eliminated and smoothing preprocessing is performed. Based on the catenary theory, a basic mathematical model of the installed cable strand alignment is constructed. The three-dimensional coordinates of the preprocessed reference targets are used as constraint points. The least squares method is used for nonlinear fitting to obtain the actual catenary alignment equation of the installed cable strands. The final positioning data of all monitoring targets on the cable strand to be installed are uniformly transformed to the overall coordinate system of the suspension bridge construction. Abnormal positioning points are eliminated and smoothing preprocessing is performed. Based on the catenary theory, a basic mathematical model of the cable strand to be installed is constructed. The three-dimensional coordinates of the preprocessed monitoring targets are used as constraint points. The least squares method is used for nonlinear fitting to obtain the actual catenary line shape equation of the cable strand to be installed.

[0034] Specifically, the equation for the catenary curve is expressed as: ;in, Let the x-axis and y-axis coordinates be taken with the mid-span as the origin. The weight per unit length of the strand. For horizontal tension of the cable, It is a hyperbolic cosine function.

[0035] In one specific embodiment, temperature data is used for correction to obtain the relative height difference between the installed strands and the strands to be installed, specifically including: Calculate the length of the strand to be installed based on the actual cable profile. The specific formula is as follows: ;in, For the length of the strand, The weight per unit length of the strand. For horizontal tension of the cable, It is a hyperbolic sine function. The span of the bridge; Based on the actual line shape of the cable strand to be installed, generate a coordinate data sequence that corresponds to the temperature data sequence (i.e., the coordinate data sequence is determined based on the sampling point position of the temperature data sequence); The temperature data sequences of the installed and uninstalled cable strands, the coordinate data sequence of the uninstalled cable strands, and the length of the uninstalled cable strands, the actual span of the bridge, and the self-weight per unit length of the cable strands are input into the correction model to obtain the corrected alignment of the uninstalled cable strands. Along the length of the cable strand, at preset intervals, extract the height difference between the corresponding coordinate points on the equation of the actual alignment of the installed cable strand and the corrected alignment of the cable strand to be installed, and use the height difference sequence as the relative height difference between the installed cable strand and the cable strand to be installed.

[0036] In one specific embodiment, the correction model is a time-series correction model based on an improved Transformer, designed for the temperature and coordinate data of the main cable strands to achieve high-precision correction of the measured coordinates. The model consists of a sequence input layer, an improved Transformer encoder, a static feature embedding layer, a feature fusion layer, and a regression output layer. The sequence input layer receives temperature data sequences and coordinate data sequences, performs data standardization, and outputs spatial feature sequences of fixed length and fixed dimensions. An improved Transformer encoder is constructed from multiple layers of lightweight coding units, including multi-head self-attention, layer normalization, feedforward network, and residual connection. It performs global correlation extraction on spatial feature sequences, extracts nonlinear correlation features between temperature and coordinates, and outputs global spatial features. The static feature embedding layer transforms three types of static structural parameters—the length of the cable strand to be installed, the actual span of the bridge, and the self-weight per unit length of the cable strand—into static embedding vectors of the same dimension as the global spatial features through standardization and linear mapping. The feature fusion layer concatenates global spatial features with static embedding vectors through channels to obtain fused features. The regression output layer, composed of multiple fully connected networks, obtains the formula parameters of the corrected line shape of the cable strand to be installed based on the fused feature mapping.

[0037] This invention also discloses a real-time cloud measurement system for the height difference of the main cable strands of a suspension bridge, comprising: Positioning module: acquires measurement data of the reference target and monitoring target measured by the measuring device, and calculates the positioning data of each target; Temperature detection module: acquires temperature data of installed and uninstalled cable strands; Transmission module: transmits positioning data and temperature data to the cloud processing module; Cloud processing module: Fits the actual line shape of the installed cable strands based on the positioning data of each benchmark target, fits the actual line shape of the cable strands to be installed based on the positioning data of each monitoring target, and makes corrections in conjunction with temperature data to obtain the relative height difference between the installed cable strands and the cable strands to be installed.

[0038] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0039] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for real-time cloud measurement of the height difference of the main cable strands of a suspension bridge, characterized in that, include: Acquire the measurement data of the reference target and monitoring target measured by the measuring device, and calculate the positioning data of each target; Obtain temperature data for installed and uninstalled cable strands; The location data and temperature data are transmitted to the cloud processing module; The cloud processing module fits the actual line shape of the installed cable strands based on the positioning data of each benchmark target, fits the actual line shape of the cable strands to be installed based on the positioning data of each monitoring target, and makes corrections based on the temperature data to obtain the relative height difference between the installed cable strands and the cable strands to be installed.

2. The method for real-time cloud measurement of the height difference of the main cable strands of a suspension bridge according to claim 1, characterized in that, The acquisition of measurement data from the reference target and monitoring target measured by the measuring device specifically includes: A reference target is set on the installed cable strand, and a monitoring target is set on the cable strand to be installed; both the reference target and the monitoring target are microprism retroreflective targets, and the target body is a flexible magnetic structure that is bent to fit the curved surface of the cable strand and is fixed to the outside of the cable strand by magnetic attraction. Several laser ranging and point cloud acquisition devices are deployed around the suspension bridge. The laser ranging and point cloud acquisition devices emit lasers to each of the reference targets and each of the monitoring targets, and collect the laser reflection point cloud data corresponding to each target in real time. The three-dimensional coordinates of the laser ranging and point cloud acquisition devices themselves, the three-dimensional coordinates of the laser reflection point cloud of each target, and the reflected light intensity data are used as measurement data.

3. The method for real-time cloud measurement of the height difference of the main cable strands of a suspension bridge according to claim 2, characterized in that, The calculation yields target positioning data for each target, specifically including: Step 1: Calculate edge parameters for the collected laser reflection point cloud data and determine the edge parameter threshold. Perform edge detection on the point cloud data based on the edge parameter threshold to extract the target region point cloud. Step 2: Based on the reflected light intensity data of the point cloud, perform light intensity filtering, denoising and normalization on the point cloud of the target area after edge detection, and remove points with abnormal light intensity and invalid point clouds. Step 3: Perform spatial interpolation to complete the target region point cloud after removing outliers and invalid point clouds; Step 4: Based on the interpolated and completed 3D coordinates of the target area point cloud, calculate and determine the direction vector from the laser ranging and point cloud acquisition equipment to the corresponding target; Step 5: Construct an input matrix using the interpolated target area point cloud 3D coordinates and reflected light intensity. Input the input matrix and the direction vector into the pre-trained positioning and detection model, and output the single positioning data of the target relative to the laser ranging and point cloud acquisition device. Step 6: Summarize the single positioning data of the same target under all laser ranging and point cloud acquisition devices, and obtain the final positioning data of the target by fitting the data using the least squares method.

4. The method for real-time cloud measurement of the height difference of the main cable strands of a suspension bridge according to claim 3, characterized in that, The edge parameter calculation includes: Based on the three-dimensional coordinates of the laser reflection point cloud and the reflected light intensity, a two-dimensional reflection intensity map is generated by projection. , , These are the coordinates of the two-dimensional reflectance intensity map; edge parameters are calculated pixel-by-pixel based on the improved Zernike moments, specifically: First, the two-dimensional reflectance intensity map of the target area. Step Tzernik Defined as: ； in, The radial distance is in polar coordinates. This represents the distance from the calculation point to the center of the unit circle; The polar coordinate azimuth angle represents the angle between the line connecting the calculation point and the center of the unit circle and the x-axis. Zernike kernel function The conjugate of complex numbers, , For Zernike radial polynomials, It is a natural constant. The imaginary unit; Then, taking advantage of the rotational invariance of Zernike moments, we obtain lower-order moments. , Calculate edge parameters: ； in, For background reflection intensity, The edge step intensity; The determination of the edge parameter threshold includes: Based on the differences in reflection intensity between the target, the cable strands, and the background, as well as geometric constraints, adaptive edge detection conditions are constructed: ;in, The threshold for edge step intensity. , The maximum reflection intensity of the point cloud in the target area; This is the reflection intensity range coefficient; To determine the background reflection intensity threshold, a binary search method is used for iterative optimization, specifically: First, calculate the average reflectance of the point cloud in the target area. Set the search range , This is a value used to adjust the search range. Then, the point clouds of the two arc-shaped boundary contours inside and outside the target are extracted, and the least squares method is used to perform ring curve fitting to construct the curve fitting loss function: ;in and These are the fitting residuals for the inner arc-shaped profile and the fitting residuals for the outer arc-shaped profile, respectively. Binary search within the interval, with loss function An intensity value less than the preset residual threshold is determined to be the background reflection intensity threshold. .

5. The method for real-time cloud measurement of the height difference of the main cable strands of a suspension bridge according to claim 3, characterized in that, The localization detection model includes an input layer, an encoding layer, a feature alignment layer, and a decoding regression layer; The input layer receives and standardizes the input data, constructs a 4-channel matrix from the interpolated target point cloud 3D coordinates and reflected light intensity, and inputs it into the encoding layer. The coding layer consists of three levels of sampling, grouping, and convolution. The first level performs FPS farthest point sampling, local neighborhood grouping, and extracts coarse geometric features of the target on the point cloud. The second level fuses spatial coordinates and reflected light intensity features through an MLP multilayer perceptron. The third level performs downsampling and aggregates global features to output a global feature vector of the target. The feature alignment layer aligns the direction vector with the global feature vector in spatial pose to eliminate viewpoint differences and obtain the aligned feature vector. The decoding regression layer consists of a flattening layer and three fully connected layers. The flattening layer flattens the aligned feature vectors in one dimension. Then, the three fully connected layers perform nonlinear mapping and position fitting on the flattened features to obtain the final three-dimensional coordinates.

6. The method for real-time cloud measurement of the height difference of the main cable strands of a suspension bridge according to claim 1, characterized in that, The acquisition of temperature data of installed and uninstalled cable strands specifically includes: using a drone equipped with an infrared thermal imager to conduct a flight path inspection along the extension direction of the installed and uninstalled cable strands, continuously imaging the entire length of the cable strands with the infrared thermal imager, collecting infrared temperature information of each measuring point on the surface of the cable strands in real time, and obtaining a temperature data sequence of each point distributed along the length of the cable strands.

7. The method for real-time cloud measurement of the height difference of the main cable strands of a suspension bridge according to claim 1, characterized in that, The actual line shapes of the installed strands and the strands to be installed are obtained by fitting the following method: The final positioning data of all reference targets on the installed cable strands are uniformly transformed to the overall coordinate system of the suspension bridge construction. Abnormal positioning points are eliminated and smoothing preprocessing is performed. Based on the catenary theory, a basic mathematical model of the installed cable strand alignment is constructed. The three-dimensional coordinates of the preprocessed reference targets are used as constraint points. The least squares method is used for nonlinear fitting to obtain the actual catenary alignment equation of the installed cable strands. The final positioning data of all monitoring targets on the cable strand to be installed are uniformly transformed to the overall coordinate system of the suspension bridge construction. Abnormal positioning points are eliminated and smoothing preprocessing is performed. Based on the catenary theory, a basic mathematical model of the cable strand to be installed is constructed. The three-dimensional coordinates of the preprocessed monitoring targets are used as constraint points. The least squares method is used for nonlinear fitting to obtain the actual catenary line shape equation of the cable strand to be installed.

8. A method for real-time cloud measurement of the height difference of the main cable strands of a suspension bridge according to claim 1, characterized in that, The step of correcting the temperature data to obtain the relative height difference between the installed cable strands and the cable strands to be installed specifically includes: Calculate the length of the strand to be installed based on the actual cable profile. The specific formula is as follows: ;in, For the length of the strand, The weight per unit length of the strand. For horizontal tension of the cable, It is a hyperbolic sine function. The span of the bridge; Based on the actual line shape of the cable strands to be installed, generate a coordinate data sequence corresponding to the temperature data sequence; The temperature data sequences of the installed and uninstalled cable strands, the coordinate data sequence of the uninstalled cable strands, and the length of the uninstalled cable strands, the actual span of the bridge, and the self-weight per unit length of the cable strands are input into the correction model to obtain the corrected alignment of the uninstalled cable strands. Along the length of the cable strand, at preset intervals, extract the height difference between the corresponding coordinate points on the equation of the actual alignment of the installed cable strand and the corrected alignment of the cable strand to be installed, and use the height difference sequence as the relative height difference between the installed cable strand and the cable strand to be installed.

9. A method for real-time cloud measurement of the height difference of the main cable strands of a suspension bridge according to claim 8, characterized in that, The corrected model is a temporal correction model based on Transformer improvement. The model consists of a sequence input layer, an improved Transformer encoder, a static feature embedding layer, a feature fusion layer, and a regression output layer. The sequence input layer receives the temperature data sequence and the coordinate data sequence, performs data standardization, and outputs a spatial feature sequence of fixed length and fixed dimensions. The improved Transformer encoder is composed of multiple layers of lightweight coding units stacked together. It performs global correlation extraction on spatial feature sequences and outputs global spatial features. The static feature embedding layer converts the length of the cable strand to be installed, the actual span of the bridge, and the self-weight per unit length of the cable strand into a static embedding vector through standardization and linear mapping. The feature fusion layer concatenates the global spatial features with the static embedding vector through channels to obtain fused features. The regression output layer consists of multiple fully connected networks, and the formula parameters for the corrected line type of the cable strand to be installed are obtained based on the fusion feature mapping.

10. A real-time cloud measurement system for the height difference of the main cable strands of a suspension bridge, employing the real-time cloud measurement method for the height difference of the main cable strands of a suspension bridge as described in any one of claims 1 to 9, characterized in that, include: Positioning module: acquires measurement data of the reference target and monitoring target measured by the measuring device, and calculates the positioning data of each target; Temperature detection module: acquires temperature data of installed and uninstalled cable strands; Transmission module: transmits the positioning data and temperature data to the cloud processing module; Cloud processing module: Fits the actual line shape of the installed cable strands based on the positioning data of each benchmark target, fits the actual line shape of the cable strands to be installed based on the positioning data of each monitoring target, and corrects it in combination with the temperature data to obtain the relative height difference between the installed cable strands and the cable strands to be installed.

Images