Highway engineering road and bridge measurement method based on BIM

Abstract

The invention discloses a highway engineering road and bridge measurement method based on BIM, and belongs to the technical field of highway engineering measurement. According to the method, the problems that the modeling efficiency and accuracy are difficult to consider, the modeling precision of a non-important area is redundant and the like in the existing road and bridge measurement method for the highway engineering are solved. The method comprises the following steps: acquiring satellite positioning data by using a satellite positioning system in a measurement area, acquiring point cloud data by using laser scanning equipment, and acquiring road and bridge images by using image acquisition equipment; converting satellite positioning data into unified coordinate system data information, carrying out spatial registration on the unified coordinate system data information and the point cloud data, and dividing an important region and a non-important region; constructing a precise three-dimensional point cloud model for the important area, constructing a path point cloud model for the non-important area, and integrating road and bridge images; and finally fusing the model to obtain road and bridge parameters. The method can improve the model efficiency and reduce the model resource overhead while effectively improving the measurement precision, and is suitable for the measurement work of roads and bridges of highway engineering in various environments.

Description

Technical Field

[0001] The present invention relates to the technical field of digital measurement in highway engineering, and particularly to a road and bridge measurement method for highway engineering based on BIM. Background Art

[0002] In the construction of highway engineering, road and bridge measurement is a crucial link, and the accuracy and integrity of its measurement results directly affect the quality, safety, and subsequent design, construction, and maintenance of the entire highway project. There are many problems with traditional road and bridge measurement methods in highway engineering.

[0003] Firstly, in terms of data acquisition, the previous measurement means were relatively single, and it was difficult to comprehensively obtain various information in the road and bridge area. For example, relying solely on traditional measurement instruments for on-site measurement, it was impossible to quickly and accurately obtain the topographic features of a large-area road and bridge area and the detailed information of the road and bridge structure, resulting in low data acquisition efficiency and easy omission.

[0004] Secondly, in terms of data processing and integration, data from different sources were often difficult to be effectively fused and analyzed due to problems such as inconsistent coordinate systems. Traditional methods lacked effective data conversion and registration means, resulting in errors between measurement data and unable to establish an accurate road and bridge model, thus affecting the accurate acquisition of road and bridge parameters.

[0005] Furthermore, for areas of different importance levels, traditional measurement methods did not have targeted processing strategies. Whether it was an important area or a non-important area, the same measurement and modeling methods were used, which not only wasted a large amount of time and resources, but also might lead to insufficient measurement accuracy in important areas, while non-important areas consumed excessive resources.

[0006] In addition, when constructing a point cloud model, traditional methods had limited capabilities in noise reduction, processing of point cloud data, and processing of areas with different densities, and were unable to generate a high-quality point cloud model. In terms of extracting path feature points and integrating road and bridge images, traditional methods also had technical defects, resulting in inaccurate and incomplete road and bridge parameters finally obtained. In terms of model fusion and parameter acquisition, the accuracy and efficiency of traditional methods were difficult to meet the requirements of modern highway engineering construction. Summary of the Invention

[0007] The present invention overcomes the problems of traditional measurement methods, such as the need for manual registration of multi-source data, the model accuracy being limited by the scanning density, it being difficult to balance the modeling efficiency and accuracy, and the modeling accuracy redundancy in non-important areas. It provides a BIM-based road and bridge measurement method for highway engineering. Data is obtained through satellite positioning, laser scanning, and image acquisition. After unifying the coordinate system, important and non-important areas are divided, and a precise three-dimensional point cloud model and a path point cloud model are constructed and fused respectively. Through zoning modeling, while ensuring the modeling accuracy of important areas, the data processing volume of non-important areas is reduced, and high-precision fusion of the model is achieved through methods such as regional unit division.

[0008] To achieve the above object, the present invention adopts the following solutions: A BIM-based road and bridge measurement method for highway engineering, which includes the following steps: S1: In the measurement area, satellite positioning data of the road and bridge area of highway engineering is obtained through the satellite positioning system. A laser scanning device is used to scan the road and bridge area to obtain point cloud data, and an image acquisition device is used to collect road and bridge images in the road and bridge area; S2: The satellite positioning data is converted into data information in a unified coordinate system, the satellite positioning data and the point cloud data are spatially registered, and the measurement area is divided into an important area and a non-important area according to the importance of the road and bridge of highway engineering; S3: For the important area, the precise three-dimensional point cloud model of the important area is constructed using the point cloud data; S4: For the non-important area, path feature points for characterizing the road and bridge extension path are extracted from the point cloud data to construct a path point cloud model. A distance step length is set, and the road and bridge extension path is divided into regional units according to the step length; The corresponding number of road and bridge images is selected according to the length of the unit and integrated into the regional unit; S5: The precise three-dimensional point cloud model and the path point cloud model are fused to obtain a road and bridge data model for highway engineering, and road and bridge parameters are obtained according to the road and bridge data model for highway engineering.

[0009] Preferably, in the step S2, the following method is used to convert the satellite positioning data into data information in a unified coordinate system: At least three ground reference points are selected and the high-precision satellite positioning coordinates of these ground reference points in the WGS84 coordinate system are obtained. The coordinates of the ground reference points are converted into coordinates in the plane rectangular coordinate system through the Gauss-Kruger projection; The method of selecting ground reference points is: Obtain the key nodes of the road and bridge structure through the road and bridge structure design data. Centering on the key structure nodes, select multiple evenly distributed ground reference points within a predetermined range. Use a total station to verify the coordinates of the selected ground reference points through secondary measurement, and perform real-time kinematic differential correction on the coordinates of the ground reference points by using a dual-frequency GNSS receiver.

[0010] Preferably, in step S2, the satellite positioning data and the point cloud data are spatially registered by using the weighted average method, which specifically includes the following steps: Extract the point cloud coordinates corresponding to each ground reference point from the point cloud data through a feature matching algorithm. Set the allowable error range, and perform iterative calculation and correction on the point cloud data corresponding to the point cloud coordinates until the registration error of the point cloud data falls within the allowable error range, and complete the matching between the point cloud data and the ground reference points; For each ground reference point, calculate the satellite positioning weight w according to the real-time signal-to-noise ratio of the dual-frequency GNSS receiver GNSS , calculate the point cloud weight w according to the local density and curvature stability of the point cloud in the area where the ground reference point is located Cloud , calculate the comprehensive weight , and perform normalization processing on the comprehensive weight; Set a rigid transformation matrix to minimize the residual before and after the weighted change of the point cloud, that is: where R and T are the rotation matrix and translation vector in the rigid transformation matrix respectively, w i is the comprehensive weight of the i-th ground reference point, n is the number of ground reference points, q i is the point cloud data of the ground reference point to be transformed, p i is the reference point cloud data after transformation; Calculate the optimal rotation matrix R and translation vector T to minimize the residual through the method of singular value decomposition; weight the point cloud data in the area where each ground reference point is located according to the rotation matrix R and translation vector T respectively, and complete the spatial registration of the point cloud data.

[0011] Preferably, in step S3, when constructing a precise three-dimensional point cloud model, a curvature analysis algorithm is used to perform noise reduction processing on the point cloud data, and a clustering algorithm is used for the high-density area and the scanning blind area holes are filled in the low-density area, including the following steps: For each point c in the point cloud data, search for the neighboring point set within the range of its neighborhood radius r, and use the PCA principal component analysis method to calculate the normal vector and curvature corresponding to the point c. If the calculated curvature is greater than the set curvature threshold, mark this point c as a noise point to be processed, otherwise mark it as a non-noise point; For non-noise points, perform smoothing processing by using a bilateral filtering algorithm; For a noise point, if the non-noise points in the point cloud data within the range of its neighborhood radius r exceed 70%, then update the coordinates of this noise point to the weighted average of the non-noise points in the neighborhood; otherwise, directly delete this noise point. Divide the point cloud data into voxel grids, count the number of point clouds in each voxel and calculate the point cloud data density. If the point cloud data density of a voxel is less than the set density threshold, then divide this voxel into a low-density area; otherwise, divide it into a high-density area. For the high-density area, use the density clustering algorithm to obtain representative core points within the area. For the low-density area, extract the set of intersection points with the high-density area. Use the Poisson surface reconstruction algorithm for the set of intersection points to generate a local surface, and sample on the local surface to obtain a set of filling points for the low-density area; for each filling point in the set of filling points, calculate its normal vector and perform consistency optimization with the normal vectors of its neighboring points.

[0012] Preferably, in step S4, the extraction of path feature points adopts an improved Douglas-Peucker algorithm. Set the curvature change threshold parameter of the algorithm and extract the curvature extreme points in the road and bridge extension direction, and combine the analysis of the point cloud normal vector to determine the center line of the path.

[0013] Preferably, in step S4, the setting of the distance step size adopts a dynamic adjustment mechanism: If the path curvature radius 100m ≤ R ≤ 300m, then the step size is set to 1m; if the path curvature radius R < 100m, then shorten the step size to 0.5m; if R > 300m, then extend the step size to 2m; and construct an exponential correlation function between the step size and the terrain complexity of the road and bridge according to the set step size.

[0014] Preferably, in step S4, the step of integrating the road and bridge image into the regional unit includes: Perform SIFT feature point matching on the collected road and bridge images, establish the spatial projection relationship of adjacent images, and use the perspective transformation matrix to map the two-dimensional image to the three-dimensional point cloud surface to generate a texture mapping model.

[0015] Preferably, when integrating the road and bridge image into the regional unit in step S4, dynamically adjust the integration quantity of the road and bridge image according to the topological distance between the regional unit and the important area, and screen out the optimal image set for integration, which specifically includes the following steps: Calculate the number of images N j to be allocated for the regional unit U j : where N max is the set maximum number of images, N base is the basic number of images, and dj is the minimum Euclidean distance from the regional unit U j to the nearest important regional boundary, d0 is the set distance threshold, η is the attenuation factor, and , where η0 is the basic attenuation factor, p r is the adjustment coefficient, and R is the curvature radius of the road and bridge; Within the regional unit, for the candidate image set I, calculate the priority score of the image: where S k is the priority score of the k-th image I k , SSIM(I k ) is the structural similarity index of the image I k , α is the clarity influence coefficient, SNR(I k ) is the signal-to-noise ratio of the image I k , β is the uniformity influence coefficient, M k is the feature matching rate between the image I k and its adjacent images, and γ is the matching rate influence coefficient; Arrange the images in the candidate image set I in descending order of the priority score, select the top N images and integrate them into the corresponding regional unit, and use the adaptive Gaussian weighting method to fuse the image overlapping areas of adjacent regional units.

[0016] Preferably, in the step S5, the model fusion adopts a multi-resolution hierarchical registration method. First, rough registration is achieved through the ICP algorithm, then the transformation matrix is optimized by the RANSAC algorithm, and the nearest neighbor matching based on the kd-tree is implemented for the point cloud in the overlapping area.

[0017] Preferably, in the step S5, the method for obtaining the road and bridge parameters includes: generating the road cross-section parameters based on the point cloud slicing technology, calculating the structural deformation amount through model comparison, and outputting the BIM component attribute report according to the IFC standard requirements. The key parameters in the report include elevation deviation, flatness index, and curvature continuity analysis results.

[0018] The present invention has at least the following beneficial effects: (1) Through the collaborative acquisition of satellite positioning data, laser point cloud, and image data, using a hierarchical modeling strategy, combining the precise three-dimensional point cloud of important areas with the path point cloud and image integration method of unimportant areas, while ensuring the accuracy of key structures, reducing data redundancy in non-critical areas, significantly improving the modeling efficiency, and reducing storage and computing costs; (2) Through multi-stage verification of ground reference points, ensuring the reliability of coordinate system conversion; adopting a weighted registration algorithm and rigid transformation matrix optimization to solve the problem of spatial alignment of multi-source data, reducing registration errors, and improving the overall consistency of the model; (3) Noise filtering and hole filling based on curvature threshold and density analysis, extracting path features, enhancing the integrity of the point cloud model and the accuracy of the center line, and being applicable to complex terrains and structures; (4) Through the dynamic step size adjustment driven by the curvature radius and the image quantity allocation based on topological distance, realizing the adaptive data acquisition density, balancing details and efficiency; adopting a priority scoring mechanism to ensure the high-quality fusion of texture mapping, and improving the visualization and information richness of the model; (5) Ensuring the model fusion accuracy through multi-resolution hierarchical registration, and generating parametric reports based on the IFC standard, directly supporting BIM collaborative design and construction acceptance, and improving the compatibility and practicality of engineering data. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 FIG. is a principle flow chart of a road and bridge measurement method for highway engineering based on BIM provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] The following further describes the present invention in detail with reference to the drawings, so that those skilled in the art can implement it according to the description in the specification.

[0021] As Figure 1 shown, the road and bridge measurement method for highway engineering based on BIM provided by the present invention includes the following steps: S1: In the measurement area, obtain satellite positioning data of the road and bridge area of the highway engineering through a satellite positioning system, scan the road and bridge area with a laser scanning device to obtain point cloud data, and collect road and bridge images in the road and bridge area with an image acquisition device.

[0022] By accessing the satellite positioning system platform, the required satellite positioning data can be quickly obtained. Through the GPS or Beidou system, the satellite positioning data tags the road and bridge area with geographical information, recording topographical information such as the longitude, latitude, and altitude of each key point in the road and bridge area. Laser scanning equipment is a common device for obtaining point cloud data. By quickly scanning the surface of the road and bridge area with lidar, a "digital contour" composed of millions of 3D points is quickly generated. Information such as each crack in the bridge and the unevenness of the road surface can be obtained and recorded in real time. Image acquisition is carried out by using on-vehicle or airborne photography equipment to photograph the road and bridge area, forming clear images, which can be used to supplement the texture details of the road and bridge area, or to extract information such as the contour dimensions of the road and bridge through image processing. And by viewing the image data, the visual information of the road and bridge area can be intuitively understood. The satellite positioning data can determine the global position, the laser scanning can provide the fine structure of the road and bridge, and the image can supplement the visual information. The combination of the three ensures comprehensive and accurate data, and the constructed model is both accurate and realistic.

[0023] S2: Convert the satellite positioning data into data information in a unified coordinate system, perform spatial registration on the satellite positioning data and the point cloud data, and divide the measurement area into important areas and non-important areas according to the importance of the road and bridge in highway engineering.

[0024] Satellite positioning data is usually data in the global coordinate system, and there are deviations in the data of different positioning systems. Therefore, when using it, the obtained satellite positioning data is converted into the local coordinate system required in road and bridge engineering. For example, data reorganization and offset correction are carried out with a certain pier as the origin to ensure that all data can be used collaboratively in one system. Spatial registration of the point cloud and satellite positioning data to align their benchmarks is to place the point cloud data in the correct position in the coordinate system by dividing it into blocks by region. For example, the scanned bridge deck point cloud needs to match the bridge deck coordinates measured by GPS. The measurement area in highway engineering is usually set according to the planning and blueprint of road and bridge design, and usually needs to cover the positions of all road and bridge paths, road facilities and surrounding buildings. According to the structural design of the road and bridge, important areas and non-important areas can be quickly determined. Key parts with complex structures, high traffic flow and high safety requirements can be designated as important areas according to the design blueprint and statistical traffic flow, such as bridge bearings, tunnel entrances, transportation hubs, etc. High-precision modeling is required for these positions to obtain more accurate data to ensure the quality of the road and bridge at these key positions can be monitored in real time. For the middle section of a straight and extended road section, the road surface green belt area, and road sections with low traffic flow and good surrounding environment, these areas are designated as non-important areas and do not require very fine modeling operations. Only the road conditions need to be distinguished by combining simple path contours and image information, which can avoid data redundancy in the refined model and save the cost of model resources. The division of important areas and non-important areas is achieved by setting label parameters for key areas such as traffic intersections, bridges, and high-traffic sections in the software, and adding a method to distinguish the importance level according to the label parameters in the software to automatically identify and divide important areas and non-important areas. The system can also manually adjust and mark the area boundaries according to the drawing design and experience judgment. During the use of the model, the system can also convert non-important areas with high viewing frequencies into important areas, or automatically shrink and expand the area boundaries by recording the viewing times of each area.

[0025] S3: For important areas, use the point cloud data to construct a precise three-dimensional point cloud model of the important areas.

[0026] For important areas, all the collected point cloud data is used for modeling, and data processing operations such as denoising and splicing are performed on the dense point cloud data to construct a 3D model with extremely high reduction. In practice, the setting can reach millimeter-level accuracy. For important areas, problems that may occur in the road and bridge structure, such as the width and depth of cracks in bridge roads, can be accurately presented. Real-time data of the road and bridge project can be measured with an accurate model to timely discover safety problems and existing hidden dangers. During the model construction process, through BIM integrated design, such as importing the model into common BIM software such as Revit, adding material properties and structural information, an intelligent model for mechanical analysis can also be obtained.

[0027] S4: For non-critical areas, extract path feature points from the point cloud data to construct a path point cloud model for characterizing the extension path of the road and bridge. Set a distance step size and divide the extension path of the road and bridge into regional units according to the step size. Integrate the corresponding number of road and bridge images into the regional units according to the length of the unit.

[0028] By extracting path feature points, key points representing information such as the road alignment and width (such as the center line and edge line of the road) are captured from the point cloud and connected into the skeleton line of the path to represent the extension direction and edge contour. When dividing the regional units, the road is divided into small segments according to the set step size (such as every 20 meters). The shorter the step size, the finer the model, but the greater the computational amount. Therefore, a reasonable step size should be set in combination with the complexity of the actual environment. Image integration matches photos at the corresponding positions for each unit. For example, the image of a certain section of the road surface will be "pasted" onto the model to endow the model with real textures, similar to the texture mapping of a game map, and the actual image information can be viewed. Further division processing can also be carried out in non-critical areas. For example, if the regional unit is close to the critical area, the division step size can be appropriately reduced, and the number and clarity of the images used for image integration can be increased.

[0029] S5: Integrate the precise three-dimensional point cloud model and the path point cloud model to obtain the road and bridge data model for highway engineering, and obtain the road and bridge parameters according to the road and bridge data model for highway engineering.

[0030] Existing mature BIM software usually provides model integration and splicing functions. By seamlessly splicing the high-precision model of the critical area and the simplified model of the non-critical area in the BIM platform, a complete road and bridge model is formed. When viewing the model according to requirements, if viewing the information of the critical area, a high-precision three-dimensional model is presented, and the required road and bridge feature information can be quickly calculated and analyzed; if viewing the information of the non-critical area, the basic dimension data and image information of the road can be known through the simplified model. The road and bridge parameters that can be obtained by viewing and automatically analyzing the model data through BIM software include road and bridge geometric parameters, engineering quantities, construction simulation parameters, etc. Among them, the geometric parameters can further include road slope, turning radius, bridge span, etc., the engineering quantities can further include concrete consumption, earthwork excavation volume, etc., and the construction simulation parameters can further include detecting whether the construction equipment will cause collision damage to the bridge structure, etc.

[0031] As an example, for satellite positioning data acquisition, a dual-frequency GNSS receiver (such as Trimble R12) is used to collect coordinates at a frequency of 1 Hz, and the single-point positioning accuracy is ±10 mm. For laser scanning equipment, a terrestrial three-dimensional laser scanner (such as Faro Focus S350) is used, and the scanning resolution is set to 5 mm@10 m, that is, the scanning resolution at a position 10 meters away from the scanning equipment can reach 5 mm. For image acquisition, a drone is equipped with a 5K resolution camera (such as Sony A7R IV), and the shooting interval can be dynamically adjusted according to the set step size. Four reference points are arranged within a range of 10 m around key nodes such as bridge bearings and expansion joints to form a regular quadrilateral. Projection conversion is carried out through the Gaussian projection module of the CGCS2000 coordinate system, and the accuracy of the central meridian is set to the average longitude of the project area. For important areas, modeling is carried out, such as using a 0.5 mm voxel grid to reduce noise for bridge piers and abutments, and the Poisson surface reconstruction depth is set to 10. For unimportant areas, when extracting path feature points, the curvature threshold is set to 0.05, and the step size is dynamically adjusted according to the radius of curvature. When fusing models, after ICP coarse registration, RANSAC is used to eliminate mismatched points, and the final registration residual is controlled to be ≤3 mm.

[0032] This method synergistically collects satellite positioning data, laser point clouds, and image data, uses a hierarchical modeling strategy, combines precise three-dimensional point clouds in important areas with path point clouds and image integration methods in unimportant areas, reduces data redundancy in non-critical areas while ensuring the accuracy of key structures, significantly improves modeling efficiency, and reduces storage and computing costs. The accuracy of model fusion is ensured through multi-resolution hierarchical registration, and parameterized reports are generated based on the IFC standard, directly supporting BIM collaborative design and construction acceptance, and improving the compatibility and practicality of engineering data.

[0033] In another technical solution, in step S2, the following method is used to convert the satellite positioning data into data information in a unified coordinate system: Select at least three ground reference points and obtain the high-precision satellite positioning coordinates of these ground reference points in the WGS84 coordinate system, and convert the coordinates of the ground reference points into coordinates in the plane rectangular coordinate system through the Gauss-Krüger projection. The method of selecting ground reference points is as follows: Obtain the key nodes of the road and bridge structure through the road and bridge structure design data. Centered on the structural key nodes, select multiple evenly distributed ground reference points within a predetermined range, use a total station to measure the coordinates of the selected ground reference points for verification, and perform real-time kinematic differential correction on the coordinates of the ground reference points using a dual-frequency GNSS receiver.

[0034] Considering that the influence of the uniformity of fiducial point distribution on the registration error will further affect the global accuracy of the spatial transformation, if the fiducial points are concentrated in a local area, it may lead to a significant increase in the registration error at positions far from this area. If the fiducial points do not cover the geometric center or edge of the measurement area, it will cause a systematic deviation of the registration model in the uncovered area. Provide an optional optimization scheme to ensure that the uniformity of fiducial point distribution meets the requirements. Use the uniformity evaluation based on Thiessen polygons, including the following steps: Initial fiducial point selection: Centered on the key structural nodes, randomly select candidate fiducial points within a predetermined radius range; Construct Thiessen polygons: Generate Thiessen polygons from the candidate fiducial points, and each polygon represents the coverage area of a fiducial point; Uniformity quantification: Calculate the standard deviation of the areas of all Thiessen polygons. The smaller the standard deviation, the more uniform the distribution; Dynamic adjustment: If the standard deviation exceeds the threshold (such as an area difference > 20%), by adding or moving fiducial points (such as supplementing fiducial points to the center of large-area polygons), regenerate the Thiessen polygons until the uniformity requirement is met. Through geometric coverage analysis, the distribution uniformity can be visually quantified and the deviation can be effectively reduced.

[0035] After the fiducial points are selected, use a total station for secondary measurement verification to ensure that the coordinate error of the fiducial points ≤ 2 mm. Through multi-stage verification of the ground fiducial points, ensure the reliability of the coordinate system transformation; adopt a weighted registration algorithm and rigid transformation matrix optimization to solve the problem of spatial alignment of multi-source data, reduce the registration error, and improve the overall consistency of the model.

[0036] In step S2, the satellite positioning data and the point cloud data are spatially registered using the weighted average method, specifically including the following steps: Extract the point cloud coordinates corresponding to each ground fiducial point from the point cloud data through the feature matching algorithm, set the allowable error range, and perform iterative calculation and correction on the point cloud data corresponding to the point cloud coordinates until the registration error of the point cloud data falls within the allowable error range to complete the matching of the point cloud data and the ground fiducial points; For each ground fiducial point, calculate the satellite positioning weight w according to the real-time signal-to-noise ratio of the dual-frequency GNSS receiver GNSS , calculate the point cloud weight w according to the local density and curvature stability of the point cloud in the area where the ground fiducial point is located Cloud , calculate the comprehensive weight , perform normalization processing on the comprehensive weight; As an example: Satellite positioning weight: Dynamically calculated according to the signal-to-noise ratio (SNR) of the dual-frequency GNSS receiver. If the SNR range is 30 - 80 dB·Hz, the calculation formula used is , if the SNR is less than 30 dB·Hz, then w GNSS = 1.

[0037] Point cloud weight: Based on the local point cloud density (number of points per unit volume) and curvature stability (reciprocal of curvature variance), the formula is w Cloud = 0.6×A + 0.4×B, where A is the density normalization value and B is the curvature stability normalization value.

[0038] Set the rigid transformation matrix to minimize the residual before and after the weighted change of the point cloud, that is: where R and T are the rotation matrix and translation vector in the rigid transformation matrix respectively, w i is the comprehensive weight of the i-th ground reference point, n is the number of ground reference points, q i is the point cloud data of the ground reference point to be transformed, p i is the reference point cloud data after transformation; Calculate the optimal rotation matrix R and translation vector T by the method of singular value decomposition to minimize the residual; weight the point cloud data in the region where each ground reference point is located according to the rotation matrix R and translation vector T respectively to complete the spatial registration of the point cloud data.

[0039] Use singular value decomposition (SVD) to calculate the rotation matrix R and translation vector T. The specific steps include: Calculate the centroids of the source point cloud and the target point cloud and , using the formula: , ; Construct the covariance matrix H, using the formula: ; Perform SVD decomposition on the covariance matrix, , to obtain the rotation matrix ; here Σ is the singular value matrix, and U and V are orthogonal matrices; Calculate the translation vector .

[0040] The iteration termination condition for iterative calculation and correction of the point cloud data corresponding to the point cloud coordinates can be set as: the maximum number of iterations is 50, or the registration error ϵ i < 2 mm, .

[0041] In another technical solution, in step S3, when constructing a precise three-dimensional point cloud model, a curvature analysis algorithm is used to perform noise reduction processing on the point cloud data, and a clustering algorithm is used for the high-density area and the scanning blind area holes are filled in the low-density area, including the following steps: For each point c in the point cloud data, search for the neighboring point set within the range of its neighborhood radius r, and use the PCA principal component analysis method to calculate the normal vector and curvature corresponding to point c. If the calculated curvature is greater than the set curvature threshold, mark this point c as a noise point to be processed, otherwise mark it as a non-noise point; For non-noise points, use the bilateral filtering algorithm for smoothing; For noise points, if the non-noise points in the point cloud data within the range of its neighborhood radius r exceed 70%, update the coordinates of this noise point to the weighted average of the non-noise points in the neighborhood, otherwise directly delete this noise point; After deleting the noise points, if local holes are generated at this position, fill them by using the method of local surface fitting. Detect holes in the point cloud to identify the hole area and extract the boundary points of the hole. Select polynomial fitting, radial basis function interpolation or Poisson surface reconstruction, etc. for local surface fitting, and obtain the filling points by sampling. For scenarios with high real-time requirements, a filling method based on neighborhood point cloud expansion can also be used. First, expand the hole area, and then use methods such as linear interpolation, Kriging interpolation, and moving least squares for point cloud interpolation. Considering the model calculation cost and the importance of data at the hole, it is not recommended to use the method of training a deep learning model to predict the filling points.

[0042] Divide the point cloud data into voxel grids, count the number of point clouds in each voxel and calculate the point cloud data density. If the point cloud data density of a voxel is less than the set density threshold, divide this voxel into a low-density area, otherwise divide it into a high-density area; For the high-density area, use the density clustering algorithm to obtain the representative core points in the area; the density clustering algorithm used can be the DBSCAN algorithm. Set the neighborhood radius to 0.2m and the minimum number of points to 10. Divide the density-connected points into the same cluster and mark it as the effective point cloud area. Determine the unclustered points as noise points or blind area boundary points.

[0043] For the low-density area, extract the set of intersection points with the high-density area, use the Poisson surface reconstruction algorithm to generate a local surface for the intersection point set, and sample on the local surface to obtain the filling point set for the low-density area; for each filling point in the filling point set, calculate its normal vector and optimize its consistency with the normal vectors of its neighboring points. Check the density of the newly generated point cloud to ensure that its density is consistent with the surrounding area.

[0044] Noise filtering and hole filling based on curvature threshold and density analysis, extract path features, enhance the integrity of the point cloud model and the accuracy of the centerline, and can be applied to complex terrains and structures.

[0045] In another technical solution, in step S4, the path feature points are extracted using an improved Douglas-Peucker algorithm. The curvature change threshold parameter of the algorithm is set, and the curvature extreme points in the extending direction of the road and bridge are extracted. The center line of the path is determined by combining the analysis of the point cloud normal vector.

[0046] In the improved Douglas-Peucker algorithm used, the initial path point set is the two-dimensional polyline obtained by projecting the point cloud onto the horizontal plane, and the curvature change threshold is set to 0.03. The determination method of the curvature extreme points is as follows: for each point, calculate the average curvature difference between it and the 5 points before and after it. If the difference exceeds the threshold, the point is retained.

[0047] The method for determining the center line of the path is as follows: perform normal vector analysis on the path point cloud, screen out the points whose included angle between the normal vector and the horizontal plane is less than 5°, and fit a cubic B-spline curve as the center line.

[0048] In step S4, a dynamic adjustment mechanism is adopted for setting the distance step size: If the path curvature radius 100m ≤ R ≤ 300m, the step size is set to 1m; if the path curvature radius R < 100m, the step size is shortened to 0.5m; if R > 300m, the step size is extended to 2m; and an exponential correlation function between the step size and the terrain complexity of the road and bridge is constructed according to the set step size.

[0049] An example of the exponential correlation function: introduce a step size correction factor , and the final step size , where L0 is the basic step size set by the initial judgment.

[0050] In another technical solution, in step S4, the steps of integrating the road and bridge image into the regional unit include: Perform SIFT feature point matching on the collected road and bridge images, establish the spatial projection relationship between adjacent images, map the two-dimensional image to the surface of the three-dimensional point cloud using a perspective transformation matrix, and generate a texture mapping model. As an example, for SIFT feature matching, the OpenCV library is used to extract SIFT features, and the matching threshold is set to 0.7 (Lowe's ratio test). The perspective transformation matrix is calculated by solving the homography matrix H through the RANSAC algorithm, the inlier threshold is set to 2 pixels, and the maximum number of iterations is 1000 times. When generating the texture mapping model, after projecting the image onto the surface of the point cloud, the UV unwrapping algorithm is used to generate texture coordinates, and the basic setting of the texture resolution is 4096×4096 pixels and the resolution is adjusted according to the adaptability of the regional unit.

[0051] In step S4, when integrating the road and bridge image into the regional unit, the integration quantity of the road and bridge image is dynamically adjusted according to the topological distance between the regional unit and the important area, and the optimal image set for integration is screened out. The specific steps are as follows: Calculate the number of images N to be allocated for the regional unit U j : j : where N max is the set maximum number of images, N base is the base number of images, d j is the minimum Euclidean distance from the regional unit U j to the nearest important regional boundary, d0 is the set distance threshold, η is the attenuation factor, and , where η0 is the basic attenuation factor, p r is the adjustment coefficient, R is the road and bridge curvature radius; an example of a set of parameter combinations: N max = 10, N base = 3, d0 = 20m, η0 = 0.5, p r = 0.1.

[0052] Within the regional unit, for the candidate image set I, calculate the priority score of the images: where S k is the priority score of the k-th image I k , SSIM(I k ) is the structural similarity index of the image I k , α is the sharpness influence coefficient, SNR(I k ) is the signal-to-noise ratio of the image I k , β is the uniformity influence coefficient, M k is the feature matching rate of the image I k with its adjacent images, γ is the matching rate influence coefficient; an example of a set of coefficient combinations: α = 0.4, β = 0.3, γ = 0.3. The SSIM calculation window size is 11×11 pixels, and the signal-to-noise ratio (SNR) is evaluated through the image gray variance.

[0053] Arrange the images in the candidate image set I in descending order of the priority score, select the top N images and integrate them into the corresponding regional unit, and use the adaptive Gaussian weighting method to fuse the image overlapping regions of adjacent regional units. The standard deviation σ of the parameter setting of the adaptive Gaussian weighting method is 0.5×W, where W is the overlapping width.

[0054] Achieve adaptive data acquisition density through the dynamic step adjustment driven by the curvature radius and the image number allocation based on the topological distance, and balance details and efficiency.

[0055] In another technical solution, in step S5, the model fusion adopts a multi-resolution hierarchical registration method. First, rough registration is achieved through the ICP algorithm, and then the RANSAC algorithm is used to optimize the transformation matrix, and the nearest neighbor matching based on the kd-tree is implemented for the point cloud in the overlapping area. The maximum number of iterations of the ICP algorithm is 100 times, the matching distance threshold is 5 cm, and the convergence residual is <1 cm. The number of samplings of the RANSAC algorithm is 500 times, the inlier determination threshold is 3 mm, and finally the transformation matrix with an inlier rate >90% is retained. When constructing the kd-tree, the maximum number of points in the leaf node is 50, and the search radius is 10 cm.

[0056] In step S5, the method for obtaining the road and bridge parameters includes: generating road cross-section parameters based on the point cloud slicing technology, calculating the structural deformation amount through model comparison, and outputting a BIM component attribute report according to the IFC standard requirements. The key parameters in the report include elevation deviation, flatness index, and curvature continuity analysis results. The point cloud slicing technology generates cross-sections every 1 m along the road centerline, the slicing thickness is 2 cm, and parameters such as elevation and width are extracted. Calculating the structural deformation amount is achieved by comparing the design model and the measured model, and calculating the Hausdorff distance between the point clouds after ICP registration, with a threshold of 10 mm. The mapping parameters include component ID, material properties, and geometric deviation, and the output format conforms to the IFC4 standard.

[0057] As an example in specific use: In an actual road and bridge survey project of a highway, it is necessary to survey a section of highway with a length of 5 kilometers, including multiple bridges and curves.

[0058] First, determine the survey area and prepare devices such as high-precision GPS receivers, 3D laser scanners, and high-definition cameras. Use the GPS receiver to stably obtain satellite positioning data within the survey area. Use the 3D laser scanner to set the scanning range to 50 meters on each side centered on the road centerline, with a resolution of 0.05 meters, and conduct a comprehensive scan of the road and bridge area to obtain point cloud data. According to the pre-established shooting plan, use the high-definition camera to take pictures at different positions and angles in the road and bridge area to collect road and bridge images.

[0059] After obtaining the satellite positioning data, according to the road and bridge structure design drawings, find key nodes such as bridge piers and abutments, and evenly select 5 ground reference points within a radius of 40 meters centered on these nodes. Use a total station to accurately measure these reference points, and then use a dual-frequency GNSS receiver for real-time kinematic differential correction. Convert the coordinates of the ground reference points in the WGS84 coordinate system to the coordinates in the plane rectangular coordinate system through the Gauss-Krüger projection to complete the conversion of satellite positioning data.

[0060] The weighted average method is used for the spatial registration of satellite positioning data and point cloud data. The point cloud coordinates are extracted through the feature matching algorithm, and the allowable error range is set to 0.04 meters for iterative calculation and correction. The weights are calculated based on the real-time signal-to-noise ratio of the dual-frequency GNSS receiver, the local density of the point cloud, and the curvature stability. The rigid transformation matrix is set, and the optimal rotation matrix R and translation vector T are calculated through singular value decomposition to complete the spatial registration of the point cloud data.

[0061] According to the importance of the road and bridge in highway engineering, the key parts, curves, etc. of the bridge are divided into important areas, and other areas are divided into non-important areas. For important areas, a precise three-dimensional point cloud model is constructed using point cloud data. The curvature analysis algorithm is used to denoise the point cloud data, the voxel grid is divided, the clustering algorithm is used for high-density areas, and the scanning blind area holes are filled for low-density areas. For non-important areas, the improved Douglas-Peucker algorithm is used to extract path feature points, the curvature change threshold parameter is set to 0.012, and the path centerline is determined by combining the point cloud normal vector analysis. The distance step size is set and dynamically adjusted according to the path curvature radius. For example, in a section with a curvature radius of 200m, the step size is set to 1m; at a curve with a curvature radius of 80m, the step size is shortened to 0.5m.

[0062] The road and bridge extension path is divided into regional units according to the step size. The SIFT feature point matching is performed on the collected road and bridge images, the spatial projection relationship is established, and the texture mapping model is generated on the surface of the three-dimensional point cloud. The integration quantity of the road and bridge images is dynamically adjusted according to the topological distance between the regional unit and the important area, the image priority score is calculated, and the optimal image set is selected for integration.

[0063] Finally, the precise three-dimensional point cloud model and the path point cloud model are fused. First, rough registration is achieved through the ICP algorithm, then the transformation matrix is optimized using the RANSAC algorithm, and the nearest neighbor matching based on the kd-tree is implemented for the point cloud in the overlapping area. The road cross-section parameters are generated based on the point cloud slicing technology, the structural deformation amount is calculated through model comparison, and the BIM component attribute report is output according to the IFC standard requirements.

[0064] Through this embodiment, various parameters of the road and bridge of the highway project are successfully obtained, and the measurement accuracy meets the engineering requirements, providing accurate and reliable data support for the subsequent highway engineering design, construction, and maintenance.

[0065] It should be noted that although the above steps are described in a specific order, it does not mean that the steps must be executed in the above specific order. In fact, some of these steps can be executed concurrently or even in a different order, as long as the required functions can be achieved. The number of devices and the processing scale described here are used to simplify the description of the present invention, and the application, modification, and variation of the present invention will be obvious to those skilled in the art.

[0066] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and embodiments. It can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and the illustrated examples described herein.

Claims

1. The highway engineering road and bridge measurement method based on BIM is characterized by: The following steps are involved: S1: In the measurement area, the satellite positioning data of the road and bridge area of ​​the highway project is obtained through the satellite positioning system, the road and bridge area is scanned by the laser scanning equipment to obtain the point cloud data, and the road and bridge images are collected in the road and bridge area by the image acquisition equipment; S2: converting the satellite positioning data into data information of a unified coordinate system, spatially registering the satellite positioning data with the point cloud data, and dividing the measurement area into important areas and non-important areas according to the importance of the highway engineering road and bridge; S3: For the important area, construct a precise three-dimensional point cloud model of the important area using the point cloud data; S4: For non-important areas, path feature points used to characterize the road and bridge extension path are extracted from the point cloud data to construct a path point cloud model, and the distance step length is set and the road and bridge extension path is divided into regional units according to the step length; a corresponding number of road and bridge images are selected according to the length of the unit and integrated into the regional unit; S5: The precise three-dimensional point cloud model and the path point cloud model are integrated to obtain a highway engineering road and bridge data model, and road and bridge parameters are obtained according to the highway engineering road and bridge data model.

2. The highway engineering road and bridge measurement method based on BIM according to claim 1 is characterized in that: In step S2, the satellite positioning data is converted into data information of a unified coordinate system using the following method: Select at least three ground reference points and obtain the high-precision satellite positioning coordinates of these ground reference points in the WGS84 coordinate system, and convert the coordinates of the ground reference points into coordinates in a plane rectangular coordinate system through Gauss-Krüger projection; The method for selecting ground reference points is: The key nodes of the road and bridge structure are obtained through the road and bridge structure design data. With the key nodes of the structure as the center, multiple evenly distributed ground reference points are selected within a predetermined range. The coordinates of the selected ground reference points are verified by a total station, and the coordinates of the ground reference points are corrected in real time by using a dual-frequency GNSS receiver.

3. The highway engineering road and bridge measurement method based on BIM according to claim 1 is characterized in that: In step S2, the satellite positioning data and the point cloud data are spatially aligned using a weighted average method, which specifically includes the following steps: The point cloud coordinates corresponding to each ground reference point are extracted from the point cloud data through the feature matching algorithm, the allowable error range is set, and the point cloud data corresponding to the point cloud coordinates are iteratively calculated and corrected until the registration error of the point cloud data falls within the allowable error range, thus completing the matching of the point cloud data with the ground reference point; For each ground reference point, the satellite positioning weight w is calculated based on the real-time signal-to-noise ratio of the dual-frequency GNSS receiver. GNSS , the point cloud weight w is calculated based on the local density and curvature stability of the point cloud in the area where the ground reference point is located Cloud , calculate the comprehensive weight , normalize the comprehensive weight; The rigid transformation matrix is ​​set to minimize the residual error before and after the weighted change of the point cloud, that is: Where R and T are the rotation matrix and translation vector in the rigid transformation matrix, respectively, i is the comprehensive weight of the i-th ground reference point, n is the number of ground reference points, q i is the point cloud data of the ground reference point to be transformed, p i is the transformed reference point cloud data; The optimal rotation matrix R and translation vector T are calculated by the singular value decomposition method to minimize the residual error; the point cloud data in the area where each ground reference point is located is weighted according to the rotation matrix R and the translation vector T to complete the spatial alignment of the point cloud data.

4. The highway engineering road and bridge measurement method based on BIM according to claim 1 is characterized in that: In step S3, when constructing a precise three-dimensional point cloud model, a curvature analysis algorithm is used to perform noise reduction processing on the point cloud data, and a clustering algorithm is used for high-density areas and a scanning blind spot is filled for low-density areas, including the following steps: For each point c in the point cloud data, search for adjacent point sets within the range of its neighborhood radius r, and use the PCA principal component analysis method to calculate the normal vector and curvature corresponding to point c. If the calculated curvature is greater than the set curvature threshold, the point c is marked as a noise point to be processed, otherwise it is marked as a non-noise point; For non-noise points, bilateral filtering algorithm is used for smoothing; For a noise point, if the non-noise points in the point cloud data within the range of its neighborhood radius r exceed 70%, the coordinates of this noise point are updated to the weighted average of the non-noise points in the neighborhood, otherwise the noise point is directly deleted; Divide the point cloud data into voxel grids, count the number of point clouds in each voxel and calculate the point cloud data density. If the point cloud data density of a voxel is less than the set density threshold, the voxel is divided into a low-density area, otherwise it is divided into a high-density area. For high-density areas, a density clustering algorithm is used to obtain representative core points in the area; For the low-density area, the boundary point set between it and the high-density area is extracted, and the local surface is generated by the Poisson surface reconstruction algorithm for the boundary point set. The filling point set for the low-density area is sampled on the local surface; for each filling point in the filling point set, its normal vector is calculated and the consistency is optimized with the normal vectors of its neighboring points.

5. The highway engineering road and bridge measurement method based on BIM according to claim 1 is characterized in that: In step S4, the improved Douglas-Peucker algorithm is used to extract the path feature points, the curvature change threshold parameter of the algorithm is set, and the curvature extreme points in the extension direction of the road and bridge are extracted, and the center line of the path is determined by combining the point cloud normal vector analysis.

6. The highway engineering road and bridge measurement method based on BIM according to claim 1 is characterized in that: In step S4, the distance step length is set using a dynamic adjustment mechanism: If the path curvature radius is 100m≤R≤300m, the step length is set to 1m; if the path curvature radius R<100m, the step length is shortened to 0.5m; if R>300m, the step length is extended to 2m; and an exponential correlation function between the step length and the complexity of the road and bridge terrain is constructed based on the set step length.

7. The highway engineering road and bridge measurement method based on BIM according to claim 1 is characterized in that: In step S4, the step of integrating the road and bridge image into the regional unit includes: SIFT feature point matching is performed on the collected road and bridge images to establish the spatial projection relationship between adjacent images. The perspective transformation matrix is ​​used to map the two-dimensional image to the three-dimensional point cloud surface to generate a texture mapping model.

8. The BIM-based highway engineering road and bridge measurement method according to claim 7 is characterized in that: In step S4, when integrating the road and bridge images into the regional unit, the number of integrated road and bridge images is dynamically adjusted according to the topological distance between the regional unit and the important area, and the optimal image set for integration is screened out, which specifically includes the following steps: The calculation needs to be for the area unit U j The number of images N allocated j : Among them, N max To set the maximum number of images, N base is the number of base images, d j The area unit U j The minimum Euclidean distance to the nearest important area boundary, d0 is the set distance threshold, η is the attenuation factor, and , where η0 is the basic attenuation factor, p r is the adjustment coefficient, R is the curvature radius of the road bridge; In the regional unit, for the candidate image set I, calculate the image priority score: Among them, S k is the kth image I k The priority score, SSIM(I k ) is image I k The structural similarity index of , α is the clarity influence coefficient, SNR(I k ) is image I k The signal-to-noise ratio, β is the uniformity influence coefficient, M k For image I k The feature matching rate of its adjacent images, γ is the matching rate influence coefficient; The images in the candidate image set I are arranged in descending order according to the priority scores, the first N images are selected and integrated into the corresponding regional units, and the image overlapping areas of adjacent regional units are fused using the adaptive Gaussian weighting method.

9. The highway engineering road and bridge measurement method based on BIM according to claim 1, characterized in that: In step S5, the model fusion adopts a multi-resolution hierarchical registration method, firstly implements coarse registration through the ICP algorithm, then uses the RANSAC algorithm to optimize the transformation matrix, and implements the nearest neighbor matching based on kd-tree for the overlapping area point clouds.

10. The highway engineering road and bridge measurement method based on BIM according to claim 1, characterized in that: In step S5, the method for obtaining road and bridge parameters includes: generating road cross-section parameters based on point cloud slicing technology, calculating structural deformation through model comparison, and outputting a BIM component property report according to IFC standard requirements. The key parameters in the report include elevation deviation, flatness index and curvature continuity analysis results.

Images