A method and system for measuring the main cable shape of a suspension bridge using a drone

Abstract

The present invention discloses a method for measuring the linear shape of a suspension bridge's main cables using a drone. The method comprises the following steps: drone data acquisition, where the drone carries a laser radar scanning device to directly acquire point cloud data, or uses a drone-mounted RGB camera for oblique photogrammetry; point cloud model reconstruction, where the data is post-processed and a point cloud model of the suspension bridge is reconstructed; point cloud extraction of the suspension bridge's main cables, where the main cable point cloud is extracted from point cloud data of the entire suspension bridge and its surroundings by analyzing the spatial distribution characteristics of the main cables; and density clustering to remove noise and handrail point clouds to accurately extract the main cable point cloud and calculate the main cable linear shape. The present invention also proposes a system for measuring the linear shape of suspension bridge main cables using a drone, which can improve the efficiency of measuring the linear shape of suspension bridge main cables while ensuring a certain level of accuracy, and has broad application prospects.

Description

Technical Field

[0001] The present invention belongs to the technical field of bridge detection, and in particular relates to a method for measuring the linear shape of a main cable of a suspension bridge using an unmanned aerial vehicle (UAV). Background Art

[0002] The main components of a suspension bridge are the main cable, suspenders, stiffening beams, towers, anchors, piers, abutments, foundations, and supports. The main cable is the primary, exposed, irreplaceable, and nearly impossible to reinforce metal component. Traditionally, the main cable alignment is measured using a total station, with a measuring station set up on one side of the main cable and reflective prisms installed at observation points along the main cable. This results in a large amount of observation data, intensive measurements, and complex methods.

[0003] As a new tool, drones offer flexibility and stable flight. They can carry cameras or other equipment to replace manual inspections and can reach difficult-to-reach locations like towering towers, improving inspection efficiency. While drones have been used to detect surface defects in bridges, there is limited research on the alignment of suspension bridge main cables. Therefore, a method for measuring the alignment of suspension bridge main cables using drones is needed. Summary of the Invention

[0004] The technical problem to be solved by the present invention is: in order to overcome the deficiencies of the above-mentioned prior art, the present invention provides a method for measuring the linear shape of the main cable of a suspension bridge by using a drone.

[0005] The present invention adopts the following technical solutions to solve the above technical problems:

[0006] The present invention first proposes a method for measuring the main cable shape of a suspension bridge using a drone, which specifically includes the following steps:

[0007] S1. UAV data collection. Drones can be equipped with lidar scanning equipment to directly acquire point cloud data, or they can be equipped with RGB cameras to perform oblique photogrammetry to reconstruct point cloud models.

[0008] S2. Point cloud model reconstruction. Whether it is lidar scanning or oblique photogrammetry, after the UAV obtains the suspension bridge data, the data needs to be post-processed to reconstruct the suspension bridge point cloud model.

[0009] S2.1. UAV-mounted LiDAR scanning. Using a UAV-mounted LiDAR, integrating multiple cutting-edge technologies such as laser ranging, positioning and attitude determination, GPS differential positioning, and computer technology, a single flight can yield high-precision, high-density three-dimensional point cloud information of a suspension bridge.

[0010] S2.2. UAV oblique photogrammetry. For oblique photogrammetry 3D model reconstruction, as long as adjacent photos share sufficient common feature points, a 3D real-world point cloud model can be constructed from image data with position and attitude information. Each point in the model contains 3D geographic location information.

[0011] S3. Extracting the point cloud of the main cables of the suspension bridge. By analyzing the spatial distribution characteristics of the main cables of the suspension bridge, the point cloud of the main cables is extracted from the point cloud data of the entire suspension bridge and its surrounding environment.

[0012] The original point cloud data of a suspension bridge mainly includes surrounding points, main cable point clouds, suspender point clouds, bridge tower point clouds, and bridge deck point clouds. Extracting the main cable point cloud line shape from the original point cloud data of a suspension bridge involves two steps: slicing and segmentation, and coarse extraction based on point cloud dimensional features.

[0013] For the main cable of a suspension bridge, its spatial characteristics roughly conform to the distribution in the vertical plane. The main cable point cloud is sliced ​​and segmented in the open source software CloudCompare. The segmented point cloud contains the tower point cloud, suspender point cloud, main cable point cloud and noise points.

[0014] Within the local neighborhood of the point cloud space, the main cables and slings exhibit spatial linear distribution characteristics and can be considered one-dimensional linear features. The tower point cloud exhibits a point-surface distribution on the concrete surface and can be considered a two-dimensional planar feature. Other noise points exhibit a three-dimensional scattered distribution and can be considered irregular spherical features. The main cables and slings can be extracted from the unclassified point cloud slices based on their spatial dimensional characteristics. Target points with dimensional characteristics that are linear features are selected as the coarse extraction results of the main cable point cloud. Considering the vertical linear characteristics of the slings, a horizontal constraint is imposed on the direction vectors, requiring that the eigenvector corresponding to the maximum eigenvalue be parallel to the horizontal plane to ensure the accuracy of the main cable extraction results.

[0015] S4. Calculate the main cable shape. The roughly extracted point cloud also contains handrails, noise points, etc. In order to segment the main cable from the roughly extracted point cloud, the density clustering method is used to extract the main cable point cloud and calculate the main cable shape. The bridge deck point cloud is transformed, and the x-axis or y-axis in the transformed o'x'y'z' coordinates along the bridge direction coincides. The points are sliced ​​at a specific step length along the bridge direction and clustered step by step. After the secondary clustering is completed, the projection coordinates of the main cable centerline perpendicular to the bridge direction are a certain type of cluster set of the clustering results. The noise points and handrail line shape can be removed from these cluster sets to identify and extract the main cable line shape.

[0016] The present invention also provides a system for measuring the main cable shape of a suspension bridge using a drone, comprising:

[0017] The UAV data acquisition module is used to directly acquire point cloud data using a UAV-mounted laser radar scanning device, or to reconstruct a point cloud model using oblique photogrammetry using a UAV-mounted RGB camera.

[0018] The point cloud model reconstruction module is used to post-process the data collected by the UAV data acquisition module to reconstruct the point cloud model of the suspension bridge;

[0019] The suspension bridge main cable point cloud extraction module is used to extract the main cable point cloud from the point cloud data of the entire suspension bridge and its surrounding environment by analyzing the spatial distribution characteristics of the suspension bridge main cables;

[0020] The main cable shape calculation module is used to extract the main cable point cloud and calculate the main cable shape using the density clustering method. The present invention adopts the above technical solution and has the following beneficial effects compared with the existing technology:

[0021] The present invention uses a new tool such as a drone, which can be equipped with a camera or lidar equipment to replace manual measurement of the linear shape of the main cable of the suspension bridge. It can achieve a certain accuracy and does not require the use of a total station for measurement. The inspection personnel do not need to arrange a large number of observation points on the main cable of the suspension bridge, thereby improving the inspection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a result diagram of density clustering segmentation of a slice point cloud set under an implementation example.

[0023] Figure 2 The present invention is a flowchart for finely extracting and calculating the linear shape of the main cable of a suspension bridge according to an embodiment of the present invention.

[0024] Figure 3 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0025] The present invention will be further described below with reference to the accompanying drawings. It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art and, unless defined as such, should not be interpreted in an idealized or overly formal sense.

[0026] refer to Figure 3 As shown, the specific process of the present invention is as follows:

[0027] To measure the main cable shape of a suspension bridge, we first need to obtain point cloud data of the suspension bridge. This data can be obtained through drone-mounted oblique photogrammetry or drone-mounted LiDAR scanning equipment.

[0028] Drone oblique photogrammetry. Reconstructing 3D models from oblique photography can create a 3D real-world point cloud model from image data containing position and attitude information, as long as adjacent photos have enough common feature points. Each point in the model contains 3D geographic location information.

[0029] Drone-mounted LiDAR scanning. Drone-mounted LiDAR scanning integrates multiple cutting-edge technologies, including laser ranging, positioning and attitude determination, GPS differential positioning, and computer technology. High-precision, high-density three-dimensional point cloud information of the suspension bridge can be obtained in a single flight.

[0030] Extracting point clouds of suspension bridge main cables. By analyzing the spatial distribution characteristics of the suspension bridge main cables, the main cable point clouds are extracted from the point cloud data of the entire suspension bridge and its surroundings.

[0031] The original point cloud data of a suspension bridge mainly includes surrounding points, main cable point clouds, suspender point clouds, bridge tower point clouds, and bridge deck point clouds. Extracting the main cable point cloud line shape from the original point cloud data of a suspension bridge involves two steps: slicing and segmentation, and coarse extraction based on point cloud dimensional features.

[0032] For the main cable of a suspension bridge, its spatial characteristics roughly conform to the distribution in the vertical plane. The main cable point cloud is sliced ​​and segmented in the open source software CloudCompare. The segmented point cloud contains the tower point cloud, suspender point cloud, main cable point cloud and noise points.

[0033] Within the local neighborhood of the point cloud space, the main cables and slings exhibit a linear distribution and can be considered one-dimensional linear features. The tower point cloud exhibits a surface distribution of points on the concrete surface and can be considered two-dimensional planar features. Other noise points exhibit a three-dimensional scattered distribution and can be considered irregular spherical features. The main cables and slings can be extracted from the unclassified point cloud slices based on their spatial dimensional characteristics.

[0034] In order to determine the dimension of the unclassified point cloud, the principal component analysis method is used to analyze the point cloud neighborhood. The principal component analysis method is to fit the unclassified point area with a plane, and the three eigenvalues ​​of its covariance matrix λ0≥λ1≥λ2 are the square errors of its fitting in three directions. Representing the fitting residuals in three directions respectively, we have: when δ0>>δ1,δ2, the fitting area has a large fitting residual in only one direction, and the point is a scattered point. Similarly, when δ0,δ1>>δ2, the point is a surface point. The corresponding eigenvector of λ2 is the normal vector of the point. When δ0≈δ1≈δ2, the point is a scattered point. Use the fitting residual to assign the possibility of the dimension to which the point belongs:

[0035]

[0036] Where a1D 、a 2D 、a 3D Respectively represent the probability that the unclassified points belong to the above three dimensions, where a 1D +a 2D +a 3D =1, the dimension d to which the point belongs V It can be defined as:

[0037] d V =arg d∈{1,2，3} max[a dD ]

[0038] Due to the differences in spatial structure size and point cloud density of different geographic identifications, the scale of the neighborhood radius will affect the calculation of dimensional features, resulting in increased uncertainty in the dimensional characteristics of unclassified points and reduced accuracy in main cable identification. For different points, a suitable neighborhood radius should be selected. According to the entropy minimization theory, the optimal neighborhood radius can be calculated by minimizing the information entropy contained in the neighborhood point set as a constraint condition:

[0039]

[0040] Where, E f Indicates the entropy value contained in the neighborhood of the point, E f The smaller the value, the less information the neighborhood of the point contains, that is, the more single the dimensional features of the point are. i When E changes gradually f will also change accordingly. When E f The smallest one, the simplest feature, and the smaller the uncertainty of the dimensional characteristics of the point. best Indicates the optimal neighborhood radius of the current point, [R min ,R max ] represents the lower and upper bounds of the neighborhood radius, E f (R i ) represents the entropy value at this radius. The dimensional characteristics shown at the optimal radius are the true dimensional characteristics of the point.

[0041] The target points with linear features were selected as the main cable point cloud coarse extraction results. Considering the vertical linear characteristics of the sling, a horizontal constraint was added to the direction vector, that is, the eigenvector corresponding to the maximum eigenvalue was parallel to the horizontal plane to ensure the accuracy of the main cable extraction results.

[0042] The coarsely extracted point cloud also contains handrails, noise points, and other features. To segment the main cable from the coarsely extracted point cloud, a density clustering method is used to extract the main cable point cloud and calculate the main cable shape. DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a representative density-based clustering algorithm. This algorithm uses sample density as a metric to measure the connectivity between samples, continuously expanding the cluster set based on connected samples to obtain the final clustering result. Its core idea is to start from a seed core point, continuously extend it to the density-reachable area (core points and boundary points), and add more points to the cluster, thereby obtaining a maximum sample subset containing both core points and boundary points. This method has two core input parameters: the neighborhood radius eps and the neighborhood density threshold MinPts. Starting from any unvisited point, all nearby points within a distance eps (inclusive of eps) are found.

[0043] If the number of nearby points is greater than or equal to MinPts, the current point and its nearby points form a cluster, and the starting point is marked as visited. Then, recursively, all points in the cluster that have not been marked as visited are processed in the same way, thus expanding the cluster. If the number of nearby points is less than MinPts, the point is temporarily marked as a noise point. If the cluster is fully expanded, that is, all points in the cluster are marked as visited, the same algorithm is used to process the unvisited points.

[0044] The bridge deck point cloud is transformed, and the x-axis or y-axis in the transformed o'x'y'z' coordinates are aligned along the bridge direction. The main cable point cloud line shape is extracted by slicing at a specific step length along the bridge direction and clustering in steps. The implementation steps are as follows:

[0045] (1) Assuming that the direction along the bridge is the x-axis direction and the direction perpendicular to the bridge is the y-axis direction, extract the minimum value X in the x-coordinate direction. min With the maximum value X max , and set the appropriate step size Δ=(X max -X min ) / m, where m is a positive integer, create (X min +(k-1)Δ,X min +kΔ) range of m point sets (where k∈[1,m]), input domain radius eps and neighborhood density MinPts.

[0046] (2) Input slice point cloud dataset P k ={(x i ,y i ,z i ,Cluster i)}, project the three-dimensional point set onto the o'y'z' two-dimensional plane, P k '={(y i ,z i ,Cluster i )}, initialize the parameters eps and MinPts, mark all laser point cloud states to be classified as unvisited, that is, unprocessed state, and assign the category label Cluster to 0.

[0047] (3) Select points sequentially from the point cloud dataset. If the point status is marked as visited, it means that the point already has a category label (belongs to a certain cluster or is marked as noise), and reselect the next point.

[0048] (4) If the point status is marked as unvisited, search for the point set R within the area radius eps of the point in the two-dimensional plane. If the number of points contained in R is less than MinPts, the point is marked as a noise point, the point status is marked as visited and the category label Cluster is assigned a value of -1. If the number of points contained in R is greater than or equal to MinPts, the point is marked as a core object point and a new class cluster set C is generated. Cluster+1 And search the neighborhood of the core object point, and classify all points that can be reached from this point into the cluster set C Cluster+1 In the Cluster+1 All points are marked as visited and the category label Cluster is assigned a new category label, that is, Cluster = Cluster + 1.

[0049] (5) Repeat (3) and (4) until all points in all point cloud datasets are traversed. Figure 1 shown.

[0050] (6) The two-dimensional point set P k 'Map the labels to the 3D point cloud, calculate the cluster centers of each cluster set, and output the cluster center point set Q k .

[0051] (7) Repeat (2) to (6) and output m slice point sets, cluster center point set Q = {Q1, Q2, Q3, ..., Q m}, project the three-dimensional point set onto the o'x'z' two-dimensional plane, Q'={(x i ,z i ,Cluster i )}, re-enter the parameters eps and MinPts and initialize them, mark all laser point cloud states to be classified as unvisited, that is, unprocessed state, and assign the category label Cluster to 0.

[0052] (8) Execute (3) to (5) and output the cluster set C i , i = -1, 1, 2, 3…ClusterID, ClusterID is the number of clusters.

[0053] As attached Figure 2 As shown in the flowchart, after clustering, the cluster center point set is divided into ClusterID+1 subsets, among which the identified outlier noise points are clustered into the subset C with the category label of -1. -1 In the figure, the projection coordinates of the main cable centerline perpendicular to the bridge direction are also a certain type of cluster set in step (8). Noise points and handrail line shapes can be removed from these cluster sets to identify and extract the main cable line shape.

[0054] The present invention also provides a system for measuring the main cable shape of a suspension bridge using a drone, comprising:

[0055] The UAV data acquisition module is used to directly acquire point cloud data using a UAV-mounted laser radar scanning device, or to reconstruct a point cloud model using oblique photogrammetry using a UAV-mounted RGB camera.

[0056] The point cloud model reconstruction module is used to post-process the data collected by the UAV data acquisition module to reconstruct the point cloud model of the suspension bridge;

[0057] The suspension bridge main cable point cloud extraction module is used to extract the main cable point cloud from the point cloud data of the entire suspension bridge and its surrounding environment by analyzing the spatial distribution characteristics of the suspension bridge main cables;

[0058] The main cable linear shape calculation module is used to extract the main cable point cloud and calculate the main cable linear shape using the density clustering method.

[0059] It should be noted that the description of the system of the embodiment of the present application is similar to the description of the above-mentioned method embodiment, and has similar beneficial effects as the method embodiment, so it will not be repeated.

[0060] The above descriptions are only partial embodiments of the present invention. It should be pointed out that ordinary technicians in this technical field can make several improvements and modifications without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for measuring the main cable shape of a suspension bridge using an unmanned aerial vehicle, characterized by: S1, UAV data collection; S2, point cloud model reconstruction: post-processing the data collected in step S1 to reconstruct the point cloud model of the suspension bridge; S3. Extraction of main cable point cloud of suspension bridge: By analyzing the spatial distribution characteristics of the main cables of the suspension bridge, the main cable point cloud is extracted from the point cloud data of the entire suspension bridge and its surrounding environment; S4. Calculate the main cable shape: Use density clustering to extract the main cable point cloud and calculate the main cable shape. Specifically, the roughly extracted main cable point cloud is sliced ​​along the bridge direction. The slice point set is projected onto a plane perpendicular to the bridge direction and density clustered. The cluster centers are calculated. All the slice cluster centers are projected onto a vertical plane along the bridge direction and density clustered again to identify and extract the main cable shape. The main cable shape is calculated as follows: coordinate transformation is performed on the bridge deck point cloud, the x-axis or y-axis in the transformed o'x'y'z' coordinates along the bridge direction coincides, and slices are performed at a specific step length along the bridge direction and clustered step by step; After the secondary clustering is completed, the projection coordinates of the main cable centerline in the perpendicular direction of the bridge are a certain type of cluster set of clustering results. Noise points and handrail line shapes are removed from these cluster sets to identify and extract the main cable line shape.

2. The method for measuring the main cable shape of a suspension bridge using an unmanned aerial vehicle according to claim 1, characterized in that: In step S1, point cloud data is directly acquired by using a UAV equipped with a laser radar scanning device, or a point cloud model is reconstructed by using an RGB camera equipped with a UAV for oblique photogrammetry.

3. The method for measuring the main cable shape of a suspension bridge using an unmanned aerial vehicle according to claim 1, characterized in that: In step S3, the main cable point cloud is roughly extracted through dimensional features, including two steps: slicing segmentation and rough extraction based on point cloud dimensional features.

4. The method for coarsely extracting a main cable point cloud using dimensional features according to claim 3, characterized in that: For the main cable of the suspension bridge, the main cable point cloud is sliced ​​and segmented. Specifically, the main cable and suspenders are extracted from the unclassified point cloud slices based on the spatial dimensional features, and the target point whose dimensional features are linear feature objects is selected as the rough extraction result of the main cable point cloud; the horizontal constraint condition of the direction vector is added, that is, the eigenvector corresponding to the maximum eigenvalue is parallel to the horizontal plane.

5. A system for measuring the main cable shape of a suspension bridge using a drone, characterized in that: include: The UAV data acquisition module is used to directly acquire point cloud data using a UAV-mounted laser radar scanning device, or to reconstruct a point cloud model using oblique photogrammetry using a UAV-mounted RGB camera. The point cloud model reconstruction module is used to post-process the data collected by the UAV data acquisition module to reconstruct the point cloud model of the suspension bridge; The suspension bridge main cable point cloud extraction module is used to extract the main cable point cloud from the point cloud data of the entire suspension bridge and its surrounding environment by analyzing the spatial distribution characteristics of the suspension bridge main cables; The main cable linear shape calculation module is used to extract the main cable point cloud and calculate the main cable linear shape using the density clustering method. Specifically, the roughly extracted main cable point cloud is sliced ​​along the bridge direction, the slice point set is projected onto the vertical plane along the bridge direction and density clustered, the cluster center is calculated, and all the slice cluster centers are projected onto the vertical plane along the bridge direction and density clustered again to identify and extract the main cable linear shape. The main cable shape is calculated as follows: coordinate transformation is performed on the bridge deck point cloud, the x-axis or y-axis in the transformed o'x'y'z' coordinates along the bridge direction coincides, and slices are performed at a specific step length along the bridge direction and clustered step by step; After the secondary clustering is completed, the projection coordinates of the main cable centerline in the perpendicular direction of the bridge are a certain type of cluster set of clustering results. Noise points and handrail line shapes are removed from these cluster sets to identify and extract the main cable line shape.

Images