Highway bridge deformation monitoring method and system based on laser measurement

Abstract

The application relates to the technical field of bridge deformation monitoring, in particular to a highway bridge deformation monitoring method and system based on laser measurement, which comprises the following steps: removing noise from various point clouds based on the slope difference of a fitting straight line obtained from the x and y coordinates of all edge points between two side edge clusters and the curvature difference of all edge points and combining the local change trend of the curvature of all edge points in each side edge cluster; fitting all plane points and all plane points in the neighborhood of each plane point, analyzing the error condition of the fitting process, combining the discrete degree of the curvature of all plane points and all plane points in the neighborhood of each plane point, and screening out feature clusters; obtaining all matching point pairs between the feature clusters of different feature point clouds, determining key indexes by analyzing the distance between each matching point pair and the difference of concave-convex coefficients between each matching point pair, and monitoring the deformation of a highway bridge. The application aims to improve the precision of monitoring the deformation of a highway bridge.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of bridge deformation monitoring, in particular to a highway bridge deformation monitoring method and system based on laser measurement. BACKGROUND

[0002] Highway bridges are key nodes in the transportation network, and their structural safety performance directly affects regional economy and public safety. During long-term use, bridges are susceptible to structural deformation and even damage accumulation due to environmental erosion, material aging, traffic load overloading, and natural disasters. Traditional deformation monitoring methods such as total station, fiber Bragg grating sensors, or GPS positioning technology have low measurement efficiency, poor environmental adaptability, and insufficient dynamic response, making it difficult to meet the needs of high-precision and real-time monitoring of bridges. In recent years, laser measurement technology has gradually become a research hotspot in the field of structural health monitoring due to its non-contact, high precision, and strong anti-interference advantages, providing a new technical path for multi-dimensional information perception of bridge deformation.

[0003] Currently, in the research of using a three-dimensional laser scanner to detect bridge deformation, the change of a certain point on the bridge before and after a certain time is usually compared to determine whether the bridge has deformed. However, point cloud data is easily affected by external objective environments, and existing technologies usually do not analyze the impact of environmental factors and vehicles driving on the bridge on point cloud quality in depth. Therefore, there is a large error in determining the comparison point, which reduces the precision of monitoring highway bridge deformation. SUMMARY

[0004] To solve the above technical problems, the purpose of the present application is to provide a highway bridge deformation monitoring method and system based on laser measurement, and the technical solution adopted is as follows:

[0005] In the first aspect, the present application provides a highway bridge deformation monitoring method based on laser measurement, which includes the following steps:

[0006] S1: Use the point cloud acquisition module of the system to acquire various point clouds of the bridge road surface to be measured, including standard point clouds and ordinary point clouds;

[0007] S2: Compare the ordinary point clouds with the standard point clouds to determine the key index, specifically:

[0008] All edge points in various point clouds are obtained, all edge points are clustered, positive and negative differences between x coordinate values of all edge points in each cluster are analyzed, side confidence of each cluster is determined, two side clusters are screened out from all cluster of various point clouds; based on slope difference of fitting straight line obtained by (x, y) coordinates of all edge points between two side clusters and curvature difference of all edge points, difference index between two side clusters is determined, and various point clouds are denoised by combining local change trend of curvature of all edge points in each side cluster, and after denoising, various point clouds are recorded as feature point clouds.

[0009] Point cloud data of all edge points in various feature point clouds is recorded as plane points, each plane point and all plane points in the neighborhood of the plane point are fitted, error condition of the fitting process is analyzed, and concave-convex coefficient of each plane point is determined by combining discrete degree of curvature of all plane points in the neighborhood of the plane point; all plane points in various feature point clouds are clustered, and by analyzing proportion of the number of all plane points in each cluster in total number of plane points in the corresponding feature point cloud and average distribution of concave-convex coefficient of all plane points in each cluster, feature cluster of various feature point clouds is screened out.

[0010] All matching point pairs between feature clusters of different feature point clouds are obtained, key index between each matching point pair is determined by analyzing distance between each matching point pair and difference of concave-convex coefficient between each matching point pair.

[0011] S3: Deformation monitoring of the highway bridge is performed based on the key index.

[0012] Preferably, the determination method of the side confidence of each cluster is as follows:

[0013] Total number of edge points with negative x coordinate value in each cluster and total number of edge points with positive x coordinate value in each cluster are divided by total number of edge points in the corresponding cluster respectively, and the division results are recorded as positive ratio and negative ratio respectively, and absolute value of difference between the positive ratio and the negative ratio is taken as the side confidence of each cluster.

[0014] Preferably, the two side clusters are clusters corresponding to the first two values in descending order arrangement results of the side confidence of all clusters of various point clouds.

[0015] Preferably, the difference index between the two side clusters is the sum of slope difference of fitting straight line obtained by (x, y) coordinates of all edge points between the two side clusters and curvature difference of all edge points.

[0016] Preferably, the denoising of various point clouds comprises:

[0017] The curvature of all edge points within each side cluster is used as the input of the mutation point detection algorithm, and the output is the mutation curvature. The edge point corresponding to the mutation curvature is denoted as the mutation point. The curvature of each mutation point and all edge points in its neighborhood are arranged in ascending order of their respective y-coordinates. In the arrangement result, the curvature of all edge points before each mutation point is used to form the pre-curvature sequence of each mutation point, and the curvature of all edge points after each mutation point is used to form the post-curvature sequence of each mutation point.

[0018] First-order difference sequences of the front and back curvature sequences are obtained respectively and denoted as front curvature change sequence and back curvature change sequence. The ratio of the number of negative elements to the total number of elements in the front curvature change sequence is denoted as the increasing proportion, and the ratio of the number of positive elements to the total number of elements in the back curvature change sequence is denoted as the decreasing proportion. The average of the increasing proportion and the decreasing proportion is used as the deformation index of each side cluster.

[0019] The ratio of the difference index between two side clusters to the average deformation index of the two side clusters is used as the noise interference coefficient between the two side clusters.

[0020] The noise interference coefficient between two side clusters is used as the input of an exponential function with the natural constant as the base, and the output is used as the neighborhood parameter in the filtering algorithm. The filtering algorithm is then used to denoise various point clouds.

[0021] Preferably, the method for determining the concavity / convexity coefficient of each planar point is as follows:

[0022] Calculate the mean square error during the fitting process for each plane point and all plane points in its neighborhood;

[0023] The product of the degree of dispersion of the curvature of each plane point and all plane points in its neighborhood with the mean square error is used as the concavity / convexity coefficient of each plane point.

[0024] Preferably, the method for obtaining the feature clusters of the various feature point clouds is as follows:

[0025] In various feature point clouds, the ratio of the number of all planar points in each cluster to the total number of planar points in the corresponding feature point cloud is calculated and denoted as the proportion of each cluster. The ratio of the proportion of each cluster to the average concavity coefficient of all planar points in the corresponding cluster is denoted as the confidence ratio of each cluster.

[0026] Among various feature point clouds, the cluster with the highest confidence ratio is taken as the feature cluster of each feature point cloud.

[0027] Preferably, the key index between each pair of matching points is the result of multiplying the difference in the concavity / convexity coefficient between each pair of matching points by the distance between the three-dimensional coordinates of the corresponding pair of matching points.

[0028] Preferably, the deformation monitoring of the highway bridge includes:

[0029] Among the normal point cloud and the standard point cloud, the key indexes between all the matched point pairs are taken as the input of the clustering algorithm, a plurality of clustering clusters are output, and are recorded as a clustering set. The mean value of all the key indexes in each clustering set is calculated, recorded as a key mean value, the clustering set with the largest key mean value is taken as a key cluster, and the points belonging to the normal point cloud in the matched point pairs corresponding to all the key indexes in the key cluster are taken as monitoring points.

[0030] If the difference between the coordinate values of the monitoring points and the matched points in any coordinate axis exceeds the preset deformation threshold, it is determined that the bridge has deformed, otherwise, it is determined that the bridge has not deformed.

[0031] In a second aspect, another embodiment of the present application also provides a highway bridge deformation monitoring system based on laser measurement, which implements the highway bridge deformation monitoring method based on laser measurement described above. The system comprises a point cloud acquisition module, a point cloud data analysis module, and a bridge deformation monitoring module.

[0032] The point cloud acquisition module is used to acquire three-dimensional point cloud data of the bridge when there is no vehicle passing and three-dimensional point cloud data of the bridge road surface when there is normal vehicle passing by using a vehicle-mounted three-dimensional laser scanner, and record them as standard point cloud and normal point cloud respectively. The vehicle-mounted laser scanner travels at a constant speed along the central axis of the bridge from one end of the bridge to the other end. The driving direction is regarded as the y-axis, the lateral direction of the bridge is regarded as the x-axis, and the direction perpendicular to the x-axis and the y-axis is regarded as the z-axis. The collected standard point cloud data and normal point cloud data are transmitted to the point cloud data analysis module.

[0033] The point cloud data analysis module is used to analyze and process the data acquired by the point cloud acquisition module. The analysis and processing method adopts S2 in the highway bridge deformation monitoring method based on laser measurement. Further, the processed results are input to the bridge deformation monitoring module.

[0034] The bridge deformation monitoring module is used to monitor the bridge deformation of the results output by the point cloud data analysis module. The monitoring method adopts S3 in the highway bridge deformation monitoring method based on laser measurement.

[0035] The present application has at least the following beneficial effects:

[0036] The application considers the influence of different environments, vehicles and the like on point cloud data, and can dynamically evaluate the noise influence degree by constructing a noise interference coefficient, so as to more accurately filter the point cloud and improve the accuracy of the point cloud data. Further, by analyzing the geometric difference between the bridge road surface and other objects on the bridge, the interference of these redundant points is removed from the normal point cloud to obtain a feature cluster reflecting the characteristics of the bridge road surface, thereby providing a basis for subsequent monitoring point selection. Further, by registering the planar points in the feature clusters between the normal point cloud and the standard point cloud, a key index between the matching point pairs is constructed, so as to evaluate the key degree of each planar point and select the most critical monitoring point. The application analyzes the influence of environmental factors such as vehicles on point cloud data, thereby filtering and removing redundant points from the point cloud data, which can provide accurate data support for the selection of monitoring points. Finally, according to the influence of the vehicle load on the bridge point cloud change, the most critical monitoring point is selected, thereby improving the accuracy of highway bridge deformation monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0038] Figure 1 The step flow chart of the highway bridge deformation monitoring method based on laser measurement provided by one embodiment of the present application is shown in the figure.

[0039] Figure 2 The key index extraction process schematic diagram provided by one embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0040] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the following describes the highway bridge deformation monitoring method and system based on laser measurement according to the present application, its specific implementation, structure, features and effects in detail, as shown in the drawings and the preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0041] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0042] The application provides a highway bridge deformation monitoring method and system based on laser measurement.

[0043] Please refer to Figure 1 which shows a step flowchart of a highway bridge deformation monitoring method based on laser measurement provided by an embodiment of the application, and the method comprises the following steps.

[0044] Step S1: acquiring various point clouds of the road surface of the bridge to be measured by a point cloud acquisition module of a road surface flatness detection system based on laser measurement, wherein the various point clouds comprise standard point clouds and normal point clouds.

[0045] In this embodiment, a three-dimensional laser scanner carried by a vehicle is used to scan the highway bridge without vehicle driving and with normal vehicle passing, and the scanning mode is as follows: the driving direction is regarded as the y-axis, the lateral direction of the bridge is regarded as the x-axis, and the direction perpendicular to the x-axis and the y-axis is regarded as the z-axis, a three-dimensional scanning system is constructed, the vehicle carrying the scanner drives from one end of the bridge to the other end, and the vehicle drives at a constant speed in a straight line. The bridge is scanned without vehicle driving and with normal vehicle passing, and three-dimensional point clouds of the bridge without vehicle driving and three-dimensional point clouds of the bridge with vehicle passing are obtained.

[0046] Further, a random sample consensus (RANSAC) algorithm is used to automatically identify and segment multiple structural surfaces in the three-dimensional point clouds without vehicle driving and multiple structural surfaces in the three-dimensional point clouds with vehicle passing, the largest structural surface in all the structural surfaces of the three-dimensional point clouds without vehicle driving is taken as the road surface of the bridge to be measured without vehicle driving, and the largest structural surface in all the structural surfaces of the three-dimensional point clouds with vehicle passing is taken as the road surface of the bridge with vehicle passing, that is, the point clouds of the road surface of the bridge to be measured without vehicle passing and the point clouds of the road surface of the bridge with vehicle passing are obtained, and are denoted as standard point clouds and normal point clouds respectively. For the convenience of description, the standard point clouds and the normal point clouds are collectively referred to as various point clouds, that is, the various point clouds comprise the standard point clouds and the normal point clouds. Further, all the edge points are extracted from the road surface of the various point clouds by using an AC method in a PCL point cloud library as the edge points of the various point clouds.

[0047] The random sample consensus (RANSAC) algorithm and the AC method in the PCL point cloud library are all known technologies, and the specific principle process is not described herein.

[0048] Step S2: comparing the feature differences between the normal point clouds and the standard point clouds to determine the key index between each matched point pair.

[0049] Due to the factors influencing the reflection characteristics of the bridge surface, such as material, roughness, color contrast and the like, errors will occur in the point cloud data obtained by the three-dimensional laser scanner. At the same time, the external environment, such as the temperature, the disturbance of the vehicle and the wind, will also affect the direction of the laser propagation in the air, thereby causing errors in the collected data. Therefore, in order to more accurately monitor the deformation of the bridge and determine the accurate monitoring point of the bridge, the collected point cloud data needs to be filtered to reduce errors, and the specific process is as follows:

[0050] S201: Obtain all edge points in various point clouds, cluster the x coordinate values of all edge points, analyze the positive and negative differences between the x coordinate values of all edge points in each cluster, determine the side confidence of each cluster, and select two side clusters from all clusters; based on the local change trend of the curvature of all edge points in each side cluster, determine the deformation index of each side cluster, and combine the slope difference of the fitted straight line obtained by the (x, y) coordinates of all edge points between the two side clusters and the curvature difference of all edge points, to denoise various point clouds, and all denoised point clouds are denoted as feature point clouds.

[0051] Since the bridge structure usually has a regular geometric shape, such as a straight line, a plane, a symmetrical structure and the like, it indicates that in the point cloud data, the local area will be smooth and continuous, and the two side areas will be symmetrical. The existence of noise will cause the point cloud to become irregular, the local area to be uneven, and the symmetry of the two side areas to be destroyed. Therefore, the degree of difference between the bridge point cloud data can be evaluated to reflect the noise interference of the point cloud data, which is specifically as follows:

[0052] Considering that the vehicle-mounted three-dimensional laser scanner drives from one end of the bridge to the other end, the road surface of the bridge will have four edges. Since the point cloud coordinates of the four edges of the bridge are not consistent, while the point cloud coordinates of the same edge are relatively consistent, the clustering algorithm can be used to distinguish the four edges.

[0053] All edge points in the road surface are used as the input of the clustering algorithm, wherein the Euclidean distance between the three-dimensional coordinates of the edge points is used as the measurement distance in the clustering algorithm, the number of clustering clusters is set to 4, and finally 4 clustering clusters are output, which represent the four edges of the bridge.

[0054] It should be noted that there are many commonly used clustering algorithms, and the k-means clustering algorithm is used to cluster the edges in this embodiment. In actual application, as other implementation manners, the implementer can also use the CURE hierarchical clustering algorithm or the fuzzy C-means clustering algorithm. The selection of the clustering algorithm is not specially limited in this embodiment.

[0055] The calculation method of the Euclidean distance and the k-means clustering algorithm are known technologies, and the specific clustering principle will not be described again.

[0056] Further, the total number of edge points with a negative x coordinate value in each cluster and the total number of edge points with a positive x coordinate value in each cluster are divided by the total number of edge points in the corresponding cluster, and the division results are denoted as a positive ratio and a negative ratio, respectively. The absolute value of the difference between the positive ratio and the negative ratio is taken as the side confidence of each cluster. The greater the side confidence is, the greater the possibility that all the x coordinates of the edge points in the cluster are positive or negative, and the greater the possibility that the edge points in the cluster are left and right edge points consistent with the running direction of the vehicle on the bridge.

[0057] The two clusters corresponding to the first two values in the descending order of the side confidence of all the clusters of various point clouds are taken as the two side clusters of the various point clouds, which are used to represent the left and right side edges consistent with the running direction of the vehicle on the bridge.

[0058] Further, since the deformation of the bridge and the noise will cause the curvature of the edge points in the point cloud to fluctuate and change, thereby destroying the symmetry between the two side edges, it is impossible to screen out the edge points that need to be denoised, and thus accurately denoised. However, since the bridge is stressed in one direction when deformed, the curvature changes gradually from the deformed edge of the deformation region to the center of the edge and then gradually decreases, which has a trend. Since the noise is random, it does not have the trend of gradual change. Therefore, based on the above analysis, the point cloud is denoised, specifically as follows:

[0059] The difference between the slopes of the fitted straight lines of the x and y coordinates of all the edge points between the two side clusters and the curvature difference of all the edge points are added together, and the result is taken as the difference index between the two side clusters, which is used to judge whether the symmetry of the bridge is destroyed. Since the edge of the bridge has regularity and changes smoothly, the fluctuations of the point clouds on both sides of the bridge should also be consistent. The greater the difference index is, the greater the possibility that the symmetry of the two side edges is destroyed. Conversely, the smaller the difference index is, the smaller the possibility that the symmetry of the two side edges is destroyed.

[0060] It should be understood that in the embodiment, the DTW distance of the curvatures of all the edge points between the two side clusters is taken as the curvature difference of all the edge points between the two side clusters. In actual application, as other implementation manners, the implementer can also use other methods for measuring the difference between data groups, such as the Euclidean distance and the Manhattan distance. The selection of the method for measuring the difference between data groups is not specially limited in the embodiment.

[0061] The calculation method of the DTW distance is a known technology, and the specific calculation process will not be described again.

[0062] It should be further explained that there are many methods for measuring the difference between data. In the embodiment, the absolute value of the difference is used to calculate the difference between data.

[0063] It should be noted that there are many commonly used fitting methods. In the embodiment, the least squares fitting method is used to fit the x and y coordinates of the edge points. In actual application, as an alternative, a linear regression method or other fitting methods can be used. The selection of the fitting method is not limited in the embodiment.

[0064] Among them, the least squares fitting method and the method for obtaining the curvature of the data points in the point cloud are known technologies, and their specific principles will not be repeated.

[0065] Further, the curvatures of all edge points in each side edge cluster are taken as the input of the mutation point detection algorithm, and the output is the mutation curvature and the edge point corresponding to the mutation curvature, which is recorded as the mutation point. The curvatures of all edge points in the neighborhood of each mutation point are arranged in ascending order according to their respective y coordinates. In the arrangement result, the curvatures of all edge points before each mutation point form the front curvature sequence of each mutation point, and the curvatures of all edge points after each mutation point form the rear curvature sequence of each mutation point.

[0066] Further, the first-order difference sequences of the front and rear curvature sequences are obtained respectively, and are recorded as the front curvature change sequence and the rear curvature change sequence. The ratio of the number of negative elements in the front curvature change sequence to the total number of elements is recorded as the increasing proportion, and the ratio of the number of positive elements in the rear curvature change sequence to the total number of elements is recorded as the decreasing proportion. The average of the increasing proportion and the decreasing proportion is taken as the deformation index of each side edge cluster. The deformation index is used to represent the possibility of the curvature of the edge in the side edge cluster being mutated due to the deformation of the bridge. The greater the deformation index, the greater the possibility of the deformation of the bridge.

[0067] Among them, the method for obtaining the first-order difference sequence is a known technology, and its specific obtaining process will not be repeated.

[0068] Further, the ratio of the difference index between the two side edge clusters to the average deformation index of the two side edge clusters is taken as the noise interference coefficient between the two side edge clusters. The noise interference coefficient is used to judge the influence degree of noise on the curvature of the edge point. The greater the difference index and the smaller the average deformation index, the greater the noise interference coefficient. This indicates that the mutation of the edge point curvature is more likely to be caused by noise rather than bridge deformation. Conversely, if the noise interference coefficient is smaller, it indicates that the mutation of the edge curvature is more likely to be caused by bridge deformation. In this case, a smaller filtering window should be set to retain the details of the bridge deformation.

[0069] Further, the noise interference coefficient between the two side edge clusters is taken as the input of the exponential function with the natural constant as the base number, and the output result is taken as the neighborhood parameter in the filtering algorithm. The filtering algorithm is used to denoise various point clouds, and the denoised various point clouds are all recorded as feature point clouds.

[0070] At this point, by analyzing the spatial distribution of the side edge point cloud and the change trend of the curvature, the various point clouds are denoised to obtain the denoised various feature point clouds.

[0071] S202: The point cloud data of various feature point clouds except all edge points is recorded as a plane point. The error condition of the fitting process is analyzed by fitting each plane point and all plane points in its neighborhood, and the concave-convex coefficient of each plane point is determined in combination with the discrete degree of the curvature of each plane point and all plane points in its neighborhood. The feature clusters of various feature point clouds are selected by analyzing the proportion of the number of all plane points in each cluster in the total number of plane points in the corresponding feature point cloud and the average distribution of the concave-convex coefficients of all plane points in each cluster.

[0072] Further, when a three-dimensional laser scanner is used to scan a bridge, irrelevant objects such as vehicles, pedestrians, streetlights, and plants around the bridge will inevitably be scanned, resulting in a large number of redundant points. At this time, if the bridge monitoring point positions are directly selected, the redundant points may be mistakenly selected as monitoring point positions, thereby affecting the accuracy of subsequent monitoring of the bridge deformation. Therefore, after filtering the initial point cloud data, the redundant points in the point cloud data need to be removed to improve the accuracy of the selection of bridge monitoring point positions. The specific process is as follows:

[0073] Since the bridge has the characteristics of consistent overall color and smooth and flat road surface, the colors of the redundant points such as vehicles, pedestrians, and plants around the bridge are not consistent, and the shapes and sizes of the redundant points are also not consistent. Therefore, the bridge point cloud and the redundant point cloud can be segmented by analyzing the color difference and local difference between the point clouds.

[0074] Since the edge point cloud in the road surface has been extracted as described above, the point cloud data of various point clouds except all edge points is recorded as a plane point.

[0075] Further, by analyzing the distribution characteristics of the data points around the plane point, the concave-convex coefficient of each plane point is determined to determine whether the local area where the plane point is located corresponds to the bridge road surface area. Specifically:

[0076] The neighborhood of each planar point is divided by taking each planar point as the center. There are many methods for dividing the neighborhood, in the embodiment, the distances between each planar point and all other planar points are calculated, the distances are arranged in ascending order, and the planar points corresponding to the first preset number of distances in the arrangement result are used to form the neighborhood of each planar point.

[0077] The value of the preset number is artificially set, in the embodiment, the value of the preset number is 15, and in actual application, the implementer can also set it by himself according to the specific circumstances, and the embodiment does not make special limitation.

[0078] It should be noted that there are many commonly used methods for measuring the distance between points, in the embodiment, the Euclidean distance is used to measure the distance between two points, and in actual application, as other implementation manners, the Mahalanobis distance or Manhattan distance or other methods for measuring the distance between points can also be used, and the embodiment does not make special limitation on the selection of the method for measuring the distance between points.

[0079] Further, the planar fitting is performed on the planar point cloud data and all planar points in the neighborhood, and the mean square error in the fitting process is calculated; the greater the mean square error, the greater the difference between the geometric features of the planar point and other planar points in the neighborhood, the more the local area where the planar point is does not conform to the flat feature of the bridge road surface, and the greater the possibility of being a redundant point, on the contrary, if the mean square error is smaller, the smaller the difference between the geometric features of the planar point and other planar points in the neighborhood, the more likely the local area where the planar point is is a bridge road surface.

[0080] In the embodiment, the least square fitting method is used for fitting each planar point and all planar points in the neighborhood, in addition, as other implementation manners, in actual application, other fitting methods such as polynomial fitting can also be used, and the embodiment does not make special limitation on the selection of the fitting method.

[0081] It should be understood that the calculation method of the mean square error is a known technology, and the specific calculation process is not described again.

[0082] Further, considering that the surface of the vehicle is also relatively smooth, but the vehicle surface has a curvature change, that is, a bending feature. Therefore, by analyzing the discrete degree of the curvature of each planar point and all planar points in the neighborhood, whether the local area where the planar point is has a bending or irregular feature is measured, so as to further exclude the data points corresponding to other objects except the bridge road surface.

[0083] It should be noted that there are many ways to measure the dispersion of the curvature of each planar point and all the planar points in its neighborhood. In the present embodiment, the variance of the curvature of each planar point and all the planar points in its neighborhood is taken as the dispersion of the curvature of each planar point and all the planar points in its neighborhood. In actual application, as other embodiments, the implementer can also use other methods to measure the dispersion of data, such as the dispersion coefficient. The present embodiment does not make special restrictions on the selection of the method for measuring the dispersion of data.

[0084] Further, the product of the dispersion of the curvature of each planar point and all the planar points in its neighborhood and the mean square error is taken as the concave-convex coefficient of each planar point. The larger the concave-convex coefficient, the greater the local area fluctuation of the planar point, and the greater the possibility that the planar point cloud data does not belong to the bridge point cloud. Conversely, the smaller the concave-convex coefficient, the smaller the local area geometric feature fluctuation of the planar point, and the greater the possibility that the planar point belongs to the bridge point cloud.

[0085] Further, in order to better distinguish the bridge and other object corresponding point cloud data, the RGB information of all the planar points is obtained and converted into a gray value. The normalized value of the concave-convex coefficient of each planar point and the gray value form a binary tuple of each planar point.

[0086] All the planar points in various point clouds are clustered, wherein the Euclidean distance between the binary tuples of the planar points is taken as the measurement distance in the clustering algorithm, the elbow method is used to determine the number of clustering clusters, and all the clustering clusters are output.

[0087] In the present embodiment, the k-means clustering algorithm is used. In actual application, the implementer can also use other clustering algorithms such as fuzzy C-means clustering. The present embodiment does not make special restrictions. In addition, the elbow method is used to determine the number of clustering clusters in the clustering process, and the specific principle is not described again.

[0088] Further, in various feature point clouds, the ratio of the number of all planar points in each cluster to the total number of planar points in the corresponding feature point cloud is calculated and denoted as the number ratio of each cluster. The ratio of the average concave-convex coefficient of all planar points in each cluster to the corresponding cluster is denoted as the confidence ratio of each cluster. The greater the confidence ratio, the more the number of elements in the current cluster, and the lower the concave-convex coefficient, the more the corresponding point cloud region of the current cluster conforms to the characteristics of the bridge road surface. Conversely, the smaller the confidence ratio, the fewer the number of elements in the current cluster, and the greater the concave-convex coefficient, the less the corresponding point cloud region of the current cluster conforms to the characteristics of the bridge road surface.

[0089] Finally, in various feature point clouds, the cluster with the largest confidence ratio is taken as the feature cluster of various feature point clouds.

[0090] Thus, by analyzing the geometric characteristics of the plane points, distinguishing the point clouds corresponding to the bridge road surface and other objects on the bridge, the feature clusters are obtained, so that the monitoring points of the bridge deformation can be more accurately determined.

[0091] S203: Obtain all matching point pairs between the feature clusters of different feature point clouds, and determine the key index between each matching point pair by analyzing the distance between each matching point pair and the difference in concave-convex coefficients between each matching point pair.

[0092] Further, considering that the key areas of the bridge that are prone to deformation under external forces such as vehicles may deviate from the standard point cloud due to the load of the vehicle; therefore, the key areas of the bridge that are prone to deformation hazards can be extracted by the deviation between the normal point cloud and the standard point cloud, and monitored, specifically:

[0093] Since the standard point cloud data does not include vehicle driving, there is no phenomenon of point cloud data being unable to be obtained due to vehicle occlusion, therefore, after removing the redundant points, the data in the standard feature cluster can contain the data in the normal feature cluster.

[0094] The normal point cloud and the standard point cloud are used as inputs of a point cloud registration algorithm, and all matching point pairs between the normal point cloud and the standard point cloud are output.

[0095] It is to be supplemented that the commonly used point cloud registration algorithms include the Iterative Closest Point (ICP) algorithm, the Normal Distribution Transform (NDT) algorithm, etc. In the present embodiment, the Iterative Closest Point (ICP) algorithm is used to match the normal point cloud and the standard point cloud. In actual application process, as other implementation manners, the implementer can also use other point cloud registration algorithms such as the Normal Distribution Transform (NDT) algorithm. The selection of the point cloud registration algorithm is not specially limited in the present embodiment.

[0096] The Iterative Closest Point (ICP) algorithm is a known technology in the field of point cloud registration, and its specific principle process will not be repeated here.

[0097] Further, the difference between the concave-convex coefficients of each matching point pair and the distance between the corresponding matching pairs of three-dimensional coordinates are multiplied to obtain the key index between each matching point pair.

[0098] Thus, by comparing the normal point cloud and the standard point cloud, the key index between the matching point pairs of the normal point cloud and the standard point cloud is obtained, which further prepares for the selection of the subsequent monitoring points.

[0099] Preferably, the key index extraction process provided by the present embodiment is shown in the schematic diagram as Figure 2 .

[0100] Step S3: Deformation monitoring of the highway bridge based on the key index.

[0101] Among the normal point cloud and the standard point cloud, the key indexes between all the matched point pairs are taken as the input of the clustering algorithm, the absolute value of the key index difference between the matched point pairs is taken as the metric distance of the clustering algorithm, the elbow method is used to determine the number of clustering clusters, a plurality of clustering clusters are output and recorded as a clustering set, the mean value of all the key indexes in each clustering set is calculated and recorded as a key mean value, the clustering set with the largest key mean value is taken as a key cluster, and the points belonging to the normal point cloud in the matched point pairs corresponding to all the key indexes in the key cluster are taken as monitoring points. The points in the key cluster are the points that have a greater impact on the bridge deformation when the vehicle is normally driven.

[0102] It is to be supplemented that the k-means clustering algorithm is used to cluster the key indexes in the embodiment.

[0103] If the difference between the monitoring points and the matched points in any coordinate axis exceeds the preset deformation threshold, it is determined that the bridge has deformed, otherwise, it is determined that the bridge has not deformed.

[0104] It should be noted that the value of the preset threshold is artificially set, and in the embodiment, the value of the preset threshold is 3 mm. In actual application, the implementer can also set it by himself according to the specific situation, and the embodiment does not have special restrictions.

[0105] So far, the application evaluates the noise influence degree dynamically by analyzing the influence of environmental factors, vehicles and the like on the point cloud data, denoises various point clouds, further, compares the differences between the normal point cloud and the standard point cloud to register the two kinds of point clouds, improves the selection accuracy of the monitoring points, and further improves the monitoring accuracy of the subsequent bridge deformation.

[0106] Based on the same inventive concept as the above method, the embodiment of the application also provides a highway bridge deformation monitoring system based on laser measurement, which implements the highway bridge deformation monitoring method based on laser measurement described above. The system comprises a point cloud acquisition module, a point cloud data analysis module, and a bridge deformation monitoring module:

[0107] The point cloud acquisition module is used to acquire the three-dimensional point cloud data of the bridge when there is no vehicle passing and the three-dimensional point cloud of the bridge road surface when the vehicle is normally passing by using the vehicle-mounted three-dimensional laser scanner, and each is recorded as a standard point cloud and a normal point cloud. The vehicle-mounted laser scanner travels at a constant speed along the central axis of the bridge from one end of the bridge to the other end, the driving direction is taken as the y axis, the lateral direction of the bridge is taken as the x axis, and the direction perpendicular to the x axis and the y axis is taken as the z axis. The collected standard point cloud data and normal point cloud data are transmitted to the point cloud data analysis module;

[0108] The point cloud data analysis module is configured to analyze and process the data obtained by the point cloud acquisition module, and the analysis and processing method is S2 in the highway bridge deformation monitoring method based on laser measurement. Further, the processed result is input to the bridge deformation monitoring module.

[0109] The bridge deformation monitoring module is configured to monitor the bridge deformation based on the result output by the point cloud data analysis module, and the monitoring method is S3 in the highway bridge deformation monitoring method based on laser measurement.

[0110] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. Moreover, the above-mentioned embodiments of the present application are described. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0111] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments.

[0112] The above-mentioned is only the preferred embodiment of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for monitoring the deformation of highway bridges based on laser measurement, wherein the method is implemented by a pavement smoothness detection system based on laser measurement, characterized in that, The method includes the following steps: S1: Use the point cloud acquisition module of the system to acquire various point clouds of the road surface of the bridge to be tested, including standard point clouds and ordinary point clouds. S2: Compare ordinary point clouds with standard point clouds to determine key indices, specifically: All edge points in various point clouds are acquired, and all edge points are clustered. The positive and negative differences between the x-coordinate values ​​of all edge points in each cluster are analyzed to determine the lateral confidence of each cluster, so as to select two lateral clusters from all clusters of various point clouds. Based on the slope difference of the fitted line obtained from the x and y coordinates of all edge points between the two lateral clusters and the curvature difference of all edge points, the difference index between the two lateral clusters is determined. Combined with the local change trend of the curvature of all edge points in each lateral cluster, the various point clouds are denoised. All the denoised point clouds are recorded as feature point clouds. Point cloud data excluding all edge points in various feature point clouds are denoted as planar points. A fitting process is performed on each planar point and all planar points in its neighborhood. The error in the fitting process is analyzed, and the concavity / convexity coefficient of each planar point is determined by combining the dispersion of the curvature of each planar point and all planar points in its neighborhood. All planar points in various feature point clouds are clustered. By analyzing the proportion of all planar points in each cluster to the total number of planar points in the corresponding feature point cloud and the average distribution of the concavity / convexity coefficient of all planar points in each cluster, feature clusters for various feature point clouds are selected. Obtain all matching point pairs between feature clusters of different feature point clouds, and determine the key indices between matching point pairs by analyzing the distance between each matching point pair and the difference in the concavity and convexity coefficients between each matching point pair. S3: Based on the aforementioned key indices, conduct deformation monitoring of highway bridges; The acquisition process of the standard point cloud and the normal point cloud is as follows: the vehicle-mounted 3D laser scanner is used to acquire the 3D point cloud data of the bridge when there are no vehicles passing by, and the 3D point cloud of the bridge road surface when there are normal vehicles passing by, and these are respectively recorded as the standard point cloud and the normal point cloud. The vehicle-mounted laser scanner travels at a constant speed along the central axis of the bridge from one end of the bridge to the other end. The direction of travel is regarded as the y-axis, the transverse direction of the bridge is regarded as the x-axis, and the z-axis is set in the direction perpendicular to the x-axis and y-axis. The deformation monitoring of highway bridges includes: Between ordinary point clouds and standard point clouds, the key indices between all matching point pairs are used as input to the clustering algorithm, and multiple clusters are output and denoted as cluster sets. The mean of all key indices in each cluster set is calculated and denoted as the key mean. The cluster set with the largest key mean is taken as the key cluster. The points in the ordinary point cloud that belong to the matching point pairs corresponding to all key indices in the key cluster are taken as monitoring points. If the difference between the coordinate values ​​of the monitoring point and its matching point on any coordinate axis exceeds the preset deformation threshold, the bridge is determined to be deformed; otherwise, the bridge is determined not to be deformed.

2. The method for monitoring highway bridge deformation based on laser measurement as described in claim 1, characterized in that, The method for determining the lateral confidence of each cluster is as follows: The total number of edge points with negative x-coordinates and the total number of edge points with positive x-coordinates within each cluster are divided by the total number of edge points within the corresponding cluster. The results of the division are recorded as positive and negative proportions, respectively. The absolute value of the difference between the positive and negative proportions is taken as the lateral confidence of each cluster.

3. The method for monitoring highway bridge deformation based on laser measurement as described in claim 1, characterized in that, The two side clusters are the clusters corresponding to the first two values ​​in the descending order of the side confidence scores of all clusters in various point clouds.

4. The method for monitoring highway bridge deformation based on laser measurement as described in claim 1, characterized in that, The difference index between the two side clusters is the sum of the slope difference of the fitted line obtained from the (x, y) coordinates of all edge points between the two side clusters and the curvature difference of all edge points.

5. The method for monitoring highway bridge deformation based on laser measurement as described in claim 1, characterized in that, The denoising of various point clouds includes: The curvature of all edge points within each side cluster is used as the input of the mutation point detection algorithm, and the output is the mutation curvature. The edge point corresponding to the mutation curvature is denoted as the mutation point. The curvature of each mutation point and all edge points in its neighborhood are arranged in ascending order of their respective y-coordinates. In the arrangement result, the curvature of all edge points before each mutation point is used to form the pre-curvature sequence of each mutation point, and the curvature of all edge points after each mutation point is used to form the post-curvature sequence of each mutation point. First-order difference sequences of the front and back curvature sequences are obtained respectively and denoted as front curvature change sequence and back curvature change sequence. The ratio of the number of negative elements to the total number of elements in the front curvature change sequence is denoted as the increasing proportion, and the ratio of the number of positive elements to the total number of elements in the back curvature change sequence is denoted as the decreasing proportion. The average of the increasing proportion and the decreasing proportion is used as the deformation index of each side cluster. The ratio of the difference index between two side clusters to the average deformation index of the two side clusters is used as the noise interference coefficient between the two side clusters. The noise interference coefficient between two side clusters is used as the input of an exponential function with the natural constant as the base, and the output is used as the neighborhood parameter in the filtering algorithm. The filtering algorithm is then used to denoise various point clouds.

6. The method for monitoring highway bridge deformation based on laser measurement as described in claim 1, characterized in that, The method for determining the concavity / convexity coefficients of each plane point is as follows: Calculate the mean square error during the fitting process for each plane point and all plane points in its neighborhood; The product of the degree of dispersion of the curvature of each plane point and all plane points in its neighborhood with the mean square error is used as the concavity / convexity coefficient of each plane point.

7. The method for monitoring highway bridge deformation based on laser measurement as described in claim 1, characterized in that, The method for obtaining the feature clusters of the various feature point clouds is as follows: In various feature point clouds, the ratio of the number of all planar points in each cluster to the total number of planar points in the corresponding feature point cloud is calculated and denoted as the proportion of each cluster. The ratio of the proportion of each cluster to the average concavity coefficient of all planar points in the corresponding cluster is denoted as the confidence ratio of each cluster. Among various feature point clouds, the cluster with the highest confidence ratio is taken as the feature cluster of each feature point cloud.

8. The method for monitoring highway bridge deformation based on laser measurement as described in claim 1, characterized in that, The key index between each matching point pair is the result of multiplying the difference in the concavity / convexity coefficient between each matching point pair by the distance between the three-dimensional coordinates of the corresponding matching pair.

9. A laser-based highway bridge deformation monitoring system, implementing the laser-based highway bridge deformation monitoring method as described in any one of claims 1-8, characterized in that, The system includes a point cloud acquisition module, a point cloud data analysis module, and a bridge deformation monitoring module. The point cloud acquisition module is used to acquire 3D point cloud data of the bridge when there are no vehicles passing by, and 3D point cloud data of the bridge road surface when there are normal vehicles passing by, and respectively denoted as standard point cloud and normal point cloud. The vehicle-mounted laser scanner travels at a constant speed along the central axis of the bridge from one end to the other. The direction of travel is regarded as the y-axis, the lateral direction of the bridge is regarded as the x-axis, and the z-axis is set in the direction perpendicular to the x-axis and y-axis. The acquired standard point cloud data and normal point cloud data are transmitted to the point cloud data analysis module. The point cloud data analysis module is used to analyze and process the data acquired by the point cloud acquisition module. The analysis and processing method adopts S2 in the laser measurement-based highway bridge deformation monitoring method, and the processed results are input into the bridge deformation monitoring module. The bridge deformation monitoring module is used to monitor bridge deformation based on the results output by the point cloud data analysis module. The monitoring method adopts S3 in the laser-based highway bridge deformation monitoring method.

Images