A deformation detection method of a tunnel structure and a medium

Abstract

The application provides a deformation detection method of a tunnel structure, comprising the following steps: performing point cloud matching on point cloud maps of two time points before and after a tunnel, calculating distances between corresponding point clouds in the point cloud maps of the two time points before and after the tunnel, and performing probability density statistics on the distances to obtain rough deformation point clouds, and then performing density clustering processing on the rough deformation point clouds to obtain a plurality of clustering clusters. The number of the clustering clusters represents the number of deformations, and the distance between the centroid of each clustering cluster and the corresponding first data point is a deformation variable. The above method avoids the process of data fitting, avoids fitting errors, uses a density clustering algorithm to cluster point clouds, and then calculates data in a clustering cluster, so that the calculation steps can be simplified, the calculation complexity can be reduced, and the measurement accuracy can be improved.

Description

Technical Field

[0001] This invention generally relates to the field of tunnel deformation detection technology, and specifically to a method and medium for detecting the deformation of a tunnel structure. Background Technology

[0002] Under long-term vibration loads, the soft soil layer beneath the tunnel will continuously soften, leading to overall settlement of the tunnel wall along the axial direction and local cross-sectional deformation, which seriously affects the safe operation of the subway tunnel. Therefore, it is extremely urgent and important to develop a subway tunnel deformation detection device and adopt effective detection technology and data processing methods to detect the deformation status information of subway tunnels and provide disaster early warning.

[0003] This places higher demands on the level of intelligent safety production and the construction of safety monitoring systems in coal mines. Coal mine tunnels need to pass through mountains, and the engineering process and mining activities will affect the stress balance of the surrounding soil layers. The rock and coal around the tunnels will slowly shift, deform, or even be damaged, which will seriously affect the stability of the tunnel structure and endanger the safety of construction workers.

[0004] Current methods for detecting deformation of tunnel structures based on 3D laser scanning technology mostly employ fitting the tunnel cross-section or fitting the tunnel centerline. This approach has at least the following drawbacks:

[0005] (1) The fitting process of these methods will introduce fitting error in the deformation detection and analysis of roadway tunnels. The systematic error in the data acquisition process of the sensor is unavoidable. The interference of the two errors will inevitably reduce the accuracy of deformation detection.

[0006] (2) Data acquisition and the fitting curve of tunnel cross section and the fitting curve of tunnel centerline are generally carried out in steps. This step-by-step process will complicate the deformation detection process and result in low detection efficiency. Summary of the Invention

[0007] In view of the above-mentioned defects or deficiencies in the prior art, it is desirable to provide a method for detecting the deformation of tunnel structures.

[0008] On the one hand, this invention provides a method for detecting the deformation of a tunnel structure, including:

[0009] The initial 3D point cloud map data of the tunnel is obtained to obtain the first point cloud map; the tunnel does not exhibit deformation in the first point cloud map; the first point cloud map contains multiple first data points;

[0010] Obtain the current 3D point cloud map data of the tunnel to obtain a second point cloud map; the first point cloud map and the second point cloud map correspond to the same tunnel segment; the second point cloud map contains multiple second data points;

[0011] The nearest point iterative algorithm is used to perform point cloud matching on the first point cloud map and the second point cloud map, and the first point cloud map and the second point cloud map are transformed to the same coordinate system to obtain the third point cloud map; the third point cloud map includes multiple first data points and multiple second data points that correspond to each other after point cloud matching.

[0012] Calculate the distance between each pair of corresponding first data points and second data points in the third point cloud map to obtain the first distance set;

[0013] Perform probability density statistics on the elements in the first spacing set, and take the second data point corresponding to the first spacing within the first set interval as the rough deformation point cloud;

[0014] The coarse deformation point cloud is subjected to density clustering using a density clustering algorithm to obtain an accurate deformation point cloud; the accurate deformation point cloud contains multiple second clusters;

[0015] The number of the second cluster is counted, and this number is used as the number of tunnel deformations.

[0016] Calculate the centroid of the second cluster and use the distance between the centroid of the second cluster and the corresponding first data point as the deformation magnitude.

[0017] According to the technical solution provided by the present invention, the step of using the nearest point iteration algorithm to perform point cloud matching on multiple first local maps and multiple second local maps, and transforming the multiple first local maps and multiple second local maps to the same coordinate system includes:

[0018] Calculate the Euclidean distance between each second data point and each first data point;

[0019] Find the nearest second data point to each first data point, and then take the nearest first data point and the nearest second data point in Euclidean distance as two corresponding points;

[0020] Calculate the rotation matrix between the first point cloud map and the second point cloud map based on the corresponding first and second data points to obtain the second rotation matrix;

[0021] The coordinates of multiple second data points in the first point cloud map are calculated based on the second rotation matrix to obtain multiple transformed coordinates;

[0022] Based on the transformed coordinates, the second data point is added to the first point cloud map, and the third point cloud map is calculated.

[0023] According to the technical solution provided by the present invention, after performing point cloud matching on the first point cloud map and the second point cloud map using the nearest point iteration algorithm, and transforming the first point cloud map and the second point cloud map to the same coordinate system, and before performing the step of calculating the distance between each pair of corresponding first data points and second data points in the third point cloud map, the method further includes:

[0024] The third point cloud map is segmented to obtain multiple local point cloud maps; each local point cloud map includes multiple corresponding first data points and multiple second data points.

[0025] Based on multiple first data points and multiple second data points, calculate the rotation matrix between corresponding first data points and second data points to obtain multiple first rotation matrices;

[0026] Calculate the determinant of each of the first rotation matrices; a first rotation matrix whose determinant is not equal to 1 indicates that there is a deformation between the first data point and the second data point;

[0027] The local point cloud map containing the first and second data points corresponding to the rotation matrix whose determinant is not equal to 1 is taken as the deformation region, and multiple first deformation regions are obtained.

[0028] Multiple first deformation regions are used as the third point cloud map.

[0029] According to the technical solution provided by the present invention, the step of calculating the distance between a first data point and a second data point at the same location in a corresponding tunnel to obtain a first distance set includes:

[0030] The third point cloud map is divided into dense and sparse regions;

[0031] Using the third set value as the search radius, an octree method is used to search for multiple first data points within the search radius around each second data point in both dense and sparse regions;

[0032] Calculate the Euclidean distance between each second data point and its corresponding first data point within the dense region to obtain the spacing between the first data point and the second data point within the dense region;

[0033] Based on multiple first data points within the search radius of each second data point in the sparse region, calculate the plane where the multiple first data points are located to obtain the reference plane for each second data point;

[0034] Calculate the distance from the second data point to the corresponding reference plane to obtain the spacing between the first data point and the second data point in the sparse region;

[0035] The spacing between the first and second data points in the dense region and the spacing between the first and second data points in the sparse region are used together as elements in the first spacing set.

[0036] According to the technical solution provided by the present invention, the step of dividing the third point cloud map into dense and sparse regions includes:

[0037] Using the third set value as the search radius, search for the number of data points surrounding the second data point;

[0038] The number of data points surrounding the second data point is taken as the density of the second data point;

[0039] Second data points with a density greater than or equal to a second set value are classified as dense regions, and second data points with a density less than the second set value are classified as sparse regions.

[0040] According to the technical solution provided by the present invention, the step of searching multiple first data points within the search radius of each second data point in a sparse region using an octree method includes:

[0041] S61: The cube space centered at the second data point and with the Nth reference value as the side length is taken as the Nth search space; the initial value of N is set to 1;

[0042] S62: Search for the number of first data points in the Nth search space to obtain the number of Nth data points;

[0043] S63: Determine whether the number of the Nth data point is less than 4;

[0044] S64: When the number of the Nth data points is greater than or equal to 4, take half of the Nth reference value as the (N+1)th reference value, increment the N value by one, and repeat steps S61 to S63.

[0045] S65: When the number of the Nth data points is less than 4, proceed with steps S66 to S67;

[0046] S66: Take all the first data points in the N-1th search space as multiple first data points within the search radius of the second data point, and end the loop.

[0047] According to the technical solution provided by the present invention, the step of calculating the plane containing the plurality of first data points within the search radius of each second data point in the sparse region, and obtaining the reference plane for each second data point, includes:

[0048] Multiple first data points within the search radius of the second data point are obtained to obtain multiple third data points;

[0049] Calculate the center point of the plurality of the third data points;

[0050] The covariance matrix of the second data point is calculated based on the multiple third data points and the center point;

[0051] Calculate the eigenvectors of the covariance matrix;

[0052] Multiple components of the feature vector are used as multiple first plane parameter values ​​of the reference plane;

[0053] The inner product of the coordinates of the center point is calculated and the negative value is taken to obtain the second plane parameter value of the reference plane.

[0054] The reference plane is constructed based on a plurality of first plane parameter values ​​and second plane parameter values.

[0055] According to the technical solution provided by the present invention, the step of performing density clustering processing on the coarse deformed point cloud using a density clustering algorithm to obtain an accurate deformed point cloud includes:

[0056] Get the neighborhood radius;

[0057] Calculate the number of data points within the neighborhood radius for each of the second data points in the coarse deformation point cloud;

[0058] The second data point whose number of data points within the neighborhood radius is greater than the fourth preset value is taken as the core data point;

[0059] Core data points whose distance to each other is less than the neighborhood radius are divided into the same cluster, resulting in multiple first clusters;

[0060] The second data points within the neighborhood radius coverage of all core data points in the first cluster and the first cluster are collectively regarded as the second cluster, resulting in multiple second clusters;

[0061] Multiple second clusters are used as precise deformed point clouds.

[0062] According to the technical solution provided by the present invention, the step of calculating the centroid of the second cluster includes:

[0063] Obtain all core data points contained within the second cluster;

[0064] Calculate the average coordinates of all core data points;

[0065] The average coordinates of all core data points are used as the centroid of the second cluster.

[0066] On the other hand, this embodiment provides a storage medium storing a program for executing a deformation detection method for a tunnel structure; when the program for the deformation detection method for a tunnel structure is executed, it is used for:

[0067] Perform a deformation detection method for a tunnel structure as described in any of the above-mentioned methods.

[0068] The beneficial effects of this invention are as follows:

[0069] A method for detecting deformation of tunnel structures is designed, comprising: matching point cloud maps of the tunnel at two different time points, calculating the distance between corresponding point clouds at the two time points, performing probability density statistics on the distances to obtain a coarse deformation point cloud, and then performing density clustering on the coarse deformation point cloud to obtain multiple clusters. The number of clusters represents the number of deformations, and the distance between the centroid of each cluster and its corresponding first data point is the deformation amount. This method avoids the data fitting process, thus avoiding fitting errors; using a density clustering algorithm to cluster the point cloud and then calculating the data within a cluster simplifies the calculation steps, reduces computational complexity, and improves measurement accuracy. Attached Figure Description

[0070] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0071] Figure 1 A flowchart illustrating a method for detecting the deformation of a tunnel structure provided by the present invention;

[0072] Figure 2 This is a flowchart for searching data points using an octree. Detailed Implementation

[0073] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0074] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0075] Example 1

[0076] The implementation process of this invention can be summarized as follows: acquiring a point cloud map of the tunnel in its initial stage without deformation and the current point cloud map; performing point cloud matching on the two maps and converting them to the same coordinate system; calculating the distance between data points in the two point cloud maps; filtering the data points and performing density clustering to divide them into multiple clusters; calculating the deformation quantity and specific deformation amount based on the clusters. Various methods are used to reduce the total computational load and minimize the impact of errors.

[0077] Please refer to Figure 1 The present invention provides a method for detecting the deformation of a tunnel structure, comprising:

[0078] S1: Obtain the initial 3D point cloud map data of the tunnel to obtain the first point cloud map; the tunnel in the first point cloud map has no deformation; the first point cloud map contains multiple first data points;

[0079] S2: Obtain the current 3D point cloud map data of the tunnel to obtain a second point cloud map; the first point cloud map and the second point cloud map correspond to the same tunnel segment; the second point cloud map contains multiple second data points;

[0080] S3: Use the nearest point iteration algorithm to perform point cloud matching on the first point cloud map and the second point cloud map, and transform the first point cloud map and the second point cloud map to the same coordinate system to obtain the third point cloud map; the third point cloud map includes multiple first data points and multiple second data points that correspond to each other after point cloud matching.

[0081] S4: Calculate the distance between each pair of corresponding first data points and second data points in the third point cloud map to obtain the first distance set;

[0082] S5: Perform probability density statistics on the elements in the first spacing set, and take the second data point corresponding to the first spacing within the first set interval as the rough deformation point cloud.

[0083] S6: The coarse deformation point cloud is subjected to density clustering processing using a density clustering algorithm to obtain an accurate deformation point cloud; the accurate deformation point cloud contains multiple second clusters;

[0084] S7: Count the number of the second cluster and use the number of the second cluster as the number of tunnel deformations;

[0085] S8: Calculate the centroid of the second cluster and use the distance between the centroid of the second cluster and the corresponding first data point as the deformation magnitude.

[0086] Specifically, the first set interval is set based on the probability density statistical results, which have a mean and a standard deviation. The first set interval is set to be greater than the mean minus 2 standard deviations and less than the mean plus 4 standard deviations.

[0087] In this embodiment, the first set interval is specifically set to a range of 2cm to 8cm.

[0088] The distance between the deformed point clouds is divided with the mean value as the center and 2cm as the dividing interval. Points with the first interval located in the second, third and fourth intervals (points with deformation of 2cm to 8cm) are output as coarse deformed point clouds.

[0089] By performing probability density statistics on the distance between the first and second point clouds and then filtering the data, cases of excessively small or large distances can be eliminated. This process also removes instances where the originally undeformed location was not determined by measurement errors, where the measurement results show only minor deformation, and where the measured deformation exceeds the actual possible range due to various error factors. This method not only reduces the impact of errors but also decreases the total amount of data requiring calculation, thus improving computational efficiency.

[0090] In some implementations, a robot equipped with a lidar is used to collect data, and the FAST-LIO algorithm (FAST-LIO: A Fast, Robust LiDAR-inertial Odometry Package by Tightly-Coupled Iterated Kalman Filter, a lidar-inertial coupling mapping algorithm based on tightly coupled iterative Kalman filtering) is used to build a three-dimensional point cloud map of the tunnel, ultimately obtaining a first point cloud map and a second point cloud map.

[0091] The process of acquiring the first point cloud map is as follows: Initially after tunnel construction, a 3D point cloud map is collected and saved. After a period of time, a 3D point cloud map of the tunnel under its current condition is collected again, and the deformation detection method described above is applied to both point cloud maps. This allows us to obtain the amount of deformation and the deformation at each point in the tunnel under its current condition.

[0092] The ousterOs0-64 lidar uses 865nm wavelength infrared light as its light source and is equipped with a built-in inertial measurement unit (IMU) that supports SLAM (Simultaneous Localization and Mapping) algorithms. It outputs uniformly spaced 3D point cloud images and pixel-aligned 2D photographic images. The ousterOs0-64 lidar has a vertical field of view of 90°, a horizontal field of view of 360°, a vertical resolution of 64 lines, and can output up to 1.3 million points per second.

[0093] The FAST-LIO algorithm can use a new rule to calculate the Kalman gain, and make the computational cost of the process dependent only on the size of the state variables rather than proportional to the amount of conventional observation data, thereby solving the problem of the surge in computational cost caused by a large number of feature points.

[0094] The above method avoids the data fitting process and avoids fitting errors. By using density clustering algorithm to cluster point clouds and then calculating the data within a cluster, the calculation steps can be simplified, the computational complexity can be reduced, and the measurement accuracy can be improved.

[0095] Regularly using the method provided by this invention to perform deformation detection on tunnels enables continuous monitoring of tunnel deformation. When a tunnel experiences deformation exceeding the danger threshold, the specific location and amount of deformation can be identified promptly.

[0096] Furthermore, the step of using the nearest point iterative algorithm to perform point cloud matching on multiple first local maps and multiple second local maps, and transforming the multiple first local maps and multiple second local maps to the same coordinate system includes:

[0097] Calculate the Euclidean distance between each second data point and each first data point;

[0098] Find the nearest second data point to each first data point, and then take the nearest first data point and the nearest second data point in Euclidean distance as two corresponding points;

[0099] Calculate the rotation matrix between the first point cloud map and the second point cloud map based on the corresponding first and second data points to obtain the second rotation matrix;

[0100] The coordinates of multiple second data points in the first point cloud map are calculated based on the second rotation matrix to obtain multiple transformed coordinates;

[0101] Based on the transformed coordinates, the second data point is added to the first point cloud map, and the third point cloud map is calculated.

[0102] Specifically, the nearest point iteration algorithm is abbreviated as ICP algorithm (Iterative Closest Point).

[0103] The process of transforming the second data point into the first point cloud map based on the corresponding first and second data points includes:

[0104] First, obtain the initial rotation matrix, then calculate the Euler angles and translation between the first and second data points, and finally substitute the Euler angles and translation into the initial rotation matrix.

[0105] The initial rotation matrix is ​​shown in Equation (I);

[0106]

[0107] Where α, β, and γ represent the Euler angles between the three coordinate axes. In this embodiment, the Euler angles and translation amounts between the three coordinate axes of the two-point cloud map are measured.

[0108] Then, by substituting the Euler angles into the initial rotation matrix, the second rotation matrix is ​​obtained; finally, by substituting the appropriate translation amount and the second rotation matrix into the coordinate transformation formula, the coordinates of the second data point in the first point cloud map can be calculated, resulting in multiple transformed coordinates.

[0109] The coordinate transformation formula is shown in Formula (II);

[0110]

[0111] in, The coordinates represent the transformed coordinates; in this embodiment, they represent the coordinates of the second data point in the first point cloud map. The coordinates before transformation are represented, and in this embodiment, the coordinates of the second data point in the second point cloud map are substituted; α, β, and γ represent the Euler angles that need to be substituted. This represents the translation vector between two points on the cloud map.

[0112] Specifically, the third point cloud map contains multiple first data points and multiple second data points. After point cloud matching and transformation to the same coordinate system, it is convenient to calculate the distance between corresponding data points.

[0113] Furthermore, after performing point cloud matching on the first and second point cloud maps using the nearest-neighbor iterative algorithm to transform the first and second point cloud maps to the same coordinate system, and before performing the step of calculating the distance between each pair of corresponding first and second data points in the third point cloud map, the method further includes:

[0114] The third point cloud map is segmented to obtain multiple local point cloud maps; each local point cloud map includes multiple corresponding first data points and multiple second data points.

[0115] Based on multiple first data points and multiple second data points, calculate the rotation matrix between corresponding first data points and second data points to obtain multiple first rotation matrices;

[0116] Calculate the determinant of each of the first rotation matrices; a first rotation matrix whose determinant is not equal to 1 indicates that there is a deformation between the first data point and the second data point;

[0117] The local point cloud map containing the first and second data points corresponding to the rotation matrix whose determinant is not equal to 1 is taken as the deformation region, and multiple first deformation regions are obtained.

[0118] Multiple first deformation regions are used as the third point cloud map.

[0119] In some implementations, the third point cloud map is divided into equally spaced segments, with the specific segmentation interval set according to the tunnel length.

[0120] For example, when the tunnel being inspected is short, only 50-100 meters long, the spacing can be set to 0.5-1 meter or any value in between. When the tunnel being inspected is 1-2 kilometers long, the spacing can be set to 5-10 meters or any value in between.

[0121] In this embodiment, the first rotation matrix is ​​the same as the second rotation matrix. In this step, the specific value of the second rotation matrix can be directly obtained and substituted into the calculation.

[0122] When the determinant of the first rotation matrix between the first data point and the second data point is equal to 1, it means that there is no deformation between the two corresponding first data points and the second data point; when the determinant of the first rotation matrix is ​​not equal to 1, it means that there is deformation between them.

[0123] The presence of deformation within the segmented region is determined by checking if the determinant of the rotation matrix equals 1. Regions exhibiting deformation are then selected and used as the third point cloud map.

[0124] This method can filter out regions without deformation, and in subsequent calculations, these regions will not be calculated, which can significantly reduce the total amount of data to be calculated and improve computational efficiency.

[0125] For example, consider inspecting a 1-kilometer-long tunnel with a segmentation interval of 10 meters. For ease of understanding, assume the entire tunnel is divided into 100 local point cloud maps only along its extension direction. If the tunnel actually has twenty deformations, and these deformation locations are distributed across ten of these areas, then the above method can filter out 90 areas without deformations, leaving only 10 areas requiring further calculation. Since the point cloud is not perfectly uniformly distributed, it can be roughly estimated that this reduces the total amount of data by approximately 90%. Therefore, this method improves computational efficiency by significantly reducing the total amount of data.

[0126] This invention divides the input point cloud map into cubes according to a set segmentation method, approximating several points within each cube with its centroid. This method can downsample the point cloud without destroying its overall structural information, preserving deformation information and improving computational efficiency.

[0127] In this embodiment, tunnel deformation information is obtained by calculating and analyzing the positional relationships between corresponding points in the point cloud. This method yields relatively high accuracy in deformation detection, but the calculation process requires high computational speed and accuracy. The OusterOs0-64 lidar scans up to 1.3 million points per second, resulting in rich data density. However, excessive point cloud data can affect the computational efficiency of deformation detection, and the deformation detection process may involve repetitive calculations due to the presence of dense point clouds in small spaces, leading to wasted computing power. Therefore, it is necessary to downsample the data without affecting the accuracy of the lidar data.

[0128] Further, the step of calculating the distance between the first data point and the second data point at the same location in the corresponding tunnel to obtain the first distance set includes:

[0129] The third point cloud map is divided into dense and sparse regions;

[0130] Using the third set value as the search radius, an octree method is used to search for multiple first data points within the search radius around each second data point in both dense and sparse regions;

[0131] Calculate the Euclidean distance between each second data point and its corresponding first data point within the dense region to obtain the spacing between the first data point and the second data point within the dense region;

[0132] Based on multiple first data points within the search radius of each second data point in the sparse region, calculate the plane where the multiple first data points are located to obtain the reference plane for each second data point;

[0133] Calculate the distance from the second data point to the corresponding reference plane to obtain the spacing between the first data point and the second data point in the sparse region;

[0134] The spacing between the first and second data points in the dense region and the spacing between the first and second data points in the sparse region are used together as elements in the first spacing set.

[0135] In some implementations, the third set value is set to 1 centimeter.

[0136] Specifically, dense regions have more data points that are relatively close to each other, resulting in smaller errors. Sparse regions have fewer data points and larger distances between them, making direct calculation of Euclidean distance susceptible to various factors and leading to errors in the distance calculation.

[0137] For the reasons stated above, this embodiment directly calculates the Euclidean distance between data points in dense regions. For points in sparse regions, a reference plane is first calculated, then the distance from the data point to the reference plane is calculated, and the distance from the second data point to the reference plane is used as the distance to the first data point.

[0138] Constructing a reference plane using multiple data points can reduce the impact of errors and ensure the accuracy of the calculation results. Therefore, using different methods for calculation in sparse and dense regions ensures that the computational workload of data point spacing in dense regions does not increase, while reducing the impact of errors on the calculation of data point spacing in sparse regions.

[0139] Furthermore, the step of dividing the third point cloud map into dense and sparse regions includes:

[0140] Using the third set value as the search radius, search for the number of data points surrounding the second data point;

[0141] The number of data points surrounding the second data point is taken as the density of the second data point;

[0142] Second data points with a density greater than or equal to a second set value are classified as dense regions, and second data points with a density less than the second set value are classified as sparse regions.

[0143] In some implementations, the third setting value is set to 1 cm, and the second setting value is set to 24. A density greater than or equal to 24 indicates denseness, and a density less than 24 indicates sparseness.

[0144] In some implementations, the search process includes:

[0145] Get the coordinates of the second data point currently being calculated;

[0146] Iterate through all the data points and calculate the distance from other second data points to the second data point.

[0147] Add the second data point whose distance is less than the search radius to the neighboring point set;

[0148] After the calculation process is completed, the number of data points in the neighboring point set is counted, which gives the number of data points around the second data point.

[0149] Further, refer to Figure 2 The steps for searching multiple first data points within the search radius of each second data point in a sparse region using an octree method include:

[0150] S61: The cube space centered at the second data point and with the Nth reference value as the side length is taken as the Nth search space; the initial value of N is set to 1;

[0151] S62: Search for the number of first data points in the Nth search space to obtain the number of Nth data points;

[0152] S63: Determine whether the number of the Nth data point is less than 4;

[0153] S64: When the number of the Nth data points is greater than or equal to 4, take half of the Nth reference value as the (N+1)th reference value, increment the N value by one, and repeat steps S61 to S63.

[0154] S65: When the number of the Nth data points is less than 4, proceed with steps S66 to S67;

[0155] S66: Take all the first data points in the N-1th search space as multiple first data points within the search radius of the second data point, and end the loop.

[0156] In some implementations, the first reference value is set to 1 cm.

[0157] Based on the above method of dividing dense areas, the number of data points in 1 cubic centimeter of space within a dense area is greater than 24 (including the second data point itself). After one cycle, the search space is reduced to 1 / 8 cubic centimeter.

[0158] If the data points are evenly distributed in space, then the number of data points surrounding the second data point should be around 3 (a total of 4 points including the second data point). However, the actual data points are not evenly distributed, which means that in some cases the number of data points surrounding the second data point is greater than 3, and in other cases less than 3.

[0159] Performing the above loop can result in a number of other data points surrounding the second data point that are found to be greater than 3 but less than 3. This ensures that enough data points are found to construct the reference plane while reducing the total amount of data to be calculated.

[0160] Specifically, when the number of Nth data points is equal to 4, it means that the Nth search space includes the second data point itself and three other second data points. The analytical expression for the reference plane can be calculated based on the coordinates of the other three data points.

[0161] To avoid search errors and reduce calculation errors caused by deviations in the position of individual points, when the number of data points is less than 4, all data points in the search space obtained in the previous step are taken as the nearest multiple data points, and the reference plane is calculated by combining more data points.

[0162] The above method can further reduce computational complexity while ensuring the accuracy of the calculation.

[0163] Further, the step of calculating the plane containing the multiple first data points within the search radius of each second data point in the sparse region, and obtaining the reference plane for each second data point, includes:

[0164] Multiple first data points within the search radius of the second data point are obtained to obtain multiple third data points;

[0165] Calculate the center point of the plurality of the third data points;

[0166] The covariance matrix of the second data point is calculated based on the multiple third data points and the center point;

[0167] Calculate the eigenvectors of the covariance matrix;

[0168] Multiple components of the feature vector are used as multiple first plane parameter values ​​of the reference plane;

[0169] The inner product of the coordinates of the center point is calculated and the negative value is taken to obtain the second plane parameter value of the reference plane.

[0170] The reference plane is constructed based on a plurality of first plane parameter values ​​and second plane parameter values.

[0171] In this embodiment, since mass factors are not considered, the centroid and the centroid are consistent, that is, the center point of the plurality of third data points and the centroid of the plurality of third data points are the same point.

[0172] The covariance matrix is ​​shown in Formula (III);

[0173]

[0174] Among them, M 3×3 Let p represent the third-order covariance matrix, where n represents the total number of the third data points, and p represents the third-order covariance matrix. i This represents the i-th third data point. Indicates the center point, (·) T This represents the transpose of a matrix.

[0175] The center point is calculated using formula (iv):

[0176]

[0177] Calculate the minimum eigenvalue and eigenvector of the covariance matrix according to formula (v);

[0178]

[0179] in, Let V represent the eigenvector and V represent the eigenvalue. Multiple eigenvalues ​​are calculated; in this embodiment, the smallest eigenvalue is selected.

[0180] The eigenvector corresponding to the minimum eigenvalue is calculated. The components represent multiple first plane parameter values ​​a, b, and c respectively; the second plane parameter value d is equal to the coordinates of the center point, the inner product is calculated and the negative value is taken.

[0181] The resulting reference plane is then represented by formula (VI);

[0182] Formula (VI) ax+by+cz+d=0

[0183] Furthermore, the step of performing density clustering processing on the coarse deformed point cloud using a density clustering algorithm to obtain an accurate deformed point cloud includes:

[0184] Get the neighborhood radius;

[0185] Calculate the number of data points within the neighborhood radius for each of the second data points in the coarse deformation point cloud;

[0186] The second data point whose number of data points within the neighborhood radius is greater than the fourth preset value is taken as the core data point;

[0187] Core data points whose distance to each other is less than the neighborhood radius are divided into the same cluster, resulting in multiple first clusters;

[0188] The second data points within the neighborhood radius coverage of all core data points in the first cluster and the first cluster are collectively regarded as the second cluster, resulting in multiple second clusters;

[0189] Multiple second clusters are used as precise deformed point clouds.

[0190] In this embodiment, the neighborhood radius is set to 5cm, and the fourth setting value is set to 10.

[0191] The neighborhood radius needs to be set according to the acquisition accuracy of the lidar sensor, and the fourth setting value is set according to the data density acquired by the lidar sensor.

[0192] Specifically, if there are three core data points, they are named as: First Core Data Point, Second Core Data Point, and Third Core Data Point. Assume the distance between the First Core Data Point and the Second Core Data Point is less than the fourth preset value, the distance between the Second Core Data Point and the Third Core Data Point is less than the fourth preset value, but the distance between the First Core Data Point and the Third Core Data Point is greater than the fourth preset value. In this case, the First Core Data Point, the Second Core Data Point, and the Third Core Data Point all belong to the same cluster.

[0193] This approach removes scattered data points, further reducing the total data volume. It also divides the data points into multiple clusters. Computation is performed only on the data points within each cluster, eliminating the need to consider the relationships between data points in different clusters, thus reducing computational complexity and improving efficiency.

[0194] Further, the step of calculating the centroid of the second cluster includes:

[0195] Obtain all core data points contained within the second cluster;

[0196] Calculate the average coordinates of all core data points;

[0197] The average coordinates of all core data points are used as the centroid of the second cluster.

[0198] In some implementations, only the average coordinates of the core data points within a cluster are calculated, which can eliminate the influence of coordinate deviations of non-core data points within the cluster on the average coordinates, while also reducing the total computation and improving computational efficiency.

[0199] Example 2

[0200] This embodiment provides a storage medium storing a program for executing a deformation detection method for a tunnel structure; when the program for the deformation detection method for a tunnel structure is executed, it is used to:

[0201] Perform a deformation detection method for a tunnel structure as described in any of the above embodiments.

[0202] Specifically, the present invention also provides a storage medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to implement the method as described in the above embodiments, specifically a deformation detection method for a tunnel structure as described in any one of the above embodiments.

[0203] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention is not limited to the specific combination of the above-described technical features, but also includes other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in this invention.

Claims

1. A method for detecting the deformation of a tunnel structure, characterized in that, include: The initial 3D point cloud map data of the tunnel is obtained to obtain the first point cloud map; the tunnel does not exhibit deformation in the first point cloud map; the first point cloud map contains multiple first data points; Obtain the current 3D point cloud map data of the tunnel to obtain a second point cloud map; the first point cloud map and the second point cloud map correspond to the same tunnel segment; the second point cloud map contains multiple second data points; The nearest point iterative algorithm is used to perform point cloud matching on the first point cloud map and the second point cloud map, and the first point cloud map and the second point cloud map are transformed to the same coordinate system to obtain the third point cloud map; the third point cloud map includes multiple first data points and multiple second data points that correspond to each other after point cloud matching. Calculate the distance between each pair of corresponding first and second data points in the third point cloud map to obtain a first distance set, including: The third point cloud map is divided into dense and sparse regions; Using the third set value as the search radius, an octree method is used to search for multiple first data points within the search radius around each second data point in both dense and sparse regions; Calculate the Euclidean distance between each second data point in the dense region and multiple first data points within the surrounding search radius, and take the minimum Euclidean distance as the spacing between the first data point and the second data point; Based on multiple first data points within the search radius of each second data point in the sparse region, calculate the plane where the multiple first data points are located to obtain the reference plane for each second data point; Calculate the distance from the second data point to the corresponding reference plane to obtain the spacing between the first data point and the second data point in the sparse region; The spacing between the first data point and the second data point in the dense region and the spacing between the first data point and the second data point in the sparse region are used together as elements in the first spacing set; The deformation detection method for the tunnel structure also includes: Perform probability density statistics on the elements in the first spacing set, and take the second data point corresponding to the first spacing within the first set interval as the rough deformation point cloud; The coarse deformation point cloud is subjected to density clustering using a density clustering algorithm to obtain an accurate deformation point cloud; the accurate deformation point cloud contains multiple second clusters; The number of the second cluster is counted, and this number is used as the number of tunnel deformations. Calculate the centroid of the second cluster and use the distance between the centroid of the second cluster and the corresponding first data point as the deformation magnitude.

2. The method for detecting deformation of a tunnel structure according to claim 1, characterized in that, The steps for performing point cloud matching on multiple first local maps and multiple second local maps using the nearest point iteration algorithm, and transforming the multiple first local maps and multiple second local maps to the same coordinate system, include: Calculate the Euclidean distance between each second data point and each first data point; Find the nearest second data point to each first data point, and then take the nearest first data point and the nearest second data point in Euclidean distance as two corresponding points; Calculate the rotation matrix between the first point cloud map and the second point cloud map based on the corresponding first and second data points to obtain the second rotation matrix; The coordinates of multiple second data points in the first point cloud map are calculated based on the second rotation matrix to obtain multiple transformed coordinates; Based on the transformed coordinates, the second data point is added to the first point cloud map, and the third point cloud map is calculated.

3. The method for detecting deformation of a tunnel structure according to claim 1, characterized in that, After performing point cloud matching on the first and second point cloud maps using the nearest-neighbor iterative algorithm, and transforming the first and second point cloud maps to the same coordinate system, and before calculating the distance between each pair of corresponding first and second data points in the third point cloud map, the method further includes: The third point cloud map is segmented to obtain multiple local point cloud maps; each local point cloud map includes multiple corresponding first data points and multiple second data points. Based on multiple first data points and multiple second data points, calculate the rotation matrix between corresponding first data points and second data points to obtain multiple first rotation matrices; Calculate the determinant of each of the first rotation matrices; a first rotation matrix whose determinant is not equal to 1 indicates that there is a deformation between the first data point and the second data point; The local point cloud map containing the first and second data points corresponding to the rotation matrix whose determinant is not equal to 1 is taken as the deformation region, and multiple first deformation regions are obtained. Multiple first deformation regions are used as the third point cloud map.

4. The method for detecting deformation of a tunnel structure according to claim 1, characterized in that, The steps for dividing the third point cloud map into dense and sparse regions include: Using the third set value as the search radius, search for the number of data points surrounding the second data point; The number of data points surrounding the second data point is taken as the density of the second data point; Second data points with a density greater than or equal to a second set value are classified as dense regions, and second data points with a density less than the second set value are classified as sparse regions.

5. The method for detecting deformation of a tunnel structure according to claim 1, characterized in that, The steps for searching multiple first data points within the search radius of each second data point in a sparse region using an octree method include: S61: The cube space centered at the second data point and with the Nth reference value as the side length is taken as the Nth search space; the initial value of N is set to 1; S62: Search for the number of first data points in the Nth search space to obtain the number of Nth data points; S63: Determine whether the number of the Nth data point is less than 4; S64: When the number of the Nth data points is greater than or equal to 4, take half of the Nth reference value as the (N+1)th reference value, increment the N value by one, and repeat steps S61 to S63. S65: When the number of the Nth data points is less than 4, proceed with steps S66 to S67; S66: Take all the first data points in the N-1th search space as multiple first data points within the search radius of the second data point, and end the loop.

6. The method for detecting deformation of a tunnel structure according to claim 1, characterized in that, The steps for calculating the plane containing multiple first data points within the search radius of each second data point in the sparse region, and obtaining the reference plane for each second data point, include: Multiple first data points within the search radius of the second data point are obtained to obtain multiple third data points; Calculate the center point of the plurality of the third data points; The covariance matrix of the second data point is calculated based on the multiple third data points and the center point; Calculate the eigenvectors of the covariance matrix; Multiple components of the feature vector are used as multiple first plane parameter values ​​of the reference plane; The inner product of the coordinates of the center point is calculated and the negative value is taken to obtain the second plane parameter value of the reference plane. The reference plane is constructed based on a plurality of first plane parameter values ​​and second plane parameter values.

7. The method for detecting deformation of a tunnel structure according to claim 1, characterized in that, The steps for performing density clustering on the coarse deformed point cloud to obtain the accurate deformed point cloud include: Get the neighborhood radius; Calculate the number of data points within the neighborhood radius for each of the second data points in the coarse deformation point cloud; The second data point whose number of data points within the neighborhood radius is greater than the fourth preset value is taken as the core data point; Core data points whose distance to each other is less than the neighborhood radius are divided into the same cluster, resulting in multiple first clusters; The second data points within the neighborhood radius coverage of all core data points in the first cluster and the first cluster are collectively regarded as the second cluster, resulting in multiple second clusters; Multiple second clusters are used as precise deformed point clouds.

8. The method for detecting deformation of a tunnel structure according to claim 7, characterized in that, The steps for calculating the centroid of the second cluster include: Obtain all core data points contained within the second cluster; Calculate the average coordinates of all core data points; The average coordinates of all core data points are used as the centroid of the second cluster.

9. A storage medium, characterized in that, The storage medium stores a program for executing a deformation detection method for a tunnel structure; when the program for the deformation detection method for a tunnel structure is executed, it is used to: The deformation detection method for a tunnel structure as described in any one of claims 1-8 shall be implemented.

Images