A method for establishing a true model of submarine cables based on three-dimensional real-time sonar data
By dividing the data into voxels and calculating the parameters of the inscribed sphere in three-dimensional sonar data, and combining noise removal and surface density thresholding, the problem of poor data quality of submarine cable structures in complex hydrological environments was solved, and efficient and accurate submarine cable structure modeling was achieved.
Patent Information
- Application Number
- CN202411146704.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-21
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-08-21
AI Technical Summary
Existing technologies struggle to accurately reconstruct submarine cable structures from 3D sonar point cloud data in complex hydrological environments, resulting in poor data quality and difficulty in analyzing submarine cable structural characteristics.
By dividing point cloud data into voxels, calculating the center point and normal vector of the inscribed sphere, and combining the direction vectors of adjacent points and the cable radius, the coordinates of the vertices of the circular surface are established to construct the cable model. Data quality is improved through noise removal and surface density thresholding.
It enables the efficient construction of accurate 3D submarine cable structure models under low-quality data conditions, reducing data processing complexity and improving the accuracy and efficiency of the models.
Smart Images

Figure CN118657893B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of submarine cable structure detection, and in particular relates to the technical field of submarine cable structure detection based on three-dimensional real-time sonar data. Background Art
[0002] In recent years, the country has attached increasing importance to marine resources, and the survey of offshore wind farms, the survey of submarine cable routes, and the operation and maintenance inspection are in full swing. The complex hydrological conditions under the sea will cause scouring and erosion of the seabed. Coupled with some geological hazards and hydrodynamic effects, the wind farm pile foundation and connecting cables may be partially exposed and suspended, which will lead to fatigue damage to the structure and suspended cables. Therefore, we need to accurately detect the real structure of the submarine cable, and analyze whether the central angle of the submarine cable access end is detached, the relationship between the bend limiter and the bell mouth, whether the bend limiter is damaged, and whether the submarine cables at both ends are exposed.
[0003] Commonly used submarine cable structure detection methods include magnetic detection, sonar detection, underwater sound and image detection, etc. Sonar detection is a technology that uses the characteristics of sound waves propagating in water to detect and image. Among them, three-dimensional sonar technology can provide more detailed and accurate information on the structure of the ocean bottom than traditional two-dimensional sonar. Its working principle is to use a three-dimensional sonar equipment system to perform acoustic wave detection on the submarine cable structure to obtain three-dimensional sonar submarine cable point cloud data, and then use professional software (such as PDMS, AutoCAD, CloudCompare, etc.) to convert the three-dimensional sonar submarine cable point cloud data into a three-dimensional model. Extract the structural characteristics of the submarine cable from the three-dimensional model, such as diameter, shape, direction, degree of curvature, etc.
[0004] However, in actual seabed exploration with complex environments, due to the influence of various factors such as survey costs and personnel operations, the final measured three-dimensional sonar cable point cloud data is relatively messy, especially the data signal-to-noise ratio behind the detection target is relatively low, making it difficult to restore the actual cable structure in the three-dimensional modeling process. Summary of the invention
[0005] In view of the problems existing in the prior art, the present invention proposes a method for establishing a real model of a submarine cable based on three-dimensional real-time sonar data. The method can restore an accurate and realistic three-dimensional submarine cable structure model from the messy three-dimensional sonar submarine cable point cloud data in a complex hydrological environment, thereby solving the problem that the quality of the actually acquired underwater three-dimensional sonar data is poor and it is difficult to analyze the structural characteristics of the submarine cable.
[0006] In order to achieve the above-mentioned object of the invention, the present invention adopts the following technical solutions:
[0007] A method for establishing a real model of a submarine cable based on three-dimensional real-time sonar data, characterized in that it comprises the following steps:
[0008] S1. Data Division
[0009] Obtain point cloud data S (x, y, z), and divide the point cloud data S (x, y, z) into n voxels, wherein the point cloud data S (x, y, z) is three-dimensional sonar submarine cable point cloud data collected by a three-dimensional sonar device system;
[0010] S2. Calculation of the coordinates C of the circular surface vertices
[0011] S21, calculating the coordinates p of the center point of the inscribed sphere in each voxel and the corresponding normal vector F2;
[0012] S22, sequentially calculating the direction vectors of the adjacent points between the center point of each inscribed sphere and the center point of the adjacent inscribed sphere
[0013]
[0014] S23, using the coordinates p of the center point of the sphere as the starting point of the vector, calculate the direction vector of the adjacent point
[0015] A set of orthogonal vectors F1 that are orthogonal to the normal vector F2;
[0016] S24. Calculate and obtain the coordinates C of the vertices of the circular surface in each voxel, C=p+r·(F1cosθ+F2sinθ), where θ={0,π,2π,...,jπ}, j is the number of vertices of each circular surface, which is user-defined; and r is the known radius of the submarine cable;
[0017] S3. Model building
[0018] The final real submarine cable model is established based on the circular surface vertex coordinates C.
[0019] The present invention innovatively proposes a method for modeling low-quality three-dimensional sonar cable data. First, the point cloud data S (x, y, z), that is, the three-dimensional sonar cable point cloud data, is obtained based on the three-dimensional sonar system. The point cloud data S (x, y, z) is divided into n voxels, and the voxel is the data unit that constitutes the point cloud data S (x, y, z), which can be divided by the octree method in the prior art; the center point coordinates p and the corresponding normal vector F2 of the inscribed sphere in each voxel are calculated respectively, and the center point coordinates can be calculated using the least squares method, grid search method, etc. in the prior art. The normal vector F2 can be calculated using the commonly used algorithms in the prior art, such as the neighborhood method and the gradient method. After calculating the center point coordinates p and the corresponding normal vector F2 in each voxel, the formula:
[0020] , calculate the adjacent point direction vector between the coordinates of each sphere center point and the adjacent point coordinates The coordinates of the center point p of the sphere are used as the starting point of the vector, and the direction vector of the adjacent point is calculated by the cross multiplication method or the eigenvector method in the prior art. A set of orthogonal vectors F1 are orthogonal to construct the circular surface vertex coordinates C: C = p + r (F1cosθ + F2sinθ), where θ = {0, π, 2π, ..., jπ}, j is the number of vertices of each custom circular surface, and r is the known cable radius. The final real submarine cable model is established based on the circular surface vertex coordinates C. Common professional software for model construction include PDMS, AutoCAD, CloudCompare, etc.
[0021] The above-mentioned method for establishing a real model of a submarine cable based on three-dimensional real-time sonar data divides the messy three-dimensional submarine cable data into n voxels and performs local processing: by calculating the coordinates p of the center point of the inscribed sphere of the voxel and the corresponding normal vector F2, the direction vectors of adjacent points between adjacent center points of the sphere And the given real submarine cable radius r, finally the circular surface vertex coordinates C in each voxel are obtained. The construction of the submarine cable model based on the circular surface vertex coordinates C reduces the complexity of data processing and makes the modeling process more efficient. The method can more accurately capture the three-dimensional structure of the submarine cable and obtain an accurate model even when the data quality is poor. It realizes the restoration of an accurate and realistic three-dimensional submarine cable structure model from the messy point cloud data in a complex hydrological environment, and solves the problem that the quality of the actual underwater three-dimensional sonar data is poor and it is difficult to analyze the characteristics of the submarine cable structure.
[0022] As a preferred embodiment of the present invention, the S1, data division further includes a noise removal step:
[0023] Divide the point cloud data S (x, y, z) into multiple regions, and calculate the surface density ρ in each region.
[0024] ρ=N / (πR 2 ),
[0025] Where N is the number of point cloud data S (x, y, z) in the area; R is the radius of the area, which is user-defined;
[0026] Set the effective surface density threshold ρ t , determine the noise area, if the surface density ρ of one of the areas is less than the effective surface density threshold ρ t , it is determined to be a noise area; the point cloud data S (x, y, z) in the noise area is deleted.
[0027] The spatial distribution of the point cloud data S (x, y, z) obtained by the actual 3D sonar system is not uniform. Some areas are denser and some are sparser. The data in the sparser areas are mostly noise points. This technical solution calculates the surface density ρ in each area and sets the effective surface density threshold ρ t Using the effective surface density threshold ρ as a criterion to identify and remove noise is an effective point cloud data preprocessing method, which helps to improve the efficiency of point cloud data processing and the accuracy of the final result. t It can be set directly by technicians based on experience, or it can be set comprehensively based on the calculated surface density ρ in each of the regions.
[0028] In addition, many point cloud processing algorithms, such as surface reconstruction, feature extraction, classification and recognition, rely on the density and uniformity of point cloud data. After removing noise through the above method, the remaining point cloud data is more concentrated and regular, which can make these algorithms get better results and more accurate results.
[0029] As a preferred embodiment of the present invention, the effective surface density threshold value ρ t =0.1*maximum surface density ρ max , where the maximum surface density ρ max is the maximum value of the surface density ρ within each of the regions.
[0030] As a preferred embodiment of the present invention, the noise removal step further includes filtering processing, removing deviation points and standardization processing in sequence.
[0031] Filtering, removing outliers and standardization are all conventional point cloud data preprocessing methods in the prior art. Common filtering processes include statistical filtering and radius filtering, and common algorithms for removing outliers include clustering-based algorithms and RANSAC (Random Sample Consensus) algorithms.
[0032] As a preferred embodiment of the present invention, the specific calculation steps of the sphere center point coordinate p in S21 are:
[0033] S21-1, select any initial point p0 (x0, y0, z0) inside the voxel, and calculate the distance between the initial point p0 (x0, y0, z0) and each point in the point cloud data S (x, y, z) in the voxel;
[0034] S21-2, select a point v0 in the point cloud data S (x, y, z), the distance between the point v0 and the initial point p0 (x0, y0, z0) ||v0p0|| is the smallest; calculate the moving point direction vector The moving point direction vector is the direction vector between the point v0 and the initial point p0 (x0, y0, z0);
[0035]
[0036] S21-3, moving the initial point p0 along the moving point direction vector Move distance l and get the first moving point p1. Wherein, l is the spatial sampling size of the sonar data; repeat S21-2 to obtain point v1 and the distance between the point v1 and the first moving point p1 ||v1p1||; the moving point direction vector is the direction vector between the point v1 and the first moving point p1;
[0037] S21-4, repeat S21-3 until it moves to the kth moving point p k When we get point v k and ||v k p k ||, satisfies the following formula:
[0038]
[0039] Among them, ||v k p k || is the kth moving point p k With point v k The distance between the points v k is the point cloud data S (x, y, z) corresponding to the kth moving point p k The point with the smallest distance; ||v k+1 p k+1 || is the k+1th moving point p k+1 With point v k+1 The distance between the points v k+1 is the point cloud data S (x, y, z) corresponding to the k+1th moving point p k+1 The point with the smallest distance; For the point v k The area density of the area; For the point v k+1 The surface density of the area; α∈[0, l] is the control weight, which can be customized;
[0040] Determine the kth moving point p k is the coordinate p of the center point of the sphere.
[0041] At this time p k Distance k The density-weighted distance reaches the maximum value within the voxel, that is, p kis the center of the largest inscribed sphere in the voxel, ‖v k p k ‖ is the radius of the inscribed sphere. Considering the working principle and data characteristics of three-dimensional real-time sonar: the data quality is higher where the point cloud density is larger, and vice versa. Therefore, the present invention sets up a density weighting term in the distance comparison, so that the algorithm refers to more high-quality data points when calculating the inscribed sphere, and finally the center point of the inscribed sphere.
[0042] As a preferred embodiment of the present invention, the step S22 further includes a connection sequence acquisition step:
[0043] Connecting the sphere center points based on the Kruskal algorithm to construct a minimum spanning tree, and obtaining an edge list of the minimum spanning tree;
[0044] The edge list of the minimum spanning tree is the connection order of all the sphere center point coordinates p.
[0045] The minimum spanning tree (MST) is an acyclic subgraph containing all vertices in the graph, and the sum of the weights of its edges is minimal. Kruskal's algorithm is a greedy algorithm for constructing a minimum spanning tree. In this way, Kruskal's algorithm can ensure that the generated minimum spanning tree contains all the sphere centers and the total connection distance is minimal.
[0046] As a preferred embodiment of the present invention, the specific steps of constructing the minimum spanning tree in S22 are:
[0047] Initialization steps: set P, E, V, M,
[0048] P: the set of coordinates p of all the sphere center points;
[0049] E: the set of edges obtained by connecting any two of the sphere center point coordinates p;
[0050] V: Vertex set, initially containing an arbitrary coordinate p of the center point of the sphere;
[0051] M: The set of edges, initially empty, will eventually contain all the edges of the minimum spanning tree;
[0052] Kruskal algorithm execution steps:
[0053] Sort all the edges in E from small to large; initialize the forest F, where each vertex is an independent tree;
[0054] Traversal steps
[0055] For each edge (u,v) in E, perform the following steps in sorted order:
[0056] If u and v are in different trees in F, then do the following:
[0057] Add the edge (u,v) to M;
[0058] In forest F, merge the two trees containing u and v into one tree;
[0059] Repeat the traversal steps until any of the following conditions is met:
[0060] V is equal to P, that is, all the center points of the sphere are included in the minimum spanning tree;
[0061] The number of edges in M reaches n-1, that is, the number of edges in the minimum spanning tree is equal to the number of vertices minus one;
[0062] Connect the edges in M end to end in sequence to obtain the connection order of all the sphere center point coordinates p. This order is the edge list of the minimum spanning tree.
[0063] As a preferred embodiment of the present invention, the S3, model building, further comprises a triangular mesh plane building step:
[0064] The circular surface vertex coordinates C are combined in groups of three, one group is a triangular mesh plane, and multiple triangular mesh planes are obtained. All the triangular mesh planes are combined to obtain the final real submarine cable model.
[0065] As a preferred embodiment of the present invention, the triangular mesh plane establishment step is specifically as follows:
[0066] Sort the vertex coordinates C of the circular surface counterclockwise, and the serial numbers of all vertices are {0, 1, 2, ..., j, ..., nj-1}, and construct two triangular mesh surfaces consisting of three vertices each time in the manner of [aj+b, aj+c, (a+1)j+b], [(a+1)j+b, aj+c, (a+1)j+c],
[0067] Among them, a={0,1,2,…,n-1}, b={0,1,2,…,j}, and c is the remainder of (b+1) / j.
[0068] In summary, the present invention has the following beneficial effects:
[0069] 1. By calculating the coordinates p of the center point of the voxel inscribed sphere and the corresponding normal vector F2, the direction vectors of the adjacent points between the adjacent center points And given the actual submarine cable radius r, the circular surface vertex coordinates C in each voxel are finally obtained; constructing the submarine cable model based on the circular surface vertex coordinates C reduces the complexity of data processing and makes the modeling process more efficient.
[0070] 2. By calculating the surface density ρ in each area and setting the effective surface density threshold ρ t Using noise as a criterion to identify and remove noise is an effective point cloud data preprocessing method, which helps to improve the efficiency of point cloud data processing and the accuracy of the final results.
[0071] 3. Taking into account the working principle and data characteristics of three-dimensional real-time sonar: the data quality is higher where the point cloud density is larger, and vice versa; therefore, the present invention sets up a density weighting term in the distance comparison, so that the algorithm refers to more high-quality data points when calculating the inscribed sphere, and finally the center point of the inscribed sphere. BRIEF DESCRIPTION OF THE DRAWINGS
[0072] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0073] Figure 1 It is the processed three-dimensional real-time sonar point cloud data of a certain offshore pile foundation in a specific implementation manner of the present invention;
[0074] Figure 2 It is a surface density distribution diagram of submarine cable point cloud data in a specific implementation manner of the present invention;
[0075] Figure 3 A flow chart of a method for establishing a real model of a submarine cable based on three-dimensional real-time sonar data according to the present invention;
[0076] Figure 4 It is a schematic diagram of the operation of obtaining the center of the maximum inscribed sphere in each voxel of the submarine cable point cloud data in a specific implementation manner of the present invention;
[0077] Figure 5 Schematic diagram of the coordinates p of the sphere center point and the corresponding normal vector in a specific implementation manner of the present invention, wherein Figure 5 (a) Point coordinates p, Figure 5 (b) is the corresponding normal vector;
[0078] Figure 6 This is the connection method of the center points of each maximum inscribed sphere of the submarine cable point cloud data in the specific implementation mode of the present invention;
[0079] Figure 7 It is a schematic diagram of a method for constructing a circular surface vertex in a specific embodiment of the present invention;
[0080] Figure 8A schematic diagram of selecting a triangular mesh surface in a specific implementation manner of the present invention;
[0081] Fig. 9 A schematic diagram of an overall method for establishing a three-dimensional model of a submarine cable in a specific embodiment of the present invention;
[0082] Fig.10 It is an overlay diagram of the submarine cable model and point cloud data finally established in the specific implementation manner of the present invention. DETAILED DESCRIPTION
[0083] The present invention is further described below by specific examples and drawings. The examples of the present invention are intended to enable those skilled in the art to better understand the present invention, and do not limit the present invention in any way.
[0084] First, the point cloud data of a pile foundation detected by real-time three-dimensional sonar is preprocessed by conventional point cloud data preprocessing methods in the prior art, such as filtering, removing outliers, and standardization. Common filtering processes include statistical filtering, radius filtering, etc. Common algorithms for removing outliers include clustering-based algorithms, RANSAC (Random Sample Consensus) algorithms, etc. Finally, the sonar point cloud data is obtained as follows Figure 1 shown.
[0085] It can be seen from the figure that the point cloud data of the submarine cable part is relatively messy, and some point cloud data are missing. Figure 2 As shown, the point cloud data of the submarine cable part is manually intercepted, an area with a radius of R = 0.2m is selected, and the number of points N in each area is counted. The surface density ρ of each point in the point cloud data is calculated.
[0086] ρ=N / (πR 2 ),
[0087] Where N is the number of point cloud data S (x, y, z) in the area; R is the radius of the area, which is user-defined;
[0088] Set the effective surface density threshold ρ t , determine the noise area, if the surface density ρ of an area is less than the effective surface density threshold ρ t , it is judged as a noise area; the point cloud data S (x, y, z) in the noise area is deleted.
[0089] This technical solution calculates the surface density ρ in each area and sets the effective surface density threshold ρ t Using the effective surface density threshold ρ as a criterion to identify and remove noise is an effective point cloud data preprocessing method, which helps to improve the efficiency of point cloud data processing and the accuracy of the final result. tIt can be set directly by technicians based on experience, or it can be set comprehensively based on the calculated surface density ρ in each area. In this embodiment, the effective surface density threshold ρ can be set t =0.1*maximum surface density ρ max , where the maximum surface density ρ max is the maximum value of the surface density ρ in each region.
[0090] In addition, many point cloud processing algorithms, such as surface reconstruction, feature extraction, classification and recognition, rely on the density and uniformity of point cloud data. After removing noise through the above method, the remaining point cloud data is more concentrated and regular, which can make these algorithms get better results and more accurate results.
[0091] Figure 3 The present invention is a schematic diagram of a method for establishing a real model of a submarine cable based on three-dimensional real-time sonar data, which includes the following steps:
[0092] S1. Data division, S2. Calculation of circular surface vertex coordinates C, S3. Model establishment
[0093] The S1 data partitioning steps are as follows:
[0094] Obtain point cloud data S (x, y, z), and divide the point cloud data S (x, y, z) into n voxels, where the point cloud data is three-dimensional sonar cable point cloud data collected by a three-dimensional sonar device system; a voxel is a data unit constituting the point cloud data S (x, y, z), and can be divided by an octree method in the prior art;
[0095] S2. The calculation of the circular surface vertex coordinates C is as follows:
[0096] S21, calculating the coordinates p of the center point of the inscribed sphere in each voxel and the connection sequence of the corresponding normal vector F2 and the center point of the sphere;
[0097] like Figure 4 As shown, firstly, the coordinates p of the center point of the maximum inscribed sphere in each voxel of the submarine cable point cloud data are obtained.
[0098] The calculation steps of the sphere center point coordinates p are as follows:
[0099] S21-1, select any initial point p0(x0, y0, z0) inside the voxel, and calculate the distance between p0(x0, y0, z0) and each point in the point cloud data S(x, y, z) inside the voxel;
[0100] S21-2, select point v0 in the point cloud data S (x, y, z), the distance between point v0 and the initial point p0 ||v0p0|| is the smallest; calculate the direction vector of the moving point Moving point direction vector is the direction vector between point v0 and the initial point p0 (x0, y0, z0);
[0101]
[0102] S21-3, move the initial point p0 along the moving point direction vector Move distance l and get the first moving point p1. Where l is the spatial sampling size of the sonar data; repeat S21-2 to obtain point v1 and the distance between point v1 and the first moving point p1 ||v1p1||; the moving point direction vector is the direction vector between point v1 and the first moving point p1;
[0103] S21-4, repeat S21-3 until it moves to the Kth moving point p k When we get point v k and ||v k p k ||, satisfies the following formula:
[0104]
[0105] Among them, ||v k p k || is the kth moving point p k With point v k The distance between points v k is the point cloud data S (x, y, z) corresponding to the kth moving point p k The point with the smallest distance; ||v k+1 p k+1 || is the k+1th moving point p k+1 With point v k+1 The distance between points v k+1 is the point cloud data S (x, y, z) corresponding to the k+1th moving point p k+1 The point with the smallest distance; For point v k The area density of the area; For point v k+1 The surface density of the area; α∈[0, l] is the control weight, which can be customized;
[0106] Determine the Kth moving point p k is the coordinates of the sphere center point p.
[0107] At this time p k Distance k The density-weighted distance reaches the maximum value within the voxel, that is, p k is the center of the largest inscribed sphere in the voxel, ||v k p k|| is the radius of the inscribed sphere. Taking into account the working principle and data characteristics of three-dimensional real-time sonar: the data quality is higher where the point cloud density is larger, and vice versa. Therefore, the present invention establishes a density weighting term in the distance comparison, so that the algorithm can refer to more high-quality data points when calculating the inscribed sphere. α∈[0, l] is the control weight, and its function is to keep the order of magnitude of the density term consistent with the distance term, and to control the importance of the density term. If the quality of the collected sonar data is relatively messy and the density distribution is uneven, α should take a larger value, otherwise a smaller value. Finally, n inscribed sphere center points are obtained. In this embodiment, the value of α is 0.3.
[0108] like Figure 5 As shown in (a), the coordinates p of the center points of n inscribed spheres are finally obtained.
[0109] like Figure 5 As shown in (b), the corresponding normal vector F2 is calculated based on the coordinates p of the center point of the ball.
[0110] The calculation steps of the normal vector F2 corresponding to the coordinates p of the sphere center are as follows:
[0111] (1) Neighborhood search: For each point, find the set of points in its neighborhood. Usually, k-nearest neighbor search (k-NN) or fixed radius search is used.
[0112] (2) Calculate the covariance matrix: Calculate the covariance matrix for the points in the neighborhood.
[0113] For a point p, find its neighborhood point set {q1, q2, ..., q k}, whose coordinates are (x j ,y j , z j ). Compute the centroid of a set of points:
[0114]
[0115] Then calculate the covariance matrix C:
[0116]
[0117] (3) Eigenvalue decomposition: Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors. The size of the eigenvalue reflects the distribution of the point cloud in the direction of the corresponding eigenvector.
[0118] C=UΛU T
[0119] Among them, Λ is a diagonal matrix, the elements on the diagonal are eigenvalues, and U is the eigenvector matrix.
[0120] (4) Get the normal vector: Select the eigenvector corresponding to the minimum eigenvalue of the covariance matrix as the normal vector of the point. The direction of the eigenvector corresponding to the minimum eigenvalue is the direction in which the point cloud is least variable on the local plane, i.e., the normal vector: normal = u min
[0121] Since this method divides the point cloud data into several voxels and calculates them separately, it supports parallel computing and greatly improves the computing efficiency.
[0122] Get the connection mode between the coordinates p of each sphere center point. The calculation steps are as follows: connect the sphere center points of each voxel by constructing a minimum spanning tree. The minimum spanning tree is constructed using the Kruskal algorithm.
[0123] Assume G = (P, E) is all the connection modes of the inscribed sphere center points, where P is the set of n sphere center points obtained in step S2, and E is the set of edges obtained by connecting any two points; initialize V = {p}, where p is any sphere center point (starting point) in the set V; select the edge with the smallest distance each time, put the vertex into V, and put the selected edge into M, until V = P or the number of edges in M is n-1; connect the edges in M end to end in sequence, and obtain the connection sequence of the sphere center points of all submarine cable data inscribed spheres. The top view (along the X, Y coordinates) and side view (along the X, Z coordinates) of the connection result of the submarine cable data sphere center points in this embodiment are as follows: Figure 6 As shown in (a) and (b).
[0124] S22, such as Figure 7 As shown, the adjacent point squares between the coordinates of each sphere center point and the coordinates of the adjacent points are calculated in turn.
[0125] Vector
[0126]
[0127] S23, using the coordinates of the sphere center point p as the vector starting point, calculate the direction vector of the adjacent point A set of orthogonal vectors F1 that are orthogonal to the normal vector F2;
[0128] S24. Calculate the coordinates C of the circular surface vertices in each voxel, C = p + r (F1cosθ + F2sinθ), where θ = {0, π, 2π, ..., jπ}, j is the number of vertices of each circular surface, which is user-defined; r is the known submarine cable radius; the larger the j value, the more vertices of each circular surface, the smoother the submarine cable model finally established, but the larger the model file. In this embodiment, the j value is 50, and the actual submarine cable radius r is 0.35m.
[0129] S3. Model building
[0130] The final real submarine cable model is established based on the circular surface vertex coordinates C. The specific steps are as follows: Sort the circular surface vertex coordinates C counterclockwise, and the serial numbers of all vertices are {0,1,2,…,j,…,nj-1}. Select 3 points in the following manner to form a triangular mesh surface: [aj+b,aj+c,(a+1)j+b], [(a+1)j+b,aj+c,(a+1)j+c]. Construct two triangular mesh surfaces composed of 3 vertices each time. Among them, a={0,1,2,…,n-1}, b={0,1,2,…,j}, and c is the remainder of (b+1) / j. Figure 8 As shown in the figure, the vertex numbers of the first two mesh planes are [0,1,j] and [j,1,j+1] respectively. Finally, all the triangular mesh planes are combined to obtain the final real submarine cable model, which is saved locally in .obj format.
[0131] Fig. 9 The schematic diagram of the overall method for establishing a three-dimensional model of a submarine cable in a specific embodiment of the present invention is mainly to divide the three-dimensional sonar point cloud data into a number of voxels, and calculate the coordinates of the center point of the inscribed sphere of the three-dimensional submarine cable data in each voxel and the corresponding normal vector; connect the center points of the inscribed spheres in each voxel by constructing a minimum spanning tree; construct the vertices of the submarine cable model according to the center points of the inscribed spheres and the given real submarine cable radius; and establish the three-dimensional submarine cable model according to the coordinates of the circular vertices. Fig.10 It is an overlay diagram of the submarine cable model and point cloud data finally established in the specific implementation manner of the present invention.
[0132] The present invention innovatively proposes a low-quality three-dimensional sonar submarine cable data modeling method, which can restore an accurate and realistic three-dimensional submarine cable structure model from the messy point cloud data in a complex hydrological environment, solving the problem that the quality of the actual acquired underwater three-dimensional sonar data is poor and it is difficult to analyze the submarine cable structure characteristics.
[0133] It should be noted that the above embodiments are only detailed descriptions of the preferred embodiments and principles of the present invention. For ordinary technicians in this field, there will be changes in the specific implementation methods based on the ideas provided by the present invention, and these changes are also considered to be within the scope of protection of the present invention.
Claims
1. A method for establishing a real model of a submarine cable based on three-dimensional real-time sonar data, characterized in that: The following steps are included: S1. Data Division Get point cloud data , the point cloud data Divide into n Voxels, the point cloud data The three-dimensional sonar cable point cloud data collected by the three-dimensional sonar equipment system; S2, circular surface vertex coordinates C calculate S21, calculating the coordinates of the center point of the inscribed sphere in each voxel p and the corresponding normal vector F 2 ; S22, sequentially calculating the direction vectors of the adjacent points between the center point of each inscribed sphere and the center point of the adjacent inscribed sphere , S23, using the coordinates of the center point of the sphere p As the starting point of the vector, calculate the direction vector of the adjacent point and the normal vector F 2 A set of orthogonal vectors that are simultaneously orthogonal F 1 ; S24, calculate and obtain the coordinates of the circular surface vertices in each voxel C , ,in , j The number of vertices of each circular surface is user-defined; r is the known radius of the submarine cable; S3. Model building Based on the vertex coordinates of the circular surface C Build the final realistic submarine cable model; The S1, data partitioning also includes a noise removal step: The point cloud data Divide into multiple regions and calculate the surface density in each region ρ , Where N is the point cloud data in the area The number of; R is the radius of the area, which can be customized; Set effective surface density threshold ρ t , determine the noise area, if the surface density of one of the areas ρ Less than the effective surface density threshold ρ t , it is determined to be a noise area; the point cloud data in the noise area delete; The sphere center point coordinates in S21 p The specific calculation steps are: S21-1, select any initial point inside the voxel , calculate the initial point The distance from the point cloud data within the voxel The distance between the points in S21-2, in the point cloud data Select point , Said point With the initial point Distance Minimum; Calculate the moving point direction vector , the moving point direction vector For the point With the initial point The direction vector of S21-3, moving the initial point along the moving point direction vector Moving distance l , get the first moving point p 1, ,in, l is the spatial sampling size of the sonar data; repeat S21-2 to get point v 1 and the points mentioned v 1 With the first moving point p 1 distance ‖ v 1 p 1‖; moving point direction vector For the point v 1 with the first moving point p 1 is the direction vector; S21-4, repeat S21-3 until moving to the k Mobile points p k When you get v k as well as‖ v k p k ‖, satisfying the following formula: in,‖ v k p k ‖ is the first k Mobile points p k With point v k The distance between the points v k For the point cloud data In the above k Mobile points p k The point with the smallest distance; v k+1 p k+1 ‖ is the first k +1 Movement Point p k+1 With point v k+1 The distance between the points v k+1 For the point cloud data and the k+1th moving point p k+1 The point with the smallest distance; For the point The area density of the area; For the point v k+1 The area density of the area; To control weight, customize; Determine the k Mobile points p k is the coordinate of the center point of the sphere p .
2. The method for establishing a real model of a submarine cable based on three-dimensional real-time sonar data according to claim 1, characterized in that: The effective surface density threshold ρ t =0.1*maximum surface densityρ max , where the maximum surface density ρ max The surface density within each region is ρ The maximum value in .
3. The method for establishing a real model of a submarine cable based on three-dimensional real-time sonar data according to claim 1, characterized in that: The noise removal step also includes filtering processing, removing deviation points and standardization processing in sequence.
4. The method for establishing a real model of a submarine cable based on three-dimensional real-time sonar data according to claim 1, characterized in that: The step S22 also includes a connection sequence acquisition step: Connecting the sphere center points based on the Kruskal algorithm to construct a minimum spanning tree, and obtaining an edge list of the minimum spanning tree; The edge list of the minimum spanning tree is the coordinates of all the sphere center points p The connection order.
5. The method for establishing a real model of a submarine cable based on three-dimensional real-time sonar data according to claim 4, characterized in that: The specific steps of constructing the minimum spanning tree in S22 are: Initialization steps Set P, E, V, M, P: coordinates of all the sphere center points p A collection of; E: Coordinates of any two sphere center points p The set of connected edges; V: Vertex set, initially containing the coordinates of an arbitrary sphere center point p ; M: The set of edges, initially empty, will eventually contain all the edges of the minimum spanning tree; Algorithm execution steps Sort all the edges in E from small to large; initialize the forest F, where each vertex is an independent tree; Traversal steps For each edge in E ( u, v ), perform the following steps in sorted order: if u and v are in different trees in F, then do the following: replace the edge ( u, v ) is added to M; in forest F, it will contain u and v The two trees are merged into one tree; Repeat the traversal steps until any of the following conditions is met: V is equal to P, that is, all sphere center points are included in the minimum spanning tree; The number of edges in M reaches n -1, that is, the number of edges in the minimum spanning tree is equal to the number of vertices minus one; Connect the edges in M end to end to get the coordinates of all the sphere center points. p The connection order is the edge list of the minimum spanning tree.
6. A method for establishing a real model of a submarine cable based on three-dimensional real-time sonar data according to any one of claims 1 to 5, characterized in that: The step S3, model building, also includes the step of establishing a triangular mesh plane: The circular surface vertex coordinates C The three-ply triangles are combined into a triangular mesh plane, and a plurality of triangular mesh planes are obtained. All the triangular mesh planes are combined to obtain the final real submarine cable model.
7. The method for establishing a real model of a submarine cable based on three-dimensional real-time sonar data according to claim 6, characterized in that: The steps of establishing the triangular mesh plane are specifically as follows: The circular surface vertex coordinates C Sorted counterclockwise, the numbers of all vertices are {0,1,2,…, j ,…, nj -1}, follow [ aj + b , aj + c , ( a + 1) j + b ],[( a + 1) j + b , aj + c , ( a + 1) j + c ], constructing two triangular mesh faces consisting of three vertices each time, in, a ={0,1,2,…, n -1}, b ={0,1,2,…, j }, c for( b + 1) / j The remainder of .
