Multi-beam sonar pile foundation scouring detection method

Abstract

The invention discloses a multi-beam sonar pile foundation scouring detection method. The method comprises the following steps that a multi-beam pile foundation three-dimensional point cloud to be subjected to scouring detection is read; pile foundation scouring pit point cloud is extracted from the multi-beam pile foundation three-dimensional point cloud; a clustering segmentation method is adopted, pile foundation scouring pit point clouds are grouped, a plurality of clusters are formed, and each cluster represents one piece of independent scouring pit data; based on the multiple clusters, the pile foundation scouring volume is calculated to serve as the basis of pile foundation scouring detection; according to the multi-beam sonar pile foundation scouring detection method, the scouring detection precision of the pile foundation is improved, and pile foundation scouring degree information with higher quality is provided for operation and maintenance work of an offshore wind power pile foundation.

Description

Technical Field

[0001] The invention belongs to the technical field of marine engineering, and in particular relates to a multi-beam sonar pile foundation scour detection method. Background Art

[0002] With the increasing demand for global sustainable development, offshore wind power pile foundations have gradually become an emerging clean energy source. However, due to the influence of the hydrodynamic environment such as waves, currents and sediments on the seabed, the original sediment transport balance will be destroyed, resulting in local scouring around the pile foundation, which in turn affects the safety and economy of the wind power pile foundation. Therefore, the operation and maintenance of offshore wind power pile foundations is very important. It is necessary to regularly monitor the status of wind power pile foundations and changes in the surrounding environment to timely discover potential risks and take preventive measures. Multi-beam sonar is widely used in seabed exploration, channel measurement and underwater structure positioning because it can quickly and efficiently obtain high-resolution topographic point clouds of the seabed. However, multi-beam sonar is difficult to express the specific scouring degree of the pile foundation, and cannot provide data support for pile foundation operation and maintenance. It is necessary to calculate the volume of the pile foundation scouring pit. Therefore, how to calculate the volume of the scouring pit has become a research hotspot in the field of offshore wind power pile foundation scouring detection.

[0003] The existing multi-beam sonar pile foundation scour detection is mainly numerical simulation, which mainly uses formulas to calculate the scour depth or uses the pile foundation three-dimensional model for scour numerical simulation. Although numerical simulation has achieved certain predictions, due to the complexity of water patterns, the detection has large differences. Nowadays, although the actual detection based on sonar can realize the detection of scour, the targeted scour value cannot be obtained. How to obtain accurate pile foundation scour values ​​has become the key in pile foundation scour detection. In addition, due to the interference problem between multiple point clouds during the three-dimensional reconstruction of the seabed point cloud, and the three-dimensional reconstruction parameters need to be manually adjusted according to the expected results, the accuracy of the three-dimensional reconstruction is reduced, which in turn leads to low volume calculation accuracy. Therefore, the research on multi-beam sonar pile foundation scour detection method is helpful to obtain high-precision scour pit volume data and improve the reliability of wind power pile foundation operation and maintenance. Summary of the invention

[0004] Purpose of the invention: In order to solve the problem that the existing multi-beam sonar pile foundation scour detection cannot obtain accurate pile foundation scour values, the present invention proposes a multi-beam sonar pile foundation scour detection method to obtain high-precision scour pit volume data, providing reliable data support for wind power pile foundation operation and maintenance.

[0005] Technical solution: A multi-beam sonar pile foundation scour detection method, comprising the following steps:

[0006] Step 1: Read the 3D point cloud of the multi-beam pile foundation to be tested;

[0007] Step 2: Extract the pile foundation scour pit point cloud from the multi-beam pile foundation three-dimensional point cloud;

[0008] Step 3: Use clustering segmentation method to group the pile foundation scour pit point cloud into multiple clusters, each cluster represents a separate scour pit data;

[0009] Step 4: Based on multiple clusters, the pile foundation scour volume is calculated as the basis for pile foundation scour detection;

[0010] Wherein, in step 2, the extraction of the pile foundation scour pit point cloud from the multi-beam pile foundation three-dimensional point cloud specifically includes:

[0011] Calculate the Euclidean distance between two adjacent points in the multi-beam pile foundation three-dimensional point cloud, cluster the point cloud according to the Euclidean distance between the two adjacent points, and obtain the neighborhood set centered on each point;

[0012] Based on the neighborhood set centered on each point, the normal vector of each point in the multi-beam pile foundation three-dimensional point cloud is calculated, and the pile foundation scour pit point cloud is extracted according to the normal vector of each point;

[0013] Wherein, in step 4, the pile foundation scour volume is calculated based on multiple clusters, and the specific operations include:

[0014] Based on multiple clusters, the corresponding three-dimensional model is constructed, and the objective function for calculating the scour volume of the pile foundation is constructed:

[0015]

[0016] In the formula, △ represents a triangle element in each cluster. Three adjacent points in each cluster form a triangle patch. All the triangle patches in a cluster are called triangle elements. V(△) is the volume formed by the triangle △. α is the parameter for controlling the refinement of the triangle. r △ is the minimum side length of a triangle; d is the spatial dimension, d = 3;

[0017] The pile foundation scour volume is obtained by solving the objective function of calculating the pile foundation scour volume.

[0018] Furthermore, the point cloud is clustered according to the Euclidean distance between two adjacent points to obtain a neighborhood set centered on each point. The specific operations include:

[0019] Assume that the multi-beam pile foundation 3D point cloud has a total of n points, and each point p i The coordinates are expressed as (x i ,y i ,z i ), i∈(1,n), according to the Euclidean distance between two adjacent points, the point cloud is clustered according to the following formula:

[0020] N ∈ (p i )={d(pi ,p i+1 )≤η}

[0021] Where N ∈ (p i ) represents the point cloud set obtained by clustering in the multi-beam pile foundation 3D point cloud, d(p i ,p i+1 ) represents point p i With point p i+1 The previous Euclidean distance, η represents the threshold.

[0022] Furthermore, the normal vector of each point in the multi-beam pile foundation three-dimensional point cloud is calculated based on the neighborhood set centered on each point, and the pile foundation scour pit point cloud is extracted according to the normal vector of each point. The specific operations include:

[0023] The normal vector of each point in the multi-beam pile foundation 3D point cloud is calculated according to the following formula:

[0024]

[0025] Where N k (p i ) represents the point p i The neighborhood set centered on contains the point p i The nearest k points, k represents a parameter used to define the size and range of the neighborhood;

[0026] The point cloud of pile foundation scour pit is extracted according to the normal vector of each point, which is expressed as:

[0027] p c ={p i ∈P:n i <β)

[0028] Where P represents the multi-beam pile foundation 3D point cloud to be processed, p c represents the scour pit point cloud set after the normal vector is extracted, and β represents the determination parameter of the scour pit degree.

[0029] Furthermore, in step 3, the clustering segmentation method is used to group the pile foundation scour pit point cloud to obtain multiple separate scour pit data. The specific operations include:

[0030] Extracting boundary point clouds from the scour pit point clouds, and removing boundary point clouds from the scour pit point clouds;

[0031] Identify core points from the scour pit point cloud after removing the boundary point cloud to form a core point set;

[0032] According to the density clustering algorithm, the core point set is grouped to form multiple clusters.

[0033] Furthermore, the extraction of the boundary point cloud from the scour pit point cloud specifically includes:

[0034] Calculate and sort the polar angles of the scour pit point cloud;

[0035] For any point p in the scour pit point cloud i , i∈(0,n), calculated according to the following formula (p top-1 -p top )×(p i -p top )’s cross product:

[0036]

[0037] (p top-1 -p top )×(p i -p top )>0

[0038] In the formula, p top and p top-1 As the top of stack S, where p top It is the top of the stack. Stack S needs to be filled with the initial p top-1 and p top Point, p top and p top-1 The initial point is taken from the two points p0 and p1 in the scour pit point cloud;

[0039] If (p top-1 -p top )×(p i -p top ) is greater than zero, it means that point pi is outside the boundary formed by the top of the stack; if (p top-1 -p top )×(p i -p top ) is less than or equal to zero, then the top point p of the stack is top Pop the stack and put point p i Push it into the stack, making it the new top of the stack p top ;

[0040] When all points in the scour pit point cloud have been processed, the points in the stack are the extracted boundary point cloud.

[0041] Furthermore, the identification of core points from the scour pit point cloud after removing the boundary point cloud to form a core point set specifically includes:

[0042] Define point O i The neighborhood set N ∈ (O i ), the neighborhood set N ∈ (Oi ) includes all distance points O i Points that do not exceed the value of δ:

[0043]

[0044] N ∈ (O i )={O i+1 ∈O:d(O i , O i+1 )≤δ}

[0045] In the formula, the parameter δ is used to control the range of the neighborhood;

[0046] Calculate point O i The neighborhood set N ∈ (O i ) within the number of points|N ∈ (O i )|, if point O i The neighborhood set N ∈ (O i ) is greater than the preset minimum point threshold MinPts, then the point O i is regarded as a core point and saved in the core point set, otherwise the point O i Marked as noise.

[0047] Furthermore, the core point set is grouped according to the density clustering algorithm to form multiple clusters, and the specific operations include:

[0048] For a core point O i and its neighborhood set N∈(O i ), initialize cluster C = {O i};

[0049] For each point O j ∈N ∈ (O i ), if point O j is the core point, and point Then point O j Add to cluster C, expressed as: C = C ∪ {O j};

[0050] Update point O j The neighborhood set N ∈ (O j ):

[0051] N ∈ (O j )={O k ∈PO:d(O j , O k )≤δ}

[0052] To N ∈ (O j ) for each point O k Make a judgment, if O k is the core point, and point Then O k Add to cluster C, expressed as: C = C ∪ {O k};

[0053] Update point O k Neighborhood N ∈ (O k ):

[0054] N ∈ (O k )={O m ∈PO:d(O k , O m )≤δ}

[0055] Step by step process the neighborhood set N ∈ (O i ) until no more core points can be added to cluster C;

[0056] This forms multiple clusters, each cluster representing a separate flushing pit data.

[0057] Furthermore, the objective function of calculating the pile foundation scour volume is solved to obtain the pile foundation scour volume, which specifically includes:

[0058] Calculate the gradient of the objective function F(α) with respect to the parameter α gradient is the rate of change of the objective function F(α) in each parameter direction, defined as:

[0059]

[0060] In the formula, α=(α1,α2,…α n ) is a parameter vector for controlling the algorithm for calculating the scour pit volume;

[0061] Use the gradient descent method to iteratively update the parameter α:

[0062]

[0063] In the formula, α k , α k+1 are the parameter values ​​for the kth and k+1th iterations respectively; λ represents the learning step size; Indicates that the objective function F(α) is at the current parameter α k The gradient at

[0064] When the following iterative convergence conditions are met, the iterative update ends and the pile foundation scour volume is obtained:

[0065] |F(α k+1 )-F(α k )|≤γ.

[0066] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0067] (1) The method for extracting scour pits from pile foundation three-dimensional point cloud data designed by the present invention is based on the characteristics of large unevenness of the terrain after pile foundation scour, and introduces a normal vector to distinguish the scour pit point cloud on the basis of clustering the pile foundation point cloud by calculating the Euclidean distance, thereby forming a scour pit extraction method for pile foundation three-dimensional point cloud data, improving the clarity of the scour pit point cloud, and performing targeted processing on the scour pit for later detection, thereby improving the accuracy of multi-beam sonar pile foundation scour detection;

[0068] (2) The clustering and segmentation of scour pit data designed by the present invention introduces the recognition and removal of the boundaries of the scour pit point cloud on the basis of calculating the polar angles of the scour pit point cloud and sorting them, that is, the scour pit is processed using the convex hull algorithm; a method for identifying the core points of the scour pit is designed, and the scour pit point cloud is clustered and segmented into separate scour pit point clouds, thereby eliminating the interference phenomenon between each scour pit when calculating the volume, avoiding the subsequent step of calculating the volume of the interfering part of the scour pit, reducing the error of the volume calculation, and thus improving the calculation accuracy of the scour pit volume;

[0069] (3) The method for calculating the scour volume of pile foundation designed in the present invention uses a single scour pit point cloud obtained by clustering and segments and processes it using an objective function designed based on volume calculation. The result is accurately updated through iteration, and the termination of the iteration is determined by the stability of the result volume, thereby obtaining a high-precision scour pit volume, thereby improving the accuracy of multi-beam sonar pile foundation scour detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is a general flow chart of a multi-beam sonar pile foundation scour detection method of the present invention;

[0071] Figure 2 is the point cloud data of a single scour pit, and the two sets of data are different types of scour pits, respectively. Figure 2 (a) in the figure shows the point cloud containing cylindrical scour pits. Figure 2 (b) shows the point cloud containing the conical scour pit;

[0072] Figure 3 The final models of the two groups of scour pit point clouds are: Figure 3 (a) in the figure shows the final model of the cylindrical scour pit point cloud. Figure 3 (b) shows the final model of the conical scour pit point cloud;

[0073] Figure 4 There are two sets of multi-scour pit point cloud data, namely Figure 4 (a) in the figure shows the point cloud of square and cylindrical scour pits. Figure 4 (b) shows the point cloud of hemispherical and hexagonal scour pits;

[0074] Figure 5 is the final model of each cluster of point clouds in the two groups, respectively Figure 5 (a) in the figure shows the final model of the square scour pit point cloud. Figure 5 (b) in the figure shows the final model of the cylindrical scour pit point cloud. Figure 5 (c) in the figure shows the final model of the hemispherical scour pit point cloud. Figure 5 (d) in the figure shows the final model of the hexagonal scour pit point cloud;

[0075] Figure 6 These are four sets of multi-beam initial point cloud data of offshore wind power pile foundations. The four sets of data are divided into four time periods, namely Figure 6 (a) in the figure represents the data on June 26, 2021. Figure 6 (b) in the figure represents the data on January 16, 2022. Figure 6 (c) in the figure represents the data as of June 15, 2022. Figure 6 (d) in the figure represents data as of January 2, 2023;

[0076] Figure 7 The scour pit point clouds and their corresponding volumes after the four sets of data extraction are respectively Figure 7 (a) in the figure indicates that the volume of the scour pit corresponding to the data on June 26, 2021 is 53.2818 cubic meters. Figure 7 (b) in the figure indicates that the volume of the scour pit corresponding to the data on January 16, 2022 is 77.2933 cubic meters. Figure 7 (c) in the figure indicates that the scour pit volume corresponding to the data on June 15, 2022 is 77.6207 cubic meters. Figure 7 (d) in the figure indicates that the scour pit volume corresponding to the data on January 2, 2023 is 67.9871 cubic meters. DETAILED DESCRIPTION

[0077] The technical solution of the present invention is now further described in conjunction with the accompanying drawings and embodiments.

[0078] like Figure 1 As shown, the present invention discloses a multi-beam sonar pile foundation scour detection method to obtain high-precision scour pit volume data, which specifically adopts the following steps:

[0079] Step 1: Read the 3D point cloud of the multi-beam pile foundation to be tested;

[0080] Step 2: Multi-beam sonar can obtain a high-resolution three-dimensional point cloud of offshore wind turbine pile foundations, but it cannot intuitively express the seabed terrain information. Relevant studies have shown that multi-beam sonar can clearly reveal the phenomenon of seabed scour. If these data can be analyzed in depth, it will provide a more reliable basis for scour detection. In order to make up for the fact that the multi-beam sonar pile foundation point cloud cannot process the scour pit point cloud, this embodiment extracts the pile foundation scour pit point cloud from the multi-beam pile foundation three-dimensional point cloud through the normal vector discrimination method, thereby improving the clarity of the scour pit point cloud, and performing targeted processing of the scour pit for later detection, thereby improving the accuracy of multi-beam sonar pile foundation scour detection. Specific operations include:

[0081] Step 2-1: Assume that the multi-beam pile foundation 3D point cloud data has a total of n points, and each point p i The coordinates of (i∈(1,n)) are expressed as (x i ,y i ,z i ), calculate the Euclidean distance between adjacent points in the multi-beam pile foundation 3D point cloud data:

[0082]

[0083] Step 2-2: Based on the Euclidean distance, cluster the point cloud according to the following formula:

[0084] N ∈ (p i )={d(p i ,p i+1 )≤η}

[0085] Wherein, the threshold η is a key parameter adjusted according to the density of the point cloud and the analysis requirements. The appropriate selection of η value will directly affect the number of points in the neighborhood, and thus affect the calculation of the normal vector of the pile foundation point cloud.

[0086] Step 2-3: Calculate the normal vector of each point and extract the point cloud of the pile foundation scour pit based on it, and design the following formula:

[0087]

[0088] Where N k (p i ) is based on point p i The neighborhood set centered on contains the point p i The k nearest points. The parameter k is used to define the size and scope of the neighborhood.

[0089] Step 2-4: The normal vector of each point obtained reflects its corresponding point p i The average position change of surrounding points facilitates the description of the local shape and surface features of the point cloud;

[0090] p c ={p i ∈P:n i <β}

[0091] Wherein, the threshold β is a judgment parameter for the extent of the scour pit, reflecting the point showing obvious scour pit characteristics in the z direction.

[0092] Step 3: Using the clustering segmentation method, multiple separate scour pit point cloud data are formed from the pile foundation scour pit point cloud, thereby avoiding the interference phenomenon between multiple scour pits in the pile foundation scour pit point cloud, reducing the subsequent volume calculation steps for the scour pit interference part, and thus improving the calculation accuracy of the scour pit volume. This embodiment includes the following specific operations:

[0093] Step 3-1: Calculate and sort the polar angle of the pile foundation scour pit point cloud according to the following formula: Polar angle sorting can form a contour of the point cloud data, which is conducive to contour feature extraction and is a preparatory step for the subsequent scour pit boundary point cloud recognition. So I only mention it here.

[0094]

[0095] Step 3-2: Identify the boundary point cloud of the pile foundation scour pit according to the following formula:

[0096]

[0097] (p top-1 -p top )×(p i -p top )>0

[0098] If the cross product is greater than zero, then point p i Outside the boundary formed by the top of the stack, the point p at the top of the stack needs to be top Pop the stack and put point p i Push it into the stack, making it the new top of the stack p top This process will continue until all points have been processed. Finally, the points in the stack are extracted as the vertex set S, which is the boundary point cloud of the pile foundation scour pit.

[0099] Step 3-3: Remove the point cloud boundary of the pile foundation scour pit according to the following formula:

[0100] O=P c -S

[0101] In the formula, p c is the initial set of all pile foundation scour pit points, and S is the boundary point cloud of the pile foundation scour pit.

[0102] Step 3-4: Identify the core points of the pile foundation scour pit according to the following sub-steps:

[0103] (a) Computational neighborhood: Define point O i The neighborhood set N ∈ (O i ), the neighborhood set N ∈ (O i ) includes all distance points O i Points that do not exceed the value of δ:

[0104]

[0105] N ∈ (O i )={O i+1 ∈O:d(O i ,O i+1 )≤δ}

[0106] In the formula, the parameter δ is used to control the range of the neighborhood;

[0107] (b) Core point identification: Calculate each point O i Number of points in the neighborhood |N ∈ (O i )|. If point O i If the number of points in the neighborhood is greater than the preset minimum point threshold MinPts, then the point O i is regarded as a core point and saved in the core point set PO. Otherwise, the point O i Marked as noise.

[0108] (c) Core point clustering: The core point set PO is grouped according to the density clustering algorithm to form multiple clusters. These clusters represent different scour pit areas in the pile foundation scour pit point cloud. The following are the specific steps for grouping according to the density clustering algorithm:

[0109] (c.1) For a core point O i and its neighborhood N ∈ (O i ), initialize cluster C as follows:

[0110] C={O i}

[0111] (c.2) For each point O j ∈N ∈ (O i ), if O j is the core point, and O j Not in any cluster C:

[0112]

[0113] Then O j Add to cluster C:

[0114] C=C∪{O j}

[0115] Update neighborhood N ∈ (O j ):

[0116] N ∈ (O j )={O k ∈PO:d(O j ,O k )≤δ}

[0117] (c.3) for N ∈ (O j ) for each point O k Make a judgment, if O k It is the core point and Then O k Add to cluster C:

[0118] C=C∪{O k}

[0119] Update neighborhood N ∈ (O k ):

[0120] N ∈ (O k )={O m ∈PO:d(O k ,O m )≤δ}

[0121] (c.4) Process the points in the neighborhood step by step until there are no more core points to add to cluster C;

[0122] When all points in the neighborhood are processed and there are no more core points to expand cluster C, the cluster expansion process ends and the final point cloud in each cluster is obtained. i , repeat the above clustering steps ((c.1)~(c.4)) to identify all clusters, and the noise points will not be assigned to any cluster.

[0123] This embodiment converts the entire scour pit terrain point cloud into multiple separate scour pit point clouds for processing, eliminating the interference phenomenon between each scour pit when calculating the volume, avoiding the subsequent volume calculation step of the interfering part of the scour pit, reducing the error of the volume calculation, and thereby improving the calculation accuracy of the scour pit volume.

[0124] Step 4: Based on the point cloud data of multiple separate scour pits, construct the corresponding 3D model and calculate the scour volume of the pile foundation. Specifically include:

[0125] Specific formula for determining pile foundation scour volume calculation:

[0126]

[0127] In the formula, △ represents a triangle element in each cluster. Three adjacent points in each cluster form a triangle patch. All the triangle patches in a cluster are called triangle elements. V(△) is the volume formed by the triangle △. α is the parameter that controls the refinement of the triangle, which affects the shape, size and complexity of the triangle. r △ is the minimum side length of a triangle; d is the spatial dimension, d=3.

[0128] Use the following steps to determine the parameters V(△), r in the expression of the pile foundation scour volume calculation method: △ :

[0129] The volume V(△) formed by the triangle △ can be calculated from the three vertices P1=(x1,y1,z1), P2=(x2,y2,z2) and P3=(x3,y3,z3):

[0130]

[0131] Scale parameter r △ is the minimum value among the side lengths of △:

[0132] r Δ =min(||p2-p1||,||p3-p1||,||p3-p2||

[0133] In the formula, ||·|| represents the Euclidean distance.

[0134] In order to optimize the objective function F(α), it is necessary to calculate its gradient with respect to the parameter α gradient is the rate of change of the objective function in each parameter direction, defined as:

[0135]

[0136] In the formula, α=(α1,α2,…α n ) is a parameter vector that controls the algorithm for calculating the scour pit volume;

[0137] In order to find the optimal solution of the objective function F(α), the gradient descent method is used to iteratively update the parameter α:

[0138]

[0139] In the formula, α k , α k+1 are the parameter values ​​for the kth and k+1th iterations respectively; λ represents the learning step size, which is used to control the step size of each iteration; is the objective function at the current parameter α k The gradient at .

[0140] The iterative convergence condition is:

[0141] |F(α k+1 )-F(α k )|≤γ.

[0142] This embodiment adopts a separate scour pit point cloud data to adopt a pile foundation scour volume calculation method, that is, using the designed objective function for calculation, accurately updating the result through an iterative method, and using the stability of the result volume to determine the termination of the iteration, thereby obtaining a high-precision scour pit volume, improving the accuracy of obtaining the scour pit volume, and thus improving the efficiency and accuracy of multi-beam sonar pile foundation scour detection.

[0143] In order to verify the effectiveness of the multi-beam sonar pile foundation scour detection method proposed in this embodiment, a single scour pit three-dimensional point cloud, multiple scour pit three-dimensional point clouds, and actual wind turbine pile foundation point cloud were selected to perform scour pit scour detection verification.

[0144] Figure 2 and Figure 3 The results of scour detection of two groups of single scour pit point clouds are given. Figure 2 As shown in the figure, the types of the two groups of scour pits are cylindrical and conical. Figure 3 It can be seen that each group of scour pits can be completely extracted and constructed into a three-dimensional model. The final volume results are shown in Table 1. It can be seen from the percentage error data in Table 1 that the algorithm has a very ideal error control performance when processing single scour pit point cloud data. The algorithm can provide highly accurate results, and the volume accuracy reaches an average of 99%, which shows that the algorithm can meet high standards in terms of accuracy.

[0145] Table 1

[0146] name <![CDATA[Theoretical scour pit volume (m 3 )]]> <![CDATA[Test scour pit volume (m 3 )]]> <![CDATA[Difference (m 3 )]]> percentage Cylindrical 3.14 3.1381 0.0034 0.108％ Conical 1.04 1.0331 0.0069 0.663％

[0147] Figure 4 and Figure 5 The results of scour detection of two groups of multi-scour pit point clouds are given. Figure 4 As shown in the figure, one of the two groups of point cloud models contains square and cylindrical scour pits; the other group of point cloud models contains hemispherical and hexagonal scour pits. Figure 5It can be seen that each group can be completely extracted into a separate scour pit point cloud and constructed into a three-dimensional model. The final volume results are shown in Table 2. It can be clearly seen that when processing multiple scour pit point cloud data, the algorithm can still effectively obtain accurate volume data and keep the error within a small range. The average accuracy reaches 99.4%. This result fully verifies the efficiency and accuracy of the algorithm under complex terrain conditions.

[0148] Table 2

[0149]

[0150]

[0151] Through the detection of the above two types of scour pit point cloud data, the scour pit extraction method of the pile foundation three-dimensional point cloud data designed in this embodiment, based on the characteristics of large uneven terrain after pile foundation scour, introduces the normal vector to distinguish the scour pit point cloud on the basis of clustering the pile foundation point cloud by calculating the Euclidean distance, forming the scour pit extraction of the pile foundation three-dimensional point cloud data, enhancing the intuitiveness and pertinence of the scour pit, thereby improving the accuracy of multi-beam sonar pile foundation scour detection. The designed clustering and segmentation of scour pit data introduces the identification and removal of the scour pit point cloud boundary on the basis of calculating the polar angle of the scour pit point cloud and sorting it, that is, the scour pit is processed using the convex hull algorithm; the identification method of the core point of the scour pit is designed, and the scour pit is clustered and segmented into separate scour pit point clouds, eliminating the interference phenomenon between each scour pit when calculating the volume, thereby improving the calculation accuracy of the scour pit volume. The designed method for calculating the scour volume of pile foundation adopts the individual scour pit point cloud segmented by clustering, and uses the designed objective function for calculation. The result is accurately updated through iteration, and the termination of iteration is judged by the stability of the result volume, so as to obtain the high-precision scour pit volume, which improves the accuracy of obtaining the scour pit volume, thereby improving the efficiency and accuracy of multi-beam sonar pile foundation scour detection.

[0152] Figure 6 and Figure 7 Four sets of offshore wind power pile foundation point cloud data detection results at different time periods are given. Figure 6 As shown in the figure, four groups of initial point cloud data of offshore wind power pile foundations in different time periods are used to detect scour. The final three-dimensional model diagram and its corresponding scour pit volume are shown in the figure. Figure 7 As shown in the figure, it can be clearly seen that after the actual offshore wind power pile foundation point cloud extraction and segmentation, a relatively complete scour pit point cloud can still be obtained and its accurate scour pit volume can be calculated.

[0153] The formation of seabed scour pits is mainly due to the action of water flow, especially when the water flow speed exceeds a certain threshold, the sediment on the bed surface will be washed away, thus forming scour pits. Figure 7The displayed point cloud data after scouring shows that the volume of the scouring pit gradually increases, but the volume of the scouring pit in the last time period is significantly reduced. This is because the initial scouring may remove loose sediments, but as the scouring continues, changes in the direction and speed of the water flow may redeposit these sediments into the scouring pit, filling up part of the volume of the scouring pit, resulting in a decrease in the depth of the scouring pit, which in turn causes the scouring pit to become relatively shallow or reduce in volume. Based on the above actual wind turbine pile foundation detection, two sets of actual offshore wind turbine pile foundation scouring part volume detections were also conducted, as shown in Table 3.

[0154] Table 3

[0155]

[0156]

[0157] Therefore, the multi-beam sonar pile foundation scour detection method of this embodiment helps to obtain high-precision scour pit volume data, provides more intuitive and reliable pile foundation scour degree values, and improves the reliability and safety of offshore wind power pile foundation operation and maintenance.

Claims

1. A multi-beam sonar pile foundation scour detection method, characterized in that: The following steps are involved: Step 1: Read the 3D point cloud of the multi-beam pile foundation to be tested; Step 2: Extract the pile foundation scour pit point cloud from the multi-beam pile foundation three-dimensional point cloud; Step 3: Use clustering segmentation method to group the pile foundation scour pit point cloud into multiple clusters, each cluster represents a separate scour pit data; Step 4: Based on multiple clusters, the pile foundation scour volume is calculated as the basis for pile foundation scour detection; Wherein, in step 2, the extraction of the pile foundation scour pit point cloud from the multi-beam pile foundation three-dimensional point cloud specifically includes: Calculate the Euclidean distance between two adjacent points in the multi-beam pile foundation three-dimensional point cloud, cluster the point cloud according to the Euclidean distance between the two adjacent points, and obtain the neighborhood set centered on each point; Based on the neighborhood set centered on each point, the normal vector of each point in the multi-beam pile foundation three-dimensional point cloud is calculated, and the pile foundation scour pit point cloud is extracted according to the normal vector of each point; Wherein, in step 4, the pile foundation scour volume is calculated based on multiple clusters, and the specific operations include: Based on multiple clusters, the corresponding three-dimensional model is constructed, and the objective function for calculating the scour volume of the pile foundation is constructed: In the formula, △ represents a triangle element in each cluster. Three adjacent points in each cluster form a triangle patch. All the triangle patches in a cluster are called triangle elements. V(△) is the volume formed by the triangle △. α is the parameter for controlling the refinement of the triangle. r △ is the minimum side length of a triangle; d is the spatial dimension, d = 3; The pile foundation scour volume is obtained by solving the objective function of calculating the pile foundation scour volume.

2. A multi-beam sonar pile foundation scour detection method according to claim 1, characterized in that: The point cloud is clustered according to the Euclidean distance between two adjacent points to obtain a neighborhood set centered on each point. The specific operations include: Assume that the multi-beam pile foundation 3D point cloud has a total of n points, and each point p i The coordinates are expressed as (x i ,y i ,z i ), i∈(1,n), according to the Euclidean distance between two adjacent points, the point cloud is clustered according to the following formula: N ∈ (p i )={d(p i ,p i+1 )≤η} Where N ∈ (p i ) represents the point cloud set obtained by clustering in the multi-beam pile foundation 3D point cloud, d(p i ,p i+1 ) represents point p i With point p i+1 The previous Euclidean distance, η represents the threshold.

3. A multi-beam sonar pile foundation scour detection method according to claim 1, characterized in that: The above-mentioned calculation of the normal vector of each point in the multi-beam pile foundation three-dimensional point cloud based on the neighborhood set centered on each point, and the extraction of the pile foundation scour pit point cloud according to the normal vector of each point, specifically includes: The normal vector of each point in the multi-beam pile foundation 3D point cloud is calculated according to the following formula: Where N k (p i ) represents the point p i The neighborhood set centered on contains the distance point p i The nearest k points, k represents a parameter used to define the size and range of the neighborhood; The point cloud of pile foundation scour pit is extracted according to the normal vector of each point, which is expressed as: p c ={p i ∈P:n i <b} Where P represents the multi-beam pile foundation 3D point cloud to be processed, p c represents the scour pit point cloud set after the normal vector is extracted, and β represents the determination parameter of the scour pit degree.

4. A multi-beam sonar pile foundation scour detection method according to claim 1, characterized in that: In step 3, the clustering segmentation method is used to group the pile foundation scour pit point cloud to obtain multiple separate scour pit data. The specific operations include: Extracting boundary point clouds from the scour pit point clouds, and removing boundary point clouds from the scour pit point clouds; Identify core points from the scour pit point cloud after removing the boundary point cloud to form a core point set; According to the density clustering algorithm, the core point set is grouped to form multiple clusters.

5. A multi-beam sonar pile foundation scour detection method according to claim 4, characterized in that: The specific operations of extracting the boundary point cloud from the scour pit point cloud include: Calculate and sort the polar angles of the scour pit point cloud; For any point p in the scour pit point cloud i , i∈(0,n), calculated according to the following formula (p top-1 -p top )×(p i -p top )’s cross product: (p top-1 -p top )×(p i -p top )>0 In the formula, p top and p top-1 As the top of stack S, where p top It is the top of the stack. Stack S needs to be filled with the initial p top-1 and p top Point, p top and p top-1 The initial point is taken from the two points p0 and p1 in the scour pit point cloud; If (p top-1 -p top )×(p i - ptop ) is greater than zero, which means that point p i Outside the boundary formed by the top of the stack; if (p top-1 -p top) ×(p i -p top ) is less than or equal to zero, then the top point p of the stack is top Pop the stack and put point p i Push it into the stack, making it the new top of the stack p top ; When all points in the scour pit point cloud have been processed, the points in the stack are the extracted boundary point cloud.

6. A multi-beam sonar pile foundation scour detection method according to claim 4, characterized in that: The identification of core points from the scour pit point cloud after removing the boundary point cloud to form a core point set specifically includes: Define point O i The neighborhood set N ∈ (O i ), the neighborhood set N ∈ (O i ) includes all distance points O i Points that do not exceed the value of δ: N ∈ (The i )={O i+1 ∈O:d(O i ,O i+1 )≤δ} In the formula, the parameter δ is used to control the range of the neighborhood; Calculate point O i The neighborhood set N ∈ (O i ) |N ∈ (O i )|, if point O i The neighborhood set N ∈ (O i ) is greater than the preset minimum point threshold MinPts, then the point O i is regarded as a core point and saved in the core point set, otherwise the point O i Marked as noise.

7. A multi-beam sonar pile foundation scour detection method according to claim 4, characterized in that: The core point set is grouped according to the density clustering algorithm to form multiple clusters. The specific operations include: For a core point O i and its neighborhood set N ∈ (O i ), initialize cluster C = {O i }; For each point O j ∈N ∈ (O i ), if point O j is the core point, and point Then point O j Add to cluster C, expressed as: C = C ∪ {O j }; Update point O j The neighborhood set N ∈ (O j ): N ∈ (The j )={O k ∈PO:d)O j ,O k )≤δ} To N ∈ (O j ) for each point O k Make a judgment, if O k is the core point, and point Then O k Add to cluster C, expressed as: C = C ∪ {O k }; Update point O k Neighborhood N ∈ (O k ): N ∈ (The k )={O m ∈PO:d(O k ,O m )≤δ} Step by step process the neighborhood set N ∈ (O i ) until no more core points can be added to cluster C; This forms multiple clusters, each cluster representing a separate flushing pit data.

8. The multi-beam sonar pile foundation scour detection method according to claim 1, characterized in that: The objective function of calculating the pile foundation scour volume is solved to obtain the pile foundation scour volume, which specifically includes: Calculate the gradient of the objective function F(α) with respect to the parameter α gradient is the rate of change of the objective function F(α) in each parameter direction, defined as: In the formula, α=(α1,α2,…α n ) is a parameter vector for controlling the algorithm for calculating the scour pit volume; Use the gradient descent method to iteratively update the parameter α: In the formula, α k , α k+1 are the parameter values ​​for the kth and k+1th iterations respectively; λ represents the learning step size; Indicates that the objective function F(α) is at the current parameter α k The gradient at When the following iterative convergence conditions are met, the iterative update ends and the pile foundation scour volume is obtained: |F(a k+1 )-F(a k )|≤γ。

Images