Geometric feature quantification method, system and equipment for underwater structure surface attachments and medium

Abstract

The invention discloses a geometric feature quantification method, system, equipment and medium for surface attachments of an underwater structure, and the method comprises the following steps: collecting an image of an underwater pile foundation containing the surface attachments, and reconstructing and correcting a three-dimensional point cloud model of the underwater pile foundation containing the surface attachments; measuring to obtain an underwater pile foundation three-dimensional point cloud model without surface attachments, and registering the two point cloud models to obtain a reconstruction point cloud and a reference surface point cloud with uniform coordinates; performing coarse segmentation on the reconstructed point cloud based on point cloud distribution characteristics to obtain a point cloud after coarse segmentation, and performing fine segmentation on the point cloud after coarse segmentation based on point cloud RGB color information to obtain a surface attachment point cloud; and performing down-sampling simplification on the surface attachment point cloud and the reference surface point cloud by adopting a cylindrical surface point cloud down-sampling method, and performing feature quantization to obtain the attachment thickness, the attachment area and the attachment volume of the surface attachment. According to the invention, high-efficiency and high-precision extraction of quantitative features from disordered images to attachments is realized.

Description

Technical Field

[0001] This invention belongs to the field of underwater pile foundation monitoring technology, and particularly relates to a method, system, equipment and medium for quantifying the geometric characteristics of underwater structural surface attachments. Background Technology

[0002] Underwater bridge foundations and other structures are exposed to complex aquatic environments for extended periods, making their surfaces susceptible to the adhesion of algae, shellfish, and other organisms or debris. These deposits alter the hydrodynamic properties of the structure, accelerate corrosion, and pose a potential threat to structural safety. Existing detection methods primarily rely on manual diving observation or acoustic testing, which suffers from low accuracy, discontinuous data, and the inability to perform quantitative analysis. In recent years, underwater inspection technologies based on optical images have been developed. However, the underwater environment presents challenges such as light attenuation, scattering, and limited field of view, making it difficult to accurately quantify the three-dimensional morphological characteristics (such as thickness and volume) of deposits using two-dimensional images.

[0003] Image-based 3D reconstruction technology has shown great potential in the field of underwater fouling morphology detection. However, due to the similarity in color and texture between the surface deposits and the structure itself, and the problems of scale uncertainty and noise interference in the reconstructed point cloud, it is particularly difficult to accurately segment the deposits and calculate their geometric features from the point cloud. Existing underwater point cloud processing methods mostly rely on manual point cloud segmentation or threshold-based classification, which are highly sensitive to changes in illumination, background noise, and image quality. Moreover, most algorithms are designed for rigid geometries and are difficult to accurately represent the complex and irregular morphology of fouling. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to provide a method, system, device and medium for quantifying the geometric features of underwater structure surface attachments, which achieves high efficiency and high accuracy in extracting quantitative features of attachments from disordered images.

[0005] Technical solution: To achieve the above objectives, the present invention provides a method for quantifying the geometric features of underwater structure surface attachments, comprising the following steps: S1. Acquire images of underwater pile foundations containing surface attachments, reconstruct and correct the 3D point cloud model of the underwater pile foundations containing surface attachments. Simultaneously, a 3D point cloud model of the underwater pile foundation without surface attachments was obtained by measurement. Then, the two point cloud models were registered to obtain a reconstructed point cloud of the underwater pile foundation with surface attachments and unified coordinates. Reference point cloud of underwater pile foundations without surface attachments ; S2, Based on the reconstructed point cloud and reference surface point cloud Based on the distribution characteristics of point clouds, the reconstruction of point clouds is performed. Perform coarse segmentation to obtain coarsely segmented point clouds. Then, based on the RGB color information of the point cloud, the coarsely segmented point cloud is further processed. Fine segmentation was performed to obtain the point cloud of surface attachments. ; S3. Use cylindrical point cloud downsampling method to analyze the point cloud of surface attachments. and reference surface point cloud Simplify by downsampling, and then analyze the point cloud of the downsampled surface attachments. and downsampled reference surface point cloud Feature quantization is performed to obtain the adhesion thickness, adhesion area, and adhesion volume of the surface deposits.

[0006] Optionally, step S1 specifically includes the following steps: S1.1. Underwater pile foundation image datasets were obtained by taking close-up, full-view photos using an underwater camera. Zhang's calibration method was used to correct distortion in the image datasets. Based on the corrected image datasets, a 3D point cloud model of the underwater pile foundations, including surface attachments, was reconstructed. , of which The points are in For the first The three-dimensional coordinates of the points For the first The RGB color channel values ​​of each point; S1.2. The global scale factor is solved using a least-squares optimization method based on multiple known dimensions. This enables the physicalization of point cloud units and the correction of point cloud scale. Specifically, this involves constructing the objective function. , Optimize the scaling factor using the least squares method. , , In the formula, A 3D point cloud model of underwater pile foundations containing surface attachments. The Middle The distance between points, For underwater 3D point cloud models containing surface attachments The Middle The true distance between points Point-to-point quantity; Based on global scale factor Perform coordinate correction. , Three-dimensional point cloud model of underwater pile foundation containing surface attachments The underwater pile foundation 3D point cloud model was updated to the corrected model, resulting in a reconstructed point cloud of the underwater pile foundation including surface attachments. Among them, the corrected first The points are ; S1.3. A three-dimensional point cloud model of the underwater pile foundation without surface attachments is obtained by measuring with a laser scanner, thus obtaining the reference surface point cloud. The underwater pile foundation axis was determined using the RANSAC cylindrical fitting algorithm. Then, a point cloud was reconstructed using a coordinate rigid body transformation with the axis as the Z-axis, and a feature point-based registration method was employed. Point cloud of reference surface Registration.

[0007] Optionally, the registration method in step S1.3 specifically involves: based on the uniqueness of the intersection points of the three planes in the point cloud (plane-plane-surface), extracting the corrected and reconstructed point cloud respectively. Point cloud of reference surface The three intersection points in the diagram are used as salient feature matching point pairs. Then, the optimal spatial transformation parameters are estimated using the salient feature matching point pairs, where the transformation matrix is... The function is: , In the formula, This represents finding the spatial transformation parameters that minimize the function; the spatial transformation parameters include the rotation matrix. Translation matrix , To reconstruct salient feature points within the point cloud, These are the significant feature points within the reference surface element. and Constitute a pair of points that match significant features; Registration yields a reconstructed point cloud of underwater pile foundations containing surface attachments, with coordinates unified. Reference point cloud of underwater pile foundations without surface attachments Reconstructing point clouds It includes surface attachment points, underwater pile foundation points, noise points, and background points.

[0008] Optionally, step S2 specifically includes the following steps: S2.1 Calculate the reference surface point cloud using the RANSAC cylindrical fitting function in Matlab. axial direction vector and a point on the axis Calculate and reconstruct point clouds Distance from the axis : , In the formula, Let the magnitude of the vector be . Will Sort in descending order, denoted as ;based on Point clouds are classified into surface attachment point clouds and background point clouds; a brute-force search method is used to find... The location of the two optimal inflection points is determined by iterating through the two possible inflection point locations and... Divide the data into three segments, perform linear fitting on each segment, calculate the overall fitting error, and find the two indices of the inflection points where the error is minimized. These indices are then denoted by their ordinates from largest to smallest. and Reconstructing point clouds middle satisfy The points are segmented to obtain a coarsely segmented point cloud. Point cloud after coarse segmentation This includes surface attachment points, noise points, and underwater pile foundation points; S2.2, For point clouds after coarse segmentation An adaptive surface attachment point cloud segmentation method based on RGB color statistical characteristics is adopted. The statistical features of the RGB color values ​​of all points are calculated, including the mean vector. and standard deviation vector Define an adaptive color threshold range for identifying point cloud regions containing surface attachments: , Finally, points within the adaptive color threshold range are considered as surface attachments, and their corresponding spatial coordinates and color information are extracted to construct a surface attachment point cloud. .

[0009] Optionally, step S3 specifically includes the following steps: S3.1. Use cylindrical point cloud downsampling method to sample the point cloud of surface attachments. and reference surface point cloud To simplify the sampling process, a cylindrical coordinate system is established, the point cloud is meshed, and sampling is performed within the mesh cells to obtain the downsampled point cloud of surface attachments. and the downsampled reference surface point cloud ; S3.2, Point cloud based on downsampled surface attachments and the downsampled reference surface point cloud For each grid cell Feature quantification is performed to obtain the adhesion thickness, adhesion area, and adhesion volume of the surface adherents.

[0010] Optionally, step S3.1 specifically includes the following steps: S3.1.1, Points in Cartesian coordinates Convert to cylindrical coordinates (lower point) The radial distance is azimuth angle is , S3.1.2, Mesh the point cloud. For each mesh, the first... Each angle grid for: , In the formula, For the number of angle grids, For angular resolution; No. Each height grid for: , In the formula, For the number of height grids, For height resolution; This represents the minimum Z-axis coordinate of the point cloud in polar coordinates. For the Points , No. Angle index of the grid cell to which the point belongs for: , No. Height index of the grid cell to which each point belongs for: , Each grid cell Defined as Each cylindrical voxel The radius parameter is downsampled by mean, and the calculation formula is as follows: , Obtain the point cloud of surface attachments after downsampling. and the downsampled reference surface point cloud .

[0011] Optionally, step S3.2 specifically includes the following steps: S3.2.1, For each grid cell The thickness of the surface attachment is defined as the distance between the point of attachment on the surface and the reference surface. The calculation formula is: , In the formula, For grid cells The radius in cylindrical coordinates corresponding to the point cloud of surface attachments. For grid cells The radius of the cylindrical coordinate system corresponding to the reference surface point element; S3.2.2 Calculate the effective area occupied by the surface attachment region on the cylindrical development diagram, remove background points and defect voids, and obtain the actual attachment area of ​​the surface attachment. The calculation formula is: , In the formula, The radius of the underwater pile foundation reference surface. The height of the underwater pile foundation reference surface. The number of grid cells occupied by the point cloud of surface attachments. The number of grids on the underwater pile foundation reference surface. , For the number of angle grids, The number of grid cells at height; S3.2.3. Using the cylindrical area integral method, the attached volume of the surface deposit is obtained by multiplying the local thickness of each grid cell by its projected area and summing the results. The calculation formula is as follows: , For high resolution, The radius of the underwater pile foundation reference surface. For angular resolution; For each grid cell Thickness of the material adhering to the lower surface.

[0012] The geometric feature quantification system for underwater structure surface attachments according to the present invention includes: The pile foundation point cloud model reconstruction module is used to acquire underwater pile foundation images containing surface attachments, and to reconstruct and correct the 3D point cloud model of the underwater pile foundation containing surface attachments. Simultaneously, a 3D point cloud model of the underwater pile foundation without surface attachments was obtained by measurement. Then, the two point cloud models were registered to obtain a reconstructed point cloud of the underwater pile foundation with surface attachments and unified coordinates. Reference point cloud of underwater pile foundations without surface attachments ; The attachment point cloud segmentation module is used to segment the reconstructed point cloud. and reference surface point cloud Based on the distribution characteristics of point clouds, the reconstruction of point clouds is performed. Perform coarse segmentation to obtain coarsely segmented point clouds. Then, based on the RGB color information of the point cloud, the coarsely segmented point cloud is further processed. Fine segmentation was performed to obtain the point cloud of surface attachments. ; The attachment geometric feature extraction module is used to extract surface attachment point clouds using a cylindrical point cloud downsampling method. and reference surface point cloud Simplify by downsampling, and then analyze the point cloud of the downsampled surface attachments. and downsampled reference surface point cloud Feature quantization is performed to obtain the adhesion thickness, adhesion area, and adhesion volume of the surface deposits.

[0013] Based on the same inventive concept, the present invention provides an electronic device including a processor and a storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method described above.

[0014] Based on the same inventive concept, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0015] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: (1) The present invention can accurately construct the three-dimensional morphology of surface attachments and accurately and efficiently quantify the geometric features of surface attachments; (2) This invention proposes an optimization correction method based on multiple known physical dimensions. The least squares method is used to solve the optimal scale factor and perform scale correction on the three-dimensional point cloud reconstructed from the image, converting its units into real physical units, thereby overcoming the scale ambiguity problem and improving the correction accuracy and robustness. (3) In view of the problem of low overlap between the reconstructed point cloud and the reference point cloud, the present invention adopts a registration method based on feature points to replace the traditional iterative nearest point algorithm, thereby achieving high-precision spatial alignment between the corrected point cloud and the reference point cloud. (4) This invention proposes a two-level segmentation strategy for accurate extraction of attachment point clouds: First, coarse segmentation is performed. The structural axis is determined based on the RANSAC cylindrical fitting algorithm. The distance from each point in the point cloud to the axis is calculated. Based on the distance sequence, a brute-force search method based on inflection point theory is used to initially classify the point cloud into attachments and background noise. Then, for the coarse segmentation results, an adaptive fine segmentation method based on RGB color statistical characteristics is proposed. By calculating the mean and standard deviation of the RGB channels of the point cloud, an adaptive color threshold range dominated by the green channel is constructed to accurately extract the green attachment point cloud and effectively eliminate the basic structure of the pile and residual noise. (5) This invention proposes a fixed object feature quantization algorithm based on cylindrical point cloud downsampling. It adopts an improved radius mean downsampling strategy, that is, only the radius parameters of points in each grid cell are averaged, while the original distribution of angle and height information is preserved, which is conducive to the feature alignment of time series data. It can directly analyze the point cloud, avoiding the errors and complexity of traditional surface reconstruction methods. It has more direct and reliable results, high computational efficiency, and can reduce the complexity of cylindrical pile foundation point cloud analysis. It is particularly suitable for rapid evaluation and long-term monitoring in engineering sites. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the process of the present invention; Figure 2 This is a flowchart illustrating the application process of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0018] like Figure 1 and Figure 2 As shown, the present invention provides a method for quantifying the geometric features of underwater structure surface attachments, comprising the following steps: S1. Acquire images of underwater pile foundations containing surface attachments, reconstruct and correct the 3D point cloud model of the underwater pile foundation containing surface attachments, and simultaneously measure and obtain the 3D point cloud model of the underwater pile foundation without surface attachments. Then, register the two point cloud models to obtain a reconstructed point cloud of the underwater pile foundation containing surface attachments with unified coordinates. Reference point cloud of underwater pile foundations without surface attachments .

[0019] Step S1 specifically includes the following steps: S1.1. Underwater pile foundation image datasets were obtained by taking close-up, full-view photos using an underwater camera. Zhang's calibration method was used to correct distortion in the image dataset. Based on the corrected image dataset, the Colmap open-source program was used to reconstruct a 3D point cloud model of the underwater pile foundation including surface attachments. , of which The points are ,in For the first The three-dimensional coordinates of the points For the first The RGB color channel values ​​of each point.

[0020] S1.2 To address the scale ambiguity issue in Colmap reconstruction, a least-squares optimization method based on multiple known physical dimensions is employed to solve for the global scale factor. This enables the physicalization of point cloud units and the correction of point cloud scale. Specifically, this involves constructing the objective function. , Optimize the scaling factor using the least squares method. , , In the formula, A 3D point cloud model of underwater pile foundations containing surface attachments. The Middle The distance between points, 3D point cloud model The Middle The actual distance between each pair of points can be obtained by querying. Point-to-point quantity; Based on global scale factor Perform coordinate correction. , Three-dimensional point cloud model of underwater pile foundation containing surface attachments The corrected 3D point cloud model of the underwater pile foundation is updated to obtain the corrected reconstructed point cloud. Among them, the corrected first The points are .

[0021] S1.3. A three-dimensional point cloud model of the underwater pile foundation without surface attachments is obtained by measuring with a laser scanner, thus obtaining the reference surface point cloud. The underwater pile foundation axis was determined using the RANSAC cylindrical fitting algorithm. A coordinate rigid body transformation was then performed with the axis as the Z-axis, and a feature point-based registration method was used to reconstruct the point cloud after correction. Point cloud of reference surface Registration.

[0022] The registration method is as follows: based on the uniqueness of the intersection points of the three planes in the point cloud, the corrected and reconstructed point cloud is extracted. Point cloud of reference surface The three intersection points in the diagram are used as salient feature matching point pairs. Then, the optimal spatial transformation parameters are estimated using the salient feature matching point pairs, where the transformation matrix is... The function is: , In the formula, This represents finding the spatial transformation parameters that minimize the function; the spatial transformation parameters include the rotation matrix. Translation matrix , To identify salient feature points within the reconstructed point cloud after correction. These are the significant feature points within the reference surface element. and These constitute a pair of points that match significant features.

[0023] Registration yields a reconstructed point cloud of underwater pile foundations containing surface attachments, with coordinates unified. Reference point cloud of underwater pile foundations without surface attachments Reconstructing point clouds It includes surface attachment points, underwater pile foundation points, noise points, and background points.

[0024] S2, Based on the reconstructed point cloud and reference surface point cloud Based on the distribution characteristics of point clouds, the reconstruction of point clouds is performed. Perform coarse segmentation to obtain coarsely segmented point clouds. Then, based on the RGB color information of the point cloud, the coarsely segmented point cloud is further processed. Fine segmentation was performed to obtain the point cloud of surface attachments. .

[0025] Step S2 specifically includes the following steps: S2.1 Calculate the reference surface point cloud using the RANSAC cylindrical fitting function pcfitcylinder in the Matlab program. axial direction vector and a point on the axis Calculate and reconstruct point clouds Distance from the axis : , In the formula, Let the magnitude of the vector be . Will Sort in descending order, denoted as Since the adhesion thickness of the underwater pile foundation surface is less than the distance from the background point cloud to the underwater pile foundation surface, therefore based on Point clouds are classified into surface attachment point clouds and background point clouds; a brute-force search method is used to find... The location of the two optimal inflection points is determined by iterating through the two possible inflection point locations and... Divide the data into three segments, perform linear fitting on each segment, calculate the overall fitting error, and find the two indices of the inflection points where the error is minimized. These indices are then denoted by their ordinates from largest to smallest. and ,Will middle satisfy The points are segmented to obtain a coarsely segmented point cloud. Point cloud after coarse segmentation This includes surface attachment points, noise points, and underwater pile foundation points.

[0026] S2.2, For point clouds after coarse segmentation An adaptive surface attachment point cloud segmentation method based on RGB color statistical characteristics is adopted. The statistical features of the RGB color values ​​of all points are calculated, including the mean vector. and standard deviation vector Define an adaptive color threshold range for identifying point cloud regions containing surface attachments: , Finally, points within the adaptive color threshold range are considered as surface attachments, and their corresponding spatial coordinates and color information are extracted to construct a surface attachment point cloud. This method has good adaptability and can autonomously adjust the green extraction boundary according to the statistical characteristics of the point cloud color distribution under different specimens or shooting conditions, effectively improving the accuracy and robustness of point cloud extraction of algae surface attachments.

[0027] S3. Use cylindrical point cloud downsampling method to analyze the point cloud of surface attachments. and reference surface point cloud Simplify by downsampling, and then analyze the point cloud of the downsampled surface attachments. and downsampled reference surface point cloud Feature quantization is performed to obtain the adhesion thickness, adhesion area, and adhesion volume of the surface deposits.

[0028] Step S3 specifically includes the following steps: S3.1. Use cylindrical point cloud downsampling method to sample the point cloud of surface attachments. and reference surface point cloud To simplify the sampling process, a cylindrical coordinate system is established, the point cloud is meshed, and sampling is performed within the mesh cells to obtain the downsampled point cloud of surface attachments. and the downsampled reference surface point cloud .

[0029] Step S3.1 specifically includes the following steps: S3.1.1, Points in Cartesian coordinates Convert to cylindrical coordinates (lower point) The radial distance is azimuth angle is .

[0030] S3.1.2. Mesh the point cloud. For each mesh, the angle mesh is as follows: , In the formula, For the number of angle grids, For angular resolution; The height grid is: , In the formula, For the number of height grids, For height resolution; This represents the minimum Z-axis coordinate of the point cloud in polar coordinates. For the Points , No. Angle index of the grid cell to which the point belongs for: , No. Height index of the grid cell to which each point belongs for: , Each grid cell Defined as Each cylindrical voxel Traditional methods use mean downsampling to generate representative points, but this method is difficult to align voxel features in temporal point clouds. Therefore, this invention only performs mean downsampling on the radius parameter, and the calculation formula is as follows: , Obtain the point cloud of surface attachments after downsampling. and the downsampled reference surface point cloud .

[0031] S3.2, Point cloud based on downsampled surface attachments and the downsampled reference surface point cloud For each grid cell Feature quantification is performed to obtain the adhesion thickness, adhesion area, and adhesion volume of the surface adherents.

[0032] Step S3.2 specifically includes the following steps: S3.2.1, For each grid cell The thickness of the surface attachment is defined as the distance between the point of attachment on the surface and the reference surface. The calculation formula is: , In the formula, For grid cells The radius in cylindrical coordinates corresponding to the point cloud of surface attachments. For grid cells The radius in the cylindrical coordinate system corresponding to the reference surface point element.

[0033] S3.2.2 Calculate the effective area occupied by the surface attachment region on the cylindrical development diagram, remove background points and defect voids, and obtain the actual attachment area of ​​the surface attachment. The calculation formula is: , In the formula, The radius of the underwater pile foundation reference surface. The height of the underwater pile foundation reference surface. The number of grid cells occupied by the point cloud of surface attachments. The number of grids on the underwater pile foundation reference surface. , For the number of angle grids, This represents the number of grid cells at height.

[0034] S3.2.3. Using the cylindrical area integral method, the attached volume of the surface deposit is obtained by multiplying the local thickness of each grid cell by its projected area and summing the results. The calculation formula is as follows: , For high resolution, The radius of the underwater pile foundation reference surface. For angular resolution; For each grid cell Thickness of the material adhering to the lower surface.

[0035] This invention proposes a processing flow from disordered images to quantitative features of attachments. Through a technical chain of 3D reconstruction, point cloud segmentation, and cylindrical quantization, it effectively solves the shortcomings of existing methods in terms of accuracy, efficiency, and adaptability. It is particularly suitable for the accurate quantification of the 3D morphology of attachments such as algae and shellfish on the surface of underwater structures such as bridge pile foundations.

[0036] Example 2: A geometric feature quantification system for underwater structure surface attachments according to the present invention includes: The pile foundation point cloud model reconstruction module is used to acquire underwater pile foundation images containing surface attachments, reconstruct and correct the 3D point cloud model of the underwater pile foundation containing surface attachments, and simultaneously measure and obtain the 3D point cloud model of the underwater pile foundation without surface attachments. The two point cloud models are then registered to obtain a reconstructed point cloud of the underwater pile foundation containing surface attachments with unified coordinates. Reference point cloud of underwater pile foundations without surface attachments .

[0037] The pile foundation point cloud model reconstruction module specifically includes: A dataset of underwater pile foundation images was obtained by capturing close-up, full-view images using an underwater camera. Zhang's calibration method was used to correct distortion in the image dataset. Based on the corrected image dataset, a 3D point cloud model of the underwater pile foundation, including surface attachments, was reconstructed using the Colmap open-source program. , of which The points are ,in For the first The three-dimensional coordinates of the points For the first The RGB color channel values ​​of each point.

[0038] To address the scale ambiguity issue in Colmap reconstruction, a least-squares optimization method based on multiple known dimensions is employed to solve for the global scale factor. This enables the physicalization of point cloud units and the correction of point cloud scale. Specifically, this involves constructing the objective function. , Optimize the scaling factor using the least squares method. , , In the formula, A 3D point cloud model of underwater pile foundations containing surface attachments. The Middle The distance between points, 3D point cloud model The Middle The actual distance between each pair of points can be obtained by querying. Point-to-point quantity; Based on global scale factor Perform coordinate correction. , Three-dimensional point cloud model of underwater pile foundation containing surface attachments The corrected 3D point cloud model of the underwater pile foundation is updated to obtain the corrected reconstructed point cloud. Among them, the corrected first The points are .

[0039] A three-dimensional point cloud model of the underwater pile foundation without surface attachments is obtained by measuring with a laser scanner, thus obtaining the reference surface point cloud. The underwater pile foundation axis was determined using the RANSAC cylindrical fitting algorithm. A coordinate rigid body transformation was then performed with the axis as the Z-axis, and a feature point-based registration method was used to reconstruct the point cloud after correction. Point cloud of reference surface Registration; The registration method is as follows: based on the uniqueness of the intersection points of the three planes in the point cloud, the corrected and reconstructed point cloud is extracted. Point cloud of reference surface The three intersection points in the diagram are used as salient feature matching point pairs. Then, the optimal spatial transformation parameters are estimated using the salient feature matching point pairs, where the transformation matrix is... The function is: , In the formula, This represents finding the spatial transformation parameters that minimize the function; the spatial transformation parameters include the rotation matrix. Translation matrix , To identify salient feature points within the reconstructed point cloud after correction. These are the significant feature points within the reference surface element. and Constitute a pair of points that match significant features; Registration yields a reconstructed point cloud of underwater pile foundations containing surface attachments, with coordinates unified. Reference point cloud of underwater pile foundations without surface attachments Reconstructing point clouds It includes surface attachment points, underwater pile foundation points, noise points, and background points.

[0040] The attachment point cloud segmentation module is used to segment the reconstructed point cloud. and reference surface point cloud Based on the distribution characteristics of point clouds, the reconstruction of point clouds is performed. Perform coarse segmentation to obtain coarsely segmented point clouds. Then, based on the RGB color information of the point cloud, the coarsely segmented point cloud is further processed. Fine segmentation was performed to obtain the point cloud of surface attachments. .

[0041] The attachment point cloud segmentation module specifically includes: The reference surface point cloud was calculated using the RANSAC cylindrical fitting function pcfitcylinder in Matlab. axial direction vector and a point on the axis Calculate and reconstruct point clouds Distance from the axis : , In the formula, Let the magnitude of the vector be . Will Sort in descending order, denoted as Since the adhesion thickness of the underwater pile foundation surface is less than the distance from the background point cloud to the underwater pile foundation surface, therefore based on Point clouds are classified into surface attachment point clouds and background point clouds; a brute-force search method is used to find... The location of the two optimal inflection points is determined by iterating through the two possible inflection point locations and... Divide the data into three segments, perform linear fitting on each segment, calculate the overall fitting error, and find the two indices of the inflection points where the error is minimized. These indices are then denoted by their ordinates from largest to smallest. and ,Will middle satisfy The points are segmented to obtain a coarsely segmented point cloud. Point cloud after coarse segmentation This includes surface attachment points, noise points, and underwater pile foundation points; For point clouds after coarse segmentation An adaptive surface attachment point cloud segmentation method based on RGB color statistical characteristics is adopted. The statistical features of the RGB color values ​​of all points are calculated, including the mean vector. and standard deviation vector Define an adaptive color threshold range for identifying point cloud regions containing surface attachments: , Finally, points within the adaptive color threshold range are considered as surface attachments, and their corresponding spatial coordinates and color information are extracted to construct a surface attachment point cloud. This method has good adaptability and can autonomously adjust the green extraction boundary according to the statistical characteristics of the point cloud color distribution under different specimens or shooting conditions, effectively improving the accuracy and robustness of point cloud extraction of algae surface attachments.

[0042] The attachment geometric feature extraction module is used to extract surface attachment point clouds using a cylindrical point cloud downsampling method. and reference surface point cloud Simplify by downsampling, and then analyze the point cloud of the downsampled surface attachments. and downsampled reference surface point cloud Feature quantization is performed to obtain the adhesion thickness, adhesion area, and adhesion volume of the surface deposits.

[0043] The attachment geometric feature extraction module specifically includes: Cylindrical point cloud downsampling method for surface attachment point clouds and reference surface point cloud To simplify the sampling process, a cylindrical coordinate system is established, the point cloud is meshed, and sampling is performed within the mesh cells to obtain the downsampled point cloud of surface attachments. and the downsampled reference surface point cloud Specifically, it includes: Points in Cartesian coordinates Convert to cylindrical coordinates (lower point) The radial distance is azimuth angle is , The point cloud is meshed, and for each mesh, its angle mesh is as follows: , In the formula, For the number of angle grids, For angular resolution; The height grid is: , In the formula, For the number of height grids, For height resolution; This represents the minimum Z-axis coordinate of the point cloud in polar coordinates. For the Points , No. Angle index of the grid cell to which the point belongs for: , No. Height index of the grid cell to which each point belongs for: , Each grid cell Defined as Each cylindrical voxel Traditional methods use mean downsampling to generate representative points, but this method is difficult to align voxel features in temporal point clouds. Therefore, this invention only performs mean downsampling on the radius parameter, and the calculation formula is as follows: , Obtain the point cloud of surface attachments after downsampling. and the downsampled reference surface point cloud .

[0044] Based on the downsampled surface attachment point cloud and the downsampled reference surface point cloud For each grid cell Feature quantization is performed to obtain the adhesion thickness, adhesion area, and adhesion volume of the surface deposits; specifically including: For each grid cell The thickness of the surface attachment is defined as the distance between the point of attachment on the surface and the reference surface. The calculation formula is: , In the formula, For grid cells The radius in cylindrical coordinates corresponding to the point cloud of surface attachments. For grid cells The radius of the cylindrical coordinate system corresponding to the reference surface point element; On the cylindrical unfolded diagram, the effective area occupied by the surface attachment region is counted, and background points and defect voids are removed to obtain the actual attachment area of ​​the surface attachment. The calculation formula is: , In the formula, The radius of the underwater pile foundation reference surface. The height of the underwater pile foundation reference surface. The number of grid cells occupied by the point cloud of surface attachments. The number of grids on the underwater pile foundation reference surface. , For the number of angle grids, The number of grid cells at height; Using the cylindrical area integral method, the attached volume of the surface deposits is obtained by summing the local thickness of each grid cell multiplied by its projected area. The calculation formula is as follows: , For high resolution, The radius of the underwater pile foundation reference surface. For angular resolution; For each grid cell Thickness of the material adhering to the lower surface.

[0045] Example 3: An electronic device according to the present invention includes a processor and a storage medium;

[0046] Storage media are used to store instructions;

[0047] The processor is configured to operate according to the instructions to perform the steps of the method described above.

[0048] Example 4: The computer-readable storage medium of the present invention stores a computer program thereon, which, when executed by a processor, implements the steps of the method described above.

Claims

1. A method of quantifying geometric features of surface fouling on underwater structures, characterized by, The method comprises the following steps: S1, collect the image of the underwater pile foundation containing surface attachments, reconstruct and correct the three-dimensional point cloud model of the underwater pile foundation containing surface attachments , simultaneously measure the three-dimensional point cloud model of the underwater pile foundation not containing surface attachments, and then register the two point cloud models to obtain the reconstructed point cloud of the underwater pile foundation containing surface attachments with unified coordinates , and the reference surface point cloud of the underwater pile foundation not containing surface attachments ; S2, reconstructing the point cloud based on the reference surface point cloud and the reference surface point cloud , performing coarse segmentation on the reconstructed point cloud based on a distribution feature of the point cloud to obtain a coarse segmented point cloud , and performing fine segmentation on the coarse segmented point cloud based on RGB color information of the point cloud to obtain the surface attached object point cloud ;​​ S3, using a cylindrical point cloud downsampling method to downsample the surface attachment point cloud and the reference surface point cloud S4, simplifying the downsampled surface attachment point cloud and the downsampled reference surface point cloud S5, performing feature quantization to obtain the attachment thickness, attachment area, and attachment volume of the surface attachment.

2. The method of quantifying geometric features of surface fouling on underwater structures of claim 1, wherein, The step S1 specifically comprises the following steps: S1.1, underwater pile image dataset is obtained by close-range full-view shooting with an underwater camera, Zhang calibration method is used for distortion correction of the image dataset, and based on the corrected image dataset, a three-dimensional point cloud model of the underwater pile containing surface attachments is reconstructed wherein the th point is wherein the th point is th S1.2, adopt a least square optimization method based on multiple known sizes to solve the global scale factor , realize the physicalization of the point cloud unit, and correct the scale of the point cloud; Specifically, a target function is constructed, ， Optimizing scale factors by least squares , ， wherein A three-dimensional point cloud model of an underwater pile foundation with surface attachments In the distance between the A three-dimensional point cloud model of an underwater pile foundation with surface attachments In the true distance between the is the number of point pairs; Based on a global scale factor , a coordinate correction is performed, ， Three-dimensional point cloud model of underwater pile foundation with surface attachments updating the corrected three-dimensional point cloud model of underwater pile foundation to obtain a reconstructed point cloud of the underwater pile foundation with surface attachments wherein the corrected i th point is ; S1.3, obtain the three-dimensional point cloud model of the underwater pile foundation without surface attachments by laser scanner, that is, obtain the datum plane point cloud , determine the axis of the underwater pile foundation by the RANSAC cylindrical fitting algorithm, and take the axis as the coordinate axis Z by the coordinate rigid body transformation, and then realize the registration of the reconstructed point cloud by the feature point-based registration method. with the datum plane point cloud .

3. A method of quantifying geometric features of surface fouling on underwater structures according to claim 2, characterised in that, The registration method in step S1.3 is specifically: according to the uniqueness of the three-plane intersection of the plane-plane-curve in the point cloud, respectively extracting the three-plane intersection in the corrected reconstruction point cloud and the three-plane intersection in the reference plane point cloud as the significant feature matching point pair, and then estimating the optimal spatial transformation parameter by using the significant feature matching point pair, wherein the function of the transformation matrix is: ， wherein represents finding spatial transformation parameters that minimize the function; the spatial transformation parameters include a rotation matrix and a translation matrix , are significant feature points within the reconstructed point cloud, are significant feature points within the reference surface point cloud, and constitute a significant feature matching point pair; Reconstructed point cloud of a submerged pile foundation with surface attachments in coordinate-unified registration and a reference surface point cloud of a submerged pile foundation without surface attachments , the reconstructed point cloud comprising surface attachment points, submerged pile foundation points, noise points, and background points.

4. The method of quantifying geometric features of surface fouling on underwater structures of claim 3, wherein, The step S2 specifically comprises the following steps: S2.1 Calculate the reference surface point cloud using the RANSAC cylindrical fitting function in Matlab. axial direction vector and a point on the axis Calculate and reconstruct point clouds Distance from the axis : ， wherein is the norm of the vector, Will Sort in descending order, denoted as ;based on Point clouds are classified into surface attachment point clouds and background point clouds; a brute-force search method is used to find... The location of the two optimal inflection points is determined by iterating through the two possible inflection point locations and... Divide the data into three segments, perform linear fitting on each segment, calculate the overall fitting error, and find the two indices of the inflection points where the error is minimized. These indices are then denoted by their ordinates from largest to smallest. and Reconstructing point clouds middle satisfy The points are segmented to obtain a coarsely segmented point cloud. Point cloud after coarse segmentation This includes surface attachment points, noise points, and underwater pile foundation points; S2.2, for the point cloud after rough segmentation An adaptive surface attachment point cloud segmentation method based on RGB color statistical characteristics is adopted to calculate the statistical characteristics of the RGB color values of all points respectively. The statistical features include a mean vector and a standard deviation vector ; defining adaptive color threshold intervals for identifying surface deposit point cloud regions: ， Finally, the points in the adaptive color threshold interval are considered as surface attachments, and their corresponding spatial coordinates and color information are extracted to form a surface attachment point cloud .

5. A method of quantifying geometric features of surface fouling on underwater structures according to claim 4, characterised in that, The step S3 specifically comprises the following steps: S3.1, using a cylindrical point cloud downsampling method to the surface attached point cloud and the reference surface point cloud Downsampling simplification, by establishing a cylindrical coordinate system, the point cloud is gridded, and sampling is carried out in the grid unit to obtain the downsampling surface attached point cloud and the downsampling reference surface point cloud ; S3.2, surface attachment point cloud after downsampling and the reference surface point cloud after downsampling , for each grid cell , the feature quantification is performed to obtain the attachment thickness, attachment area and attachment volume of the surface attachment.

6. A method of quantifying geometric features of surface fouling on underwater structures according to claim 5, wherein, The step S3.1 specifically comprises the following steps: S3.1.1, converting a point in Cartesian coordinates to a point in cylindrical coordinates with a radial distance of and an azimuth angle of , S3.1.2, meshing the point cloud, for each mesh, the first angle mesh is: ， wherein is the number of angular bins, is the angular resolution; No. Each height grid for: ， In the formula, is the number of high grids, is the high resolution; is the minimum value of the Z coordinate of the point cloud in the polar coordinate system. For the first point , the angle index of the grid cell to which the first point belongs is: ​ ， No. Height index of the grid cell to which each point belongs for: ， Each grid cell , defined as Each cylinder voxel Down-sampling the radius parameter by averaging, calculated as: ， down-sampled surface attachment point cloud and a down-sampled reference surface point cloud .

7. A method of quantifying geometric features of surface fouling on underwater structures according to claim 6, characterised in that, The step S3.2 specifically comprises the following steps: S3.2.1, for each grid cell The distance between the surface attachment point and the reference plane is the surface attachment thickness The calculation formula is: ， In the formula, is the grid unit is the radius in the cylindrical coordinate system corresponding to the point cloud of the surface attachment in the middle, is the grid unit is the radius in the cylindrical coordinate system corresponding to the point cloud of the surface attachment in the middle, S3.2.2, on the cylindrical development map, the area of the surface attached object is counted, the background points and defect empty areas are removed, and the actual attached area of the surface attached object is obtained The calculation formula is: ， wherein, R is the radius of the underwater pile foundation reference surface, H is the height of the underwater pile foundation reference surface, N is the number of meshes occupied by the surface attachment point cloud, N is the number of meshes occupied by the underwater pile foundation reference surface, , N is the number of angle meshes, N is the number of height meshes; S3.2.3, using the column area integral method, the attachment volume of the surface attachment is accumulated according to the local thickness of each grid unit multiplied by its projected area, and the calculation formula is: ， is a high resolution, is a radius of the underwater pile foundation reference surface, is an angular resolution; is a thickness of the lower surface attachment for each grid cell lower surface attachment thickness.

8. A system for quantifying geometric features of surface fouling on underwater structures, the system comprising: The method comprises the following steps: The pile foundation point cloud model reconstruction module is used for collecting underwater pile foundation images containing surface attachments, reconstructing and correcting underwater pile foundation three-dimensional point cloud models containing surface attachments , simultaneously measuring underwater pile foundation three-dimensional point cloud models not containing surface attachments, registering the two point cloud models, and obtaining a reconstructed point cloud of the underwater pile foundation containing surface attachments with unified coordinates and the reference surface point cloud of the underwater pile foundation not containing surface attachments ; The attachment point cloud segmentation module is used to segment the reconstructed point cloud. and reference surface point cloud Based on the distribution characteristics of point clouds, the reconstruction of point clouds is performed. Perform coarse segmentation to obtain coarsely segmented point clouds. Then, based on the RGB color information of the point cloud, the coarsely segmented point cloud is further processed. Fine segmentation was performed to obtain the point cloud of surface attachments. ; The attachment geometry extraction module is configured to adopt a cylindrical point cloud down-sampling method to down-sample the surface attachment point cloud and the reference surface point cloud to obtain the attachment thickness, the attachment area and the attachment volume of the surface attachment. and the down-sampled reference surface point cloud to obtain the attachment thickness, the attachment area and the attachment volume of the surface attachment.

9. An electronic device, comprising: The device comprises a processor and a storage medium; The storage medium is used for storing instructions; The processor is used for operating according to the instructions to execute the steps of the method in any one of claims 1-7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to realize the steps of the method in any one of claims 1-7.

Images