Underwater concrete structure surface defect measuring method, system and device

Abstract

The invention discloses an underwater concrete structure surface defect measurement method, system and device, and the method comprises the steps: carrying out the underwater calibration of a camera, and building a mapping relation between underwater three-dimensional entity points and two-dimensional image pixel points; shooting a defect image of the underwater concrete structure; carrying out distortion correction on the defect image; performing point cloud processing on the point cloud data of the corrected defect image, reconstructing a three-dimensional model of the concrete defect image according to the processed point cloud data, and calculating a defect feature size; and performing definition processing on the defect image, fusing the defect feature size with the processed defect image, and marking the defect type and position. According to the method, the corrected defect image is subjected to point cloud processing, and through point cloud filtering, segmentation and registration, noise data interference in the defect image can be avoided to the greatest extent, and the defect part in the image is accurately extracted for positioning and measurement, so that the method is more suitable for positioning the surface defect of the underwater concrete structure, and the definition and the accuracy of the surface defect are improved.

Description

Technical Field

[0001] The present invention relates to the field of underwater image processing and defect measurement, and in particular to a method, system and device for measuring surface defects of underwater concrete structures. Background Art

[0002] Underwater structures such as bridge piers and dams are typically constructed of reinforced concrete. Over long periods of service, these structures are susceptible to serious defects such as cracks, exposed steel, and surface spalling due to factors such as material aging, water erosion, temperature fluctuations, and accidental impacts. If these surface defects are not discovered and properly addressed in a timely manner, they can worsen and even lead to major safety accidents.

[0003] Therefore, it is extremely important to inspect and maintain the safety and health status of water conservancy project structures, especially to conduct detailed inspections and assessments of the underwater parts of these engineering structures, to promptly discover and evaluate potential structural problems and the development of diseases, and then to implement necessary repair and maintenance measures. This is the basis for ensuring their long-term safe operation and can significantly reduce the occurrence of large-scale safety accidents.

[0004] At present, the inspection of underwater engineering structures focuses on the qualitative analysis of defects, that is, the identification of underwater defects, to determine whether there are surface defects such as cracks and peeling on the surface of underwater structures; however, there is less research on the quantitative analysis of underwater defects. Quantitative analysis of defects can effectively help technicians judge the development of underwater diseases and then evaluate and analyze the safety of underwater structures.

[0005] Existing methods for measuring surface defects on underwater structures often use monocular cameras, which severely lose depth information. These factors significantly impact the extraction and measurement of target defect feature information. Furthermore, existing techniques calculate the 3D coordinates of an object using the left-right parallax of a binocular camera, and employ the SIFT algorithm for stereo matching. While robust, the SIFT algorithm is computationally intensive and can affect real-time performance, particularly underwater, where light fluctuations and suspended particles can lead to feature point matching errors, affecting the accuracy of the disparity map and 3D coordinates, and consequently, the accuracy of 3D reconstruction. Summary of the Invention

[0006] To address the deficiencies of the prior art, the present invention proposes a method, system, and device for measuring surface defects of underwater concrete structures, so as to improve the measurement accuracy of defective parts in defect images of underwater concrete structures.

[0007] To this end, the technical solution adopted in the present invention is: The present invention provides a method for measuring surface defects of an underwater concrete structure, the method comprising: Perform underwater camera calibration and establish a mapping relationship between the 3D entity points of the underwater concrete structure and the 2D image pixels based on the calibration results; Using the calibrated camera to capture defect images of the underwater concrete structure at different positions and / or angles, and performing distortion correction on the defect images; Performing point cloud processing on the point cloud data of the corrected defect image, reconstructing a three-dimensional model of the concrete defect image based on the processed point cloud data, and calculating the defect characteristic size of the three-dimensional model; the point cloud processing includes point cloud filtering, point cloud segmentation, and point cloud registration; The defect image is processed for clarity, and the defect feature size is fused with the processed defect image to mark the defect type and location of the concrete structure.

[0008] According to the above scheme, the mapping relationship between underwater three-dimensional entity points and two-dimensional image pixel points is established by establishing a coordinate system, and the distortion correction of the defective image includes correction based on radial distortion and correction based on tangential distortion; wherein, the corrected distortion point of the correction based on radial distortion is calculated by the distance between the coordinate origin and the distortion point and the radial distortion coefficient, and the corrected distortion point of the correction based on tangential distortion is calculated by the distance between the coordinate origin and the distortion point and the tangential distortion coefficient; the radial distortion coefficient and the tangential distortion coefficient are obtained according to calibration.

[0009] According to the above scheme, point cloud filtering specifically includes: obtaining the distance between each point in the point cloud data of the corrected defect image and its neighboring points, and calculating the mean and standard deviation of the distance; The distance threshold is calculated based on the distance mean and standard deviation, and it is determined whether the distance between each point and its neighboring points exceeds the distance threshold. All the area points whose distance to the point exceeds the distance threshold are eliminated to obtain the filtered point cloud data.

[0010] According to the above scheme, point cloud segmentation specifically includes: separating defect point cloud data from filtered point cloud data, clustering and screening the defect point cloud data, and retaining the largest defect point cloud cluster; Separating the defective point cloud data from the filtered point cloud data includes: selecting part of the point cloud data in the filtered point cloud data to form a minimum data set, and using the minimum data set to calculate a defect model; using the defect model to test all the data in the data set to determine the number of inliers that meet the defect model; iterating the defect model: expanding the number of point cloud data in the data set, and calculating a new defect model, comparing it with the previous better defect model, and setting the defect model with a larger number of inliers as the better defect model; repeating the iterative process until the number of iterations reaches a certain number or the number of inliers of the better defect model reaches a certain threshold, and obtaining the optimal defect model, at which time the inliers included in the optimal defect model are set as the defective point cloud data; wherein the inliers refer to the point cloud data that meet the conditions of the defect model; Among them, clustering screening of defective point cloud data includes: randomly selecting a point cloud in the defective point cloud data as the starting point of the current cluster; performing a neighborhood search on the starting point to find all neighboring points whose distance to the starting point is less than a threshold; for each neighboring point found, repeating the neighborhood search process, and adding the searched point cloud to the current cluster; when no more point clouds can be added to the current cluster, ending the cluster construction; reselecting a new point cloud that has not been assigned to any cluster, repeating the neighborhood search and cluster construction process until all point clouds are assigned to a cluster; comparing and analyzing each point cloud cluster, retaining the largest cluster as the defective point cloud cluster.

[0011] According to the above scheme, point cloud registration specifically includes: using a random sampling consensus algorithm to perform coarse registration on the point cloud, and using an iterative closest point algorithm to perform fine registration on the coarsely registered point cloud.

[0012] According to the above scheme, the three-dimensional model of the concrete defect image is reconstructed by using the rolling ball method and smoothing the three-dimensional model of the concrete defect image; the defect characteristic size of the three-dimensional model is calculated specifically including: calculating the length, width, depth, and three-dimensional volume of the defect, and performing a planar projection of the defect of the three-dimensional model and calculating the two-dimensional area of ​​the defect.

[0013] According to the above scheme, the clarity processing of the defect image specifically includes: performing preliminary processing on the underwater defect image of the concrete structure through the dark channel prior algorithm, and scoring the defect image after preliminary processing according to the underwater image quality evaluation standard, and reprocessing the defect image with a score lower than the set threshold until the score of all defect images is higher than the set threshold, thereby obtaining the processed defect image.

[0014] According to the above scheme, the defective image after preliminary processing is scored according to the underwater image quality evaluation standard by recalculating the sum of the weighted coefficient of the chromaticity standard deviation and the weighted coefficient of the saturation average value of the defective image after preliminary processing based on the calculation of UCIQE.

[0015] The present invention also provides a system for measuring surface defects of underwater concrete structures, the system comprising: The calibration module is used to calibrate the camera underwater and establish the mapping relationship between the 3D entity points of the underwater concrete structure and the 2D image pixels based on the calibration results; a distortion correction module, configured to capture defect images of underwater concrete structures at different positions and / or angles using a calibrated camera and perform distortion correction on the defect images; A point cloud processing and 3D reconstruction module is used to perform point cloud processing on the point cloud data of the corrected defect image, reconstruct a 3D model of the concrete defect image based on the processed point cloud data, and calculate the defect feature size of the 3D model; the point cloud processing includes point cloud filtering, point cloud segmentation, and point cloud registration; The defect marking module is used to process the clarity of the defect image, fuse the defect feature size with the processed defect image, and mark the defect type and location of the concrete structure.

[0016] The present invention also provides an image acquisition device, including a test water tank, a camera, a waterproof device, a clamping device and a lighting device; the concrete structure containing defects is placed on a boss on one side of the water tank, and a scale groove is provided at the upper end of the test water tank in the longitudinal direction; the camera is sealed in the waterproof device, and a detachable polarization lens is installed at the front end of the camera; the waterproof device is fixed by two F-type trigger-type clamping devices, a water tank cross beam is provided in the test water tank, and a water tank longitudinal beam that can move along the length direction of the water tank cross beam is provided on the water tank cross beam, the clamping device is provided on the water tank longitudinal beam and can move up and down along the water tank longitudinal beam; the lighting device is arranged on one side of the waterproof device.

[0017] The beneficial effects of the present invention are as follows: the present invention obtains underwater concrete structure defect images and corrects their distortion, then performs point cloud processing on the corrected defect images, and reconstructs the concrete structure defects in three dimensions. Through point cloud filtering, segmentation and alignment, the present invention can avoid noise data interference in the defect images to the greatest extent, and accurately extract the defective parts in the underwater concrete structure defect images for positioning and measurement. The present invention is more suitable for locating irregular structural surface defects of underwater concrete, and improves its clarity and accuracy.

[0018] Furthermore, the present invention evaluates the processed images through underwater image quality evaluation standards, which can accurately reflect the defect image quality of underwater concrete structures in an objective manner, thereby obtaining high-quality processed defect images, which is conducive to the precise marking of the obtained defect feature sizes. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a schematic diagram of a method flow for measuring surface defects of an underwater concrete structure according to an embodiment of the present invention; Figure 2 is a schematic diagram of the three-dimensional structure of an image acquisition device according to an embodiment of the present invention; Figure 3 is a schematic cross-sectional view of an image acquisition device according to an embodiment of the present invention; Figure 4 is a schematic top view of an image acquisition device according to an embodiment of the present invention; Figure 5This is a schematic diagram of a flow chart of a method for point cloud processing and defect feature size according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the overall flow chart of the surface defect measurement method according to an embodiment of the present invention; In the figure: 1-1: test water tank; 1-2: depth camera; 1-3: waterproof device; 1-4: clamping device; 1-5: lighting device; 1-6: concrete structure; 1-1-1: boss; 1-1-2: scale groove; 1-1-3: beam; 1-2-1: polarization lens. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0021] The embodiment of the present invention provides a method for measuring surface defects of underwater concrete structures to solve the problem of measurement accuracy of defective parts in defect images of underwater concrete structures in the prior art. Figure 1 As shown, the method includes: S1. Calibrate the camera underwater and establish a mapping relationship between the three-dimensional entity points of the underwater concrete structure and the two-dimensional image pixels based on the calibration results.

[0022] S2. Using the calibrated camera to capture defect images of the underwater concrete structure at different positions and / or angles, and performing distortion correction on the defect images.

[0023] S3. Performing point cloud processing on the point cloud data of the corrected defect image, reconstructing a three-dimensional model of the concrete defect image based on the processed point cloud data, and calculating the defect feature size of the three-dimensional model; the point cloud processing includes point cloud filtering, point cloud segmentation, and point cloud registration.

[0024] S4. Process the clarity of the defect image, fuse the defect feature size with the processed defect image, and mark the defect type and location of the concrete structure.

[0025] Specifically, in this embodiment, the Zhang Zhengyou calibration method is used to calibrate the camera used for shooting. The calibration plate is placed in a water tank, and calibration photos are taken from different angles and distances using the camera to obtain the internal and external parameters of the camera. Preferably, the camera shooting and calibration are achieved through an image acquisition device, wherein, Figure 2 、 Figure 3 and Figure 4The three-dimensional structure, cross-section and top view schematic diagrams of the image acquisition device used in this embodiment are respectively shown, including a test water tank 1-1, a depth camera 1-2, a waterproof device 1-3, a clamping device 1-4 and a lighting device 1-5; the test water tank is used to simulate an underwater environment, and a concrete structure 1-6 containing defects is placed on a boss 1-1-1 on one side of the water tank. A scale groove 1-1-2 is provided at the upper end of the water tank, and the distance between the camera and the defective surface can be changed by adjusting the position of the crossbeam 1-1-3; the depth camera 1-2 is sealed in the waterproof device 1-3, and the waterproof device 1-3 is made of a highly transparent acrylic plate to ensure the transmittance of the device. Two detachable polarized lenses 1-2-1 are installed at the front end; the waterproof device is fixed by two F-type trigger-type clamping devices 1-4, the upper end of the clamping device 1-4 is fixed to the crossbeam of the water tank, and the pitch angle between the camera and the defective surface can be changed by adjusting the clamping height; the lighting device 1-5 is preferably symmetrically distributed on both sides of the waterproof device 1-3, and the illumination angle and aperture size can be changed.

[0026] The camera's internal parameters, including focal length, principal point position, and distortion coefficients, describe the optical properties of the camera lens. External parameters, such as translation and rotation matrices, describe the camera's position and attitude in three-dimensional space. Accurate camera calibration is crucial for improving the accuracy of image analysis, especially in applications requiring highly precise measurements.

[0027] Specifically, the mapping relationship between underwater three-dimensional entity points and two-dimensional image pixel points is established by establishing a coordinate system, and the two-dimensional defect image and the three-dimensional structure entity point are mapped by taking defect images at different positions and angles; in addition, the distortion correction of the defect image includes correction based on radial distortion and correction based on tangential distortion; wherein, the corrected distortion point of the correction based on radial distortion is calculated by the distance between the coordinate origin and the distortion point and the radial distortion coefficient, and the corrected distortion point of the correction based on tangential distortion is calculated by the distance between the coordinate origin and the distortion point and the tangential distortion coefficient; the radial distortion coefficient and the tangential distortion coefficient are obtained according to calibration.

[0028] Among them, the correction based on radial distortion can be expressed as:

[0029] The correction based on tangential distortion can be expressed as:

[0030] in, , is the distance from the origin of the image coordinate system to the distortion point, where 、 are the horizontal and vertical coordinates of the coordinate origin respectively, 、 are the horizontal and vertical coordinates of the distortion point, 、 are the horizontal and vertical coordinates of the distortion point after correction; , , is the radial distortion parameter; and is the tangential distortion coefficient; the radial distortion parameter and tangential distortion coefficient can be obtained from the camera calibration results.

[0031] Specifically, point cloud processing of the corrected image point cloud data includes point cloud filtering, point cloud segmentation and point cloud registration in sequence. The specific process is as follows: Figure 5 shown.

[0032] Among them, point cloud filtering uses statistical filtering technology to remove noise and abnormal points in the original data; by analyzing the distance distribution between each point and its neighboring points, points that deviate from the average distance by a certain threshold are regarded as outliers and removed, thereby optimizing the point cloud data set and ensuring the accuracy and reliability of subsequent processing. Specifically, it includes: obtaining each point in the point cloud data of the corrected image and its neighboring points distance , and calculate the mean distance between all points and their neighboring points and standard deviation ; Calculate the distance threshold based on the distance mean and standard deviation ,in To adjust the parameters, The selection is directly related to the setting of the threshold, which in turn affects the processing results of the point cloud data. Different point cloud structures and noise levels require adjustment. The value is adapted to calculate the distance threshold, and then the distance between each point and its neighboring points is determined. Whether the distance threshold is exceeded , and remove all distances to this point Exceeding distance threshold Neighborhood points , that is, if Greater than , the point The outliers are determined to be deleted, otherwise they are regarded as key feature points and retained; the above steps are repeated until all point cloud data are traversed to obtain the filtered point cloud data.

[0033] Among them, point cloud segmentation is based on the Random Sample And Consensus (RANSAC) algorithm and the Euclidean Cluster Extraction algorithm to extract the target defect point cloud from the surrounding environment. Specifically, it includes separating the defect point cloud data from the filtered point cloud data, clustering and screening the defect point cloud data, and retaining the largest defect point cloud cluster.

[0034] Separating the defect point cloud data from the filtered point cloud data includes: selecting the part of the point cloud data that can estimate the defect model from the filtered point cloud data to form a minimum data set. (i ≥ 0), and using the minimum data set Calculate the defect model (i≥0); using defect model For the dataset All data in the dataset are tested to determine the number of inliers that meet the defect model, and the secondary defect model is first set as the better defect model; the defect model is iterated: the number of point cloud data in the dataset is expanded, and a new defect model is calculated. Compared with the previously obtained better defect model, the defect model with more inliers is set as the better defect model; the iterative process is repeated until the number of iterations reaches a certain number or the number of inliers of the defect model reaches a certain threshold. The better defect model at this time is the optimal defect model, and all inliers contained in the optimal defect model are set as defect point cloud data; wherein, the inliers refer to point cloud data that meet the conditions of the defect model.

[0035] In addition, clustering screening of defective point cloud data includes: randomly selecting a point cloud in the defective point cloud data as the starting point of the current cluster; using a spatial search data structure (such as a KD tree) to find all neighboring points whose distance to the starting point is less than a threshold; for each neighboring point found, repeating the above neighborhood search process, and adding the searched point cloud to the current cluster; when no more point clouds can be added to the current cluster, ending the construction of the cluster; reselecting a new point cloud that has not been assigned to any cluster, and repeating the above process until all points are assigned to a cluster; comparing and analyzing each point cloud cluster, and retaining the largest cluster as the defective point cloud cluster.

[0036] Preferably, in this embodiment, point cloud registration includes coarse registration based on a random sampling consensus algorithm and fine registration based on an iterative closest point (ICP) algorithm. These methods can accurately register point cloud data acquired at different angles and positions for the same defect, enhancing spatial cognition of the defect area. The coarse registration phase occurs when the relative position and posture of the point clouds are completely unknown. The primary goal is to determine a rough translation and rotation matrix that allows the two sets of point clouds to be roughly aligned, thereby mapping the point cloud data to be registered into a common coordinate system. Fine registration is performed on the basis of coarse registration, minimizing the differences between corresponding points in the point clouds through careful adjustments to achieve high-precision registration results. The iterative closest point algorithm optimizes the transformation matrix by iteratively searching for the closest corresponding point pairs and minimizing the distance between them.

[0037] Preferably, in this embodiment, the 3D model of the concrete defect image is reconstructed based on a rolling ball algorithm. The rolling ball method is an algorithm for surface reconstruction of 3D point cloud data. When a sphere rolls along two determined points and searches for a third point, a triangle is formed, based on which the entire surface can be gradually constructed. The implementation process of the rolling ball algorithm is as follows: Initialization: Select an appropriate sphere radius (the sphere radius is one of the key parameters of the algorithm and needs to be reasonably selected based on the density and distribution of the point cloud) and find an initial triangle in the point cloud. The initial triangle can be constructed by finding its nearest neighbor pairs and using them as the basis. Rolling the sphere: Place the sphere on one edge of the initial triangle and let it touch two vertices of this edge. Then find a new point in the point cloud so that the sphere can touch these three points at the same time (two starting points and one new point), and construct a new triangle based on these three points. Repeat the process of rolling the sphere, each time trying to find new points to construct a new triangle, until no more points can be found or the sphere returns to the starting position; Generate mesh: Through the above steps, the algorithm gradually constructs a triangular mesh covering the surface of the point cloud; Model reconstruction: Based on the obtained triangular mesh, a three-dimensional model of the surface defects is reconstructed and the three-dimensional model is smoothed.

[0038] Specifically, calculating the characteristic size of the defect of the three-dimensional model includes: calculating the length, width, depth, and three-dimensional volume of the defect, and performing a plane projection on the defect of the three-dimensional model and calculating the two-dimensional area of ​​the defect.

[0039] Specifically, the clarity processing of the defect image includes: performing preliminary processing on the underwater defect image of the concrete structure through the dark channel prior algorithm, and scoring the defect image after preliminary processing according to the underwater image quality evaluation standard, and reprocessing the defect image with a score lower than the set threshold until the score of all defect images is higher than the set threshold, thereby obtaining the processed defect image.

[0040] Preferably, the captured defect images are processed using a fusion imaging method of polarization imaging technology based on a physical model and a dark channel prior algorithm based on a digital image. The fusion imaging method includes using a professional underwater camera equipped with a polarization lens to filter out light scattering caused by water impurities in the image acquisition stage and applying a DCP (dark channel prior) algorithm in the image post-processing stage to further highlight the visual characteristics of the defects. This method combines the advantages of physical imaging technology and computer image processing technology, providing an effective solution for the detection of surface defects of underwater structures, which helps to improve the accuracy and efficiency of underwater detection. In addition, polarization imaging technology can invert the target morphology and material information, and while achieving image clarity, it can also realize target detection, identification and three-dimensional reconstruction. Therefore, by applying polarization imaging technology, the quality of the image can be enhanced, and detailed information on the defect surface can be revealed or hidden features can be discovered.

[0041] Among them, the scoring of the defective image after preliminary processing according to the underwater image quality evaluation standard is specifically obtained by calculating the sum of the weighted coefficients of the chromaticity standard deviation and the saturation average value of the defective image after preliminary processing. The underwater image quality evaluation standard UDIQE is an objective evaluation method without reference. It eliminates the global contrast based on UCIQE and recalculates the weighted coefficients of the chromaticity standard deviation and the saturation average value. It does not rely on the high-quality original image prepared in advance for comparison and can directly judge the quality of the image. The underwater image quality evaluation standard can be specifically expressed as:

[0042] in, 、 Represents the UCIQE linear weighting coefficient; 、 Indicates the existing linear weighting coefficient of UDIQE; Indicates the standard deviation of the image's chromaticity; Indicates the average saturation value of the image.

[0043] In addition, preferably, the defect type and location of the marked concrete structure can be saved, and the severity of the defect can be recorded, and the severity adopts a four-color marking method: red represents the most urgent situation, followed by orange, and then yellow and green, which respectively represent the severity of the defect decreasing step by step.

[0044] like Figure 6 FIG. 1 shows the overall measurement process of surface defect measurement of underwater concrete structures according to an embodiment of the present invention.

[0045] In addition, an embodiment of the present invention further provides an underwater concrete structure surface defect measurement system for implementing the underwater concrete structure surface defect measurement method described above in an embodiment of the present invention. The system includes: The calibration module is used to calibrate the camera underwater and establish the mapping relationship between the 3D entity points of the underwater concrete structure and the 2D image pixels based on the calibration results; a distortion correction module, configured to capture defect images of underwater concrete structures at different positions and / or angles using a calibrated camera and perform distortion correction on the defect images; A point cloud processing and 3D reconstruction module is used to perform point cloud processing on the point cloud data of the corrected defect image, reconstruct a 3D model of the concrete defect image based on the processed point cloud data, and calculate the defect feature size of the 3D model; the point cloud processing includes point cloud filtering, point cloud segmentation, and point cloud registration; The defect marking module is used to process the clarity of the defect image, fuse the defect feature size with the processed defect image, and mark the defect type and location of the concrete structure.

[0046] The various modules or mechanisms of the system are mainly used to implement the various steps of the above method embodiments, which will not be described in detail here.

[0047] In addition, an embodiment of the present invention further provides an image acquisition device, comprising a test water tank 1-1, a camera, a waterproof device 1-3, a clamping device 1-4 and a lighting device 1-5; a concrete structure 1-6 containing defects is placed on a boss 1-1-1 on one side of the water tank, and a graduated groove 1-1-2 is longitudinally provided at the upper end of the test water tank 1-1; the camera is sealed in the waterproof device 1-3, and a detachable polarizing lens 1-2-1 is installed at the front end of the camera; the waterproof device is fixed by two F-type trigger-type clamping devices 1-4, a water tank cross beam 1-1-3 is provided in the test water tank 1-1, and a water tank longitudinal beam that can move along the length direction of the water tank cross beam 1-1-3 is provided on the water tank cross beam 1-1-3, and the clamping device 1-4 is provided on the water tank longitudinal beam and can move up and down along the water tank longitudinal beam; the lighting device 1-5 is arranged on one side of the waterproof device 1-3; the image acquisition device is used for a method for measuring surface defects of an underwater concrete structure as described above.

[0048] The embodiments of the present invention provide a method, system, and device for measuring surface defects of underwater concrete structures. The method obtains an underwater concrete structure defect image and corrects its distortion, then performs point cloud processing on the corrected defect image to perform three-dimensional reconstruction of the concrete structure defects. Through point cloud filtering, segmentation, and alignment, the noise data interference in the defect image can be avoided to the greatest extent, and the defective part in the underwater concrete structure defect image can be accurately extracted for positioning and measurement. The method is more suitable for locating irregular structural surface defects of underwater concrete, and its clarity and accuracy are improved.

[0049] Furthermore, the embodiment of the present invention evaluates the processed image through the underwater image quality evaluation standard, which can accurately reflect the defect image quality of the underwater concrete structure in an objective manner, thereby obtaining a high-quality processed defect image, which is conducive to the accurate marking of the obtained defect feature size.

[0050] It should be pointed out that, according to the needs of implementation, the various steps / components described in this application can be split into more steps / components, or two or more steps / components or partial operations of steps / components can be combined into new steps / components to achieve the purpose of the present invention.

[0051] The size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0052] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. A method for measuring surface defects of underwater concrete structures, characterized in that: The method comprises: Perform underwater camera calibration and establish a mapping relationship between the 3D entity points of the underwater concrete structure and the 2D image pixels based on the calibration results; Using the calibrated camera to capture defect images of the underwater concrete structure at different positions and / or angles, and performing distortion correction on the defect images; Performing point cloud processing on the point cloud data of the corrected defect image, reconstructing a three-dimensional model of the concrete defect image based on the processed point cloud data, and calculating the defect characteristic size of the three-dimensional model; the point cloud processing includes point cloud filtering, point cloud segmentation, and point cloud registration; The defect image is processed for clarity, and the defect feature size is fused with the processed defect image to mark the defect type and location of the concrete structure.

2. The method for measuring surface defects of underwater concrete structures according to claim 1, characterized in that: The mapping relationship between underwater three-dimensional entity points and two-dimensional image pixel points is established by establishing a coordinate system, and the distortion correction of the defective image includes correction based on radial distortion and correction based on tangential distortion; among them, the corrected distortion point of the correction based on radial distortion is calculated by the distance between the coordinate origin and the distortion point and the radial distortion coefficient, and the corrected distortion point of the correction based on tangential distortion is calculated by the distance between the coordinate origin and the distortion point and the tangential distortion coefficient; the radial distortion coefficient and the tangential distortion coefficient are obtained according to calibration.

3. The method for measuring surface defects of underwater concrete structures according to claim 1, characterized in that: Point cloud filtering specifically includes: obtaining the distance between each point in the point cloud data of the corrected defect image and its neighboring points, and calculating the mean and standard deviation of the distance; The distance threshold is calculated based on the distance mean and standard deviation, and it is determined whether the distance between each point and its neighboring points exceeds the distance threshold. All neighboring points whose distance to the point exceeds the distance threshold are removed to obtain the filtered point cloud data.

4. A method for measuring surface defects of underwater concrete structures according to claim 1 or 3, characterized in that: Point cloud segmentation specifically includes: separating defect point cloud data from filtered point cloud data, clustering and screening the defect point cloud data, and retaining the largest defect point cloud cluster; Separating the defective point cloud data from the filtered point cloud data includes: selecting part of the point cloud data in the filtered point cloud data to form a minimum data set, and using the minimum data set to calculate a defect model; using the defect model to test all the data in the data set to determine the number of inliers that meet the defect model; iterating the defect model: expanding the number of point cloud data in the data set, and calculating a new defect model, comparing it with the previous better defect model, and setting the defect model with a larger number of inliers as the better defect model; repeating the iterative process until the number of iterations reaches a certain number or the number of inliers of the better defect model reaches a certain threshold, and obtaining the optimal defect model, at which time the inliers included in the optimal defect model are set as the defective point cloud data; wherein the inliers refer to the point cloud data that meet the conditions of the defect model; Clustering screening of defective point cloud data includes: randomly selecting a point cloud in the defective point cloud data as the starting point of the current cluster; performing a neighborhood search on the starting point to find all neighboring points whose distance to the starting point is less than a threshold; for each neighboring point found, repeating the neighborhood search process and adding the searched point cloud to the current cluster; when no more point clouds can be added to the current cluster, ending the cluster construction; reselecting a new point cloud that has not been assigned to any cluster, repeating the neighborhood search and cluster construction process until all point clouds are assigned to a cluster; comparing and analyzing each point cloud cluster, retaining the largest cluster as the defective point cloud cluster.

5. The method for measuring surface defects of underwater concrete structures according to claim 1, characterized in that: Point cloud registration specifically includes: using the random sampling consensus algorithm to perform coarse registration on the point cloud, and using the iterative closest point algorithm to perform fine registration on the coarsely registered point cloud.

6. The method for measuring surface defects of underwater concrete structures according to claim 1, characterized in that: The three-dimensional model of the concrete defect image is reconstructed by using the rolling ball method and smoothing the three-dimensional model of the concrete defect image; the defect characteristic size of the three-dimensional model is calculated, which specifically includes: calculating the length, width, depth, and three-dimensional volume of the defect, and performing a planar projection of the defect of the three-dimensional model and calculating the two-dimensional area of ​​the defect.

7. The method for measuring surface defects of underwater concrete structures according to claim 1, characterized in that: The clarity processing of the defect image specifically includes: performing preliminary processing on the underwater defect image of the concrete structure through the dark channel prior algorithm, and scoring the defect image after preliminary processing according to the underwater image quality evaluation standard, and reprocessing the defect image with a score lower than the set threshold until the score of all defect images is higher than the set threshold, thereby obtaining the processed defect image.

8. The method for measuring surface defects of underwater concrete structures according to claim 7, characterized in that: The defect image after preliminary processing is scored according to the underwater image quality evaluation standard by recalculating the sum of the weighted coefficient of the chromaticity standard deviation and the weighted coefficient of the saturation average value of the defect image after preliminary processing based on the calculation of UCIQE.

9. A surface defect measurement system for underwater concrete structures, characterized in that: The system comprises: The calibration module is used to calibrate the camera underwater and establish the mapping relationship between the 3D entity points of the underwater concrete structure and the 2D image pixels based on the calibration results; a distortion correction module, configured to capture defect images of underwater concrete structures at different positions and / or angles using a calibrated camera and perform distortion correction on the defect images; A point cloud processing and 3D reconstruction module is used to perform point cloud processing on the point cloud data of the corrected defect image, reconstruct a 3D model of the concrete defect image based on the processed point cloud data, and calculate the defect feature size of the 3D model; the point cloud processing includes point cloud filtering, point cloud segmentation, and point cloud registration; The defect marking module is used to process the clarity of the defect image, fuse the defect feature size with the processed defect image, and mark the defect type and location of the concrete structure.

10. An image acquisition device for implementing the underwater concrete structure surface defect measurement method according to any one of claims 1 to 8, characterized in that: The invention comprises a test water tank (1-1), a camera, a waterproof device (1-3), a clamping device (1-4) and a lighting device (1-5); a concrete structure (1-6) containing defects is placed on a boss (1-1-1) on one side of the water tank, and a scale groove (1-1-2) is provided on the upper end of the test water tank (1-1) along the longitudinal direction; the camera is sealed in the waterproof device (1-3), and a detachable polarization lens (1-2-1) is installed at the front end of the camera; the waterproof device is fixed by two F-type trigger-type clamping devices (1-4); a water tank crossbeam (1-1-3) is provided in the test water tank (1-1), and a water tank longitudinal beam that can move along the length direction of the water tank crossbeam (1-1-3) is provided on the water tank crossbeam (1-1-3); the clamping device (1-4) is provided on the water tank longitudinal beam and can move up and down along the water tank longitudinal beam; the lighting device (1-5) is arranged on one side of the waterproof device (1-3).

Images