Bridge crack detection method and system

Abstract

The present invention discloses a bridge crack detection method and system, which relate to the field of bridge detection technology. High-precision three-dimensional point cloud data P is generated through laser radar scanning. By aligning the three-dimensional point cloud data P and establishing a three-dimensional model C of the crack, the spatial accuracy of crack assessment is further improved, and the impact of cracks on the stability of the bridge structure can be accurately assessed. The comprehensive assessment index ZRZ is calculated and compared with the preset maintenance assessment threshold Tthe to help determine whether repair is needed and how to repair it. This full-process detection method solves the problems of insufficient accuracy and information fragmentation existing in traditional single detection methods, making the crack assessment results more comprehensive and reliable, providing more powerful technical support for bridge safety, and being able to detect potential crack risks in advance. It can also extend the service life of the bridge through scientific assessment and repair plans, ensuring that the bridge safety is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge detection, and in particular to a bridge crack detection method and system. Background Art

[0002] As global infrastructure continues to develop and age, bridges, as critical transportation facilities, face increasingly serious structural problems. Bridge health monitoring has become a key research topic in engineering. Bridge cracks are a common cause of structural failure, and their occurrence and expansion directly impact the service life and safety of bridges. Therefore, the detection and assessment of bridge cracks has become a core task in bridge maintenance and safety management.

[0003] In the invention of Chinese patent application No. CN202010174163.7, a bridge crack detection method is disclosed, which belongs to the field of bridge detection technology and aims to improve the detection accuracy and efficiency of bridge crack detection. The method comprises the following steps: crack segmentation of a set of collected bridge images; based on the crack segmentation results, detecting and classifying bridge cracks using a pre-built bridge crack classification model. The segmentation processing of bridge cracks adopts an improved GAC algorithm model, which can segment visible cracks in the image of the bottom of the bridge taken by the high-definition camera of the drone; the construction of the bridge crack classification model adopts a deep learning method, and a deep convolutional neural network model is designed for bridge identification; the three-dimensional reconstruction of bridge cracks and crack information detection adopts a moving cube algorithm to determine the number of cracks, average width, geometric properties and their spatial relationship with the whole, so that professionals can perform qualitative or quantitative analysis of the cracks. The present invention realizes the use of computer detection technology based on deep learning to solve corresponding crack detection and other construction problems.

[0004] As can be seen, image detection technology has been widely used for preliminary detection of cracks on bridge surfaces due to its ease of operation and low cost. However, relying solely on image processing technology also has significant limitations. Image detection typically only provides information on the surface morphology of cracks, but cannot effectively assess crack depth, extension, or internal damage to the bridge. Furthermore, image detection technology is susceptible to lighting conditions, weather factors, and image resolution, resulting in unstable detection results. Especially in complex environments, noise and ambiguity in the image can lead to inaccurate crack detection results. Summary of the Invention

[0005] In view of the deficiencies in the prior art, the present invention provides a bridge crack detection method and system, which solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A bridge crack detection method includes the following steps:

[0007] S1. Using a camera to photograph the bridge surface, preprocessing the acquired bridge surface image, and then detecting and extracting the processed bridge surface image to obtain a crack position feature vector V;

[0008] S2. Scanning the crack area of ​​the bridge using a laser radar device based on the obtained crack location feature vector V to obtain the three-dimensional coordinates of the bridge surface and generate three-dimensional point cloud data P. Preliminary analysis is performed based on the three-dimensional point cloud data P to determine whether there are structural plane faults and crack features, and obtain a determination result R.

[0009] S3. Triggering a crack verification mechanism based on the judgment result R, including registering the three-dimensional point cloud data P and then establishing a three-dimensional model C of the crack;

[0010] S4. Perform risk assessment of cracks on bridge outcomes based on the established three-dimensional crack model C, and obtain a comprehensive assessment index ZR of the impact of cracks on bridge stability;

[0011] S5. Compare the obtained comprehensive evaluation index ZR with the preset maintenance evaluation threshold Tthe to generate a maintenance repair plan and a detection plan for the bridge cracks.

[0012] Preferably, said S1 includes S11, S12 and S13;

[0013] S11. Use a high-definition camera and an infrared camera to capture images of the bridge surface to obtain a bridge surface image set Iraw. Then perform image preprocessing on the bridge surface image set Iraw, including performing denoising preprocessing using Gaussian filtering to obtain a denoised image set Ipre after the denoising preprocessing, and perform image enhancement preprocessing using histogram equalization to improve image contrast to obtain an enhanced image set Ib after the image enhancement preprocessing.

[0014] Preferably, S12, performing crack detection on the enhanced image set Ib by using an edge detection algorithm to obtain a crack edge region Iedges;

[0015] The crack edge region Iedges is specifically obtained through a gradient calculation step, a non-maximum suppression step, and an edge connection step;

[0016] The gradient calculation step calculates the gradient of each pixel in the image, including using the Sobel operator to calculate the gradient magnitude and direction of the image;

[0017] The non-maximum suppression step is used to refine the edge. Each pixel is detected along the gradient direction to see if it is a local maximum. If the pixel is not a local maximum, it is set to 0. Otherwise, the gradient value of the pixel is retained.

[0018] The edge connection step uses a dual threshold method to connect edges and distinguish whether the edge is a strong edge or a weak edge. Specifically, the distinction is made by setting a distinction threshold T1 and a distinction threshold T2, and the distinction threshold T1 is less than the distinction threshold T2.

[0019] When the gradient amplitude is greater than the discrimination threshold T2, the pixel value is determined to be a strong edge;

[0020] When the gradient amplitude is less than the distinction threshold T1, the pixel value is determined to be non-edge;

[0021] When the distinction threshold T1 < gradient amplitude < distinction threshold T2, the pixel value is determined to be a weak edge. If it is connected to a strong edge, it is determined to be a valid edge.

[0022] S13, then extracting features from the crack edge region Iedges, obtaining the spatial position features of the crack, and constructing a feature vector V, including the center coordinates (x, y), length Lmax, and width Wmax of the crack;

[0023] Among them, the center coordinates (x, y) are obtained by calculating the geometric center of the crack area, where x and y represent the horizontal and vertical coordinate values ​​of the center point, respectively; the length Lmax is obtained by calculating the maximum extension length of the crack; and the width Wmax is obtained by calculating the regional histogram of the crack.

[0024] Preferably, said S2 includes S21 and S22;

[0025] S21. Generate a laser radar scanning area Rscan based on the length Lmax and width Wmax of the center coordinate (x, y) in the crack location feature vector V. Then use the laser radar device to scan the crack area of ​​the bridge to obtain the three-dimensional coordinates (x, y, z) of the bridge surface, where x, y, and z represent the horizontal, vertical, and longitudinal coordinate values ​​of the three-dimensional coordinates, respectively. Then, integrate the three-dimensional coordinates (x, y, z) to form three-dimensional point cloud data P.

[0026] The laser radar scanning area Rscan is obtained by the following calculation formula:

[0027]

[0028] The three-dimensional point cloud data P is specifically P={(x1, y1, z1), (x2, y2, z2), ..., (xn, yn, zn)}.

[0029] Preferably, S22, performing denoising processing on the generated three-dimensional point cloud data P, including using a statistical outlier removal method and a voxel grid filtering method for denoising, and then performing a preliminary analysis on the denoised three-dimensional point cloud data P to determine whether there are structural plane faults and crack features, including using a plane fitting method to identify structural faults in the crack area, and obtaining a determination result R;

[0030] The judgment result R is obtained by the following analysis formula:

[0031]

[0032] When the judgment result R=1, it indicates that there are structural plane faults and cracks;

[0033] When the judgment result R=0, it means that there are no structural plane faults and cracks;

[0034] Where FE(P) represents the fitting error index, Fthe represents the preset deviation evaluation threshold between the cloud data and the fitting surface; Ki represents the local curvature of the i-th cloud point of the 3D point cloud data P, and Kthe represents the preset cloud point curvature evaluation threshold;

[0035] The fitting error index FE(P) is used to fit the crack area in the three-dimensional point cloud data P using the least squares method to obtain a plane. Specifically, the plane fitting is performed using the following calculation formula:

[0036] ax+by+cz+d=0;

[0037] Where a, b, and c represent the normal vector components of the horizontal axis x, vertical axis y, and longitudinal axis z in the plane, respectively, and d represents the distance between the plane and the origin;

[0038] The square error is then used to calculate the fitting error index FE(P) for the plane obtained by fitting, which reflects the degree of deviation between the point cloud point and the plane. The specific calculation formula is as follows:

[0039]

[0040] Wherein, (xi, yi, zi) represents the i-th point cloud in the three-dimensional point cloud data P, and n represents the total number of point clouds in the three-dimensional point cloud data P.

[0041] Preferably, said S3 includes S31 and S32;

[0042] S31. Triggering a crack verification mechanism based on a judgment result R. When the judgment result R=1, the crack verification mechanism is triggered, and the three-dimensional point cloud data P is registered using the crack position feature vector V. The point cloud data from different perspectives and at different times are merged into the same coordinate system to obtain the registered point cloud data Pa, including using a rigid transformation method for merging.

[0043] The point cloud data Pa is obtained by the calculation formula Pa=XZ*P+PY;

[0044] The rigid transformation method includes a rotation matrix XZ and a translation vector PY, wherein the rotation matrix XZ represents the rotation state of the point cloud data in the three-dimensional point cloud data P, and the translation vector PY represents the translation information of the point cloud data in the three-dimensional point cloud data P.

[0045] Preferably, S32, using a clustering algorithm and a point cloud filtering method to extract point cloud areas related to the crack on the registered point cloud data Pa, including using voxel grid filtering and region growing method for extraction, and then using a Poisson reconstruction algorithm to construct a three-dimensional model C of the crack based on the extracted point cloud area, and performing geometric analysis to obtain the length CD, width KD, and depth SD of the crack in the three-dimensional model C, and marking the crack in the three-dimensional model C;

[0046] The three-dimensional model C is obtained by a Poisson reconstruction algorithm formula;

[0047] The length CD is obtained by the calculation formula CD=max(||Ci-Cj||), i, j∈{1, 2, 3, ..., n};

[0048] Where max represents the maximum value function, ||Ci-Cj|| represents the Euclidean distance, specifically the straight-line distance between point cloud i and point cloud j in the three-dimensional model C;

[0049] The width KD is obtained by the calculation formula KD=max(||Ci-Cj||), Ci, Cj⊥CD;

[0050] Where Ci, Cj⊥CD represents two points of point cloud i and point cloud j perpendicular to the length CD direction in the three-dimensional model C;

[0051] The depth SD is obtained by the calculation formula SD=max(||Ci-Cbottom||), i∈{1, 2, 3, ..., n};

[0052] Where Cbottom represents the lower limit point of the three-dimensional model C.

[0053] Preferably, the S4 includes S41;

[0054] S41. Perform risk assessment of cracks on bridge outcomes based on the established three-dimensional crack model C, and obtain a comprehensive assessment index ZR of crack impact on bridge stability.

[0055] The comprehensive evaluation index ZR is obtained by the following calculation formula:

[0056] ZR=z1*CD+z2*KD+z3*SD+z4*QYL+z5*QCL;

[0057] Where QYL represents the bridge stress, which is obtained by the bridge stress state monitoring sensor to measure the bridge stress state under load. QCL represents the bridge material property, which is determined based on the bridge material and physical properties, including material tests and standard values ​​specified in the design manual. z1, z2, z3, z4, and z5 represent the preset weights of length CD, width KD, depth SD, bridge stress QYL, and bridge material property QCL, respectively. z1 + z2 + z3 + z4 + z5 = 1, and the specific values ​​are set by the user.

[0058] Preferably, the bridge crack repair plan and detection plan are obtained by the following comparison method:

[0059] When the comprehensive assessment index ZR ≥ the preparation assessment threshold Tthe, a preparation and repair plan for the bridge crack is generated, including sending the bridge crack location information to the relevant preparation and repair department's pending list for processing, and simultaneously generating an inspection plan, including adjusting the inspection frequency of the bridge crack location to twice the preset frequency;

[0060] When the comprehensive evaluation index ZR is less than the preparation evaluation threshold Tthe, no preparation and repair plan for the bridge crack is generated, and the detection plan is initialized, including adjusting the current detection frequency of the bridge crack position to the preset frequency.

[0061] A bridge crack detection system includes a surface image acquisition module, a laser scanning module, a verification module, a risk assessment module, and a decision generation module;

[0062] The surface image acquisition module uses a camera to shoot the bridge surface, pre-processes the acquired bridge surface image, and then detects and extracts the processed bridge surface image to obtain a crack position feature vector V;

[0063] The laser scanning module uses a laser radar device to scan the crack area of ​​the bridge according to the obtained crack position feature vector V, obtains the three-dimensional coordinates of the bridge surface, and generates three-dimensional point cloud data P. The three-dimensional point cloud data P is then used for preliminary analysis to determine whether there are structural plane faults and crack features, and obtain a judgment result R.

[0064] The verification module triggers the crack verification mechanism according to the judgment result R, including registering the three-dimensional point cloud data P and then establishing a three-dimensional model C of the crack;

[0065] The risk assessment module performs risk assessment of cracks on bridge results based on the established three-dimensional crack model C, and obtains a comprehensive assessment index ZR of the impact of cracks on bridge stability;

[0066] The decision generation module compares the obtained comprehensive evaluation index ZR with a preset maintenance evaluation threshold Tthe to generate a maintenance and repair plan and a detection plan for bridge cracks.

[0067] The present invention provides a bridge crack detection method and system, which have the following beneficial effects:

[0068] (1) High-precision three-dimensional point cloud data P is generated through laser radar scanning. By aligning the three-dimensional point cloud data P and establishing a three-dimensional model C of the crack, the spatial accuracy of the crack assessment is further improved, and the impact of cracks on the stability of the bridge structure can be accurately assessed. The comprehensive assessment index ZRZ is calculated and compared with the preset maintenance assessment threshold Tthe to help determine whether repair is needed and how to repair it. This full-process detection method solves the problems of insufficient accuracy and information fragmentation in traditional single detection methods, making the crack assessment results more comprehensive and reliable, providing more powerful technical support for bridge safety, and being able to detect potential crack risks in advance. It can also extend the service life of the bridge through scientific assessment and repair plans, ensuring that the safety of the bridge is greatly improved, thereby effectively preventing structural failures and safety accidents caused by cracks.

[0069] (2) By judging the result R = 1, the crack verification mechanism is activated, and the crack position feature vector V is combined with the three-dimensional point cloud data P to obtain the registered point cloud data Pa. This step effectively avoids data inconsistencies caused by point cloud data misalignment or perspective differences, ensuring that the geometric features of the crack can be accurately restored after multiple scans. The Poisson reconstruction algorithm is used to successfully construct the three-dimensional model C of the crack. Through the accurate extraction of crack geometric features, data support can be provided for bridge reinforcement and repair, significantly improving the reliability and timeliness of bridge safety assessments and reducing the uncertainty caused by manual inspection and empirical judgment.

[0070] (3) Based on the comparison between the comprehensive evaluation index ZR and the preset maintenance evaluation threshold Tthe, a maintenance and repair plan or an adjustment detection plan can be automatically generated. This effectively establishes dynamic feedback between crack repair and monitoring, which can reduce unnecessary waste of maintenance resources while ensuring bridge safety, achieve efficient resource allocation and management, and improve the intelligence level of crack detection and repair decision-making. It also optimizes the utilization of monitoring resources by dynamically adjusting the monitoring frequency, effectively reducing maintenance costs. At the same time, the quantitative evaluation based on the comprehensive evaluation index ensures the scientific nature and pertinence of the maintenance and repair plan, thereby providing more accurate and real-time decision support in ensuring bridge safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 This is a schematic diagram of the steps of a bridge crack detection method according to the present invention;

[0072] Figure 2 This is a schematic block diagram of a bridge crack detection system according to the present invention. DETAILED DESCRIPTION

[0073] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0074] Example 1

[0075] The present invention provides a bridge crack detection method, please refer to Figure 1 , including the following steps:

[0076] S1. Using a camera to photograph the bridge surface, preprocessing the acquired bridge surface image, and then detecting and extracting the processed bridge surface image to obtain a crack position feature vector V;

[0077] S2. Scanning the crack area of ​​the bridge using a laser radar device based on the obtained crack location feature vector V to obtain the three-dimensional coordinates of the bridge surface and generate three-dimensional point cloud data P. Preliminary analysis is performed based on the three-dimensional point cloud data P to determine whether there are structural plane faults and crack features, and obtain a determination result R.

[0078] S3. Triggering a crack verification mechanism based on the judgment result R, including registering the three-dimensional point cloud data P and then establishing a three-dimensional model C of the crack;

[0079] S4. Perform risk assessment of cracks on bridge outcomes based on the established three-dimensional crack model C, and obtain a comprehensive assessment index ZR of the impact of cracks on bridge stability;

[0080] S5. Compare the obtained comprehensive evaluation index ZR with the preset maintenance evaluation threshold Tthe to generate a maintenance repair plan and a detection plan for the bridge cracks.

[0081] In this embodiment, a camera captures and preprocesses images of the bridge surface, effectively extracting crack location feature vectors V. LiDAR scanning then generates high-precision 3D point cloud data P, providing richer crack information. This process not only ensures more accurate crack location but also integrates multi-source information fusion between image data and 3D point cloud data, providing a more precise basis for analyzing crack depth, morphology, and topology. Subsequently, by registering the 3D point cloud data P and constructing a 3D crack model C, the spatial accuracy of crack assessment is further improved. This allows for a precise assessment of the impact of cracks on bridge structural stability, calculating a comprehensive assessment index ZRZ, and comparing it with a preset maintenance assessment threshold Tthe to determine whether repairs are necessary and how they should be performed. This comprehensive inspection method addresses the inaccuracies and information fragmentation inherent in traditional single-step inspection methods, resulting in more comprehensive and reliable crack assessment results. This provides stronger technical support for bridge safety, enables early detection of potential crack risks, and extends bridge service life through scientific assessment and repair plans, significantly improving bridge safety and effectively preventing structural failures and accidents caused by cracks.

[0082] Example 2

[0083] This embodiment is explained in Example 1, please refer to Figure 1 , specifically: said S1 includes S11, S12 and S13;

[0084] S11. Use a high-definition camera and an infrared camera to capture images of the bridge surface to obtain a bridge surface image set Iraw. Then perform image preprocessing on the bridge surface image set Iraw, including performing denoising preprocessing using Gaussian filtering to obtain a denoised image set Ipre after the denoising preprocessing, and perform image enhancement preprocessing using histogram equalization to improve image contrast to obtain an enhanced image set Ib after the image enhancement preprocessing.

[0085] S12, performing crack detection on the enhanced image set Ib by using an edge detection algorithm to obtain a crack edge region Iedges;

[0086] The crack edge region Iedges is specifically obtained through a gradient calculation step, a non-maximum suppression step, and an edge connection step;

[0087] The gradient calculation step calculates the gradient of each pixel in the image, including using the Sobel operator to calculate the gradient magnitude and direction of the image;

[0088] The non-maximum suppression step is used to refine the edge. Each pixel is detected along the gradient direction to see if it is a local maximum. If the pixel is not a local maximum, it is set to 0. Otherwise, the gradient value of the pixel is retained.

[0089] The edge connection step uses a dual threshold method to connect edges and distinguish whether the edge is a strong edge or a weak edge. Specifically, the distinction is made by setting a distinction threshold T1 and a distinction threshold T2, and the distinction threshold T1 is less than the distinction threshold T2.

[0090] When the gradient amplitude is greater than the discrimination threshold T2, the pixel value is determined to be a strong edge;

[0091] When the gradient amplitude is less than the distinction threshold T1, the pixel value is determined to be non-edge;

[0092] When the distinction threshold T1 < gradient amplitude < distinction threshold T2, the pixel value is determined to be a weak edge. If it is connected to a strong edge, it is determined to be a valid edge.

[0093] S13, then extracting features from the crack edge region Iedges, obtaining the spatial position features of the crack, and constructing a feature vector V, including the center coordinates (x, y), length Lmax, and width Wmax of the crack;

[0094] Among them, the center coordinates (x, y) are obtained by calculating the geometric center of the crack area, where x and y represent the horizontal and vertical coordinate values ​​of the center point, respectively; the length Lmax is obtained by calculating the maximum extension length of the crack; and the width Wmax is obtained by calculating the regional histogram of the crack.

[0095] In this example, a bridge surface image set Iraw captured by a high-definition camera and an infrared camera is subjected to Gaussian filtering for denoising and histogram equalization for contrast enhancement, resulting in a denoised image set Ipre and an enhanced image set Ib. This provides a high-quality image foundation for subsequent crack detection. An edge detection algorithm is used to further extract the crack edge region Iedges. Gradient calculation, non-maximum suppression, and edge connection are employed to effectively remove noise and ensure accurate identification of crack edge features. In particular, a dual-threshold method is used to finely distinguish strong and weak edges, enhancing the robustness of crack detection. Finally, based on the crack edge region Iedges, the spatial location features of the crack are extracted and a crack feature vector V is constructed, comprising the crack's center coordinates (x, y), maximum extension length Lmax, and maximum width Wmax. These feature vectors provide precise crack location information for subsequent 3D modeling and risk assessment. This significantly improves the accuracy and reliability of crack detection, enabling a more comprehensive assessment of the health of bridge structures and avoiding potential safety risks caused by inaccurate crack location or missed detection. This method not only ensures the accuracy of crack detection, but also effectively reduces the impact of environmental factors or equipment problems on the results, providing reliable technical support for the long-term monitoring and maintenance of bridges.

[0096] Example 3

[0097] This embodiment is explained in Example 2, please refer to Figure 1 Specifically: S2 includes S21 and S22;

[0098] S21. Generate a laser radar scanning area Rscan based on the length Lmax and width Wmax of the center coordinate (x, y) in the crack location feature vector V. Then use the laser radar device to scan the crack area of ​​the bridge to obtain the three-dimensional coordinates (x, y, z) of the bridge surface, where x, y, and z represent the horizontal, vertical, and longitudinal coordinate values ​​of the three-dimensional coordinates, respectively. Then, integrate the three-dimensional coordinates (x, y, z) to form three-dimensional point cloud data P.

[0099] The laser radar scanning area Rscan is obtained by the following calculation formula:

[0100]

[0101] The three-dimensional point cloud data P is specifically P={(x1, y1, z1), (x2, y2, z2), ..., (xn, yn, zn)}.

[0102] S22, performing denoising processing on the generated three-dimensional point cloud data P, including using a statistical outlier removal method and a voxel grid filtering method for denoising, and then performing a preliminary analysis on the denoised three-dimensional point cloud data P to determine whether there are structural plane faults and crack features, including using a plane fitting method to identify structural faults in the crack area, and obtaining a determination result R;

[0103] The judgment result R is obtained by the following analysis formula:

[0104]

[0105] When the judgment result R=1, it indicates that there are structural plane faults and cracks;

[0106] When the judgment result R=0, it means that there are no structural plane faults and cracks;

[0107] Where FE(P) represents the fitting error index, Fthe represents the preset deviation evaluation threshold between the cloud data and the fitting surface; Ki represents the local curvature of the i-th cloud point of the 3D point cloud data P, and Kthe represents the preset cloud point curvature evaluation threshold;

[0108] The fitting error index FE(P) is used to fit the crack area in the three-dimensional point cloud data P using the least squares method to obtain a plane. Specifically, the plane fitting is performed using the following calculation formula:

[0109] ax+by+cz+d=0;

[0110] Where a, b, and c represent the normal vector components of the horizontal axis x, vertical axis y, and longitudinal axis z in the plane, respectively. The normal vector is a vector perpendicular to the plane, specifically indicating the position and orientation of the plane. d represents the distance between the plane and the origin.

[0111] The square error is then used to calculate the fitting error index FE(P) for the plane obtained by fitting, which reflects the degree of deviation between the point cloud point and the plane. The specific calculation formula is as follows:

[0112]

[0113] Wherein, (xi, yi, zi) represents the i-th point cloud in the three-dimensional point cloud data P, and n represents the total number of point clouds in the three-dimensional point cloud data P.

[0114] In this embodiment, the LiDAR scanning area Rscan is determined based on the crack's center coordinates (x, y), maximum length Lmax, and maximum width Wmax. A LiDAR scan is then performed within this area, acquiring complete three-dimensional coordinate data (x, y, z). This process ensures comprehensive capture of the spatial information of the crack area. Next, the generated three-dimensional point cloud data P is denoised using statistical outlier removal and voxel grid filtering to remove noise points and irrelevant data, ensuring the accuracy and reliability of the point cloud data. Furthermore, a plane fitting method is used to analyze the three-dimensional point cloud of the crack area to identify the presence of structural plane faults or crack features, resulting in a judgment result R. A judgment result of R = 1 indicates the presence of structural plane faults and crack features in the crack area. This judgment method based on three-dimensional point cloud data not only improves the spatial accuracy of crack detection but also effectively identifies cracks and their structural features, avoiding the misjudgment and omission issues that can occur with traditional two-dimensional image analysis. Accurate three-dimensional scanning and data analysis can provide more reliable basic data for subsequent risk assessment and repair, thereby improving the safety and scientific nature of bridge inspections.

[0115] Example 4

[0116] This embodiment is explained in Example 3, please refer to Figure 1 , specifically: said S3 includes S31 and S32;

[0117] S31. Triggering a crack verification mechanism based on a judgment result R. When the judgment result R=1, the crack verification mechanism is triggered, and the three-dimensional point cloud data P is registered using the crack position feature vector V. The point cloud data from different perspectives and at different times are merged into the same coordinate system to obtain the registered point cloud data Pa, including using a rigid transformation method for merging.

[0118] The point cloud data Pa is obtained by the calculation formula Pa=XZ*P+PY;

[0119] The rigid transformation method includes a rotation matrix XZ and a translation vector PY, wherein the rotation matrix XZ represents the rotation state of the point cloud data in the three-dimensional point cloud data P. Specifically, each point in the three-dimensional point cloud data P is rotated by the least squares method. After each point of the point cloud data is rotated, multiple point cloud data are aligned in the same coordinate system for acquisition. The translation vector PY represents the translation information of the point cloud data in the three-dimensional point cloud data P. Specifically, the least squares method is used to translate the entire point cloud along a certain direction so that the point cloud can be aligned with the target position to obtain a three-dimensional vector, which specifically represents the translation amount on the three-dimensional coordinates (x, y, z).

[0120] S32. Using a clustering algorithm and a point cloud filtering method to extract point cloud regions related to the crack from the registered point cloud data Pa, including using voxel grid filtering and region growing methods for extraction, and then using a Poisson reconstruction algorithm to construct a three-dimensional model C of the crack based on the extracted point cloud regions. Performing geometric analysis to obtain the length CD, width KD, and depth SD of the crack in the three-dimensional model C, and marking the crack in the three-dimensional model C.

[0121] The three-dimensional model C is obtained by a Poisson reconstruction algorithm formula;

[0122] The length CD is obtained by the calculation formula CD=max(||Ci-Cj||), i, j∈{1, 2, 3, ..., n};

[0123] Where max represents the maximum value function, ||Ci-Cj|| represents the Euclidean distance, specifically the straight-line distance between point cloud i and point cloud j in the three-dimensional model C;

[0124] The width KD is obtained by the calculation formula KD=max(||Ci-Cj||), Ci, Cj⊥CD;

[0125] Where Ci, Cj⊥CD represents two points of point cloud i and point cloud j perpendicular to the length CD direction in the three-dimensional model C;

[0126] The depth SD is obtained by the calculation formula SD=max(||Ci-Cbottom||), i∈{1, 2, 3, ..., n};

[0127] Where Cbottom represents the lower limit point of the bottom of the three-dimensional model C, specifically the minimum point of the z-axis in the three-dimensional model C.

[0128] In this embodiment, the crack verification mechanism is activated by checking the result R = 1. The crack location feature vector V is combined with the 3D point cloud data P. The point cloud data is then registered using a rigid transformation method to ensure that point cloud data from different viewpoints and at different times are accurately aligned to the same coordinate system, resulting in the registered point cloud data Pa. This step effectively avoids data inconsistencies caused by point cloud data misalignment or perspective differences, ensuring that the geometric features of the crack can be accurately restored across multiple scans. Clustering algorithms and point cloud filtering methods are then used to further extract crack-related point cloud regions. Combining voxel grid filtering and region growing, noise and irrelevant regions are filtered out, retaining only the valid crack point cloud. Based on these accurately extracted point cloud regions, a Poisson reconstruction algorithm is applied to successfully construct a 3D crack model C. In this model, the crack's geometric features, including length CD, width KD, and depth SD, are precisely calculated and labeled. This not only improves crack detection accuracy but also provides detailed spatial data for subsequent bridge stability assessment and repair planning. Through accurate extraction of crack geometric features, data support can be provided for bridge reinforcement and repair, significantly improving the reliability and timeliness of bridge safety assessments and reducing the uncertainty caused by manual inspection and empirical judgment.

[0129] Example 5

[0130] This embodiment is explained in Example 4. Please refer to Figure 1 Specifically: the S4 includes S41;

[0131] S41. Perform risk assessment of cracks on bridge outcomes based on the established three-dimensional crack model C, and obtain a comprehensive assessment index ZR of crack impact on bridge stability.

[0132] The comprehensive evaluation index ZR is obtained by the following calculation formula:

[0133] ZR=z1*CD+z2*KD+z3*SD+z4*QYL+z5*QCL;

[0134] Where QYL represents the bridge stress, which is obtained by the bridge stress state monitoring sensor to measure the bridge stress state under load. QCL represents the bridge material property, which is determined based on the bridge material and physical properties, including material tests and standard values ​​specified in the design manual. z1, z2, z3, z4, and z5 represent the preset weights of length CD, width KD, depth SD, bridge stress QYL, and bridge material property QCL, respectively. z1 + z2 + z3 + z4 + z5 = 1, and the specific values ​​are set by the user.

[0135] The bridge crack repair and inspection solutions are obtained by comparing the following:

[0136] When the comprehensive assessment index ZR ≥ the preparation assessment threshold Tthe, a preparation and repair plan for the bridge crack is generated, including sending the bridge crack location information to the relevant preparation and repair department's pending list for processing, and simultaneously generating an inspection plan, including adjusting the inspection frequency of the bridge crack location to twice the preset frequency;

[0137] When the comprehensive evaluation index ZR is less than the preparation evaluation threshold Tthe, no preparation and repair plan for the bridge crack is generated, and the detection plan is initialized, including adjusting the current detection frequency of the bridge crack position to the preset frequency.

[0138] In this embodiment, based on the comparison of the comprehensive evaluation index ZR with the preset maintenance assessment threshold Tthe, a maintenance and repair plan can be automatically generated or the inspection plan adjusted. This effectively establishes dynamic feedback between crack repair and monitoring, ensuring bridge safety while reducing unnecessary waste of maintenance resources, achieving efficient resource allocation and management, and improving the intelligence level of crack detection and repair decision-making. Dynamic adjustment of monitoring frequency also optimizes monitoring resource utilization, effectively reducing maintenance costs. Furthermore, quantitative evaluation based on the comprehensive evaluation index ensures the scientific and targeted nature of the maintenance and repair plan, thereby providing more accurate and real-time decision-making support for ensuring bridge safety.

[0139] Example 6

[0140] A bridge crack detection system, please refer to Figure 2 ,Specifically: including surface image acquisition module, laser scanning module, ,verification module, risk assessment module and decision generation module;

[0141] The surface image acquisition module uses a camera to shoot the bridge surface, pre-processes the acquired bridge surface image, and then detects and extracts the processed bridge surface image to obtain a crack position feature vector V;

[0142] The laser scanning module uses a laser radar device to scan the crack area of ​​the bridge according to the obtained crack position feature vector V, obtains the three-dimensional coordinates of the bridge surface, and generates three-dimensional point cloud data P. The three-dimensional point cloud data P is then used for preliminary analysis to determine whether there are structural plane faults and crack features, and obtain a judgment result R.

[0143] The verification module triggers the crack verification mechanism according to the judgment result R, including registering the three-dimensional point cloud data P and then establishing a three-dimensional model C of the crack;

[0144] The risk assessment module performs risk assessment of cracks on bridge results based on the established three-dimensional crack model C, and obtains a comprehensive assessment index ZR of the impact of cracks on bridge stability;

[0145] The decision generation module compares the obtained comprehensive evaluation index ZR with a preset maintenance evaluation threshold Tthe to generate a maintenance and repair plan and a detection plan for bridge cracks.

[0146] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A bridge crack detection method, characterized by: The following steps are involved: S1. Using a camera to photograph the bridge surface, preprocessing the acquired bridge surface image, and then detecting and extracting the processed bridge surface image to obtain a crack position feature vector V; S2. Scanning the crack area of ​​the bridge using a laser radar device based on the obtained crack location feature vector V to obtain the three-dimensional coordinates of the bridge surface and generate three-dimensional point cloud data P. Preliminary analysis is performed based on the three-dimensional point cloud data P to determine whether there are structural plane faults and crack features, and obtain a determination result R. Said S2 includes S21 and S22; S21. Generate a laser radar scanning area Rscan based on the length Lmax and width Wmax of the center coordinate (x, y) in the crack location feature vector V. Then use the laser radar device to scan the crack area of ​​the bridge to obtain the three-dimensional coordinates (x, y, z) of the bridge surface, where x, y, and z represent the horizontal, vertical, and longitudinal coordinate values ​​of the three-dimensional coordinates, respectively. Then, integrate the three-dimensional coordinates (x, y, z) to form three-dimensional point cloud data P. The laser radar scanning area Rscan is obtained by the following calculation formula: The three-dimensional point cloud data P is specifically P={(x1, y1, z1), (x2, y2, z2), ..., (xn, yn, zn)}; S22, performing denoising processing on the generated three-dimensional point cloud data P, including using a statistical outlier removal method and a voxel grid filtering method for denoising, and then performing a preliminary analysis on the denoised three-dimensional point cloud data P to determine whether there are structural plane faults and crack features, including using a plane fitting method to identify structural faults in the crack area, and obtaining a determination result R; The judgment result R is obtained by the following analysis formula: When the judgment result R=1, it indicates that there are structural plane faults and cracks; When the judgment result R=0, it means that there are no structural plane faults and cracks; Where FE(P) represents the fitting error index, Fthe represents the preset deviation evaluation threshold between the cloud data and the fitting surface; Ki represents the local curvature of the i-th cloud point of the 3D point cloud data P, and Kthe represents the preset cloud point curvature evaluation threshold; The fitting error index FE(P) is used to fit the crack area in the three-dimensional point cloud data P using the least squares method to obtain a plane. Specifically, the plane fitting is performed using the following calculation formula: ax+by+cz+d=0; Where a, b, and c represent the normal vector components of the horizontal axis x, vertical axis y, and longitudinal axis z in the plane, respectively, and d represents the distance between the plane and the origin; The square error is then used to calculate the fitting error index FE(P) for the plane obtained by fitting, which reflects the degree of deviation between the point cloud point and the plane. The specific calculation formula is as follows: Where (xi, yi, zi) represents the i-th point cloud in the three-dimensional point cloud data P, and n represents the total number of point clouds in the three-dimensional point cloud data P; S3. Triggering a crack verification mechanism based on the judgment result R, including registering the three-dimensional point cloud data P and then establishing a three-dimensional model C of the crack; Said S3 includes S31; S31. Triggering a crack verification mechanism based on a judgment result R. When the judgment result R=1, the crack verification mechanism is triggered, and the three-dimensional point cloud data P is registered using the crack position feature vector V. The point cloud data from different perspectives and at different times are merged into the same coordinate system to obtain the registered point cloud data Pa, including using a rigid transformation method for merging. The point cloud data Pa is obtained by the calculation formula Pa=XZ*P+PY; The rigid transformation method includes a rotation matrix XZ and a translation vector PY, wherein the rotation matrix XZ represents the rotation state of the point cloud data in the three-dimensional point cloud data P, and the translation vector PY represents the translation information of the point cloud data in the three-dimensional point cloud data P; S4. Perform risk assessment of cracks on bridge outcomes based on the established three-dimensional crack model C, and obtain a comprehensive assessment index ZR of the impact of cracks on bridge stability; S5. Compare the obtained comprehensive evaluation index ZR with the preset maintenance evaluation threshold Tthe to generate a maintenance repair plan and a detection plan for the bridge cracks.

2. A bridge crack detection method according to claim 1, characterized in that: Said S1 includes S11, S12 and S13; S11. Use a high-definition camera and an infrared camera to capture images of the bridge surface to obtain a bridge surface image set Iraw. Then perform image preprocessing on the bridge surface image set Iraw, including performing denoising preprocessing using Gaussian filtering to obtain a denoised image set Ipre after the denoising preprocessing, and perform image enhancement preprocessing using histogram equalization to improve image contrast to obtain an enhanced image set Ib after the image enhancement preprocessing.

3. A bridge crack detection method according to claim 2, characterized in that: S12, performing crack detection on the enhanced image set Ib by using an edge detection algorithm to obtain a crack edge region Iedges; The crack edge region Iedges is specifically obtained through a gradient calculation step, a non-maximum suppression step, and an edge connection step; The gradient calculation step calculates the gradient of each pixel in the image, including using the Sobel operator to calculate the gradient magnitude and direction of the image; The non-maximum suppression step is used to refine the edge. Each pixel is detected along the gradient direction to see if it is a local maximum. If the pixel is not a local maximum, it is set to 0. Otherwise, the gradient value of the pixel is retained. The edge connection step uses a dual threshold method to connect edges and distinguish whether the edge is a strong edge or a weak edge. Specifically, the distinction is made by setting a distinction threshold T1 and a distinction threshold T2, and the distinction threshold T1 is less than the distinction threshold T2. When the gradient amplitude is greater than the discrimination threshold T2, the pixel value is determined to be a strong edge; When the gradient amplitude is less than the distinction threshold T1, the pixel value is determined to be non-edge; When the distinction threshold T1 < gradient amplitude < distinction threshold T2, the pixel value is determined to be a weak edge. If it is connected to a strong edge, it is determined to be a valid edge. S13, then extracting features from the crack edge region Iedges, obtaining the spatial position features of the crack, and constructing a feature vector V, including the center coordinates (x, y), length Lmax, and width Wmax of the crack; Among them, the center coordinates (x, y) are obtained by calculating the geometric center of the crack area, where x and y represent the horizontal and vertical coordinate values ​​of the center point, respectively; the length Lmax is obtained by calculating the maximum extension length of the crack; and the width Wmax is obtained by calculating the regional histogram of the crack.

4. The bridge crack detection method according to claim 1, characterized in that: Said S3 also includes S32; S32. Using a clustering algorithm and a point cloud filtering method to extract point cloud regions related to the crack from the registered point cloud data Pa, including using voxel grid filtering and region growing methods for extraction, and then using a Poisson reconstruction algorithm to construct a three-dimensional model C of the crack based on the extracted point cloud regions. Performing geometric analysis to obtain the length CD, width KD, and depth SD of the crack in the three-dimensional model C, and marking the crack in the three-dimensional model C. The three-dimensional model C is obtained by a Poisson reconstruction algorithm formula; The length CD is obtained by the calculation formula CD=max(||Ci-Cj||), i, j∈{1, 2, 3, ..., n}; Where max represents the maximum value function, ||Ci-Cj|| represents the Euclidean distance, specifically the straight-line distance between point cloud i and point cloud j in the three-dimensional model C; The width KD is obtained by the calculation formula KD=max(||Ci-Cj||), Ci, Cj⊥CD; Where Ci, Cj⊥CD represents two points of point cloud i and point cloud j perpendicular to the length CD direction in the three-dimensional model C; The depth SD is obtained by the calculation formula SD=max(||Ci-Cbottom||), i∈{1, 2, 3, ..., n}; Where Cbottom represents the lower limit point of the three-dimensional model C.

5. A bridge crack detection method according to claim 4, characterized in that: Said S4 includes S41; S41. Perform risk assessment of cracks on bridge outcomes based on the established three-dimensional crack model C, and obtain a comprehensive assessment index ZR of crack impact on bridge stability. The comprehensive evaluation index ZR is obtained by the following calculation formula: ZR=z1*CD+z2*KD+z3*SD+z4*QYL+z5*QCL; Where QYL represents the bridge stress, which is obtained by the bridge stress state monitoring sensor to measure the bridge stress state under load. QCL represents the bridge material property, which is determined based on the bridge material and physical properties, including material tests and standard values ​​specified in the design manual. z1, z2, z3, z4, and z5 represent the preset weights of length CD, width KD, depth SD, bridge stress QYL, and bridge material property QCL, respectively. z1 + z2 + z3 + z4 + z5 = 1, and the specific values ​​are set by the user.

6. The bridge crack detection method according to claim 1, characterized in that: The bridge crack repair and inspection solutions are obtained by comparing the following: When the comprehensive assessment index ZR ≥ the preparation assessment threshold Tthe, a preparation and repair plan for the bridge crack is generated, including sending the bridge crack location information to the relevant preparation and repair department's pending list for processing, and simultaneously generating an inspection plan, including adjusting the inspection frequency of the bridge crack location to twice the preset frequency; When the comprehensive evaluation index ZR is less than the preparation evaluation threshold Tthe, no preparation and repair plan for the bridge crack is generated, and the detection plan is initialized, including adjusting the current detection frequency of the bridge crack position to the preset frequency.

7. A bridge crack detection system, applied to a bridge crack detection method according to any one of claims 1 to 6, characterized in that: It includes surface image acquisition module, laser scanning module, verification module, risk assessment module and decision making module; The surface image acquisition module uses a camera to shoot the bridge surface, pre-processes the acquired bridge surface image, and then detects and extracts the processed bridge surface image to obtain a crack position feature vector V; The laser scanning module uses a laser radar device to scan the crack area of ​​the bridge according to the obtained crack position feature vector V, obtains the three-dimensional coordinates of the bridge surface, and generates three-dimensional point cloud data P. The three-dimensional point cloud data P is then used for preliminary analysis to determine whether there are structural plane faults and crack features, and obtain a judgment result R. The verification module triggers the crack verification mechanism according to the judgment result R, including registering the three-dimensional point cloud data P and then establishing a three-dimensional model C of the crack; The risk assessment module performs risk assessment of cracks on bridge results based on the established three-dimensional crack model C, and obtains a comprehensive assessment index ZR of the impact of cracks on bridge stability; The decision generation module compares the obtained comprehensive evaluation index ZR with a preset maintenance evaluation threshold Tthe to generate a maintenance and repair plan and a detection plan for bridge cracks.

Images