Dam crack measurement and analysis method based on double laser modeling and imu attitude compensation

Abstract

The application discloses a dam crack measurement and analysis method based on double laser modeling and IMU posture compensation, and relates to the technical field of underwater dam crack measurement. The method comprises the following steps: calibrating the intrinsic matrix and distortion coefficient of an underwater camera in a water medium through Zhang Zhengyou calibration method, and realizing distortion correction of an image according to the same; acquiring the extrinsic matrix of the camera in real time by using an IMU, and calibrating the fixed posture offset between the camera and the IMU; constructing an inverse proportional calibration model between the distance of double laser points and the depth, and realizing non-contact depth acquisition; realizing reverse projection restoration based on depth constraint based on a forward projection equation and depth information, and solving the three-dimensional coordinates of crack points; and realizing millimeter-level high-precision measurement of crack width by combining image segmentation and geometric calculation. The crack measurement and analysis method effectively overcomes problems such as underwater refraction, camera posture disturbance and depth information loss, and is suitable for structural safety detection in scenes such as underwater dams and marine engineering.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of underwater dam crack measurement, in particular to a dam crack measurement and analysis method based on double laser modeling and IMU attitude compensation. BACKGROUND

[0002] At present, about 95,000 large dams in China bear the core functions of flood control, water supply, power generation and navigation, and their long-term service performance and structural safety are of great concern. In the complex and harsh underwater environment, the concrete structures of dam body, stilling basin and gate pier are prone to cracks and erosion under the combined action of long-term water erosion, freeze-thaw cycle, alkali aggregate reaction and load fatigue. Once the structural damage accumulates to the critical point of instability, it may cause a catastrophic dam failure accident, with disastrous consequences.

[0003] Current underwater dam crack measurement technology mainly includes manual direct measurement and non-contact detection based on robots, but both have significant technical bottlenecks, especially in actual size quantification, which has not yet formed a mature solution.

[0004] Manual measurement is the traditional means of underwater crack detection, which relies on divers carrying contact instruments to complete data acquisition. Common technical methods include mechanical measurement, ultrasonic depth measurement technology and geophysical auxiliary method: mechanical measurement uses underwater vernier caliper, depth gauge and other tools to directly measure the width and depth of the crack surface. The principle is simple but limited by the operation space and hand stability of the diver, and it is not suitable for narrow and deep cracks. Ultrasonic depth measurement technology is based on the principle of vibration energy diffraction, using a FFC 1002 crack depth meter to realize depth quantification. Although it has high precision in a controlled laboratory environment (measurement range 0~100mm, expanded uncertainty ≤1mm), it is easily affected by water flow disturbance when applied underwater, resulting in poor probe coupling and significant measurement error. Geophysical auxiliary method combines land sonar method, resistivity CT and other geophysical technologies to indirectly invert the crack distribution characteristics by divers laying sensor arrays. It is mainly used for deep hidden danger detection, but it is difficult to accurately size a single crack due to its low spatial resolution (usually centimeter level).

[0005] Overall, manual measurement has three major bottlenecks: first, poor safety, low underwater visibility, complex water flow and unstable dam structure can easily lead to diver accidents; second, low efficiency, limited scope of a single operation, unable to meet the needs of full coverage detection of large dams; third, limited precision, contact measurement is easily affected by operation method, and it is difficult to access cracks in deep or hidden areas of the dam.

[0006] Underwater robots (ROV type) have become the mainstream equipment for underwater dam detection due to their strong maneuverability and wide range of operation. Current research mainly focuses on rapid identification and segmentation of cracks, but there are still obvious shortcomings in the actual size measurement link. The mainstream technical path includes visual imaging and semantic segmentation, sonar scanning technology, and multi-sensor fusion attempts: visual imaging and semantic segmentation acquire crack images through ROV carrying high-definition cameras or binocular vision systems, and combine deep learning models (such as U-Net) to achieve crack region extraction. Although it can achieve high pixel accuracy and average intersection over union, it can effectively distinguish cracks from background interference, but it can only output two-dimensional pixel contours and cannot be directly converted to physical size; sonar scanning technology uses side-scan sonar or synthetic aperture sonar (SAS) to obtain three-dimensional point cloud data of the dam body, which is suitable for turbid water areas, but it is difficult to accurately depict the crack edge shape due to low resolution (usually 5-10 mm) and susceptibility to dam surface roughness; multi-sensor fusion attempts combine vision and inertial measurement units (IMU) to construct a three-dimensional model of the dam body through SLAM technology, but the model accuracy cannot meet the millimeter-level measurement requirements of cracks due to uneven underwater lighting, missing feature points, and other factors.

[0007] In summary, existing underwater dam crack measurement techniques still have key bottlenecks such as refractive error correction, insufficient depth acquisition accuracy, and scale distortion caused by unstable camera posture in actual size quantification. There is still no high-precision engineering solution that can operate stably in complex underwater environments. SUMMARY

[0008] The present application provides a dam crack measurement analysis method based on double laser modeling and IMU attitude compensation, which can realize accurate measurement of underwater dam crack size, and provides a non-contact, high-precision underwater dam crack measurement method for dam safety warning and precise management.

[0009] To solve the above technical problems, the technical solution proposed by the present application is:

[0010] A dam crack measurement analysis method based on double laser modeling and IMU attitude compensation, comprising the following steps:

[0011] Step S1, in a water medium environment, the intrinsic matrix and distortion coefficient of the underwater camera are obtained by Zhang Zhengyou calibration method, and the image distortion correction is carried out accordingly;

[0012] Step S2, the real-time attitude of the camera in the world coordinate system is obtained by using IMU, the fixed rotation offset matrix between the camera system and the IMU system is calibrated, and the camera extrinsic matrix is constructed;

[0013] Step S3, based on the double laser point device, an inverse proportional calibration model between the laser point pixel distance and the target surface depth is established;

[0014] Step S4, based on the forward projection equation and depth information, realizing depth constraint based back-projection restoration, solving the three-dimensional coordinates of the fracture point; combining image segmentation and geometric calculation, realizing millimeter level measurement of the fracture width.

[0015] As a further improvement of the above technical solution:

[0016] Preferably, in step S1, the intrinsic matrix K of the underwater camera is confirmed by Zhang Zhengyou calibration method:

[0017]

[0018] Wherein, , respectively represent the equivalent focal length of the camera in x and y axes, in units of pixels, and represent the ratio of the camera focal length to the pixel size; and are the coordinates of the principal point in the pixel coordinate system;

[0019] The distortion coefficient D caused by the camera lens is obtained as:

[0020]

[0021] Wherein, , , is the radial distortion coefficient, , is the tangential distortion coefficient.

[0022] The camera is corrected for distortion, and the correction formula is as follows:

[0023]

[0024] Wherein, is the ideal normalized image coordinate without distortion, is the actual distorted image coordinate, represents the decision coefficient, and .

[0025] Preferably, in step S2, the camera and the IMU are fixed to the same physical point, and there is no relative translation between the two, and there is a fixed rotation offset matrix. Using all the collected synchronous pose data, a target function based on Frobenius norm is constructed in the rotation space, and the optimal solution that minimizes the global pose residual is solved by least square optimization algorithm.

[0026] Preferably, in step S2, the extrinsic matrix of the camera in the world coordinate system at time t is :

[0027]

[0028] wherein, is a fixed rotation offset matrix of the camera coordinate system (C) relative to the IMU coordinate system (I) for transforming a spatial vector in the camera coordinate system to the IMU coordinate system; represents a rotation matrix of the IMU in the world coordinate system with the IMU geometric center as the origin at time t; 0 represents a translation vector with the IMU geometric center as the origin of the world coordinate system.

[0029] Preferably, in the step S3, the two laser points are located on the same horizontal line and have a fixed distance, and the relative positions of the laser points and the camera optical center are fixed, which satisfies the similar triangle relationship of underwater imaging.

[0030] Preferably, the inverse proportional calibration model is:

[0031]

[0032] wherein, z represents the depth, represents an inverse proportional coefficient, represents a pixel distance.

[0033] Preferably, in the step S4, the crack point depth is approximately coplanar with the laser point, and the inverse projection restoration based on the depth constraint is realized based on the forward projection equation and the depth information, so as to solve the three-dimensional coordinates of the crack point.

[0034] Preferably, the forward projection equation is:

[0035]

[0036] wherein, is the coordinate of the point on the dam crack in the camera coordinate system, is the coordinate of the point on the dam crack in the pixel coordinate system, is the camera external parameter, and let represents the rotation matrix of the point on the dam crack in the world coordinate system, K is the internal parameter matrix of the camera in the water medium, and s is a scaling factor and is equal to the depth information of the dam crack.

[0037] Preferably, the method further comprises: enhancing and segmenting the collected underwater crack image, extracting the crack region, and calculating the distance between the two end points corresponding to the maximum width in the connected region in the world coordinate system as the crack width.

[0038] The dam crack measurement and analysis method based on double laser modeling and IMU attitude compensation provided by the application has the following advantages compared with the prior art:

[0039] (1) The dam crack measurement and analysis method based on double laser modeling and IMU attitude compensation of the application, the real imaging internal distortion parameters of the underwater camera in the water medium are obtained by Zhang Zhengyou calibration method, and the real-time attitude compensation of the camera external parameter matrix is realized by using IMU measurement, which fundamentally solves the problem of spatial inconsistency caused by projection distortion and attitude change caused by underwater refraction; further, a double laser point pixel distance-depth inverse proportional calibration model is constructed, the direct quantitative calculation of the target surface depth is realized; based on the crack point depth and the approximate coplanar assumption of the laser point, combined with the forward projection equation and the depth information, the reverse projection restoration based on the depth constraint is realized, and the three-dimensional coordinates of the crack points are solved; finally, the three-dimensional space coordinates are solved by combining the forward projection equation, and the millimeter-level high-precision restoration of the crack width and other geometric sizes is realized.

[0040] (2) The dam crack measurement and analysis method based on double laser modeling and IMU attitude compensation of the application constructs a complete underwater crack size quantification link containing refraction correction, attitude compensation, depth information solving and three-dimensional reconstruction, which overcomes the problem that the existing technology cannot reliably calculate the real physical size from two-dimensional pixels, and provides a stable, fast and engineering deployable high-precision measurement method for underwater fine detection of dams and other hydraulic structures. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 It is the flowchart of the measurement method in the application.

[0042] Figure 2 It is the double laser point depth calibration experiment scene diagram in the application.

[0043] Figure 3 It is the underwater real crack and laser point imaging diagram in the application.

[0044] Figure 4 (a) is a single crack diagram of a dam photographed by an underwater ROV type robot, and obtained by an underwater image enhancement algorithm.

[0045] Figure 4 (b) is a multi-crack diagram of a dam photographed by an underwater ROV type robot, and obtained by an underwater image enhancement algorithm.

[0046] Figure 4 (c) is a segmentation effect diagram obtained after the single crack diagram of the dam is segmented by an algorithm.

[0047] Figure 4 (d) is a segmentation effect diagram obtained after the multi-crack diagram of the dam is segmented by an algorithm.

[0048] Figure 5 (a) is the diameter of the largest inscribed circle of the dam crack obtained by the application, that is, the maximum width of the crack.

[0049] Figure 5 (b) is the length of the connecting line of the center line of the dam crack obtained by the present application, that is, the length of the crack.

[0050] Figure 5 (c) is the area of the dam crack obtained by the present application. DETAILED DESCRIPTION

[0051] The specific embodiments of the present application are described in detail below. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present application, and are not used to limit the present application.

[0052] As shown in Figure 1 , the dam crack measurement and analysis method based on double laser modeling and IMU attitude compensation of the present application specifically includes the following steps:

[0053] Step S1, the Zhang Zhengyou calibration method is used to confirm the intrinsic matrix and distortion coefficient of the underwater camera in the water medium, and the distortion correction of the image is realized accordingly.

[0054] In the measurement task of underwater dam surface cracks, due to the inherent distortion coupling effect of the convex lens of the underwater camera lens, the convex lens structure of the camera lens will inherently produce radial distortion (barrel / pillow distortion) and tangential distortion. This distortion characteristic, in the water medium environment, will form a coupling effect with the refraction effect of the water medium, and the distortion law changes compared with the air medium. In addition, due to the limitation of machining tolerance and assembly precision in the camera production and assembly process, there is a deviation between the imaging principal point (projection of the optical center on the imaging plane) and the physical center of the image sensor.

[0055] Therefore, the imaging parameter calibration of the underwater camera is the premise of realizing accurate defect measurement, and the Zhang Zhengyou calibration method is used to confirm the intrinsic matrix and distortion coefficient of the underwater camera in the water medium environment. Zhang Zhengyou calibration method can solve the intrinsic matrix (equivalent focal length, principal point coordinates) and distortion coefficient (radial distortion coefficient, tangential distortion coefficient) of the camera through the matching of multiple sets of water medium environment calibration board image feature points.

[0056] The calibration result is:

[0057] Under the water medium, the intrinsic matrix K of the underwater camera is confirmed by the Zhang Zhengyou calibration method:

[0058] (1)

[0059] wherein, , respectively represent the equivalent focal length of the camera on the x and y axes, in pixels, that is, the ratio of the focal length of the camera to the pixel size; and The coordinates of the principal point in the pixel coordinate system.

[0060] At the same time, the distortion coefficient D generated by the camera lens is obtained as:

[0061] (2)

[0062] wherein, , , is a radial distortion coefficient, , is a tangential distortion coefficient.

[0063] And the camera is corrected for distortion using opencv, and the correction formula is as follows:

[0064] (3)

[0065] wherein, is the ideal normalized image coordinate without distortion, is the actual distorted image coordinate, represents the determination coefficient, and .

[0066] Step S2, determine the extrinsic matrix of the underwater camera using IMU.

[0067] The equipment used is: 9-axis IMU gyroscope + accelerometer + magnetometer (Mag), underwater camera and checkerboard calibration board.

[0068] S2-1, preliminary preparation:

[0069] First, calibrate the IMU for zero offset and magnetism, then fix the camera optical center (lens optical center) and IMU geometric center at the same physical point and weld them, and fix them in front of the underwater robot. After welding, there is no relative motion. The IMU and the camera are completely fixed, and the posture and position of the two are completely consistent during underwater movement.

[0070] S2-2, model establishment:

[0071] The world system (W), IMU system (I) and camera system (C) are defined respectively. The origin of the world system (W) is the initial underwater position of the robot, the axis is the forward direction, the axis is the lateral direction, the axis is perpendicular to the water surface upward; the origin of the IMU system (I) is the IMU center (coinciding with the camera), the axis is the forward direction of the robot, the axis is the lateral direction, the axis is perpendicular to the motion plane; the origin of the camera system (C) is the camera optical center (coinciding with the IMU), is the optical axis direction, The axial imaging plane is downward, The axial imaging plane is rightward, R is a rotation matrix, and T is a translation matrix.

[0072] Since the camera optical center and the IMU geometric center are fixed at the same physical point without relative position offset, the translation vector between the camera system and the IMU system is satisfies: 0, indicating that the origin of the camera system and the origin of the IMU system completely coincide (without spatial position offset) and are at the same reference point at the coordinate system level.

[0073] S2-3, calibration:

[0074] is a fixed rotation offset matrix of the camera coordinate system (C) relative to the IMU coordinate system (I). In engineering practice, due to the angle error of the installation posture, the inconsistency of the IMU coordinate axis convention and the camera coordinate axis, there is an inherent rotation deviation between the two in actual engineering, which is denoted as

[0075] (4)

[0076] In the formula, denotes a three-dimensional special orthogonal group, which is used to represent that the matrix is an orthogonal matrix containing only rotation, without scaling and mirror transformation.

[0077] If the offset is not calibrated, it will cause the crack width measurement error to be amplified through the imaging depth Z, so it needs to be calibrated first. is a fixed rotation offset matrix of the camera coordinate system (C) relative to the IMU coordinate system (I).

[0078] The specific calibration process is as follows:

[0079] S2-3-1, the IMU can provide its own rotation matrix in the world coordinate system , which represents the rotation matrix of the IMU in the world coordinate system at time t. represents the rotation matrix of the IMU in the world coordinate system in the kth group of data.

[0080] S2-3-2, by identifying the pixel coordinates of the chessboard corner points in the image and combining the known physical geometric dimensions of the calibration board, a mapping relationship between three-dimensional space points and two-dimensional pixel points is established, and the PnP algorithm is used to solve the pose of the camera relative to the calibration board coordinate system, and then the rotation matrix of the camera in the world coordinate system in the kth group of data is obtained as .

[0081] S2-3-3, since single pose measurement is susceptible to noise, the application keeps stable conditions for detecting platform at each pose, and performs multiple synchronous acquisition of pose data of the camera and the IMU.

[0082] The collected data is first removed of abnormal values, and then multiple sets of pose matrices at the same pose are averaged to obtain an effective pose matrix at the pose.

[0083] The "pose matrix" includes a rotation matrix of the IMU coordinate system relative to the world coordinate system and a rotation matrix of the camera coordinate system relative to the world coordinate system The kth set of data is .

[0084] S2-3-4, a target function based on Frobenius norm is constructed in the rotation space using all data (single data is the average at a pose) of the camera and the IMU. An optimal solution that minimizes the global pose residual is solved by a least square optimization algorithm. This method can effectively suppress the influence of underwater environmental noise on the calibration result through the joint constraint of multiple sets of observation values. The least square optimization method jointly solves all samples, and the solving formula is as follows:

[0085] (5)

[0086] In the formula, R represents an optimization variable, and at the optimal point N represents the total number of calibration data, indicates the index of the kth set of synchronous data collected in the calibration process, indicates the rotation matrix of the IMU in the world coordinate system in the kth set of data, indicates the rotation matrix of the camera in the world coordinate system in the kth set of data, and F represents the Frobenius norm. The optimal fixed rotation offset matrix

[0087] obtained is the fixed rotation offset matrix of the camera relative to the IMU coordinate system. Only depends on the physical installation relationship between sensors, and has coordinate system independence. Once the calibration is completed, the redefinition or drift of the world coordinate system in the subsequent measurement process will not affect the accuracy of the parameter.

[0088] S2-3-5, in the application, in order to improve the operation efficiency and eliminate the influence of displacement drift, only the rotation part is retained when the camera external parameter matrix is constructed, and the translation of the IMU relative to the world coordinate system is not introduced.

[0089] ​Specifically, the origin of the IMU coordinate system is taken as the reference coordinate system, and the camera coordinate system is mapped to the IMU coordinate system through a fixed rotation offset matrix , thereby establishing a rotation constraint relationship between the camera coordinate system and the IMU coordinate system. Thus, the extrinsic matrix of the camera in the IMU coordinate system at time t is obtained , which is simplified as

[0090] (6)

[0091] wherein, is a fixed value, represents the rotation matrix of the IMU in the IMU coordinate system at time t.

[0092] Step S3, double-laser-point pixel distance-depth inverse proportion calibration.

[0093] S3-1, modeling conditions:

[0094] As shown in Figure 2 , two laser point devices are installed on the underwater camera and kept on the same horizontal line. The laser preferably adopts a medium wavelength of about 520-532 nm, and the two laser points are fixed in physical space with a fixed distance and a fixed relative position between the laser points and the camera optical center, so that the underwater imaging satisfies the similar triangle relationship.

[0095] The underwater camera with the installed laser points is fixed, and the laser points are projected onto a plane perpendicular to the camera optical axis, ensuring that when the underwater camera moves along the direction perpendicular to the laser point projection plane, the obtained depth z is the vertical distance from the camera to the wall.

[0096] The depth z at different positions is measured by a high-precision ranging tool.

[0097] S3-2, modeling process:

[0098] S3-2-1, the image collected by the camera is precisely positioned using OpenCV with circle point detection, color detection combined with ROI (region of interest) to locate the underwater laser points, and the Euclidean distance of the two laser points in the pixel coordinate system is calculated.

[0099] (7)

[0100] wherein, the points , and , are the coordinates of the laser points in the pixel coordinate system, represents the Euclidean distance of the two laser points in the pixel coordinate system, i.e., the pixel distance.

[0101] Since underwater imaging satisfies the similar triangle relationship, i.e. the ratio of the laser point spacing L to the depth z in the physical space is equal to the ratio of the pixel distance to the equivalent intrinsic of the underwater camera , i.e.

[0102] (8)

[0103] Taking the laser point spacing L as the input and the depth z as the output, we have:

[0104] (9)

[0105] It is shown that, under the condition that the laser point spacing L and the equivalent intrinsic are fixed, the pixel distance is strictly inversely proportional to the depth z.

[0106] S3-2-2, fix the underwater camera with laser points installed, and move the underwater camera along the direction perpendicular to the laser point projection plane, record the pixel distance and the depth z at different positions. If a certain group of data (x, y) is obviously deviated from the trend, check whether the image has laser point occlusion, reflection, or the depth measurement is wrong, and remove the abnormal points to complete the data preprocessing. Transform the inverse proportional function into , where a is the inverse proportional coefficient, which avoids many problems of nonlinear fitting. Using linear regression algorithm, we get the unbiased estimate of the parameters, which improves the accuracy, stability and efficiency of parameter estimation. Finally, calculate the determination coefficient r², and get r²≥0.95, which shows that the fitting result is effective within the range of data.

[0107] Thus, the inverse proportional calibration model is obtained, taking the pixel distance as the input and the depth z as the output:

[0108] (10)

[0109] Step S4, inverse projection restoration based on depth constraint and crack geometric parameter calculation.

[0110] S4-1, inverse projection restoration based on depth constraint:

[0111] When the IMU geometric center is taken as the origin of the world coordinate system, the forward projection equation of the camera (projection from 3D world points to 2D pixel points) is:

[0112] (11)

[0113] where, is the coordinate of the point on the dam crack in the camera coordinate system,​​ These are the coordinates of a point on the dam crack in the pixel coordinate system. For camera external parameters, let Let K represent the rotation matrix of a point on the dam crack in the world coordinate system, K be the intrinsic parameter matrix of the camera in the water medium, and s be the scaling factor, which is equal to the depth information of the dam crack.

[0114] (12)

[0115] in, This represents the depth component of a feature point of a dam crack in the camera coordinate system at time t. Its value is equivalent to the projected distance of the feature point along the camera's optical axis.

[0116] like Figure 3 As shown, the laser point emitted from the underwater camera lens is approximately on the same plane as the dam crack. Considering the local continuity in physical space between the dam crack area and its adjacent laser projection area, this invention introduces the local plane assumption, treating this area as a local spatial plane.

[0117] Because the camera may tilt during underwater movement, the actual depths (i.e., projections in the Z-axis direction of the camera coordinate system) of each feature point of the crack are not absolutely equal. Therefore, this invention no longer uses a single depth assignment, but instead uses the depth z obtained from the dual-laser calibration model in step S3 as the reference depth factor for the local plane.

[0118] Deriving the back projection transformation equation: By performing the inverse operation on the forward equation, a back projection model is constructed that maps from the two-dimensional pixel coordinate system to the three-dimensional world coordinate system.

[0119] (13)

[0120] in, equal Let be the coordinates of a point on the dam crack in the camera coordinate system. Since the inverse of the rotation matrix is ​​equal to its transpose, this equation establishes the directional mapping relationship between pixel coordinates and three-dimensional spatial vectors.

[0121] Introducing local plane depth constraints: In the scenario of dam crack detection, based on the physical characteristic that the crack area and the laser projection area are in the same local spatial plane, the length of the back projection ray is determined using the depth z obtained in step S3.

[0122] S4-2, Confirm crack geometry parameters:

[0123] like Figure 5 As shown, Figure 5 In the diagram (a), the diameter of the largest inscribed circle of the dam crack obtained by the present invention is the maximum width of the crack.Figure 5 (b) is the length of the connecting line of the center line of the dam crack obtained by the present application, that is, the length of the crack; Figure 5 (c) is the area of the dam crack obtained by the present application.

[0124] The three-dimensional coordinates of the crack feature points are calculated by substituting the homogeneous pixel coordinates of the feature points into the back projection equation and reconstructing in combination with the depth z, and the calculation formula is:

[0125] (14)

[0126] In the formula, the inverse of the intrinsic matrix is used to normalize the pixel, and then multiplied by the depth z to restore its physical scale in the camera system, and finally the real three-dimensional space coordinates of the feature points in the reference coordinate system are solved through the rotation matrix And , , , , , .

[0127] The laser point emitted by the underwater camera is in the same plane as the dam crack, and the underwater image enhancement algorithm is used in combination with the image segmentation algorithm to segment the dam surface crack, as shown in Figure 4 . Figure 4 (a) is a single crack image of a dam obtained by underwater ROV type robot shooting and underwater image enhancement algorithm; Figure 4 (b) is a multi-crack image of a dam obtained by underwater ROV type robot shooting and underwater image enhancement algorithm; Figure 4 (c) is a segmentation effect diagram obtained by the segmentation algorithm of the single crack image of the dam; Figure 4 (d) is a segmentation effect diagram obtained by the segmentation algorithm of the multi-crack image of the dam.

[0128] Then, the segmented crack image is converted into a binary image, and the largest inscribed circle that can be completely contained in the connected region of the crack is searched for, and the diameter of the largest inscribed circle is taken as the maximum width of the crack, and the two points of the maximum width on the crack in the world coordinate system are obtained by formula (14) And .

[0129] Therefore, the maximum width of the crack is:

[0130] (15).

[0131] Experimental cases

[0132] In the implementation process of the real dam scene, the camera carried by the ROV underwater robot is used to obtain dam surface images, and the deployment of the algorithm is completed by Jetson TX2 NX. According to the above maximum width calculation logic, the system automatically locks the pixel pair of the widest part of the crack, and maps it to the world coordinate system with the IMU geometric center as the coordinate origin for Euclidean distance solution.

[0133] To verify the accuracy of the measurement method in the real underwater environment, the embodiment carries out quantitative comparison experiment on 20 groups of typical crack samples in the dam site environment, and part of the measurement data is shown in Table 1, wherein the actual crack width is obtained by manual measurement of vernier caliper with measurement accuracy of 0.01mm.

[0134] Table 1 Comparison of crack measurement results after underwater camera calibration

[0135]

[0136] The experimental results show that under the complex illumination and dynamic water flow interference of the real dam, the method described in the application can accurately capture the widest feature of the crack. In the test of 20 groups of samples, the measurement value of the method described in the application is highly consistent with the actual manual measurement value, the average relative error is only 3.12%, the maximum relative error is less than 5, and the maximum absolute error is limited within 1mm. It has very high geometric fidelity and measurement reliability, and has engineering application feasibility.

[0137] The above implementation cases are only preferred embodiments of the application, and do not limit the application in any form. Although the application has been disclosed as above with preferred embodiments, it is not intended to limit the application. Therefore, any simple modification, equivalent change and modification of the above embodiments without departing from the technical solution of the application, according to the technical essence of the application, should fall within the protection scope of the technical solution of the application.

Claims

1. A method for measuring and analyzing dam cracks based on dual-laser modeling and IMU attitude compensation, characterized in that, Includes the following steps: Step S1: In an aqueous environment, the intrinsic parameter matrix and distortion coefficients of the underwater camera are obtained using the Zhang Zhengyou calibration method, and the distortion correction of the image is performed accordingly. Step S2: Use the IMU to obtain the real-time attitude of the camera in the world coordinate system, calibrate the fixed rotation offset matrix between the camera system and the IMU system, and construct the camera extrinsic parameter matrix; Step S3: Based on the dual laser point device, establish an inverse proportional calibration model between the distance between laser point pixels and the depth of the target surface; Step S4: Based on the forward projection equation and depth information, back projection reconstruction based on depth constraints is realized to solve the three-dimensional coordinates of the crack points; combined with image segmentation and geometric calculation, millimeter-level measurement of crack width is achieved. In step S4, the crack point depth and the laser point are approximately coplanar. Based on the forward projection equation and depth information, the back projection reconstruction based on depth constraints is realized to solve the three-dimensional coordinates of the crack point. The forward projection equation is: ； in, Let these be the coordinates of a point on the dam crack in the camera coordinate system. These are the coordinates of a point on the dam crack in the pixel coordinate system. For camera external parameters, let The rotation matrix of the point on the dam crack in the world coordinate system is represented by K, the intrinsic parameter matrix of the camera in the water medium is K, and s is the scaling factor, which is equal to the depth information of the dam crack.

2. The dam crack measurement and analysis method based on dual-laser modeling and IMU attitude compensation according to claim 1, characterized in that, In step S1, the intrinsic parameter matrix K of the underwater camera is confirmed using the Zhang Zhengyou calibration method: ； in, , These represent the equivalent focal length of the camera on the x and y axes, respectively, expressed in pixels, representing the ratio of the camera's focal length to the pixel size. and The coordinates of the principal point in the pixel coordinate system; The distortion coefficient D caused by the camera lens is obtained as follows: ； in, , , The radial distortion coefficient is... , The tangential distortion coefficient; The camera distortion is corrected using the following formula: ； in, For ideal, distortion-free, normalized image coordinates, These are the coordinates of the distorted image actually captured. Denotes the coefficient of determination, and .

3. The dam crack measurement and analysis method based on dual-laser modeling and IMU attitude compensation according to claim 2, characterized in that, In step S2, the camera and IMU are fixed at the same physical point, with no relative translation between them and a fixed rotation offset matrix. Using all the acquired synchronous attitude data, an objective function based on the Frobenius norm is constructed in the rotation space. The optimal solution that minimizes the global attitude residual is solved by the least squares optimization algorithm.

4. The dam crack measurement and analysis method based on dual-laser modeling and IMU attitude compensation according to claim 3, characterized in that, In step S2, the extrinsic parameter matrix of the camera in the world coordinate system at time t. for: ； in, This is a fixed rotation offset matrix for the camera coordinate system relative to the IMU coordinate system, used to transform spatial vectors in the camera coordinate system to the IMU coordinate system; represents the rotation matrix of the IMU in the world coordinate system at time t, with the IMU geometric center as the origin; 0 indicates that the translation vector is zero when the IMU geometric center is the origin of the world coordinate system.

5. The dam crack measurement and analysis method based on dual-laser modeling and IMU attitude compensation according to claim 1, characterized in that, In step S3, in the underwater camera coordinate system, the two laser points are located on the same horizontal line with a fixed spacing, and the relative positions of the laser points and the camera optical center are fixed, satisfying the similar triangle relationship for underwater imaging.

6. The dam crack measurement and analysis method based on dual-laser modeling and IMU attitude compensation according to claim 5, characterized in that, In step S3, the inverse proportional calibration model is: ； Where z represents depth. Indicates the inverse proportionality coefficient. Indicates pixel distance.

7. The dam crack measurement and analysis method based on dual-laser modeling and IMU attitude compensation according to claim 1, characterized in that, The method further includes: enhancing and segmenting the acquired underwater crack image, extracting the crack region, and calculating the distance between the two endpoints corresponding to the maximum width in the connected region in the world coordinate system as the crack width.

Images