A dam surface settlement measurement method based on natural markers

Abstract

The present application belongs to the technical field of dam safety monitoring, and provides a dam surface settlement measurement method based on natural markers, comprising: reference image and detection image acquisition, radial distortion correction and tangential distortion correction, image correction, template matching, feature point extraction and matching, original marker displacement calculation, camera settlement data acquisition and displacement correction. The present application solves the relative displacement of the feature point pair of the marker based on the natural texture features or natural features of the existing objects on the dam surface as the measurement markers, realizes the contactless dam surface settlement statistics, and reduces the dependence on the field environment; through the proposed image correction formula, the accurate registration between the reference image and the detection image is realized, and the monitoring accuracy is significantly improved; through the acquisition of camera settlement data and displacement correction, the measurement accuracy is further improved.

Description

Technical Field

[0001] This invention relates to the field of dam safety monitoring technology, and in particular to a method for measuring dam surface settlement based on natural markers. Background Technology

[0002] Global Navigation Satellite System (GNSS) is a crucial infrastructure component of water conservancy projects. The long-term safe operation of dams directly impacts the safety of life and property in downstream inundated areas, as well as social stability. Therefore, monitoring and assessing the health status of dams is paramount, and surface deformation monitoring is a core indicator reflecting dam safety. Deformation monitoring can promptly identify potential hazards in dam operation, providing a scientific basis for disaster prevention. Based on the measurement methods used, deformation monitoring is currently mainly divided into two categories: contact and non-contact. Contact monitoring primarily includes traditional measurement methods such as hydrostatic leveling and tension wires. Its advantages lie in high measurement accuracy, intuitive and reliable data, and suitability for long-term monitoring of deformation in critical areas. However, contact methods have significant limitations, such as high instrument installation costs, difficult maintenance, and difficulty in covering large monitoring areas. In recent years, non-contact monitoring technologies have gradually become a research hotspot due to their flexibility and efficiency. These include machine vision, GNSS, total stations, and laser displacement sensors. Among these, image processing technology based on machine vision is increasingly widely used in dam construction. These methods can be further divided into monitoring based on manually placed targets and monitoring based on natural markers. The artificial target placement technology achieves monitoring by placing reflective marks or other artificial markers on the dam surface, combined with photogrammetry or laser scanning technology.

[0003] Although the above methods can achieve high-precision measurements in specific environments, they require the deployment of targets on-site, which is costly. The number of measurement points is limited, and changes in camera position can lead to excessive errors, thus failing to fully leverage the technical advantages of wide-area measurement using machine vision. Summary of the Invention

[0004] In order to overcome the shortcomings of the existing technology, the purpose of this invention is to provide a method for measuring dam surface settlement based on natural markers, which solves the problems of high cost of setting up targets on site, limited measurement points, inability to fully utilize the technical advantages of wide-area measurement of machine vision, and significant increase in error caused by camera attitude deflection.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A method for measuring dam surface settlement based on natural markers includes:

[0007] Use image acquisition equipment to periodically acquire baseline and detection images of the target area;

[0008] The detection image is subjected to image capture distortion elimination using radial distortion correction formulas and tangential distortion correction formulas;

[0009] The image correction formula of the simplified projection model is used to correct the calibrated detection image.

[0010] The modified detection map and the baseline map are template matched using a normalized correlation coefficient matching method to obtain matching region data.

[0011] Based on the matching region data, feature points are extracted for each template region using the SURF algorithm, and the extracted feature points are matched using the nearest neighbor search algorithm to obtain matching pairs. The matching pairs are then filtered and optimized using the LO-RANSAC algorithm to obtain feature point pairs.

[0012] Calculate the mean of the pixel differences of all the feature point pairs to obtain the pixel displacement, and calculate the physical distance corresponding to the pixel displacement according to the preset ratio to obtain the original marker displacement;

[0013] The settlement of the image acquisition device relative to a preset stable area is collected to obtain camera settlement data;

[0014] The original marker displacement is corrected based on the camera settlement data to obtain the true marker displacement.

[0015] Preferably, the radial distortion correction formula includes:

[0016] x d1 =x(1+k1r) 2 +k2r 4 +k3r 6 ),

[0017] y d1 =y(1+k1r 2 +k2r 4 +k3r 6 );

[0018] in, x d1 y d1 x and y are the x and y coordinates of the pixel after radial distortion correction, respectively; k1, k2, and k3 are the first, second, and third radial distortion coefficients, respectively; and r is the distance from the pixel to the calibration center.

[0019] The tangential distortion correction formula includes:

[0020] x d2 =x+[2p1xy+p2(r)2 +2x 2 )]、

[0021] y d2 =y+[p1(r 2 +2y 2 )+2p2xy];

[0022] Where, x d2 y d2 p1 and p2 are the pixel x-coordinate and y-coordinate after tangential distortion correction, respectively; p1 and p2 are the first and second tangential distortion coefficients, respectively.

[0023] Preferably, the image correction formula includes:

[0024]

[0025] Where u and ν are the U-axis and V-axis coordinates of the original pixel coordinates, respectively; R 11 is the element corresponding to the first row and first column of the rotation matrix; u′ and ν′ are the U-axis and V-axis coordinates of the corrected pixel coordinates, respectively; f is the camera focal length.

[0026] Preferably, the optimization formula of the LO-RANSAC algorithm is:

[0027]

[0028] Where θ′ represents the model parameters after local optimization; y i f(x) represents the true output value of the i-th observation. i ;θ) represents the predicted value of the i-th observation calculated based on the current model parameters θ; x i represents the input value of the i-th observation; S is the set of interior points of the current model; argmin represents the sum of squared residuals.

[0029] Preferably, the displacement correction is expressed as:

[0030]

[0031] Where X is the displacement of the actual marker; ΔX is the displacement of the original marker; ΔXp is the camera settlement data; Lp is Lp; and L is the distance from the camera optical center to the target area.

[0032] The present invention discloses the following technical effects:

[0033] This invention provides a method for measuring dam surface settlement based on natural markers. By using the natural texture features of existing objects on the dam surface or natural features as measurement markers to solve the relative displacement of marker feature point pairs, it solves the problems of high cost of on-site target deployment, limited measurement points, and inability to fully utilize the technical advantages of wide-area measurement by machine vision. It achieves non-contact dam surface settlement statistics. Through the proposed image correction formula, it solves the problem of significantly increased error caused by camera attitude deflection, and achieves accurate registration between the reference image and the detection image. By collecting camera settlement data and displacement correction, it solves the problem that traditional methods do not detect the settlement of the equipment itself, and further improves the measurement accuracy. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a schematic diagram of the dam surface settlement measurement process based on natural markers provided in an embodiment of the present invention;

[0036] Figure 2 This is an overall flowchart of the settlement displacement algorithm provided in the embodiments of the present invention;

[0037] Figure 3 This is a schematic diagram of coordinate changes provided in an embodiment of the present invention;

[0038] Figure 4 This is a schematic diagram of displacement correction provided in an embodiment of the present invention;

[0039] Figure 5 A schematic diagram of the experimental apparatus provided in an embodiment of the present invention;

[0040] Figure 6 This is a schematic diagram comparing the correction effects provided in the embodiments of the present invention;

[0041] Figure 7 This is a schematic diagram of settlement displacement results provided in an embodiment of the present invention;

[0042] Figure 8 This is a schematic diagram of the settlement displacement difference provided in an embodiment of the present invention;

[0043] Figure 9 Error statistics chart provided for embodiments of the present invention. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] The purpose of this invention is to provide a method for measuring dam surface settlement based on natural markers, which solves the problems of high cost of setting up targets on site, limited measurement points, inability to fully utilize the technical advantages of wide-area measurement by machine vision, and significant increase in error due to camera attitude deflection.

[0046] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0047] Figure 1 This is a schematic diagram of the dam surface settlement measurement process based on natural markers provided in an embodiment of the present invention, as shown below. Figure 1 As shown, this invention provides a method for measuring dam surface settlement based on natural markers, including:

[0048] Step 100: Use image acquisition equipment to periodically acquire baseline and detection images of the target area;

[0049] Step 200: Use the radial distortion correction formula and the tangential distortion correction formula to perform image capture distortion elimination on the detection image;

[0050] Step 300: Perform image correction on the calibrated detection image using the image correction formula of the simplified projection model;

[0051] Step 400: Use the normalized correlation coefficient matching method to perform template matching on the corrected detection map and the baseline map to obtain the matching region data;

[0052] Step 500: Based on the matching region data, use the SURF algorithm to extract feature points for each template region, use the nearest neighbor search algorithm to match the extracted feature points to obtain matching pairs, and use the LO-RANSAC algorithm to filter and optimize the matching pairs to obtain feature point pairs.

[0053] Step 600: Calculate the mean of the pixel differences of all feature point pairs to obtain the pixel displacement, and calculate the physical distance corresponding to the pixel displacement according to the preset ratio to obtain the original marker displacement;

[0054] Step 700: Acquire the settlement of the image acquisition device relative to the preset stable area to obtain camera settlement data;

[0055] Step 800: Correct the displacement of the original marker based on the camera settlement data to obtain the true displacement of the marker.

[0056] Furthermore, the radial distortion correction formula includes:

[0057] x d1 =x(1+k1r) 2 +k2r 4 +k3r 6 ),

[0058] y d1 =y(1+k1r 2 +k2r 4 +k3r 6 );

[0059] in, x d1 y d1 x and y are the x and y coordinates of the pixel after radial distortion correction, respectively; k1, k2, and k3 are the first, second, and third radial distortion coefficients, respectively; and r is the distance from the pixel to the calibration center.

[0060] The tangential distortion correction formula includes:

[0061] x d2 =x+[2p1xy+p2(r) 2 +2x 2 )]、

[0062] y d2 =y+[p1(r 2 +2y 2 )+2p2xy];

[0063] Where, x d2 y d2 p1 and p2 are the pixel x-coordinate and y-coordinate after tangential distortion correction, respectively; p1 and p2 are the first and second tangential distortion coefficients, respectively.

[0064] Specifically, the image correction formula includes:

[0065]

[0066] Where u and ν are the U-axis and V-axis coordinates of the original pixel coordinates, respectively; R 11 is the element corresponding to the first row and first column of the rotation matrix; u′ and ν′ are the U-axis and V-axis coordinates of the corrected pixel coordinates, respectively; f is the camera focal length.

[0067] Furthermore, the optimization formula for the LO-RANSAC algorithm is as follows:

[0068]

[0069] Where θ′ represents the model parameters after local optimization; y i f(x) represents the true output value of the i-th observation. i ;θ) represents the predicted value of the i-th observation calculated based on the current model parameters θ; x i represents the input value of the i-th observation; S is the set of interior points of the current model; argmin represents the sum of squared residuals.

[0070] Specifically, the displacement correction is expressed as:

[0071]

[0072] Where X is the actual displacement of the marker; ΔX is the original displacement of the marker; ΔXp is the camera settlement data; Lp is Lp; and L is the distance from the camera optical center to the target area.

[0073] Preferably, image correction utilizes a reference region to correct the detection image. When detecting targets using natural markers, traditional projection models require the reference region to be at the same depth as the target region, i.e., within the same plane, to ensure model effectiveness. However, the simplified projection model developed in this embodiment can ignore depth differences, mapping points in the detection image to the coordinate system of the reference image, thereby eliminating geometric distortions caused by shooting angle, position, and other external factors. The normalized correlation coefficient matching method matches the target positions in the reference and detection images, thus locking the regions of interest in both images and ignoring interference from complex environments. SURF is a feature detection algorithm that accelerates feature point detection through Hessian matrix approximation and integral images. It is characterized by high computation speed and robustness to various image transformations. It is suitable for some complex environments in dam monitoring and is effectively applicable to natural marker environments where targets are difficult to install. LO-RANSAC is an improved RANSAC algorithm that improves the accuracy of mismatch point removal through local optimization steps. Compared to the traditional RANSAC algorithm, LO-RANSAC optimizes locally after each round of random sampling to maximize the number of matching inliers, thereby finding the optimal geometric transformation model. The relative pixel displacement of feature point pairs is converted into actual distance using monocular ranging principles.

[0074] Specifically, Figure 2 The following is a flowchart of the settlement displacement algorithm:

[0075] Image acquisition: Use image acquisition equipment to acquire baseline and detection images. The baseline image serves as a standard that no settlement has occurred in the target. The detection images are taken periodically and compared with the baseline image for calculation.

[0076] Furthermore, camera calibration: To reduce the adverse effects of image distortion on calculations, calibration is performed using radial and tangential distortion correction formulas to eliminate radial and tangential distortion generated during image capture. Radial distortion is caused by the lens shape, and the correction formula is:

[0077] x d1 =x(1+k1r) 2 +k2r 4 +k3r 6 )

[0078] y d1 =y(1+k1r 2 +k2r 4 +k3r 6 )

[0079] The tangential distortion correction formula includes:

[0080] x d2 =x+[2p1xy+p2(r) 2 +2x 2 )]

[0081] y d2 =y+[p1(r 2 +2y 2 )+2p2xy]

[0082] refer to Figure 3 Image Correction: To eliminate geometric distortions caused by shooting angle, position, and other external factors, this embodiment proposes an image correction technique based on a simplified projection model. A stable reference point O is defined in space, with its initial position in the camera coordinate system being (x1, y1, z1). After camera movement, its position becomes (x2, y2, z2), and the corresponding pixel coordinates on the image plane are (u, v). After movement, the image coordinates are (u', v'). The rotation matrix relative to the point before movement is R, which is a 3×3 matrix. The camera translation vectors are t1, t2, and t3, and the camera focal length is f. According to the imaging principle, the mapping relationship between the coordinates of a point and the image can be expressed by the following formula:

[0083]

[0084]

[0085] Combining the above equations, we can obtain the relationship between the camera before and after the movement:

[0086]

[0087] Replacing x² and y² with z² respectively, we get the formula:

[0088]

[0089] As shown in the above formula, solving for the coordinate relationship before and after the change requires not only obtaining the camera's rotation matrix and focal length, but also knowing the depth information of the reference point. That is, the target area needs to be on the same plane as the stable area, which greatly limits its application. Considering that in practical engineering, the distance from the monitoring area to the camera's projection center is usually much greater than the camera translation caused by human or natural factors, t can be approximated as... m If / z2(m=1,2,3) is 0, then the previous equation can be simplified as follows:

[0090]

[0091] At this point, the changes in image coordinates are only related to the camera's rotation and focal length. Therefore, during the use of the model, a reference point with a different depth from the target point can be selected for image correction.

[0092] refer to Figure 3 To further improve measurement accuracy, the settlement displacement of the target point is corrected by calculating the settlement displacement of the camera relative to the stable area. This minimizes the impact of camera attitude changes. If the measured settlement of the target area relative to the camera is ΔX, and the settlement of the camera relative to the stable area is ΔXp, then the corrected displacement X of the target area relative to the stable area can be calculated using the following formula:

[0093]

[0094] Furthermore, template matching: In order to eliminate interference information in the detection image, the normalized cross-correlation (NCC) method is used to achieve accurate template matching, thereby locking the feature point extraction and matching region.

[0095] Preferably, feature point matching: After the template matching process, the template regions of the natural identifiers have been matched one-to-one. The SURF algorithm is used to extract feature points from each template region, improving the adaptability to changes in viewing angle and different lighting conditions, and enhancing the robustness of the algorithm. This embodiment uses the nearest neighbor search algorithm for initial feature point matching, filters reliable matching pairs through distance ratio testing, and integrates the LO-RANSAC algorithm for further filtering and optimization of matching point pairs. LO-RANSAC is an improved RANSAC algorithm that improves the accuracy of removing mismatched points through local optimization steps. Compared with the traditional RANSAC algorithm, LO-RANSAC maximizes the number of matching intrapoints through local adjustment optimization after each round of random sampling, thereby finding the optimal geometric transformation model. The specific formula of its optimization function is as follows:

[0096]

[0097] Further, displacement calculation: After the feature points of the natural marker are successfully paired and filtered, the average of the pixel differences of all feature point pairs is taken as the pixel displacement of the target area. By obtaining the ratio between pixel distance and physical distance, the true displacement of the marker is calculated.

[0098] Specifically, the implementation example is verified. The test site for this example was a stadium in a certain city. The image acquisition device was a Canon D3100 SLR camera with 14.2 megapixels and a 55mm focal length. The shooting angle was always kept horizontal. Red and white striped traffic barriers were used to replace the crash barriers on the highway in the actual project. A lifting slide combined with an electronic dial indicator was used to measure the actual settlement of the traffic barriers. The test setup is as follows: Figure 5 As shown.

[0099] Furthermore, an image correction comparison experiment was conducted. A barrier was placed at the test site, and the wall of the building behind the stadium was used to simulate the bedrock in the actual project for image correction. The shooting distance was 70m. The specific experimental design is shown in Table 1.

[0100] Table 1

[0101]

[0102] Specifically, the test results are as follows: Figure 6 As shown, the results indicate that the results without image correction deviate significantly from the actual displacement, while the experimental results with the corrected image are closer to the actual displacement, verifying the reliability and effectiveness of image correction using the distant bedrock as a fixed point.

[0103] Furthermore, accuracy verification experiments were conducted. To verify the accuracy and stability of the algorithm under different settlement variations, an experimental scheme was designed as shown in Table 1. The experimental results are as follows: Figure 7As shown, the results indicate that the measured displacement and the actual change trend are consistent, and the mean absolute errors of the three sets of tests are 0.64 mm, 0.47 mm and 0.42 mm, respectively.

[0104] Specifically, a measurement range verification experiment was conducted. The experimental plan for verifying the applicability of the algorithm by adjusting different shooting distances is shown in Table 2. Shooting began at a distance of 35m, and the experimental results are as follows: Figure 8 As shown. By Figure 8 It can be seen that when the test distance reaches 70m, the maximum error is 0.88mm, and when the test distance reaches 80m, the maximum error is 1.38mm, and the error increases significantly. Therefore, the measurement error of this embodiment within the range of 70m is ≤±1.00mm, the fluctuation range of the measured value is small, and the average absolute error of the three groups is 0.55mm, 0.75mm, and 0.49mm, respectively.

[0105] Table 2

[0106]

[0107] Furthermore, robustness tests were conducted under different environmental conditions. The robustness of the algorithm was tested under different lighting conditions. Tests were performed during normal daylight hours, with the baseline image not exposed to direct sunlight, while the detection images were exposed to direct sunlight at different times and angles. The test plan is shown in Table 2, and the data are as follows: Figure 9 As shown in the figure, the maximum error was 0.97 mm, and the minimum was 0.14 mm. The overall difference trend fluctuated around 0 mm, and there was no significant deviation due to changes in illumination. Although all four sets of test results showed varying degrees of error, the measurement results still demonstrated high accuracy and stability, meeting the needs of engineering monitoring.

[0108] The beneficial effects of this invention are as follows:

[0109] This invention achieves non-contact settlement statistics of the dam surface by using the natural texture features or natural features of existing objects on the dam surface as measurement markers to solve for the relative displacement of marker feature point pairs, thus reducing dependence on the on-site environment. Through the proposed image correction formula, it achieves accurate registration between the reference image and the detection image, significantly improving monitoring accuracy. Furthermore, by collecting camera settlement data and performing displacement correction, it further improves measurement accuracy.

[0110] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0111] This embodiment uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for measuring dam surface settlement based on natural markers, characterized in that, include: Use image acquisition equipment to periodically acquire baseline and detection images of the target area; The detection image is subjected to image capture distortion elimination using radial distortion correction formulas and tangential distortion correction formulas; The image correction formula of the simplified projection model is used to correct the calibrated detection image. The modified detection map and the baseline map are template matched using a normalized correlation coefficient matching method to obtain matching region data. Based on the matching region data, feature points are extracted for each template region using the SURF algorithm, and the extracted feature points are matched using the nearest neighbor search algorithm to obtain matching pairs. The matching pairs are then filtered and optimized using the LO-RANSAC algorithm to obtain feature point pairs. Calculate the mean of the pixel differences of all the feature point pairs to obtain the pixel displacement, and calculate the physical distance corresponding to the pixel displacement according to the preset ratio to obtain the original marker displacement; The settlement of the image acquisition device relative to a preset stable area is collected to obtain camera settlement data; The original marker displacement is corrected based on the camera settlement data to obtain the true marker displacement. The image correction formula includes: ； in, , These are the U-axis and V-axis coordinates of the original pixel coordinates, respectively; This refers to the element in the i-th row and j-th column of the rotation matrix; , These are the U-axis and V-axis coordinates for correcting pixel coordinates, respectively. The focal length of the camera; The optimization formula for the LO-RANSAC algorithm is as follows: ； in, These are the model parameters after local optimization; For the first The actual output value of each observation; Indicates based on the current model parameters Calculated Predicted values ​​for each observation; Indicates the first Input values ​​for each observation; This is the set of interior points of the current model; Represents the sum of squared residuals; The displacement correction is expressed as follows: ； in, The displacement of the actual marker; The displacement of the original marker; The camera's settlement data; for ; This is the distance from the camera's optical center to the target area.

2. The method for measuring dam surface settlement based on natural markers according to claim 1, characterized in that, The radial distortion correction formula includes: 、 ； in, ; , These are the pixel's x-coordinate and y-coordinate after radial distortion correction, respectively; , These are the x-coordinate and y-coordinate of a pixel without distortion correction; , , These are the first radial distortion coefficient, the second radial distortion coefficient, and the third radial distortion coefficient, respectively. The distance from the pixel to the calibration center; The tangential distortion correction formula includes: 、 ； in, , These are the pixel's x-coordinate and y-coordinate after tangential distortion correction, respectively; , These are the first tangential distortion coefficient and the second tangential distortion coefficient, respectively.

Images