Cable-stayed cable vibration identification method and system based on unmanned aerial vehicle video and image rectification

Abstract

The application discloses a kind of based on unmanned aerial vehicle video and image correction's cable-stayed cable vibration identification method and system, method includes: S100, through unmanned aerial vehicle shooting cable-stayed cable under natural excitation dynamic response video;S200, cable-stayed cable region in video image is carried out semantic segmentation, obtains clear cable pixel region;S300, to cable-stayed cable target area and surrounding fixed reference point area are carried out pixel displacement tracking, extract time series displacement data including cable-stayed cable vibration response and unmanned aerial vehicle disturbance component;S400, realize image sequence geometry correction, eliminate the overall jitter caused by unmanned aerial vehicle attitude change;S500, extract inherent mode component;By comparing the modal characteristic identification common mode disturbance component, and from cable-stayed cable response stripping;S600, identification and early warning to slack, broken strand, excitation abnormality and other abnormal vibration behavior.The technical closed loop from video acquisition to abnormal early warning is realized, and the high-precision, automatic monitoring of cable-stayed cable vibration state is realized.

Description

Technical Field

[0001] This invention belongs to the field of bridge monitoring technology, and more specifically, relates to a method and system for identifying cable-stayed bridge vibration based on UAV video and image correction. Background Technology

[0002] As the core load-bearing component of a cable-stayed bridge, the structural health of the stay cables directly affects the overall safety and durability of the bridge. In the natural environment, stay cables are subjected to complex excitations such as vehicle loads, wind loads, and temperature changes over long periods, making them prone to various vibration phenomena. Excessive vibration can lead to serious problems such as fatigue damage, cable tension decay, and even strand breakage. If these issues are not detected and addressed in a timely manner, they may cause bridge structural failure and threaten traffic safety. Therefore, accurate identification and monitoring of the vibration state of stay cables is a crucial link in ensuring the safe operation of cable-stayed bridges throughout their entire lifecycle.

[0003] Existing cable-stayed bridge vibration identification technologies are mainly divided into two categories: contact and non-contact methods. Contact methods typically involve installing devices such as accelerometers and strain gauges on the cable surface to directly collect vibration data for analysis. Non-contact methods primarily rely on visual monitoring, with early approaches often using fixed cameras to capture images of the cable and combining this with digital image processing technology to extract vibration characteristics. Some studies have also introduced devices such as laser vibrometers to assist in data acquisition.

[0004] Existing technologies have significant limitations: contact methods require physical modifications to the cables, resulting in high installation and maintenance costs, and sensors exposed to the outdoors for extended periods are susceptible to environmental interference, leading to data drift; in traditional non-contact solutions, fixed cameras have limited monitoring range, making it difficult to cover all the cables of long-span bridges, and are easily affected by factors such as changes in lighting and obstruction; at the same time, most existing visual recognition technologies do not consider the impact of camera shake, and when cameras are displaced due to wind vibration or unstable installation, the vibration data will be mixed with a large amount of noise, severely reducing recognition accuracy. Summary of the Invention

[0005] To address the aforementioned shortcomings or improvement needs of existing technologies, this invention provides a method and system for identifying cable-stayed bridge vibration based on UAV video and image correction. This system integrates UAV non-contact monitoring with multi-dimensional image processing technology, overcoming the installation limitations and coverage shortcomings of traditional contact monitoring. Through a closed-loop technology process from video acquisition to anomaly warning, it achieves high-precision and automated monitoring of cable-stayed bridge vibration status, providing an efficient and low-cost new technical approach for bridge structural health assessment, combining flexibility and practicality.

[0006] To achieve the above objectives, according to a first aspect of the present invention, a method for identifying cable-stayed bridge vibration based on UAV video and image correction is provided, comprising the following steps:

[0007] S100: Control the drone to fly around the target cable-stayed bridge from multiple angles via a preset flight path, and capture video of the dynamic response of the cable-stayed cable under natural excitation, ensuring that the footage includes the target cable-stayed cable and the surrounding fixed environment.

[0008] S200. Perform semantic segmentation on the cable-stayed bridge region in the video image to distinguish the cable-stayed bridge from unstructured objects in a complex background and obtain clear cable-stayed bridge pixel regions.

[0009] S300 uses a digital image correlation algorithm to perform frame-by-frame pixel displacement tracking of the cable target area and the surrounding fixed reference point area, and extracts time-series displacement data containing cable vibration response and UAV disturbance components.

[0010] S400: Construct a stable image reference coordinate system based on reference points, and realize image sequence geometric correction by calculating the homography matrix between images to eliminate the overall jitter caused by UAV attitude changes;

[0011] S500: Perform variational mode decomposition on the displacement time series of the cable region and the reference point region respectively to extract the intrinsic mode components; identify the common mode disturbance components by comparing the modal characteristics of the two and remove them from the cable response to obtain the pure absolute vibration response signal of the cable.

[0012] The S600 performs time-frequency joint analysis on pure vibration signals to identify atypical vibration characteristics. Combined with the normal state vibration mode library, it can identify and warn of abnormal vibration behaviors such as relaxation, strand breakage, and abnormal excitation.

[0013] Furthermore, in step S300, when performing the pixel position tracking, it is necessary to define the video's first... Frame image is , No. Frame image is ;

[0014] The rectangular sub-region where the cable stays are located is , containing pixel set ;

[0015] Select the sub-region in the image corresponding to the fixed structure of the bridge. And assume that it is only affected by drone disturbances in physical space;

[0016] For any pixel within the region , define the first The grayscale value of the frame image , No. The gray value at the corresponding position in the frame image ,in for or .

[0017] Furthermore, pixels from Move to The displacement is ,Right now , ;

[0018] The gray-level similarity between two frames of images is measured using a normalized cross-correlation coefficient, specifically:

[0019] ,

[0020] in, , , respectively, regions The average gray level in the two frames, This represents the total number of pixels within the region.

[0021] Displacement The optimal solution is the variable value that maximizes NCC:

[0022] ,

[0023] The overall motion of the sub-region is described by polynomial fitting, specifically as follows:

[0024] ,

[0025] in, , , , , , All are coefficients, solved by minimizing the grayscale error function, specifically:

[0026] Repeat the above calculation for all consecutive frames in the video to obtain:

[0027] Displacement sequence of cable region: The displacement sequence of the cable region includes vibration response and UAV disturbance;

[0028] And the displacement sequence of the reference point region: The displacement sequence of this reference point region only includes UAV disturbances.

[0029] Furthermore, the grayscale error function is:

[0030] ,

[0031] because For parameters Nonlinear functions need to be linearized using Taylor expansion. In initial parameters The first-order expansion is as follows:

[0032] ,

[0033] in, These are the coordinates after displacement corresponding to the initial parameters.

[0034] , which represents the coordinate changes caused by parameter correction.

[0035] Substituting the subregion displacements into the polynomial yields:

[0036] ,

[0037] in, All of these are parameter correction values.

[0038] Furthermore, in step S400, the specific method is as follows: from the sub-region corresponding to the bridge fixed structure... Pixels with significant features are selected as stable reference points, denoted as . And regarding the video, Frame and the Frame, of which the first The first frame is used as the initial frame and as the reference frame. A feature matching algorithm is used to find the corresponding position of the reference point between the two frames. Specifically:

[0039] Reference point in reference frame: ,

[0040] No. Corresponding points in the frame: .

[0041] Furthermore, the image homography matrix is ​​a 3×3 invertible matrix. satisfy:

[0042] ,

[0043] in, As a scale factor, ,

[0044] The homography matrix is ​​in the form of:

[0045] ,

[0046] Expanding the above equation and eliminating the scale factor This yields a system of linear equations:

[0047]

[0048] in, Solve the overdetermined system of equations using the least squares method:

[0049] ,

[0050] in, Let be the parameter vector to be determined.

[0051] For corresponding points The coefficient matrix,

[0052] For corresponding points The constant term vector.

[0053] Furthermore, regarding the first Any pixel in a frame image Corrected coordinates are mapped to the reference frame coordinate system via a homography matrix. :

[0054] ,

[0055] in, For the first The inverse transformation matrix from frame to reference frame is used to pull the jittery image back to the reference coordinate system;

[0056] Due to the corrected coordinates The values ​​may be non-integers. The pixel values ​​of the corrected image are calculated using bilinear or bicubic interpolation algorithms, specifically:

[0057] ,

[0058] in, for The surrounding integer coordinate pixels,

[0059] Weights are based on differences;

[0060] Finally, the pixel values ​​of the corrected image Obtain the corrected image sequence The fixed reference point remains stable in the sequence, and the overall jitter caused by changes in the drone's attitude is eliminated.

[0061] Furthermore, in step S500, the specific method is as follows:

[0062] Extract the time series data corrected in step S400 to obtain:

[0063] Displacement sequence of cable region ,in For time frame indexing;

[0064] and reference point region displacement sequence ;

[0065] The two sequences above are averaged to eliminate the DC component, specifically as follows:

[0066] ,

[0067] Then, variational mode decomposition is used to extract intrinsic mode components. Through iterative optimization, the signal is decomposed into multiple intrinsic mode components with specific frequency bandwidths. Each intrinsic mode component satisfies the sparsity and finite bandwidth characteristics.

[0068] For signals The purpose of variational mode decomposition is to solve the following variational problems:

[0069] ,

[0070] in, for or ,

[0071] The preset number of intrinsic modal components,

[0072] For the first The center frequency of each intrinsic modal component

[0073] It's a countdown to the end of the day. For convolution operations, For the Dirac function,

[0074] The objective function constrains the sum of the bandwidths of all intrinsic modal components to be minimized, and the superposition of all intrinsic modal components equals the original signal.

[0075] Furthermore, by solving the variational problem using the Lagrange multiplier method, we obtain:

[0076] Pull cable signal decomposition: ,in For Lasso's One intrinsic modal component;

[0077] Reference point signal decomposition: ,in The first reference point One intrinsic mode component.

[0078] According to a second aspect of the present invention, a cable-stayed bridge vibration identification system based on UAV video and image correction is provided, comprising:

[0079] Image acquisition module: Used to control the UAV to fly around the target cable-stayed bridge from multiple angles via a preset flight path, and capture video of the dynamic response of the cable-stayed cable under natural excitation, ensuring that the image includes the target cable-stayed cable and the surrounding fixed environment;

[0080] Semantic segmentation module: used to perform semantic segmentation on the cable-stayed bridge region in video images, distinguish the cable-stayed bridge from unstructured objects in complex backgrounds, and obtain clear cable-stayed bridge pixel regions;

[0081] Displacement tracking module: Used to perform frame-by-frame pixel displacement tracking of the cable target area and surrounding fixed reference point area using digital image correlation algorithms, and extract time-series displacement data containing cable vibration response and UAV disturbance components;

[0082] Geometric correction module: used to construct a stable image reference coordinate system based on reference points, and to realize geometric correction of image sequences by calculating the homography matrix between images, thereby eliminating the overall jitter caused by changes in UAV attitude;

[0083] The disturbance stripping module is used to perform variational mode decomposition on the displacement time series of the cable region and the reference point region respectively, extract the intrinsic mode components, identify the common mode disturbance components by comparing the modal characteristics of the two, and strip them from the cable response to obtain the pure absolute vibration response signal of the cable.

[0084] Joint Analysis Module: Used for time-frequency joint analysis of pure vibration signals, identifying atypical vibration characteristics, and combining with the normal state vibration mode library to identify and warn of abnormal vibration behaviors such as relaxation, strand breakage, and abnormal excitation.

[0085] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, wherein when the computer program is executed by a processor, the steps of a method for identifying cable-stayed bridge vibration based on UAV video and image correction are implemented.

[0086] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0087] 1. The cable-stayed bridge vibration identification method of this invention integrates UAV non-contact monitoring and multi-dimensional image processing technology, overcoming the installation limitations and coverage shortcomings of traditional contact monitoring. Through a closed-loop technology process from video acquisition to anomaly early warning, it achieves high-precision and automated monitoring of cable-stayed bridge vibration status, providing an efficient and low-cost new technical approach for bridge structural health assessment, combining flexibility and practicality.

[0088] 2. The cable-stayed bridge vibration identification method of the present invention employs a deep learning semantic segmentation algorithm to accurately separate the cable-stayed bridge from the complex background, and eliminates background interference through masking matrix and morphological optimization. This step lays a clean data foundation for subsequent displacement tracking, avoids misjudging vibration signals by unstructured objects, significantly improves the target area identification accuracy, and ensures that subsequent analysis focuses solely on the dynamic response of the cable-stayed bridge itself.

[0089] 3. The cable-stayed bridge vibration identification method of the present invention, based on the common-mode disturbance identification method of variational mode decomposition, can effectively remove noise components such as UAV jitter. By comparing the modal characteristics of the cable and the reference point, the pure vibration signal is accurately extracted, eliminating the influence of environmental interference on vibration analysis, making the subsequent time-frequency analysis results more consistent with the actual vibration state of the cable, and improving the accuracy of feature extraction.

[0090] 4. The cable-stayed bridge vibration identification method of this invention combines short-time Fourier transform time-frequency joint analysis to comprehensively capture the time-varying characteristics and frequency distribution of vibration signals. By constructing a normal mode library to achieve quantitative identification of abnormal behavior, it can accurately determine problems such as slack and broken strands. The multi-level early warning mechanism provides clear guidance for bridge maintenance and reduces structural safety hazards. Attached Figure Description

[0091] Figure 1 This is a flowchart illustrating a method for identifying cable-stayed bridge vibration based on UAV video and image correction, according to an embodiment of the present invention.

[0092] Figure 2 This is a schematic diagram of a drone-based method for identifying cable-stayed bridge vibration based on drone video and image correction, according to an embodiment of the present invention.

[0093] Figure 3 This is a schematic diagram of the operation interface for selecting target points in a video file of a cable-stayed bridge vibration recognition system based on UAV video and image correction, according to an embodiment of the present invention.

[0094] Figure 4 This is a schematic diagram of the operation interface of a cable displacement tracking process of a cable vibration recognition system based on UAV video and image correction according to an embodiment of the present invention;

[0095] Figure 5 This is a schematic diagram of the time history of cable displacement monitored by a cable vibration identification system based on UAV video and image correction according to an embodiment of the present invention;

[0096] Figure 6 This is a schematic diagram of the cable displacement transformation result of a cable vibration recognition system based on UAV video and image correction according to an embodiment of the present invention;

[0097] Figure 7This is a schematic diagram of the health monitoring transformation results of a cable-stayed bridge vibration identification system based on UAV video and image correction, according to an embodiment of the present invention. Detailed Implementation

[0098] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0099] Example 1

[0100] like Figure 1 , 2 As shown, this embodiment of the invention provides a method for identifying cable-stayed bridge vibration based on UAV video and image correction, including the following steps:

[0101] S100: Control the drone to fly around the target cable-stayed bridge from multiple angles via a preset flight path, and capture video of the dynamic response of the cable-stayed cable under natural excitation, ensuring that the footage includes the target cable-stayed cable and the surrounding fixed environment.

[0102] S200. Perform semantic segmentation on the cable-stayed bridge region in the video image to distinguish the cable-stayed bridge from unstructured objects in a complex background and obtain clear cable-stayed bridge pixel regions.

[0103] S300 uses a digital image correlation algorithm to perform frame-by-frame pixel displacement tracking of the cable target area and the surrounding fixed reference point area, and extracts time-series displacement data containing cable vibration response and UAV disturbance components.

[0104] S400: Construct a stable image reference coordinate system based on reference points, and realize image sequence geometric correction by calculating the homography matrix between images to eliminate the overall jitter caused by UAV attitude changes;

[0105] S500: Perform variational mode decomposition on the displacement time series of the cable region and the reference point region respectively to extract the intrinsic mode components; identify the common mode disturbance components by comparing the modal characteristics of the two and remove them from the cable response to obtain the pure absolute vibration response signal of the cable.

[0106] The S600 performs time-frequency joint analysis on pure vibration signals to identify atypical vibration characteristics. Combined with the normal state vibration mode library, it can identify and warn of abnormal vibration behaviors such as relaxation, strand breakage, and abnormal excitation.

[0107] In step S200, the image pixels are classified using an algorithm to accurately distinguish the cableway pixels from the background pixels. Specifically:

[0108] First, let the image in the video be... ,in, The number of pixels representing the image height. The number of pixels in the image width. This represents the RGB three-channel color information.

[0109] The coordinates of each pixel in the image are: ,

[0110] Pixel values ​​are: ,

[0111] in, These represent the brightness values ​​for the red, green, and blue channels, respectively.

[0112] Then, a semantic segmentation model is used to classify the images; the model is a mapping function. :

[0113] in, The number of categories is specified, and it must include at least two categories: cable-stayed bridge and background.

[0114] Then the model output For fractional graphs, where For pixel category Confidence score;

[0115] The confidence scores are converted into a probability distribution using the Softmax function, specifically as follows:

[0116]

[0117] in, For pixels Category The probability of and satisfying .

[0118] Next, the cable-stayed area is extracted, and the category of the cable-stayed area is defined as follows. The region of the cable-stayed bridge pixel is determined by thresholding, specifically as follows:

[0119]

[0120] in, , is a binary mask matrix (1 represents the cable-stayed bridge pixel, and 0 represents the background pixel). This is the segmentation threshold.

[0121] To eliminate segmentation noise, the binarized mask matrix also needs to be modified. Perform morphological operations, including corrosion and expansion.

[0122] The corrosion is performed to eliminate small-area noise points using the following formula:

[0123]

[0124] in, As a structural element,

[0125] The erosion operation is used to shrink the target area. The center pixel is only preserved when the structuring element is completely embedded inside the wooden plaque area.

[0126] The expansion is used to repair the voids in the cable area using the following formula:

[0127]

[0128] in, For dilation operations, the target region is expanded. As long as the structuring element overlaps with the target region, the center pixel is marked as the target region.

[0129] The final output is the optimized cable-stayed bridge mask. Its non-zero region is the pure cable-stayed bridge pixel region.

[0130] In step S300, when performing the pixel position tracking, it is necessary to define the video's first... Frame image is , No. Frame image is The rectangular sub-region where the stay cables are located is , containing pixel set Select the sub-region in the image corresponding to the fixed structure of the bridge. And it is assumed that it is only affected by drone disturbances in physical space.

[0131] For any pixel within the region , define the first The grayscale value of the frame image , No. The gray value at the corresponding position in the frame image ,in for or .

[0132] Pixels from Move to The displacement is ,Right now , .

[0133] The gray-level similarity between two frames of images is measured using a normalized cross-correlation coefficient, specifically:

[0134]

[0135] in, , , respectively, regions The average gray level in the two frames, This represents the total number of pixels within the region.

[0136] Displacement The optimal solution is the variable value that maximizes NCC:

[0137]

[0138] The overall motion of the sub-region is described by polynomial fitting, specifically as follows:

[0139]

[0140] in, , , , , , All are coefficients, solved by minimizing the grayscale error function, specifically:

[0141] Repeat the above calculation for all consecutive frames in the video to obtain:

[0142] Displacement sequence of cable region: The displacement sequence of the cable region includes vibration response and UAV disturbance;

[0143] And the displacement sequence of the reference point region: The displacement sequence of this reference point region only includes UAV disturbances.

[0144] The grayscale error function is:

[0145]

[0146] because For parameters Nonlinear functions need to be linearized using Taylor expansion. In initial parameters The first-order expansion is as follows:

[0147]

[0148] in, These are the coordinates after displacement corresponding to the initial parameters.

[0149] , which represents the coordinate changes caused by parameter correction.

[0150] Substituting the subregion displacements into the polynomial yields:

[0151]

[0152] in, All of these are parameter correction values.

[0153] linearized Substituting the error function and rearranging, we obtain a quadratic function with respect to the parameter corrections. To minimize the error, we set the partial derivatives of the error function with respect to each parameter correction to zero, resulting in a system of linear equations: ;

[0154] in, For parameter vectors, For the correction vector,

[0155] This is a Jacobian matrix, where each row corresponds to a pixel. The elements are:

[0156]

[0157] in, Let be the residual vector, with elements of . .

[0158] The parameter correction is obtained by solving the system of linear equations. Update parameters Repeat the following steps until convergence:

[0159] S301, Calculate the current parameter corresponding to and residual ;

[0160] S302, Solving the Jacobian matrix ;

[0161] S303, Solving the system of linear equations yields... And update the parameters;

[0162] Parameters that eventually converge That is, the optimal solution that minimizes grayscale error can completely describe the overall displacement field of the sub-region.

[0163] In step S400, the specific method is as follows: from the sub-region corresponding to the bridge fixed structure Pixels with significant features are selected as stable reference points, denoted as . And regarding the video, Frame and the Frame, of which the first The first frame is used as the initial frame and as the reference frame. A feature matching algorithm is used to find the corresponding position of the reference point between the two frames. Specifically:

[0164] Reference point in reference frame: ,

[0165] No. Corresponding points in the frame: .

[0166] The image homography matrix is ​​a 3×3 invertible matrix. satisfy:

[0167]

[0168] in, As a scale factor, ,

[0169] The homography matrix is ​​in the form of:

[0170]

[0171] Expanding the above equation and eliminating the scale factor This yields a system of linear equations:

[0172]

[0173] in, Solve the overdetermined system of equations using the least squares method:

[0174]

[0175] in, Let be the parameter vector to be determined.

[0176] For corresponding points The coefficient matrix,

[0177] For corresponding points The constant term vector.

[0178] For the first Any pixel in a frame image Corrected coordinates are mapped to the reference frame coordinate system via a homography matrix. :

[0179]

[0180] in, For the first The inverse transformation matrix from frame to reference frame is used to pull the jittery image back to the reference coordinate system.

[0181] Due to the corrected coordinates The values ​​may be non-integers. The pixel values ​​of the corrected image are calculated using bilinear or bicubic interpolation algorithms, specifically:

[0182]

[0183] in, for The surrounding integer coordinate pixels,

[0184] This is the difference weight.

[0185] Finally, the pixel values ​​of the corrected image Obtain the corrected image sequence The fixed reference point remains stable in the sequence, and the overall jitter caused by changes in the drone's attitude is eliminated.

[0186] In step S500, the specific method is as follows:

[0187] Extract the time series data corrected in step S400 to obtain:

[0188] Displacement sequence of cable region ,in For time frame indexing;

[0189] and reference point region displacement sequence .

[0190] The two sequences above are averaged to eliminate the DC component, specifically as follows:

[0191]

[0192] Then, variational mode decomposition is used to extract intrinsic mode components. Through iterative optimization, the signal is decomposed into multiple intrinsic mode components with specific frequency bandwidths. Each intrinsic mode component satisfies the sparsity and finite bandwidth characteristics.

[0193] For signals The purpose of variational mode decomposition is to solve the following variational problems:

[0194]

[0195] in, for or ,

[0196] The preset number of intrinsic modal components,

[0197] For the first The center frequency of each intrinsic modal component

[0198] It's a countdown to the end of the day. For convolution operations, For the Dirac function,

[0199] The objective function constrains the sum of the bandwidths of all intrinsic modal components to be minimized, and the superposition of all intrinsic modal components equals the original signal.

[0200] Solving the above variational problem using the Lagrange multiplier method yields:

[0201] Pull cable signal decomposition: ,in For Lasso's One intrinsic modal component;

[0202] Reference point signal decomposition: ,in The first reference point One intrinsic mode component.

[0203] It is understandable that drone disturbances will simultaneously affect the displacement signals of both the cable and the reference point, resulting in similar frequency characteristics in their intrinsic mode components. Common-mode disturbances can be identified through the following steps:

[0204] S501. Calculate the center frequencies of each intrinsic modal component: based on the output of variational mode decomposition. The center frequencies of the natural modal components of the cable and the reference point are obtained. and ;

[0205] S502, Frequency Similarity Matching: Set frequency threshold ,like Then determine and Common-mode components;

[0206] S503. Calculate the energy percentage for the matched intrinsic mode component pairs. ,like If so, it is confirmed as the main disturbance component;

[0207] S504. The identified common-mode disturbance components of the cable are stripped away to obtain the pure absolute vibration response signal of the cable. ;

[0208] in, The set of identified common-mode disturbance components of the cable. .

[0209] The time series of cable vibration after eliminating drone disturbances was obtained:

[0210] .

[0211] In step S600, the time-frequency joint analysis of the pure vibration signal involves dividing the signal into segments using a sliding time window and performing a Fourier transform on each segment.

[0212]

[0213] in, It is a window function, and the output is a time-frequency matrix. Reflecting different times The frequency below Components and energy.

[0214] The identification of atypical vibration characteristics involves extracting frequency characteristics, amplitude characteristics, and time-varying characteristics from time-frequency analysis results to distinguish between normal and abnormal vibrations.

[0215] The frequency characteristic includes the principal vibration frequency offset: ,

[0216] Frequency bandwidth: ,

[0217] Harmonic component ratio:

[0218] in, The current main frequency,

[0219] This is the normal main frequency.

[0220] For the maximum frequency,

[0221] For the minimum frequency,

[0222] Harmonic energy,

[0223] Total energy;

[0224] The amplitude characteristics include the peak value of the vibration amplitude: ,

[0225] Amplitude variation coefficient:

[0226] in, The standard deviation of the amplitude.

[0227] The mean;

[0228] The time-varying characteristics include the frequency drift rate (the change in the dominant frequency per unit time): ,

[0229] Number of transient impacts (number of sudden increases in amplitude per unit time): .

[0230] The normal state vibration mode library includes:

[0231] Sample set: Collect vibration signals of stay cables under healthy conditions (covering different environmental excitations, temperatures, traffic loads, and other working conditions);

[0232] Feature template: Extract the above time-frequency features from normal samples, and establish normal feature intervals through statistical analysis (such as mean, standard deviation, probability distribution);

[0233] Modal classification: Stored according to cable type (such as length and diameter) and bridge working condition to improve matching accuracy.

[0234] During anomaly identification, the following typical anomalies are identified by comparing the current features with the normal modality library:

[0235] Relaxation recognition: main frequency Significantly lower than the normal range, and the coefficient of variation of amplitude. An increase indicates that the decrease in cable tension leads to a decrease in stiffness;

[0236] Broken strand identification: frequency bandwidth Abnormal widening, harmonic ratio A significant increase indicates that the break in the strands has led to structural discontinuity, resulting in multi-frequency vibrations;

[0237] Excitation anomaly: number of transient impacts Exceeding the normal range, and accompanied by frequency drift rate Sudden change indicates the generation of abnormal excitations such as wind and rain vibration or vortex-induced vibration.

[0238] Example 2

[0239] like Figure 3-7 As shown, this embodiment of the invention provides a cable-stayed bridge vibration identification system based on UAV video and image correction, comprising:

[0240] Image acquisition module: Used to control the UAV to fly around the target cable-stayed bridge from multiple angles via a preset flight path, and capture video of the dynamic response of the cable-stayed cable under natural excitation, ensuring that the image includes the target cable-stayed cable and the surrounding fixed environment;

[0241] Semantic segmentation module: used to perform semantic segmentation on the cable-stayed bridge region in video images, distinguish the cable-stayed bridge from unstructured objects in complex backgrounds, and obtain clear cable-stayed bridge pixel regions;

[0242] Displacement tracking module: Used to perform frame-by-frame pixel displacement tracking of the cable target area and surrounding fixed reference point area using digital image correlation algorithms, and extract time-series displacement data containing cable vibration response and UAV disturbance components;

[0243] Geometric correction module: used to construct a stable image reference coordinate system based on reference points, and to realize geometric correction of image sequences by calculating the homography matrix between images, thereby eliminating the overall jitter caused by changes in UAV attitude;

[0244] The disturbance stripping module is used to perform variational mode decomposition on the displacement time series of the cable region and the reference point region respectively, extract the intrinsic mode components, identify the common mode disturbance components by comparing the modal characteristics of the two, and strip them from the cable response to obtain the pure absolute vibration response signal of the cable.

[0245] Joint Analysis Module: Used for time-frequency joint analysis of pure vibration signals, identifying atypical vibration characteristics, and combining with the normal state vibration mode library to identify and warn of abnormal vibration behaviors such as relaxation, strand breakage, and abnormal excitation.

[0246] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A stay cable vibration identification method based on unmanned aerial vehicle video and image rectification, characterized in that, Includes the following steps: S100: Control the drone to fly around the target cable-stayed bridge from multiple angles via a preset flight path, and capture video of the dynamic response of the cable-stayed cable under natural excitation, ensuring that the footage includes the target cable-stayed cable and the surrounding fixed environment; S200. Perform semantic segmentation on the cable-stayed bridge region in the video image to distinguish the cable-stayed bridge from unstructured objects in a complex background and obtain clear cable-stayed bridge pixel regions. S300 uses a digital image correlation algorithm to perform frame-by-frame pixel displacement tracking of the cable target area and the surrounding fixed reference point area, and extracts time-series displacement data containing cable vibration response and UAV disturbance components. S400: Construct a stable image reference coordinate system based on reference points, and realize image sequence geometric correction by calculating the homography matrix between images to eliminate the overall jitter caused by UAV attitude changes; S500: Variational mode decomposition is performed on the displacement time series of the cable region and the reference point region respectively to extract the intrinsic mode components; common mode disturbance components are identified by comparing the modal characteristics of the two and are separated from the cable response to obtain the pure absolute vibration response signal of the cable. S600 performs time-frequency joint analysis on pure vibration signals to identify atypical vibration characteristics. Combined with the normal state vibration mode library, it can identify and warn of abnormal vibration behaviors such as relaxation, strand breakage, and abnormal excitation. In step S300, when tracking the pixel position, the first frame image of the video needs to be defined as The frame image is , the first frame image is ; The rectangular sub-region where the stay cable is located is , containing a pixel set , The optimized stay cable mask, whose non-zero region is the pure stay cable pixel region; Selecting a sub-region in the image corresponding to the fixed structure of the bridge and assuming that it is only affected by the drone perturbation in physical space; For any pixel within the region , define the first The grayscale value of the frame image , No. The gray value at the corresponding position in the frame image ,in for or ; Pixels from move to a displacement of i.e. , ; The gray-level similarity between two frames of images is measured using a normalized cross-correlation coefficient, specifically: ， wherein, , , respectively, the area the average gray level in the two frames, the total number of pixels in the area; displacement The optimal solution is the variable value that maximizes the NCC: ， The overall motion of the sub-region is described by polynomial fitting, specifically as follows: ， in, , , , , , All are coefficients, solved by minimizing the grayscale error function, specifically: Repeat the above calculation for all consecutive frames in the video to obtain: Displacement sequence of cable region: The displacement sequence of the cable region includes vibration response and UAV disturbance; and a sequence of reference point region displacements: which only contain drone perturbations.

2. The cable vibration identification method based on UAV video and image rectification according to claim 1, characterized in that, The grayscale error function is: ， because For parameters Nonlinear functions need to be linearized using Taylor expansion. In initial parameters The first-order expansion is as follows: ， in, These are the coordinates after displacement corresponding to the initial parameters. a coordinate change due to parameter modification; Substituting the subregion displacements into the polynomial yields: ， wherein , are parameter correction amounts.

3. The cable vibration identification method based on UAV video and image rectification according to claim 2, characterized in that, In step S400, the specific method is as follows: from the sub-region corresponding to the bridge's fixed structure... Pixels with significant features are selected as stable reference points, denoted as . And regarding the video, Frame and the Frame, of which the first The first frame is used as the initial frame and as the reference frame. A feature matching algorithm is used to find the corresponding position of the reference point between the two frames. Specifically: Reference points in the reference frame: , First corresponding points in the frames: .

4. The cable vibration identification method based on UAV video and image rectification according to claim 3, characterized in that, The inter-image homography matrix is a 3x3 invertible matrix satisfies: ， wherein is a scale factor, , The homography matrix is ​​in the form of: ， Expanding the above equation and eliminating the scale factor yields a system of linear equations: in, Solve the overdetermined system of equations using the least squares method: ， wherein is the parameter vector to be determined, For corresponding points The coefficient matrix, corresponding point constant term vector.

5. The cable vibration identification method based on UAV video and image rectification according to claim 4, characterized in that, to the first any pixel in the frame image , mapped to a rectified coordinate in the reference frame coordinate system by the homography matrix : ， wherein, is the first is the inverse transform matrix from the frame to the reference frame, which realizes pulling the jitter image back to the reference coordinate system; Since the corrected coordinates The pixel value of the corrected image can be calculated by bilinear interpolation or bicubic interpolation algorithm, which is specifically: ， wherein is the integer coordinate pixel surrounding are interpolation weights; Finally, the pixel values ​​of the corrected image Obtain the corrected image sequence The fixed reference point remains stable in the sequence, and the overall jitter caused by changes in the drone's attitude is eliminated.

6. The method for identifying cable-stayed bridge vibration based on UAV video and image correction according to claim 5, characterized in that, In step S500, the specific method is as follows: Extract the time series data corrected in step S400 to obtain: Cable region displacement sequence wherein is a time frame index; And reference point area displacement sequence ; The two sequences above are averaged to eliminate the DC component, specifically as follows: ， Then, variational mode decomposition is used to extract intrinsic mode components. Through iterative optimization, the signal is decomposed into multiple intrinsic mode components with specific frequency bandwidths. Each intrinsic mode component satisfies the sparsity and finite bandwidth characteristics. For the signal The purpose of the variational modal decomposition is to solve the following variational problem: ， wherein is or , for a preset number of inherent modal components, the center frequency of the mth natural mode component, is the time inverse, is a convolution operation, is the Dirac function, The objective function constrains the sum of the bandwidths of all intrinsic modal components to be minimized, and the superposition of all intrinsic modal components equals the original signal.

7. The cable vibration identification method based on UAV video and image rectification according to claim 6, characterized in that, Solving the variational problem using the Lagrange multiplier method yields: Pull cable signal decomposition: ,in For Lasso's One intrinsic modal component; Reference point signal decomposition: ,in The first reference point One intrinsic mode component.

8. A cable-stayed bridge vibration identification system based on UAV video and image correction, used to implement the cable-stayed bridge vibration identification method based on UAV video and image correction as described in any one of claims 1-7, characterized in that, include: Image acquisition module: Used to control the UAV to fly around the target cable-stayed bridge from multiple angles via a preset flight path, and capture video of the dynamic response of the cable-stayed cable under natural excitation, ensuring that the image includes the target cable-stayed cable and the surrounding fixed environment; Semantic segmentation module: used to perform semantic segmentation on the cable-stayed bridge region in video images, distinguish the cable-stayed bridge from unstructured objects in complex backgrounds, and obtain clear cable-stayed bridge pixel regions; Displacement tracking module: Used to perform frame-by-frame pixel displacement tracking of the cable target area and surrounding fixed reference point area using digital image correlation algorithms, and extract time-series displacement data containing cable vibration response and UAV disturbance components; Geometric correction module: used to construct a stable image reference coordinate system based on reference points, and to realize geometric correction of image sequences by calculating the homography matrix between images, thereby eliminating the overall jitter caused by changes in UAV attitude; The disturbance stripping module is used to perform variational mode decomposition on the displacement time series of the cable region and the reference point region respectively, extract the intrinsic mode components, identify the common mode disturbance components by comparing the modal characteristics of the two, and strip them from the cable response to obtain the pure absolute vibration response signal of the cable. Joint Analysis Module: Used for time-frequency joint analysis of pure vibration signals, identifying atypical vibration characteristics, and combining with the normal state vibration mode library to identify and warn of abnormal vibration behaviors such as relaxation, strand breakage, and abnormal excitation.

Images