A full-time bridge cable visual cable force monitoring method and system

Abstract

This invention discloses a method and system for all-time visual cable force monitoring of bridge stay cables. The method includes: deploying visual targets on the bridge stay cables; acquiring target image sequences under complex lighting conditions throughout the day using a high frame rate industrial camera; performing adaptive binarization processing on the images using an improved Sauvola algorithm that incorporates adaptive circular windows, Gaussian weighted statistics, and edge consistency constraints to segment the targets; identifying target regions through connected component analysis; calculating the sub-pixel center coordinates of the targets using the gray-scale centroid method or geometric center method to generate a cable vibration displacement time history signal; performing power spectral density analysis on the displacement time history signal to obtain a candidate frequency set; determining the true fundamental frequency using cable vibration harmonic characteristics and frequency correlation analysis; and calculating the cable force value based on the true fundamental frequency. This improves the intelligence and reliability of bridge structural health monitoring.

Description

Technical Field

[0001] This invention belongs to the field of bridge structural health monitoring technology, and more specifically, relates to a method and system for visual cable force monitoring of bridge stay cables at all times. Background Technology

[0002] As a key load-bearing component of bridge structures, the stress state of bridge stay cables directly affects the overall safety and durability of the bridge. With the expansion of transportation infrastructure construction and the increase in service life, the health monitoring of stay cables has become one of the core aspects of bridge operation and maintenance management. Traditional cable stress monitoring methods mostly rely on contact sensors. These methods not only require complex installation procedures but also suffer from problems such as susceptibility to environmental corrosion, poor long-term stability, and limited monitoring range, making it difficult to meet the needs of all-time, large-scale monitoring.

[0003] Visual monitoring technology, with its advantages of being non-contact, low-cost, and remotely operable, has gradually become a research hotspot in the field of bridge monitoring. However, in practical engineering applications, outdoor lighting conditions are complex and variable. Differences in lighting at different times of day, such as low light in the early morning, strong light at noon, and cloudy / rainy weather, can easily lead to problems such as low contrast and blurred edges in target images. Existing visual monitoring algorithms often suffer from insufficient target segmentation accuracy and large positioning errors when processing such images, which in turn affects the accuracy of subsequent vibration displacement extraction and frequency analysis, making it impossible to achieve stable and reliable all-time cable force monitoring.

[0004] Furthermore, existing visual monitoring solutions lack effective coordination between different stages. Image preprocessing, target localization, and frequency analysis are relatively independent processes, failing to fully consider the cascading effects of lighting changes on each stage. This results in weak anti-interference capabilities and significant fluctuations in monitoring results under complex conditions. Therefore, developing a visual cable force monitoring method for stay cables that can adapt to complex lighting environments at all times and ensure stable coordination among all stages is of significant practical importance and application value for improving the intelligence level of bridge operation and maintenance and ensuring bridge structural safety. Summary of the Invention

[0005] This invention aims to solve the problems of traditional cable-stayed bridge tension monitoring methods, such as cumbersome contact installation, weak anti-interference ability, and the difficulty of adapting existing visual monitoring technologies to complex lighting environments at all times, insufficient target segmentation accuracy, and poor stability of monitoring results. It provides a non-contact, high-precision, and stable visual cable-stayed bridge tension monitoring method to improve the intelligence and reliability of bridge structural health monitoring.

[0006] To address the aforementioned deficiencies or improvement needs of existing technologies, as a first aspect of this invention, the present invention provides a method for all-time visual cable force monitoring of bridge stay cables, comprising: S1. Visual targets are set up on the bridge cable stays, and high frame rate industrial cameras are used to continuously acquire images of the targets under complex lighting conditions to obtain a target image sequence covering the entire time period. S2. An improved Sauvola algorithm, which incorporates an adaptive circular window, Gaussian weighted statistics, and edge consistency constraints, is used to adaptively binarize each frame of the acquired image. The response of the central region is enhanced by Gaussian weighted local mean, and the real edge response is strengthened by edge consistency measurement, thus completing the segmentation of the target under complex lighting conditions. S3. Perform connected component analysis on the binarized image to identify the target region, calculate the sub-pixel level coordinates of the target center using the gray-scale centroid method or the geometric center method, and generate the displacement time history signal of the cable vibration based on the sub-pixel level coordinates. S4. Perform power spectral density analysis on the displacement time history signal to screen out a set of candidate frequencies that meet the preset conditions; process the candidate frequencies using the harmonic characteristics of the cable vibration frequency, and determine the true fundamental frequency of the cable through frequency correlation analysis; calculate the cable force value based on the true fundamental frequency.

[0007] Furthermore, the method for selecting the visual target in S1 is as follows: Targets that meet the requirements of visual recognition adaptability, environmental interference resistance, and vibration synchronization are selected. Visual recognition adaptability is reflected in the distinguishable feature differences between the target and the cable-stayed bridge and the surrounding environment, and the target shape and size are adapted to the shooting parameters of the industrial camera to ensure that the target features can be stably identified after imaging. Environmental interference resistance is reflected in the fact that the target material and surface characteristics can resist the influence of environmental factors such as outdoor light reflection, temperature and humidity changes, and wind loads. Vibration synchronization is reflected in the fact that the target is rigidly connected to the cable-stayed bridge through a fixed structure to ensure that the target vibrates synchronously with the cable-stayed bridge without relative displacement, thus ensuring the authenticity of displacement signal acquisition.

[0008] Furthermore, the sampling rate of the high frame rate industrial camera in S1 is not less than 25Hz, and the acquisition time is not less than 5 minutes to ensure the effectiveness of subsequent signal analysis.

[0009] Furthermore, the threshold of the improved Sauvola algorithm in S2 is obtained through the collaborative calculation of Gaussian weighted local mean, weighted standard deviation, edge consistency metric, and edge term strength control parameter, specifically as follows: in, This is a Gaussian-weighted local mean, used to enhance the response to the central region; The weighted standard deviation; As an edge consistency metric, it is calculated by using a radius of... The average value of the gradient normal projection is obtained by upsampling the circular contour and used to enhance the real edge response; For edge term strength control parameters; The parameters are adjusted to regulate the influence of the weighted standard deviation on the threshold, adapting to the grayscale differentiation requirements under different light intensities. The grayscale dynamic range of the image, i.e. the maximum range of grayscale values ​​in the image, is used as the normalization benchmark for the weighted standard deviation; this step ensures that the target morphology can still be completely extracted even when the contrast between the target core region and the background is extremely weak.

[0010] Furthermore, the adaptive binarization process in S2 is as follows: To address the complex lighting conditions of outdoor environments at all times, an improved Sauvola algorithm incorporating adaptive circular windows, Gaussian weighted statistics, and edge consistency constraints is used to process each frame of the image. First, an adaptive circular window is constructed based on the region corresponding to the preset shape of the target. Let the coordinates of any pixel within the window be... The coordinates of the window center are , pixels The distance to the center of the window is Calculate the Gaussian distribution mapping value ,in This is a Gaussian diffusion parameter used to define the range of pixel grayscale influence within the window; pass Calculate the Gaussian-weighted local mean, where For pixels grayscale value, Used to enhance the grayscale feature response of the central region of the window; pass Calculate the weighted standard deviation to characterize the dispersion of gray-level distribution within the window; Select evenly on the outline of the circular window For each sampling point, calculate the gradient vector of that sampling point. Obtain the projection value of the gradient vector in the normal direction of the window contour. ,in The contour normal unit vector is obtained by... Calculate edge consistency metrics to enhance the true edge response of the target; Complete the threshold by combining the above parameters. Calculate; classify image pixels by grayscale based on this threshold, and satisfy... The pixels that are identified as target area pixels are identified as background area pixels. The process involves iterating through all target areas in the image and repeating the window construction, parameter calculation, threshold generation, and pixel classification steps to complete the automatic binarization of the entire frame image, ensuring effective separation of the target area and the background area under complex lighting conditions throughout the day.

[0011] Furthermore, the process of calculating the sub-pixel level coordinates of the target center using the gray-scale centroid method or the geometric center method in S3 is as follows: Connectivity analysis is performed on the binarized image to select a set of pixels that match the shape of a preset target. Let the coordinates of any pixel in this set be... The grayscale value corresponding to the pixel The total number of pixels is ; If the grayscale centroid method is used, through the formula The sub-pixel coordinates of the target center were calculated. ;in, It is the sum of the products of the x-coordinate of each pixel within the target region and its corresponding gray value. The sum of pixel grayscale values ​​within the target area is used to adapt to the changes in target grayscale distribution under all-time illumination by enhancing the contribution of high grayscale areas of the target to the center coordinates. If the geometric center method is used, through the formula The sub-pixel coordinates of the target center were calculated. ;in, It is the sum of the x-coordinates of the pixels within the target area. The sum of the vertical coordinates of pixels within the target area is used to calculate the geometric center of the target area by averaging the pixel coordinates, thus adapting to scenarios with regular target shapes. Sub-pixel level coordinates obtained based on any of the above methods By combining the coordinate data of the target center in the continuous frame images, the displacement time history signal corresponding to the cable vibration is generated, providing basic data for subsequent cable force calculation.

[0012] Furthermore, the process of generating the displacement time history signal of the cable vibration based on the sub-pixel level coordinates in S3 is as follows: For a sequence of continuously acquired target images, the sub-pixel coordinates of the target center in each frame are extracted in chronological order of image acquisition. ,in The image acquisition time; select the target center coordinates of the first frame image in the sequence. As a reference benchmark, the offset of the target center coordinates relative to this benchmark is calculated in each subsequent frame image, i.e.: in, for The offset of the target center in the horizontal direction at any given time. for The offset of the target center along the vertical axis at any given time; combined with the imaging parameters and shooting distance of the industrial camera, the above pixel offset is converted into an actual physical displacement, resulting in... The vibration displacement of the cable in the corresponding direction at each moment is measured; the physical displacement at each moment is arranged in the order of acquisition time to form the displacement time history signal of the cable vibration.

[0013] Furthermore, the process of determining the true fundamental frequency of the cable through frequency correlation analysis in S4 is as follows: Harmonic characteristic screening is performed on the candidate frequency set obtained from power spectral density analysis. The integer multiple relationships of each frequency in the candidate frequency set are calculated, and a frequency correlation matrix is ​​constructed. Let any frequency in the candidate frequency set be denoted as . The remaining frequencies are Through formula Calculate frequency and correlation ,in This is the rounding function. The smaller the value, the more likely it is to indicate and The stronger the harmonic correlation; Set a correlation threshold, filter out frequency pairs with a correlation less than the threshold, count the number of strongly correlated frequencies corresponding to each frequency in the candidate frequency set, and the candidate frequency with the most strongly correlated frequencies is the true fundamental frequency of the cable; verify the true fundamental frequency obtained by screening, calculate the deviation between its integer multiple frequency and the corresponding frequency in the candidate frequency set, and confirm the validity of the fundamental frequency if the deviation is within the preset range, thereby eliminating pseudo frequencies caused by interference factors including environmental noise and equipment vibration.

[0014] As a second aspect of the present invention, a visual cable force monitoring system for bridge stay cables covering all time periods is also provided, comprising: The all-time target image acquisition unit is used to deploy visual targets on the bridge cable stays. It uses a high frame rate industrial camera to continuously acquire images of the targets under complex lighting conditions, obtaining a target image sequence covering the entire time period. The image segmentation unit is used to perform adaptive binarization processing on each frame of the acquired image by adopting the improved Sauvola algorithm, which introduces an adaptive circular window, Gaussian weighted statistics and edge consistency constraints. It enhances the response of the central region by Gaussian weighted local mean and strengthens the real edge response by edge consistency measurement, so as to complete the segmentation of the target under complex lighting conditions. The displacement signal generation unit is used to perform connected component analysis on the binarized image, identify the target region, calculate the sub-pixel level coordinates of the target center using the gray-scale centroid method or the geometric center method, and generate the displacement time history signal of the cable vibration based on the sub-pixel level coordinates. The cable force calculation unit is used to perform power spectral density analysis on the displacement time history signal and screen out a set of candidate frequencies that meet preset conditions; it processes the candidate frequencies using the harmonic characteristics of the cable vibration frequency and determines the true fundamental frequency of the cable through frequency correlation analysis; based on the true fundamental frequency, it calculates the cable force value of the cable.

[0015] As a third aspect of the invention, a computer-readable storage medium is also provided, on which a computer program is stored, which is executed by a processor as described in any one of the claims: a method for visual cable force monitoring of bridge stay cables for all-time periods.

[0016] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects: 1. The present invention provides a visual cable force monitoring method for bridge stay cables under all-time lighting conditions. This method employs an improved Sauvola algorithm incorporating an adaptive circular window, Gaussian weighted statistics, and edge consistency constraints. It adaptively binarizes the acquired target images across all time periods, enhances the response of the target's central region using Gaussian weighted local means, and strengthens the true edge response through edge consistency metrics. This enables accurate segmentation of the target from the background under complex lighting conditions. This processing method is adaptable to all-time lighting scenarios, including low light at dawn and strong light at noon, solving the problem of blurred target segmentation in traditional algorithms under extreme lighting conditions. It provides clear, high-quality image support for subsequent target region identification and sub-pixel coordinate calculation, ensuring the data validity of downstream monitoring processes.

[0017] 2. The present invention provides a method for all-time visual cable force monitoring of bridge stay cables. Subsequent processes are conducted using binarized segmentation results based on an improved Sauvola algorithm. Connected component analysis is then used to accurately extract the target region, further supporting sub-pixel-level coordinate calculation and displacement time-history signal generation. Finally, cable force monitoring is completed through frequency analysis. Binarized segmentation, as a core pre-processing step, directly determines the accuracy of target positioning, thus affecting the authenticity of the displacement signal and the reliability of fundamental frequency identification. This algorithm improvement enhances the overall accuracy of all-time cable force monitoring from the source, ensuring stable and effective monitoring results even under complex lighting conditions.

[0018] 3. The present invention provides a method for all-time visual cable force monitoring of bridge stay cables. By deeply coupling the adaptive binarization processing of the improved Sauvola algorithm with the all-time image acquisition process, it automatically adjusts the circular window parameters and edge response weights according to the differences in light intensity at different times, completing target segmentation without additional manual intervention. This coupling design allows the image preprocessing stage to accurately match the characteristics of light changes throughout the time period, effectively avoiding target omissions or false detections caused by light fluctuations. This lays a stable image foundation for subsequent sub-pixel localization and displacement signal generation, ensuring the continuity and consistency of all-time cable force monitoring. Attached Figure Description

[0019] Figure 1 This is a flowchart of a method for monitoring the visual cable force of bridge stay cables at all times, according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the target deployment scheme corresponding to the super-large bridge in an embodiment of the present invention; Figure 3 This is a schematic diagram comparing the recognition failure and the improved recognition effect in an embodiment of the present invention; Figure 4 This is a system unit diagram of an embodiment of the present invention. Detailed Implementation

[0020] 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.

[0021] Example 1 Please refer to Figure 1 This embodiment 1 provides a method for visual cable force monitoring of bridge stay cables for all-time periods, including: S1. Visual targets are set up on the bridge cable stays, and high frame rate industrial cameras are used to continuously acquire images of the targets under complex lighting conditions to obtain a target image sequence covering the entire time period. S2. An improved Sauvola algorithm, which incorporates an adaptive circular window, Gaussian weighted statistics, and edge consistency constraints, is used to adaptively binarize each frame of the acquired image. The response of the central region is enhanced by Gaussian weighted local mean, and the real edge response is strengthened by edge consistency measurement, thus completing the segmentation of the target under complex lighting conditions. S3. Perform connected component analysis on the binarized image to identify the target region, calculate the sub-pixel level coordinates of the target center using the gray-scale centroid method or the geometric center method, and generate the displacement time history signal of the cable vibration based on the sub-pixel level coordinates. S4. Perform power spectral density analysis on the displacement time history signal to screen out a set of candidate frequencies that meet the preset conditions; process the candidate frequencies using the harmonic characteristics of the cable vibration frequency, and determine the true fundamental frequency of the cable through frequency correlation analysis; calculate the cable force value based on the true fundamental frequency.

[0022] This embodiment 1 further elaborates on the above steps.

[0023] (1) Full-time target image acquisition In the field of bridge cable-stayed health monitoring, complex outdoor lighting and environmental interference can easily lead to distorted monitoring data. To ensure the effectiveness of cable force monitoring throughout the entire period, it is necessary to first complete scientific and reasonable monitoring preparation and image acquisition work.

[0024] Please refer to Figure 2 The first step is to select and deploy visual targets. Target selection must comprehensively consider three core requirements to adapt to outdoor monitoring scenarios. Regarding visual recognition adaptability, it is necessary to ensure that the target, the cable-stayed bridge itself, and the surrounding environment have clearly distinguishable differences in color, texture, and other features. Simultaneously, the target's shape and size specifications must precisely match the shooting parameters of the industrial camera, including lens focal length and shooting distance, to ensure that the target features are clear and complete after camera imaging, enabling stable recognition by subsequent algorithms. Regarding environmental interference resistance, the target must be made of weather-resistant materials with specially treated surface properties to effectively resist the effects of various environmental factors such as strong outdoor light reflection, dim lighting, drastic temperature and humidity changes, and wind load impacts, preventing changes in the target's own characteristics due to environmental variations. Regarding vibration synchronization, the target is rigidly connected to the cable-stayed bridge through a dedicated fixing structure. This connection method ensures that the target and the cable-stayed bridge are completely synchronized during vibration, with no relative displacement, guaranteeing from the source that the subsequently collected displacement signals can accurately reflect the actual vibration state of the cable-stayed bridge.

[0025] After the target is deployed, a high frame rate industrial camera is used to continuously acquire images of the target to obtain a target image sequence covering the entire time period. To meet the requirements of data integrity and validity for subsequent vibration signal analysis, the sampling rate of the industrial camera must be no less than 25Hz. This sampling rate can accurately capture the vibration details of the cable stays and avoid the loss of vibration characteristics due to insufficient sampling frequency. At the same time, the acquisition time must be no less than 5 minutes. Sufficient acquisition time can ensure the acquisition of complete vibration cycle data of the cable stays, providing a comprehensive and reliable data foundation for subsequent displacement time history signal generation, frequency analysis, and cable force calculation, and ensuring the smooth progress of the all-time monitoring process.

[0026] (2) Image segmentation In outdoor, all-weather monitoring scenarios, extreme lighting conditions such as low light at dawn and strong light at noon can easily reduce the contrast and blur the edges of target images, directly affecting the accuracy of subsequent target localization and cable force calculation. Therefore, it is necessary to perform targeted adaptive binarization processing on each frame of the acquired image. The processing adopts an improved Sauvola algorithm that introduces an adaptive circular window, Gaussian weighted statistics, and edge consistency constraints to achieve accurate segmentation of the target and background under complex lighting conditions.

[0027] In practice, for complex outdoor lighting environments throughout the day, an improved Sauvola algorithm incorporating adaptive circular windows, Gaussian weighted statistics, and edge consistency constraints is used to process each frame of the image. First, an adaptive circular window is constructed based on the region corresponding to the preset shape of the target. Let the coordinates of any pixel within the window be... The coordinates of the window center are , pixels The distance to the center of the window is Calculate the Gaussian distribution mapping value ,in This is a Gaussian diffusion parameter used to define the range of pixel grayscale influence within the window; pass Calculate the Gaussian-weighted local mean, where For pixels grayscale value, Used to enhance the grayscale feature response of the central region of the window; pass Calculate the weighted standard deviation to characterize the dispersion of gray-level distribution within the window; Select evenly on the outline of the circular window For each sampling point, calculate the gradient vector of that sampling point. Obtain the projection value of the gradient vector in the normal direction of the window contour. ,in The contour normal unit vector is obtained by... Calculate edge consistency metrics to enhance the true edge response of the target; Complete the threshold by combining the above parameters. Calculate; classify image pixels by grayscale based on this threshold, and satisfy... The pixels that are identified as target area pixels are identified as background area pixels. The process involves iterating through all target areas in the image and repeating the window construction, parameter calculation, threshold generation, and pixel classification steps to complete the automatic binarization of the entire frame image, ensuring effective separation of the target area and the background area under complex lighting conditions throughout the day.

[0028] Specifically, the threshold of the improved Sauvola algorithm is obtained through the collaborative calculation of Gaussian weighted local mean, weighted standard deviation, edge consistency metric, and edge term strength control parameter, as follows: in, This is a Gaussian-weighted local mean, used to enhance the response to the central region; The weighted standard deviation; As an edge consistency metric, it is calculated by using a radius of... The average value of the gradient normal projection is obtained by upsampling the circular contour and used to enhance the real edge response; For edge term strength control parameters; The parameters are adjusted to regulate the influence of the weighted standard deviation on the threshold, adapting to the grayscale differentiation requirements under different light intensities. The grayscale dynamic range of the image, i.e. the maximum range of grayscale values ​​in the image (e.g., R=255 in an 8-bit image), is used as the normalization benchmark for the weighted standard deviation; this step ensures that the target morphology can still be completely extracted even when the contrast between the core area of ​​the target and the background is extremely weak (e.g., at dawn or dusk).

[0029] (3) Generation of displacement signal The image, after adaptive binarization, has achieved preliminary separation of the target from the background, but further target region localization is still needed to obtain effective data on cable vibration. First, connected component analysis is performed on the binarized image to select a pixel set that matches the preset target shape, eliminating pseudo-connected components caused by environmental noise and lighting interference to ensure that the pixel set used in subsequent calculations completely corresponds to the real target; where the coordinates of any pixel within this set are set as follows: The grayscale value corresponding to the pixel The total number of pixels is .

[0030] After determining the set of pixels corresponding to the target, the sub-pixel level coordinates of the target center can be calculated using either the gray-scale centroid method or the geometric center method.

[0031] When using the grayscale centroid method, the coordinates and corresponding grayscale values ​​of each pixel within the pixel set are combined for calculation. By enhancing the contribution of high grayscale areas to the center coordinates, it adapts to the differences in target grayscale distribution caused by changes in illumination throughout the day, enabling accurate target center positioning even in low-light or strong-reflection scenarios. The specific process is as follows: using the formula... The sub-pixel coordinates of the target center were calculated. ;in, It is the sum of the products of the x-coordinate of each pixel within the target region and its corresponding gray value. This is the sum of pixel grayscale values ​​within the target area. This calculation method adapts to the changes in target grayscale distribution under all-time illumination by enhancing the contribution of high grayscale areas of the target to the center coordinates.

[0032] When using the geometric center method, the average of the horizontal and vertical coordinates of all pixels within the pixel set is directly calculated. This method is simple to operate, computationally efficient, and more suitable for monitoring scenarios where the target shape is regular and the grayscale distribution is uniform. The specific process is as follows: using the formula... The sub-pixel coordinates of the target center were calculated. ;in, It is the sum of the x-coordinates of the pixels within the target area. This is the sum of the vertical coordinates of pixels within the target area. This calculation method uses the mean of pixel coordinates to locate the geometric center of the target area, thus adapting to scenarios with regular target shapes.

[0033] Sub-pixel level coordinates calculated by the two methods The accuracy is higher than that of traditional pixel-level positioning, which can effectively avoid the error of pixel-level positioning. Combined with the coordinate data of the target center in continuous frame images, the displacement time history signal corresponding to the cable vibration is generated, which provides basic data for subsequent cable force calculation.

[0034] like Figure 3 As shown, the method in this embodiment significantly improves target segmentation performance under complex lighting conditions. In the original image, due to interference from weak or strong light, the contrast between the target and the background is low, and the edges are blurred. The image processed by the traditional Sauvola algorithm suffers from the problem of the target area being misidentified as background, resulting in recognition failure and inability to effectively extract target features. However, the image processed by the improved Sauvola algorithm in this embodiment shows that the target area is completely and clearly segmented, forming a clear distinction from the background. The real edge response is effectively enhanced. Even in scenes where the contrast between the core target area and the background is extremely weak, the target morphology can still be completely extracted, providing a reliable image foundation for subsequent target localization and cable force calculation. This fully verifies the effectiveness and stability of this method under complex lighting conditions at all times.

[0035] After extracting the subpixel-level coordinates of the target center from consecutive frame images, the displacement time history signal of the cable vibration can be generated. That is, for a continuously acquired sequence of target images, the subpixel-level coordinates of the target center in each frame are extracted in chronological order of image acquisition. ,in The image acquisition time; select the target center coordinates of the first frame image in the sequence. As a reference benchmark, the offset of the target center coordinates relative to this benchmark is calculated in each subsequent frame image, i.e.: in, for The offset of the target center in the horizontal direction at any given time. for The offset of the target center along the vertical axis at any given time; combined with the imaging parameters and shooting distance of the industrial camera, the above pixel offset is converted into an actual physical displacement, resulting in... The vibration displacement of the cable in the corresponding direction at each moment is measured. The physical displacement at each moment is arranged in the order of acquisition to form the displacement time history signal of the cable vibration. This signal can reflect the vibration state of the cable throughout the entire time period and provide time domain data support for subsequent cable force calculation.

[0036] (4) Cable force calculation Displacement time history signals contain the time-domain characteristics of cable vibration, and the core basis for cable force calculation is the fundamental frequency of cable vibration. Therefore, it is necessary to extract effective frequency information from the signal through frequency analysis. However, in outdoor monitoring environments, interference factors such as environmental noise and equipment vibration can easily cause pseudo frequencies to be mixed into the signal, directly affecting the accuracy of fundamental frequency identification and thus causing deviations in cable force calculation. Therefore, it is necessary to screen the true fundamental frequency through a targeted frequency analysis process.

[0037] First, power spectral density analysis is performed on the acquired cable vibration displacement time history signal to convert the time-domain signal into a frequency-domain signal. From this, a candidate frequency set that meets preset conditions is selected, and interfering frequencies that clearly do not conform to the cable vibration frequency range are initially eliminated. Since cable vibration exhibits significant harmonic characteristics—that is, integer multiples of the fundamental frequency appear simultaneously with the fundamental frequency—the candidate frequency set can be further processed based on this characteristic, and the true fundamental frequency can be determined through frequency correlation analysis.

[0038] In the specific analysis, the correlation degree between any two frequencies in the candidate frequency set is first calculated. The core logic is to determine whether one frequency is an integer multiple of another frequency—the smaller the correlation degree value, the stronger the harmonic correlation between the two frequencies. The specific process is as follows: harmonic characteristics are screened from the candidate frequency set obtained from power spectral density analysis; the integer multiple relationships of each frequency in the candidate frequency set are calculated; and a frequency correlation matrix is ​​constructed. Let any frequency in the candidate frequency set be... The remaining frequencies are Through formula Calculate frequency and correlation ,in This is the rounding function. The smaller the value, the more likely it is to indicate and The stronger the harmonic correlation; Set a correlation threshold, filter out frequency pairs with a correlation less than the threshold, count the number of strongly correlated frequencies corresponding to each frequency in the candidate frequency set, and the candidate frequency with the most strongly correlated frequencies is the true fundamental frequency of the cable; verify the true fundamental frequency obtained by screening, calculate the deviation between its integer multiple frequency and the corresponding frequency in the candidate frequency set, and confirm the validity of the fundamental frequency if the deviation is within the preset range, thereby eliminating pseudo frequencies caused by interference factors including environmental noise and equipment vibration.

[0039] This verification step effectively eliminates spurious frequencies caused by interference factors such as environmental noise and equipment vibration, ensuring the reliability of the fundamental frequency identification results. Finally, based on the verified true fundamental frequency and combined with the physical parameters of the cable, the cable force value can be calculated, completing the core process of all-time cable force monitoring.

[0040] The cable force calculation process must proceed systematically, with the fundamental frequency as the core and the key physical parameters of the cable as the main parameters. These physical parameters include the linear density, effective length, and elastic modulus of the cable. The linear density is determined by the cable material and cross-sectional dimensions, the effective length is the actual stress length between the two fixed points of the cable, and the elastic modulus is an inherent property of the cable material itself. The accuracy of these parameters directly affects the accuracy of the cable force calculation results; therefore, the parameters must be accurately measured and verified before the calculation.

[0041] In the specific calculation process, based on the inherent mechanical relationship between the fundamental frequency of cable vibration and the cable force, the actual fundamental frequency and the aforementioned physical parameters are substituted into the corresponding mechanical calculation model. The actual cable force value of the cable is obtained through model calculation. This calculation model has been verified by engineering practice and can accurately reflect the quantitative correlation between the fundamental frequency, physical parameters and cable force, effectively avoiding cable force deviation caused by parameter errors.

[0042] After the calculations are completed, the cable force results need to be verified for reasonableness. This involves considering the design load-bearing capacity of the bridge's stay cables, historical monitoring data, and the current operational conditions of the bridge to determine if the calculated cable force values ​​are within the normal range. If abnormal fluctuations in the cable force values ​​are observed, the previous monitoring process needs to be reviewed to verify the validity of data from image acquisition, target positioning, frequency analysis, and other stages, ensuring that the abnormal results are not caused by errors in previous stages. Through this series of rigorous calculations and verification processes, accurate and reliable stay cable force values ​​are ultimately obtained, completing the core closed loop of all-time cable force monitoring and providing direct data support for bridge structural health assessment and operation and maintenance decisions.

[0043] The cable stress monitoring method proposed in this embodiment, with its non-contact monitoring advantage, effectively avoids the shortcomings of traditional contact monitoring methods, such as cumbersome installation, susceptibility to environmental corrosion, and poor long-term stability. It is particularly suitable for harsh outdoor environments with complex lighting and fluctuating temperature and humidity, and can be widely applied to health monitoring scenarios for various long-span bridge stay cables. Whether for routine operation and maintenance monitoring of newly built bridges or for defect investigation and safety assessment of bridges with longer service lives, this method can provide continuous and accurate cable stress data, providing reliable support for bridge operation and maintenance management departments to promptly grasp the stress state of stay cables and predict structural safety risks, significantly reducing bridge operation and maintenance costs and improving the level of intelligent operation and maintenance management.

[0044] With the increasing demand for intelligent upgrades to transportation infrastructure and the growing emphasis on bridge safety maintenance, the application prospects of this embodiment will be further expanded. Its core improved image segmentation algorithm and precise fundamental frequency recognition technology can be transferred to the monitoring of other bridge load-bearing components such as the main cables of suspension bridges and the tie rods of tied arch bridges, forming a more versatile bridge structural health monitoring solution. Simultaneously, the integration of this method with technologies such as remote data transmission and cloud platform data analysis can construct a full-time, all-round bridge health monitoring system, facilitating the digital and intelligent transformation of bridge operation and maintenance. This has significant practical implications and broad market application value for ensuring the safety of transportation infrastructure and extending the service life of bridges.

[0045] Example 2 Please refer to Figure 4 This embodiment 2 provides a visual cable force monitoring system for bridge stay cables that operates around the clock, including: The all-time target image acquisition unit is used to deploy visual targets on the bridge cable stays. It uses a high frame rate industrial camera to continuously acquire images of the targets under complex lighting conditions, obtaining a target image sequence covering the entire time period. The image segmentation unit is used to perform adaptive binarization processing on each frame of the acquired image by adopting the improved Sauvola algorithm, which introduces an adaptive circular window, Gaussian weighted statistics and edge consistency constraints. It enhances the response of the central region by Gaussian weighted local mean and strengthens the real edge response by edge consistency measurement, so as to complete the segmentation of the target under complex lighting conditions. The displacement signal generation unit is used to perform connected component analysis on the binarized image, identify the target region, calculate the sub-pixel level coordinates of the target center using the gray-scale centroid method or the geometric center method, and generate the displacement time history signal of the cable vibration based on the sub-pixel level coordinates. The cable force calculation unit is used to perform power spectral density analysis on the displacement time history signal and screen out a set of candidate frequencies that meet preset conditions; it processes the candidate frequencies using the harmonic characteristics of the cable vibration frequency and determines the true fundamental frequency of the cable through frequency correlation analysis; based on the true fundamental frequency, it calculates the cable force value of the cable.

[0046] Example 3 This embodiment 3 also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement any step of a method for visual cable force monitoring of bridge stay cables for all-time periods.

[0047] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0048] For a description of the computer-readable storage medium provided in this application, please refer to the above method embodiments; further details will not be repeated here.

[0049] 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 method for visual cable force monitoring of bridge stay cables across all time periods, characterized in that, include: S1. Visual targets are set up on the bridge cable stays, and high frame rate industrial cameras are used to continuously acquire images of the targets under complex lighting conditions to obtain a target image sequence covering the entire time period. S2. An improved Sauvola algorithm, which incorporates an adaptive circular window, Gaussian weighted statistics, and edge consistency constraints, is used to adaptively binarize each frame of the acquired image. The response of the central region is enhanced by Gaussian weighted local mean, and the real edge response is strengthened by edge consistency measurement, thus completing the segmentation of the target under complex lighting conditions. S3. Perform connected component analysis on the binarized image to identify the target region, calculate the sub-pixel level coordinates of the target center using the gray-scale centroid method or the geometric center method, and generate the displacement time history signal of the cable vibration based on the sub-pixel level coordinates. S4. Perform power spectral density analysis on the displacement time history signal to screen out a set of candidate frequencies that meet the preset conditions; process the candidate frequencies using the harmonic characteristics of the cable vibration frequency, and determine the true fundamental frequency of the cable through frequency correlation analysis; The cable force value of the stay cable is calculated based on the actual fundamental frequency.

2. The method for visual cable force monitoring of bridge stay cables for all-time periods as described in claim 1, characterized in that, The method for selecting the visual target in S1 is as follows: Select targets that meet the requirements of visual recognition adaptability, environmental anti-interference and vibration synchronization. The visual recognition adaptability is reflected in the distinguishable feature differences between the target and the cable and the surrounding environment, and the target shape and size are adapted to the shooting parameters of the industrial camera to ensure that the target features can be stably identified after imaging. The environmental resistance is reflected in the fact that the target material and surface characteristics can resist the influence of environmental factors such as outdoor light reflection, temperature and humidity changes and wind load. The vibration synchronization is manifested in the fact that the target is rigidly connected to the cable through a fixed structure, ensuring that the target vibrates synchronously with the cable without relative displacement, thus guaranteeing the authenticity of the displacement signal acquisition.

3. The method for visual cable force monitoring of bridge stay cables for all-time periods according to claim 1, characterized in that, The sampling rate of the high frame rate industrial camera in S1 is no less than 25Hz, and the acquisition time is no less than 5 minutes to ensure the effectiveness of subsequent signal analysis.

4. The method for visual cable force monitoring of bridge stay cables for all-time periods as described in claim 1, characterized in that, The threshold of the improved Sauvola algorithm in S2 is obtained by co-calculating the Gaussian weighted local mean, weighted standard deviation, edge consistency metric, and edge term strength control parameter. Specifically: in, This is a Gaussian-weighted local mean, used to enhance the response to the central region; The weighted standard deviation; As an edge consistency metric, it is calculated by using a radius of... The average value of the gradient normal projection is obtained by upsampling the circular contour and used to enhance the real edge response; For edge term strength control parameters; The parameters are adjusted to regulate the influence of the weighted standard deviation on the threshold, adapting to the grayscale differentiation requirements under different light intensities. The grayscale dynamic range of the image, i.e. the maximum range of grayscale values ​​in the image, is used as the normalization benchmark for the weighted standard deviation; this step ensures that the target morphology can still be completely extracted even when the contrast between the target core region and the background is extremely weak.

5. The method for visual cable force monitoring of bridge stay cables for all-time periods according to claim 4, characterized in that, The adaptive binarization process in S2 is as follows: To address the complex lighting conditions of outdoor environments at all times, an improved Sauvola algorithm incorporating adaptive circular windows, Gaussian weighted statistics, and edge consistency constraints is used to process each frame of the image. First, an adaptive circular window is constructed based on the region corresponding to the preset shape of the target. Let the coordinates of any pixel within the window be... The coordinates of the window center are , pixels The distance to the center of the window is Calculate the Gaussian distribution mapping value ,in This is a Gaussian diffusion parameter used to define the range of pixel grayscale influence within the window; pass Calculate the Gaussian-weighted local mean, where For pixels grayscale value, Used to enhance the grayscale feature response of the central region of the window; pass Calculate the weighted standard deviation to characterize the dispersion of gray-level distribution within the window; Select evenly on the outline of the circular window For each sampling point, calculate the gradient vector of that sampling point. Obtain the projection value of the gradient vector in the normal direction of the window contour. ,in The contour normal unit vector is obtained by... Calculate edge consistency metrics to enhance the true edge response of the target; Complete the threshold by combining the above parameters. Calculate; classify image pixels by grayscale based on this threshold, and satisfy... The pixels that are identified as target area pixels are identified as background area pixels. The process involves iterating through all target areas in the image and repeating the window construction, parameter calculation, threshold generation, and pixel classification steps to complete the automatic binarization of the entire frame image, ensuring effective separation of the target area and the background area under complex lighting conditions throughout the day.

6. The method for visual cable force monitoring of bridge stay cables for all-time periods according to claim 1, characterized in that, The process of calculating the sub-pixel level coordinates of the target center using the gray-scale centroid method or the geometric center method in S3 is as follows: Connectivity analysis is performed on the binarized image to select a set of pixels that match the shape of a preset target. Let the coordinates of any pixel in this set be... The grayscale value corresponding to the pixel The total number of pixels is ; If the grayscale centroid method is used, through the formula The sub-pixel coordinates of the target center were calculated. ;in, It is the sum of the products of the x-coordinate of each pixel within the target region and its corresponding gray value. The sum of pixel grayscale values ​​within the target area is used to adapt to the changes in target grayscale distribution under all-time illumination by enhancing the contribution of high grayscale areas of the target to the center coordinates. If the geometric center method is used, through the formula The sub-pixel coordinates of the target center were calculated. ;in, It is the sum of the x-coordinates of the pixels within the target area. The sum of the vertical coordinates of pixels within the target area is used to calculate the geometric center of the target area by averaging the pixel coordinates, thus adapting to scenarios with regular target shapes. Sub-pixel level coordinates obtained based on any of the above methods By combining the coordinate data of the target center in the continuous frame images, the displacement time history signal corresponding to the cable vibration is generated, providing basic data for subsequent cable force calculation.

7. The method for visual cable force monitoring of bridge stay cables for all-time periods according to claim 1, characterized in that, The process of generating the displacement time history signal of the cable vibration based on the sub-pixel level coordinates in S3 is as follows: For a sequence of continuously acquired target images, the sub-pixel coordinates of the target center in each frame are extracted in chronological order of image acquisition. ,in The image acquisition time; select the target center coordinates of the first frame image in the sequence. As a reference benchmark, the offset of the target center coordinates relative to this benchmark is calculated in each subsequent frame image, i.e.: in, for The offset of the target center in the horizontal direction at any given time. for The offset of the target center along the vertical axis at any given time; combined with the imaging parameters and shooting distance of the industrial camera, the above pixel offset is converted into an actual physical displacement, resulting in... The vibration displacement of the cable in the corresponding direction at each moment is measured; the physical displacement at each moment is arranged in the order of acquisition time to form the displacement time history signal of the cable vibration.

8. The method for visual cable force monitoring of bridge stay cables for all-time periods according to claim 1, characterized in that, The process of determining the true fundamental frequency of the cable through frequency correlation analysis in S4 is as follows: Harmonic characteristic screening is performed on the candidate frequency set obtained from power spectral density analysis. The integer multiple relationships of each frequency in the candidate frequency set are calculated, and a frequency correlation matrix is ​​constructed. Let any frequency in the candidate frequency set be denoted as . The remaining frequencies are Through formula Calculate frequency and correlation ,in This is the rounding function. The smaller the value, the more likely it is to indicate and The stronger the harmonic correlation; Set a correlation threshold, filter out frequency pairs with a correlation less than the threshold, count the number of strongly correlated frequencies corresponding to each frequency in the candidate frequency set, and the candidate frequency with the most strongly correlated frequencies is the true fundamental frequency of the cable; verify the true fundamental frequency obtained by screening, calculate the deviation between its integer multiple frequency and the corresponding frequency in the candidate frequency set, and confirm the validity of the fundamental frequency if the deviation is within the preset range, thereby eliminating pseudo frequencies caused by interference factors including environmental noise and equipment vibration.

9. A visual cable force monitoring system for bridge stay cables that operates around the clock, characterized in that, include: The all-time target image acquisition unit is used to deploy visual targets on the bridge cable stays. It uses a high frame rate industrial camera to continuously acquire images of the targets under complex lighting conditions, obtaining a target image sequence covering the entire time period. The image segmentation unit is used to perform adaptive binarization processing on each frame of the acquired image by adopting the improved Sauvola algorithm, which introduces an adaptive circular window, Gaussian weighted statistics and edge consistency constraints. It enhances the response of the central region by Gaussian weighted local mean and strengthens the real edge response by edge consistency measurement, so as to complete the segmentation of the target under complex lighting conditions. The displacement signal generation unit is used to perform connected component analysis on the binarized image, identify the target region, calculate the sub-pixel level coordinates of the target center using the gray-scale centroid method or the geometric center method, and generate the displacement time history signal of the cable vibration based on the sub-pixel level coordinates. The cable force calculation unit is used to perform power spectral density analysis on the displacement time history signal and screen out a set of candidate frequencies that meet preset conditions; the candidate frequencies are processed using the harmonic characteristics of the cable vibration frequency, and the true fundamental frequency of the cable is determined through frequency correlation analysis; The cable force value of the stay cable is calculated based on the actual fundamental frequency.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program is executed by a processor as described in any one of claims 1-8: a method for monitoring the visual cable force of bridge stay cables throughout all time periods.

Images