Method for measuring prestress of cable-stayed reticulated shell based on unmanned aerial vehicle photography and noise reduction algorithm

Abstract

The invention relates to a cable-stayed reticulated shell prestress measurement method based on unmanned aerial vehicle photography and a noise reduction algorithm, and belongs to the technical field of prestress space steel structure monitoring, and the method comprises the following operation steps: 1, carrying out the image collection of a cable-stayed reticulated shell component to be measured through unmanned aerial vehicle photography; 2, inhaul cable acceleration time history extraction based on Canny edge detection and center line recognition; and converting the input image into a grayscale image. And step 3, acceleration data noise reduction processing based on improved empirical mode decomposition. And 4, calculating the cable force of the to-be-measured component, and deducing the mapping relation between the cable force and the vibration frequency. The method overcomes the defects of poor stability and high measurement cost of the sensor in the traditional method by avoiding the arrangement of the sensor on the structure. Meanwhile, interference of mechanical vibration and environmental wind power is effectively reduced by introducing a noise reduction algorithm, and the cable force monitoring precision is improved, so that a more accurate and reliable prestress measurement result is provided for the cable-stayed reticulated shell structure.

Description

Technical Field

[0001] The present invention relates to the technical field of monitoring of prestressed space steel structures, and particularly relates to a method for measuring the prestress of a cable-stayed reticulated shell based on unmanned aerial vehicle photography and a noise reduction algorithm. Background Art

[0002] Cable-stayed reticulated shell structures are widely used in large-scale projects such as modern architecture and stadiums. Their greatest advantage is that they give full play to the characteristics of rigid and flexible components, making the best use of their advantages and avoiding their disadvantages, thereby improving the mechanical properties of the entire structural system. This structure can significantly reduce the amount of material used while ensuring the safety and stability of the structure. With the development of construction technology, such structures often bear complex load conditions. During long-term use, the prestress state of the structure may change, thereby affecting the overall performance of the structure. Therefore, accurately measuring the prestress condition of the cable-stayed reticulated shell structure, especially the prestress of the cable components, has become an important task for ensuring structural safety and realizing structural health monitoring.

[0003] In order to achieve precise measurement of the prestress of cable-stayed reticulated shells, the industry has studied various measurement methods. Traditional measurement methods mainly rely on sensors. These sensors collect data by directly contacting or being fixed to the structure and calculate the prestress value through calculation. However, these methods all have problems such as strong equipment dependence, the measurement accuracy being affected by various factors, and high costs. Therefore, developing a new type of measurement technology that can efficiently and accurately obtain the prestress data of cable-stayed reticulated shell structures has become a key direction in current technical research.

[0004] In actual measurement of the prestress of cable-stayed reticulated shells, common methods are divided into two categories: static-based and dynamic-based. Static-based methods include the jack method, the pressure ring method, the magnetic flux method, etc. These methods usually require arranging sensors on the cable components and calculating the prestress by measuring the mechanical response. However, as the scale and complexity of the structure increase, the arrangement density of sensors also gradually increases, resulting in a significant increase in equipment costs, and the long-term stability and reliability of the sensors become key factors in the measurement results. Especially in complex environmental conditions, the performance of the sensors may be unstable, thus affecting the final measurement accuracy. The dynamic-based methods mainly include the frequency method, which indirectly calculates the prestress of the cable by measuring the natural vibration frequency of the structure. Although the principle of the frequency method is clear and the implementation is convenient, its accuracy is still affected by various factors, especially how to accurately measure the natural vibration frequency. In practical applications, how to accurately control and measure the frequency of the cable components has become the key to affecting the measurement accuracy.

[0005] Whether it is a static method or a dynamic method, their common point is that sensors need to be arranged on the cable components, and data is obtained through the sensors and further prestress is calculated. However, the traditional sensor-based measurement methods have the following several limitations:

[0006] (1) Sensor stability issues: The long-term stability and reliability of sensors directly affect the measurement results. Especially in complex external environments, sensors are often affected by factors such as temperature and humidity, resulting in a decline in measurement accuracy.

[0007] (2) High measurement costs: In cable-stayed reticulated shell structures containing a large number of cable members, sensors usually need to be arranged at multiple measurement points, which not only increases the procurement cost of equipment but also significantly increases the subsequent maintenance and calibration costs.

[0008] (3) Measurement difficulties in complex environments: Many cable-stayed reticulated shell structures were built in the 1980s and 1990s, with a long service life. As time goes by, the structural deformation and prestress changes are complex. Traditional methods are difficult to accurately obtain the absolute value of the prestress of cable members. Especially in old structures, the difficulty of contacting measurement points makes it impossible for traditional methods to comprehensively and highly accurately measure the cable force.

[0009] In the dynamic-based method, the frequency method has been widely used because of its clear principle and convenient implementation. However, the accuracy of the frequency method is affected by various factors, and one of the most critical factors is how to accurately measure the natural vibration frequency of cable members. Currently, several new measurement technologies, such as fixed cameras, microwave radars, laser Doppler velocimeters, etc., can achieve non-contact cable force measurement. These methods have improved the efficiency of cable force testing for cable-stayed reticulated shell structures to a certain extent, but still require operators to manually move the measurement equipment to different measurement points for data collection, which limits its measurement efficiency and universality in practical applications.

[0010] In contrast, the measurement technology based on UAV photography provides a new idea for the prestress measurement of cable-stayed reticulated shells. This method utilizes the high-efficiency image acquisition ability of UAVs and no longer relies on traditional sensor arrangements, thus overcoming the limitations of poor sensor stability and high costs. However, this method still faces significant challenges in practical applications, especially in terms of data noise interference. When the UAV hovers for photogrammetry, its own mechanical vibration and the influence of environmental wind may cause the UAV to shake, thereby introducing noise, which in turn affects the stability and accuracy of image data. These noise interference factors may lead to the uncertainty of the final measurement results.

[0011] Techniques for non-contact measurement using UAVs and prestress calculation using the frequency method. These techniques can be roughly divided into the following categories:

[0012] (1) Measurement technology based on UAV photography.

[0013] At present, some studies have utilized drones for cable force measurement. The main advantage of drone camera technology lies in its ability to collect data at high altitudes or in inaccessible areas, avoiding the high costs and difficulties of manually deploying sensors. In these implementation schemes, drones are usually equipped with high-definition cameras or other sensors. By flying to capture panoramic images of the cable-stayed reticulated shell structure and then using image processing software to analyze the data and extract structural deformation information. However, there are still certain deficiencies in these existing schemes. For example, noise interference in the image data and the influence of mechanical vibrations lead to instability and uncertainty in the measurement results.

[0014] (2) Cable force measurement method based on the frequency method.

[0015] As a classical dynamic-based measurement method, the frequency method has a simple principle. It calculates the prestress by measuring the natural vibration frequency of the cable component. However, in practical applications, how to accurately measure the natural vibration frequency of the cable is one of the key factors affecting the measurement accuracy. Existing frequency methods usually use accelerometers arranged on the cable component to detect the change in its vibration frequency. To improve the measurement accuracy, some studies have obtained more accurate frequency data through multi-point sensors combined with vibration analysis, and then deduced the prestress of the cable. Although these methods can provide relatively accurate measurement data, they still rely on the arrangement of sensors and are difficult to measure in complex environments.

[0016] (3) Existing non-contact cable force measurement technologies.

[0017] In addition to traditional sensor arrangement methods, some non-contact cable force measurement technologies have emerged in recent years, such as using devices like laser Doppler velocimeters and microwave radars. These technologies can avoid the drawbacks of direct contact measurement. Through non-contact means, these devices can improve the measurement efficiency and show great advantages especially in the measurement of large-scale structures. Although these technologies have good application effects in certain scenarios, they still require manual operation of the devices and adjustment at different measurement points, limiting their universality and efficiency in large-scale structures.

[0018] The following are some related patents:

[0019] CN 105910743 A "Method for Measuring Cable Tension of Cable-Stayed Bridge Using Unmanned Aerial Vehicle".

[0020] CN 118111610A "Method for Measuring Cable Tension of Cable-Stayed Bridge Using Unmanned Aerial Vehicle".

[0021] CN 111259770B "Fast Cable Force Testing Method in Complex Background Based on Unmanned Aerial Vehicle Platform and Deep Learning". Summary of the Invention

[0022] The present invention mainly solves the deficiencies existing in the prior art, and provides a method for measuring the prestress of a cable-stayed reticulated shell based on UAV photography and a noise reduction algorithm. By avoiding arranging sensors on the structure, it overcomes the disadvantages of poor sensor stability and high measurement cost in traditional methods. At the same time, the introduction of the noise reduction algorithm effectively reduces the interference of mechanical vibration and environmental wind force, improves the cable force monitoring accuracy, and thus provides more accurate and reliable prestress measurement results for the cable-stayed reticulated shell structure.

[0023] The above technical problems of the present invention are mainly solved by the following technical solutions:

[0024] A method for measuring the prestress of a cable-stayed reticulated shell based on UAV photography and a noise reduction algorithm includes the following operating steps:

[0025] The first step: Using UAV photography to collect images of the cable-stayed reticulated shell components to be measured; the UAV is equipped with high-precision imaging equipment, and omnidirectional image collection is carried out by flying around the structure to obtain high-definition images of the components to be measured.

[0026] The second step: Extracting the cable acceleration time history based on Canny edge detection and centerline recognition; converting the input image into a grayscale image.

[0027] The third step: Noise reduction processing of the acceleration data based on improved empirical mode decomposition.

[0028] The fourth step: Calculating the cable force of the component to be measured. By deriving the mapping relationship between the cable force and the vibration frequency, it is necessary to consider the basic vibration equation of the beam element and analyze various factors therein, such as the influence of the flexural stiffness, support conditions, and concentrated mass on the vibration frequency, and then introduce the influence of the cable force, and finally obtain the relationship with the vibration frequency.

[0029] Preferably, by analyzing the distribution of the cable-stayed reticulated shell structure, the flight path of the UAV is set; during the flight, the image data is analyzed in real time to ensure that the UAV can accurately lock the cable and provide clear basic images for subsequent data processing; during the hovering process, the imaging equipment will record a video of the target cable, especially the vibration process of the cable; the UAV will record the dynamic response of the cable through a high-frame-rate video acquisition system to ensure that the small vibrations of the cable are captured.

[0030] Preferably, the input image I(x, y) is converted into a grayscale image G(x, y) by weighted averaging the RGB channels; a Gaussian filter is used to smooth the image to reduce the influence of noise; edge detection is very sensitive to noise, and the smoothing process helps to remove small fluctuations in the image; the image is convolved with a 5×5 or larger Gaussian kernel to obtain the smoothed image; G smoothed (x, y) = G(x, y) * G kernel , where Gkernel is a Gaussian kernel;

[0031] Calculate the gradient of each pixel in the image. The gradient represents the magnitude and direction of the change in image brightness. Use the Sobel operator to calculate the gradients in the horizontal and vertical directions; calculate the horizontal gradient Gx and the vertical gradient Gy:

[0032] Remove the noise in the gradient magnitude through non-maximum suppression, only retaining the local maximum values to ensure that the detected edges are precise; this process scans each pixel along the gradient direction and sets the pixels that are not local maximum values to 0; detect the gradient magnitude of each pixel along the gradient direction and suppress those pixels that are not local maximum values.

[0033] Preferably, double-threshold processing is used to determine the edge strength and distinguish edges from non-edges. By setting two thresholds, namely the high threshold and the low threshold; if the gradient magnitude of a certain pixel is greater than the high threshold, it is considered an edge pixel; if the gradient magnitude of a certain pixel is less than the low threshold, it is considered a non-edge pixel; if the gradient magnitude of a certain pixel is between the high threshold and the low threshold and it is connected to a strong edge, it is considered an edge pixel; edge connection is determined by tracing the pixels between the high threshold and the low threshold to obtain the final edge; only those weak edge pixels that are connected to strong edges will be retained; perform edge connection, traverse all pixels, and add the qualified pixels to the edge.

[0034] Preferably, extracting the centerline of the image is carried out on the basis of detecting the edge; remove the pixels outside the edge region and only retain the central axis part of the edge; obtain the dynamic displacement of the midpoint of the cable centerline by comparing the pixel coordinate differences between different images; use the central difference method to numerically differentiate the time history of the lateral displacement of the cable to obtain the time history of the acceleration at different positions of the cable to be measured.

[0035] Preferably, use EEMD, i.e., Ensemble Empirical Mode Decomposition, to decompose the acceleration time history data. Add white noise n(t) to the original acceleration signal a(t) to obtain a new signal:

[0036] x new (t) = x(t) + αn(t);

[0037] where α is the noise amplitude and n(t) is white noise. Decompose the noisy signal X new (t) using the EMD algorithm to obtain several Intrinsic Mode Functions - IMFs:

[0038]

[0039] where, IMF i(t) is the i-th IMF, r N (t) is the remaining signal - usually the trend term. For each set of noise amplitude and decomposition times settings, calculate its BIC value. The BIC value contains the balance between the model complexity - the number of IMFs and the model fitting error - the influence of noise; select the noise amplitude α and decomposition times k with the smallest BIC value as the optimal parameters; this step balances the noise level and model complexity by minimizing the BIC value, thus ensuring the most accurate signal decomposition and the most effective noise reduction processing; the definition formula of BIC is as follows:

[0040] BIC = nln(σ 2 ) + kln(n);

[0041] In the formula, n is the number of samples (i.e., the number of time points of the signal); σ 2 is the variance of the error, representing the fitting accuracy of the model; k is the number of free parameters of the model, representing the model complexity.

[0042] Preferably, through multiple noise addition and EMD decomposition, multiple IMFs are obtained, and the IMFs of each time are integrated to finally obtain a more stable IMF with less noise influence. For the multiple IMFs obtained by decomposition, principal component analysis (PCA) needs to be used for further noise reduction to extract the principal components that have the greatest influence on the cable vibration signal;

[0043] M IMFs are obtained from EEMD, and these IMFs form a matrix X:

[0044] X = [IMF 1 (t) IMF 2 (t) … IMF M (t)];

[0045] Calculate the covariance matrix C of the IMF matrix X:

[0046]

[0047] In the formula, M is the number of IMFs. By performing eigenvalue decomposition on the covariance matrix C, eigenvalues and corresponding eigenvectors are obtained; arrange all eigenvalues in descending order, and select the principal components with a cumulative contribution rate exceeding 90%; the cumulative contribution rate ξ is calculated as follows:

[0048]

[0049] In the formula, k is the number of selected principal components; select the smallest k that can make the cumulative contribution rate exceed 90%, and then select the corresponding eigenvectors v 1 , v 2 , …, v k; Restore the selected principal components to obtain the cable acceleration after noise reduction.

[0050] Preferably, the vibration equation of a single cable considering the flexural stiffness can be expressed as:

[0051]

[0052] In the formula, ω(x, t) is the deflection of the cable at position x; E, I, and m are the elastic modulus, moment of inertia, and mass per unit length of the cable, respectively; T is the axial force component of the cable at the equilibrium position, that is, the cable force; after substituting different boundary conditions and geometric parameters of the cable, the following unified cable force estimation formula can be obtained:

[0053]

[0054] In the formula, k 1 and k 2 are the characteristic parameters of the cable to be measured, and the main parameters affecting the values of the characteristic parameters are the slenderness ratio ψ and the boundary conditions; when ψ is greater than 150, the influence of the cable bending stiffness is ignored.

[0055] Preferably, for the hinged boundary condition: when the slenderness ratio 50 < ψ ≤ 150, the characteristic parameters k 1 = 4 and k 2 = π; when the slenderness ratio ψ > 150, the characteristic parameters k 1 = 4 and k 2 = π;

[0056] For the fixed boundary condition: when the slenderness ratio 50 < ψ ≤ 150, the characteristic parameters k 1 = 3.74 and k 2 = -59.62; when the slenderness ratio ψ > 150, the characteristic parameters k 1 = 3.74 and k 2 = 0.

[0057] The present invention can achieve the following effects:

[0058] The present invention provides a method for measuring the prestress of a cable-stayed reticulated shell based on UAV photography and a noise reduction algorithm. Compared with the prior art, by avoiding arranging sensors on the structure, the disadvantages of poor sensor stability and high measurement cost in the traditional method are overcome. At the same time, introducing a noise reduction algorithm effectively reduces the interference of mechanical vibration and environmental wind force, improves the cable force monitoring accuracy, and thus provides more accurate and reliable prestress measurement results for the cable-stayed reticulated shell structure.

[0059] The non-contact prestress measurement method based on UAV photography uses UAV photography technology to measure the prestress of cable-stayed reticulated shell structures. Different from the traditional method that requires arranging sensors on the structure, using UAVs for image acquisition can avoid the limitations of sensor arrangement, reduce equipment dependence and maintenance costs. Especially when collecting data in high-altitude or inaccessible areas, it can efficiently and comprehensively obtain structural data.

[0060] The application of the noise reduction algorithm To solve the noise interference caused by mechanical vibration and environmental wind force during the hovering process of the UAV, a noise reduction algorithm based on EEMD-BIC is introduced, which significantly improves the stability and accuracy of image data. This algorithm can effectively reduce the influence of vibration and external wind force, thereby improving the accuracy and reliability of measurement results.

[0061] The measurement system combining UAV photography and data analysis algorithm By combining UAV photography technology and data analysis algorithm, a new measurement system is formed. The UAV collects images through a high-definition camera, then extracts the deformation information of the structure through image processing technology, and combines the algorithm to deduce the prestress state of the structural cable members. It simplifies the traditional measurement means and improves the measurement efficiency. Brief Description of the Drawings

[0062] Figure 1 is the flowchart of the present invention.

[0063] Figure 2 is the specific implementation step diagram of the present invention.

[0064] Figure 3 is the true acceleration response diagram of the cable of the present invention.

[0065] Figure 4 is the cable frequency identification diagram of the present invention.

[0066] Figure 5 is the cable force identification result diagram of the present invention. Detailed Description of the Invention

[0067] Next, through embodiments and in combination with the drawings, the technical solutions of the invention will be further specifically described.

[0068] Embodiment: As Figures 1-5 shown, a cable-stayed reticulated shell prestress measurement method based on UAV photography and noise reduction algorithm includes the following operating steps:

[0069] The first step: Use UAV photography to collect images of the cable-stayed reticulated shell components to be measured; the UAV is equipped with high-precision imaging equipment and conducts omnidirectional image collection by flying around the structure to obtain high-definition images of the components to be measured.

[0070] During the flight of the unmanned aerial vehicle (UAV), appropriate flight altitude and angle are determined according to the scale and complexity of the structure. The flight altitude should generally ensure that the UAV can take all-round photos of the entire cable-stayed reticulated shell structure from multiple angles, especially for high-altitude areas or parts that are not easily accessible. By analyzing the distribution of the cable-stayed reticulated shell structure, the flight path of the UAV is set. The path planning should ensure that the UAV can cover each cable member of the structure to avoid missing or duplicate shooting. During the flight, the image data is analyzed in real time to ensure that the UAV can accurately lock the cable and provide clear basic images for subsequent data processing. The UAV uses a high-precision positioning system and an automatic control system to ensure that it hovers at the target position and tries to avoid the influence of wind or other environmental factors on its position stability. During the hovering process, the imaging device will record a video of the target cable, especially the vibration process of the cable. The UAV will record the dynamic response of the cable through a high-frame-rate video acquisition system to ensure that the minute vibrations of the cable are captured. These video data provide direct visual evidence for the subsequent extraction of acceleration time history and vibration analysis.

[0071] Step 2: Extraction of cable acceleration time history based on Canny edge detection and centerline recognition; Convert the input image into a grayscale image. The input image I(x, y) is converted into a grayscale image G(x, y) by weighted averaging the RGB channels; Smooth the image using a Gaussian filter to reduce the influence of noise; Edge detection is very sensitive to noise, and the smoothing process helps to remove small fluctuations in the image; Convolve the image with a Gaussian kernel of size 5×5 or larger to obtain the smoothed image; G smoothed (x, y) = G(x, y) * G kernel , where G kernel is the Gaussian kernel;

[0072] Calculate the gradient of each pixel point in the image. The gradient represents the magnitude and direction of the change in image brightness, and the Sobel operator is used to calculate the gradients in the horizontal and vertical directions; Calculate the horizontal gradient Gx and the vertical gradient Gy:

[0073] Remove the noise in the gradient magnitude through non-maximum suppression, and only retain the local maximum values to ensure that the detected edges are accurate; This process scans each pixel point along the gradient direction and sets the pixel points that are not local maximum values to 0; Detect the gradient magnitude of each pixel along the gradient direction and suppress those pixels that are not local maximum values.

[0074] Double - threshold processing is used to determine the edge strength and distinguish edges from non - edges. By setting two thresholds, namely the high threshold and the low threshold; if the gradient magnitude of a certain pixel is greater than the high threshold, it is considered an edge pixel; if the gradient magnitude of a certain pixel is less than the low threshold, it is considered a non - edge pixel; if the gradient magnitude of a certain pixel is between the high threshold and the low threshold and it is connected to a strong edge, it is considered an edge pixel; edge connection is determined by tracing the pixels between the high threshold and the low threshold to obtain the final edge; only those weak - edge pixels that are connected to strong edges will be retained; for edge connection, all pixels are traversed and the eligible pixels are added to the edge.

[0075] Extracting the centerline of the image is carried out based on the detected edges; pixels outside the edge region are removed, and only the central axis part of the edge is retained; the dynamic displacement of the mid - point of the cable centerline is obtained by comparing the pixel coordinate differences between different images; the central difference method is used to numerically differentiate the time - history of the lateral displacement of the cable to obtain the time - history of the acceleration at different positions of the cable to be measured.

[0076] Step 3: Denoising processing of acceleration data based on improved empirical mode decomposition. Use EEMD, i.e., ensemble empirical mode decomposition, to decompose the time - history data of the acceleration. Add white noise \(n(t)\) to the original acceleration signal \(a(t)\) to obtain a new signal:

[0077] x new (t)=x(t)+αn(t);

[0078] In the formula, \(\alpha\) is the noise amplitude and \(n(t)\) is white noise. Decompose the noisy signal \(x\) new (t) using the EMD algorithm to obtain several intrinsic mode functions - IMFs:

[0079]

[0080] Among them, \(IMF\) i (t) is the \(i\) - th IMF, and \(r\) N (t) is the remaining signal - usually the trend term. For each set of noise amplitude and decomposition times settings, calculate its BIC value. The BIC value contains the balance between the model complexity - the number of IMFs and the model fitting error - the influence of noise; select the noise amplitude \(\alpha\) and decomposition times \(k\) with the minimum BIC value as the optimal parameters; this step balances the noise level and model complexity by minimizing the BIC value, thereby ensuring the most accurate signal decomposition and the most effective denoising processing; the definition formula of BIC is as follows:

[0081] BIC = nln(\(\sigma\) 2 )+kln(n);

[0082] where n is the number of samples (i.e., the number of time points of the signal); σ 2 is the variance of the error, representing the fitting accuracy of the model; k is the number of free parameters of the model, representing the complexity of the model.

[0083] Through multiple times of adding noise and EMD decomposition, multiple IMFs are obtained, and the IMFs of each time are integrated. Finally, a more stable IMF with less influence of noise is obtained. For the multiple IMFs obtained by decomposition, principal component analysis (PCA) needs to be used for further noise reduction, so as to extract the principal components that have the greatest influence on the cable vibration signal;

[0084] M IMFs are obtained from EEMD, and these IMFs form a matrix X:

[0085] X = [IMF 1 (t) IMF 2 (t) … IMF M (t)];

[0086] Calculate the covariance matrix C of the IMF matrix X:

[0087]

[0088] where M is the number of IMFs. By performing eigenvalue decomposition on the covariance matrix C, eigenvalues and corresponding eigenvectors are obtained; all the eigenvalues are arranged in descending order, and the principal components with a cumulative contribution rate exceeding 90% are selected; the cumulative contribution rate ξ is calculated according to the following formula:

[0089]

[0090] where k is the number of selected principal components; select the smallest k that can make the cumulative contribution rate exceed 90%, and then select the corresponding eigenvectors v 1 , v 2 ,..., v k ; Restore the selected principal components to obtain the cable acceleration after noise reduction.

[0091] Fourth step: Calculation of the cable force of the component to be measured. By deriving the mapping relationship between the cable force and the vibration frequency, it is necessary to consider the basic vibration equation of the beam element and analyze each factor therein, such as the influence of the flexural stiffness, support conditions, and concentrated mass on the vibration frequency, and then introduce the influence of the cable force. Finally, the relationship with the vibration frequency is obtained. The vibration equation of a cable considering the flexural stiffness can be expressed as:

[0092]

[0093] In the formula, ω(x, t) is the deflection of the cable at position x; E, I, and m are the elastic modulus, moment of inertia, and mass per unit length of the cable, respectively; T is the axial force component of the cable at the equilibrium position, that is, the cable force. After substituting different boundary conditions and geometric parameters of the cable, the following unified formula for estimating the cable force can be obtained:

[0094]

[0095] In the formula, k 1 and k 2 are the characteristic parameters of the cable to be measured. The main parameters affecting the values of the characteristic parameters are the slenderness ratio ψ and the boundary conditions. When ψ is greater than 150, the influence of the cable bending stiffness is ignored.

[0096] Hinged boundary condition: When the slenderness ratio 50 < ψ ≤ 150, the characteristic parameters k 1 = 4 and k 2 = π; when the slenderness ratio ψ > 150, the characteristic parameters k 1 = 4 and k 2 = π;

[0097] Fixed boundary condition: When the slenderness ratio 50 < ψ ≤ 150, the characteristic parameters k 1 = 3.74 and k 2 = -59.62; when the slenderness ratio ψ > 150, the characteristic parameters k 1 = 3.74 and k 2 = 0.

[0098] Application example: See Figures 2 to 5 , the specific implementation steps of the practical measurement method for the prestress of the cable-stayed reticulated shell based on UAV photography and noise reduction algorithm are as follows ( Figure 2 ):

[0099] Step 1: Image acquisition. Use the on-board camera of the UAV to take a sequence of cable vibration images with a pixel resolution of 4096×2160 and a sampling frequency of 30 Hz. To verify the applicability of this method, the structural health monitoring system arranged on the structure is used to measure the cable vibration frequency in this actual measurement.

[0100] Step 2: Image preprocessing. Select the region of interest in the acquired image as the calculation region, segment the picture, save the calculation region in the cable-stayed reticulated shell as a picture sequence, and identify the cables in the sequence.

[0101] Step 3: Acceleration calculation. After performing edge detection on the picture sequence obtained in Step 2 to obtain the original acceleration data, perform corresponding noise reduction processing to effectively distinguish the noise components and the real cable vibration signal, remove the interference caused by mechanical vibration and environmental factors, and retain the real acceleration response of the cable ( Figure 3 ).

[0102] Step 4: Cable frequency identification. Perform Fourier spectrum analysis on the cable vibration displacement time history curve obtained in Step 3. The modal frequency can be extracted from the spectrum, as Figure 4 shown.

[0103] Step 5: Cable force calculation. Using the relationship between the natural vibration frequency and the cable force, the cable force of the cable and the modal order corresponding to the identified frequency can be calculated. Compare the calculated cable force value with the measurement result of the accelerometer, Figure 5 is the measured cable force result. It can be seen that the calculated cable force is in good agreement with the true value, and the relative error between the two is less than 5%, meeting the engineering accuracy requirements, and verifying the effectiveness and accuracy of the inventive method for calculating the cable force of the cable-stayed reticulated shell structure.

[0104] In summary, the prestress measurement method for cable-stayed reticulated shells based on UAV photography and noise reduction algorithms overcomes the disadvantages of poor sensor stability and high measurement cost in traditional methods by avoiding arranging sensors on the structure. At the same time, introducing a noise reduction algorithm effectively reduces the interference of mechanical vibration and environmental wind, improving the cable force monitoring accuracy, thereby providing more accurate and reliable prestress measurement results for cable-stayed reticulated shell structures.

[0105] For the acquisition of cable motion images, in addition to using UAVs, other types of sensors such as lidar (LiDAR) or ground sensors can be used to improve the accuracy of data acquisition. LiDAR technology is especially suitable for high-precision three-dimensional modeling at high altitudes or in complex terrains, helping to supplement the blind areas of UAV image data.

[0106] For the noise reduction processing of the original acceleration, different types of algorithms can be used, but the essence of the algorithm is to eliminate the environmental noise caused by UAV vibration and environmental wind.

[0107] For the extraction of cable lateral vibration, automatic image segmentation and recognition can be carried out by combining a convolutional neural network (CNN). Combine Canny edge detection with other traditional methods to improve the recognition accuracy, especially in the case of high noise and complex backgrounds.

[0108] By using UAVs for non-contact measurement, the present invention not only reduces the installation and maintenance costs of equipment, but also solves the problems of poor sensor stability and sensitivity to environmental changes in traditional methods. Especially in complex working environments, the present invention has significant advantages. The present invention can adapt to different types of cable-stayed reticulated shell structures, including old structures, high-altitude areas that are difficult to access, etc., providing a more efficient and accurate prestress measurement solution. This technology has strong universality and broad application prospects.

[0109] The above are only specific embodiments of the present invention, but the structural features of the present invention are not limited thereto. Any changes or modifications made by those skilled in the art within the scope of the present invention are covered by the patent scope of the present invention.

Claims

1. A method for measuring the prestress of a cable-stayed lattice shell based on drone photography and noise reduction algorithm, characterized in that The steps are as follows: Step 1: Use drone photography to collect images of the cable-stayed lattice shell components to be tested. The drone is equipped with high-precision camera equipment and can obtain high-definition images of the components to be tested by flying around the structure to collect images in all directions. Step 2: Extraction of cable acceleration time history based on Canny edge detection and centerline recognition; Convert the input image into a grayscale image; Step 3: Acceleration data denoising based on improved empirical mode decomposition; Step 4: Calculate the cable tension of the component to be tested. By deriving the mapping relationship between cable tension and vibration frequency, it is necessary to consider the basic vibration equation of the beam unit and analyze various factors therein, such as the influence of bending stiffness, support conditions and concentrated mass on the vibration frequency, and then introduce the influence of cable tension to finally derive the relationship between it and the vibration frequency.

2. The method for measuring prestress of a cable-stayed lattice shell based on drone photography and noise reduction algorithm according to claim 1 is characterized in that: The flight path of the UAV is set by analyzing the distribution of the cable-stayed lattice shell structure. During the flight, the image data is analyzed in real time to ensure that the UAV can accurately lock the cable and provide a clear basic image for subsequent data processing. During the hovering process, the camera equipment will record the target cable, especially the vibration process of the cable. The drone will record the dynamic response of the cable through a high frame rate video acquisition system to ensure that the tiny vibrations of the cable are captured.

3. The method for measuring prestress of a cable-stayed lattice shell based on drone photography and noise reduction algorithm according to claim 1 is characterized in that: The input image I(x, y) is converted to a grayscale image G(x, y) by weighted averaging the RGB channels; Use a Gaussian filter to smooth the image to reduce the impact of noise; edge detection is very sensitive to noise, and the smoothing process helps to remove small fluctuations in the image; use a 5×5 or larger Gaussian kernel to convolve the image to obtain a smoothed image; G smoothed (x, y) = G(x, y)*G kemel , where G kernel is the Gaussian kernel; Calculate the gradient of each pixel in the image. The gradient represents the magnitude and direction of the image brightness change. Use the Sobel operator to calculate the horizontal and vertical gradients. Calculate the horizontal gradient Gx and the vertical gradient Gy: Non-maximum suppression is used to remove noise from the gradient amplitude, retaining only the local maximum value to ensure that the detected edge is accurate; this process scans each pixel along the gradient direction and sets the non-maximum pixel points to 0; the gradient amplitude of each pixel is detected along the gradient direction, and those pixels that are not local maxima are suppressed.

4. The method for measuring prestress of a cable-stayed lattice shell based on drone photography and noise reduction algorithm according to claim 3 is characterized in that: Double threshold processing is used to determine edge strength and distinguish between edges and non-edges. By setting two thresholds, a high threshold and a low threshold; if the gradient magnitude of a pixel is greater than the high threshold, it is considered an edge pixel; if the gradient magnitude of a pixel is less than the low threshold, it is considered a non-edge pixel; if the gradient magnitude of a pixel is between the high threshold and the low threshold, and it is connected to a strong edge, it is considered an edge pixel; edge connection determines the final edge by tracing the pixels between the high threshold and the low threshold; Only those weak edge pixels that are connected to strong edges are retained; Perform edge connection, traverse all pixels, and add pixels that meet the conditions to the edge.

5. The method for measuring prestress of a cable-stayed lattice shell based on drone photography and noise reduction algorithm according to claim 4 is characterized in that: The center line of the image is extracted on the basis of edge detection; the pixels outside the edge area are removed, and only the central axis of the edge is retained; the dynamic displacement of the midpoint of the cable center line is obtained by comparing the pixel coordinate differences between different images; the central difference method is used to numerically differentiate the lateral displacement time history of the cable to obtain the acceleration time history of different positions of the cable to be tested.

6. The method for measuring prestress of a cable-stayed lattice shell based on drone photography and noise reduction algorithm according to claim 1 is characterized in that: EEMD, or integrated empirical mode decomposition, is used to decompose the acceleration time history data. White noise n(t) is added to the original acceleration signal a(t) to obtain a new signal: x new (t)=x(t)+αn(t); Where α is the noise amplitude and n(t) is white noise. new (t) Use the EMD algorithm to decompose and obtain several intrinsic mode functions -IMF: Among them, IMF i (t) is the i-th IMF, r N (t) is the residual signal - usually a trend term. For each set of noise amplitude and decomposition times, calculate its BIC value. The BIC value contains the balance between the model complexity - the number of IMFs and the model fitting error - the impact of noise; the noise amplitude α and the number of decompositions k with the smallest BIC value are selected as the optimal parameters; this step balances the noise level and model complexity by minimizing the BIC value, thereby ensuring the most accurate signal decomposition and the most effective noise reduction processing; the definition formula of BIC is as follows: BIC=nln(σ 2 )+cln(n); Where n is the number of samples (i.e., the number of time points of the signal); σ2 is the variance of the error, which indicates the fitting accuracy of the model; k is the number of free parameters of the model, which indicates the complexity of the model.

7. The method for measuring prestress of a cable-stayed lattice shell based on drone photography and noise reduction algorithm according to claim 6 is characterized in that: Through multiple noise addition and EMD decomposition, multiple IMFs are obtained, and the IMFs of each time are integrated to finally obtain a more stable IMF with less noise influence. For the multiple IMFs obtained by decomposition, principal component analysis (PCA) needs to be used for further noise reduction to extract the principal component that has the greatest impact on the cable vibration signal; M IMFs are obtained from EEMD, which form a matrix X: X=[IMF1(t) IMF2(t)…IMF M (t)]; Calculate the covariance matrix C of the IMF matrix X: In the formula, M is the number of IMFs. By performing eigenvalue decomposition on the covariance matrix C, the eigenvalues ​​and corresponding eigenvectors are obtained; all eigenvalues ​​are arranged in order from large to small, and the principal components with cumulative contributions exceeding 90% are selected; the cumulative contribution ξ is calculated as follows: Where k is the number of principal components selected; select the minimum k that can make the cumulative contribution exceed 90%, and then select the corresponding eigenvectors v1, v2, ..., v k ; Restore the selected principal components to obtain the denoised cable acceleration.

8. The method for measuring prestress of a cable-stayed lattice shell based on drone photography and noise reduction algorithm according to claim 1 is characterized in that: The vibration equation of the cable considering the bending stiffness can be expressed as: Where ω(x, t) is the deflection of the cable at position x; E, I, and m are the elastic modulus, moment of inertia, and mass per unit length of the cable, respectively; T is the axial force component of the cable at the equilibrium position, i.e., the cable force; after substituting different boundary conditions and geometric parameters of the cable, the following unified estimation formula for the cable force can be obtained: Where k1 and k2 are the characteristic parameters of the cable to be tested. The main parameters affecting the value of the characteristic parameters are the aspect ratio ψ and the boundary conditions. When ψ is greater than 150, the influence of the cable bending stiffness is ignored.

9. The method for measuring prestress of a cable-stayed lattice shell based on drone photography and noise reduction algorithm according to claim 8, characterized in that: Articulated boundary conditions: when the aspect ratio is 50<ψ≤150, the characteristic parameters k1=4 and k2=π; when the aspect ratio is ψ>150, the characteristic parameters k1=4 and k2=π; Fixed boundary conditions: when the aspect ratio is 50<ψ≤150, the characteristic parameters k1=3.74 and k2=-59.62; when the aspect ratio is ψ>150, the characteristic parameters k1=3.74 and k2=0.

Images