Vibration Source Testing and Identification Method and System Based on Vision System

Abstract

This application provides a vibration source testing and identification method and system based on a vision system, relating to the field of rail transit vibration monitoring technology. This application collects synchronous multi-view image sequences of vibration events of monitored objects such as key sections along the track and sensitive locations of buildings on the track. Then, it selects grating deformation images that are synchronous with the impact excitation and have a signal-to-noise ratio that meets the preset conditions from the image sequence to form an effective image subset. Through phase analysis and three-dimensional reconstruction, it obtains the displacement vector data of each monitoring point on the surface of the monitored object and maps it into physical space displacement time history data. It numerically processes the displacement time history data to generate acceleration time history and extracts visual vibration features. Combined with the vibration wave propagation model, it solves the visual vibration features in reverse to determine the spatial location, action time history, and frequency components of the vibration source, which can realize the accurate identification and parameter determination of vibration sources in the rail transit environment.

Description

Technical Field

[0001] This application relates to the field of vibration monitoring technology for rail transit, and in particular to a vibration source testing and identification method and system based on a vision system. Background Technology

[0002] Vibration source testing and identification methods are key technologies for ensuring the safety of engineering structures and reducing vibration hazards. In the rail transit sector, there is an urgent need for vibration monitoring in critical sections along the track and at sensitive locations of buildings on top of it. This method can effectively capture vibration patterns and locate vibration sources. Its application not only provides data support for structural maintenance but also reduces the impact of vibration on the surrounding environment, demonstrating broad practical application prospects.

[0003] Existing vibration source testing and identification methods based on vision systems mostly acquire vibration-related information by collecting image sequences of the monitored object, and have gained some application due to the advantage of non-contact monitoring. These methods typically extract vibration features after simple image processing, but in the complex environment of rail transit, they are susceptible to external interference, leading to poor image synchronization and insufficient signal-to-noise ratio. Furthermore, the control over phase analysis and spatial mapping during data processing is not rigorous enough, affecting the accuracy of subsequent vibration information.

[0004] Existing methods suffer from a lack of rigorous standards in image selection, making it difficult to obtain effective vibration image data and leading to errors in the calculation of key information such as displacement and acceleration at monitoring points. Furthermore, insufficient integration with the propagation characteristics of vibration waves results in significant deviations in the identification of parameters such as the spatial location and frequency components of the vibration source. These shortcomings directly affect the accuracy of vibration source identification and fail to meet the practical needs of high-precision monitoring in rail transit environments. Therefore, existing technologies suffer from insufficient accuracy and reliability in identifying vibration source parameters in rail transit environments. Summary of the Invention

[0005] The purpose of this application is to provide a vibration source testing and identification method and system based on a vision system, so as to solve the problem of insufficient accuracy and reliability of vibration source parameter identification in the existing technology in the rail transit environment.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides a vibration source testing and identification method based on a vision system, comprising:

[0007] For vibration events of monitored objects in the rail transit environment, a multi-view image sequence synchronized with the vibration events is acquired. The monitored objects include key sections along the track and sensitive locations of buildings on the track.

[0008] Multiple grating deformation images that are synchronized with the impact excitation and whose image signal-to-noise ratio meets the preset conditions are selected from the multi-view image sequence, and the multiple grating deformation images are combined into an effective image subset;

[0009] By performing phase analysis and three-dimensional reconstruction on the effective image subset, displacement vector data of each monitoring point on the surface of the monitored object is obtained, and the displacement vector data is mapped to displacement time history data of each monitoring point in physical space.

[0010] The displacement time history data is numerically processed to generate acceleration time histories corresponding to each monitoring point, and visual vibration features for characterizing the vibration source are extracted from the acceleration time histories.

[0011] By combining the propagation model of vibration waves in the structural medium, the visual vibration characteristics are solved in reverse to determine the vibration parameters of the vibration source. The vibration parameters include the spatial location of the vibration source, the duration of action, and the frequency components. The inverse solution includes wave direction estimation and propagation path deconvolution.

[0012] Optionally, it also includes:

[0013] The vibration parameters are compared with the collected reference vibration data. Based on the comparison results, the propagation model is optimized, and the optimized vibration parameters are output.

[0014] Optionally, the inverse solution of the visual vibration characteristics, based on the propagation model of vibration waves in the structural medium, to determine the vibration parameters of the vibration source includes:

[0015] Based on the structural medium properties of the monitored object, a vibration wave propagation model is established to show the vibration wave propagation from the vibration source to each monitoring point.

[0016] Based on the visual vibration characteristics, the direction of arrival is estimated using the vibration wave propagation model to determine the main propagation direction of the vibration wave, and the spatial location of the vibration source is estimated by combining the three-dimensional spatial location of each monitoring point.

[0017] Using the visual vibration characteristics as the system output, and based on the spatial location of the vibration source and the vibration wave propagation model, a deconvolution operation of the transmission path is performed to solve for the action time history and frequency components of the vibration source.

[0018] Optionally, the step of using the visual vibration characteristics as the system output and performing a transmission path deconvolution operation based on the spatial location of the vibration source and the vibration wave propagation model to solve for the action time history and frequency components of the vibration source includes:

[0019] Based on the spatial location of the vibration source and the vibration wave propagation model, the transmission path response of the vibration wave from the spatial location to each monitoring point is determined.

[0020] Based on the transmission path response, a system response function is constructed, which describes the relationship between the unit vibration input emitted from the spatial location and the expected visual vibration characteristics generated at each monitoring point.

[0021] The visual vibration characteristics actually obtained from all monitoring points are used as the known overall system output;

[0022] Based on the system response function, the overall system output is deconvolved to calculate the vibration input sequence of the vibration source.

[0023] The duration of the vibration source and its corresponding frequency components can be extracted from the vibration input sequence.

[0024] Optionally, the step of estimating the direction of arrival (ROA) using the vibration wave propagation model based on the visual vibration characteristics, determining the main propagation direction of the vibration wave, and estimating the spatial location of the vibration source by combining the three-dimensional spatial locations of each monitoring point includes:

[0025] The vibration arrival time information of each monitoring point is obtained from the visual vibration characteristics, and the vibration arrival time difference between every two monitoring points is calculated.

[0026] Based on the wave velocity in the vibration wave propagation model and the time difference of vibration arrival, calculate the distance difference of the vibration wave propagation path between every two monitoring points;

[0027] By combining the three-dimensional spatial location of each monitoring point with the distance difference of the vibration wave propagation path, the geometric relationship is solved to determine the propagation direction of the vibration wave and the spatial location of the vibration source.

[0028] Optionally, the step of numerically processing the displacement time history data to generate acceleration time histories corresponding to each monitoring point, and extracting visual vibration features from the acceleration time histories to characterize the vibration source, includes:

[0029] The displacement time history data of each monitoring point are differentiated to generate the acceleration time history of each monitoring point;

[0030] Waveform feature parameters are extracted from the acceleration time history, including the arrival time and waveform shape of the vibration at each monitoring point;

[0031] The acceleration time history is subjected to spectral transformation to extract spectral feature parameters;

[0032] The waveform feature parameters are combined with the spectral feature parameters to form the visual vibration feature.

[0033] Optionally, the step of obtaining displacement vector data of each monitoring point on the surface of the monitored object by performing phase analysis and three-dimensional reconstruction on the effective image subset, and mapping the displacement vector data to displacement time history data of each monitoring point in physical space, includes:

[0034] Phase analysis is performed on the grating deformation images in the effective image subset to obtain the phase data of each monitoring point at each acquisition time;

[0035] Based on the phase data, calculate the displacement of each monitoring point on the image plane at each acquisition time;

[0036] By combining the displacement from multiple perspectives and the spatial parameters of the visual acquisition system, the three-dimensional spatial position of each monitoring point at each acquisition time is reconstructed.

[0037] Based on the three-dimensional spatial position, calculate the displacement vector data of each monitoring point relative to the fixed reference system;

[0038] The displacement vector data are arranged in chronological order of acquisition time to generate displacement time history data for each monitoring point.

[0039] Secondly, this application provides a vibration source testing and identification system based on a vision system, comprising:

[0040] The acquisition module is used to acquire multi-view image sequences synchronized with vibration events of monitored objects in the rail transit environment. The monitored objects include key sections along the track and sensitive locations of covered buildings.

[0041] The filtering module is used to filter out multiple grating deformation images that are synchronized with the impact excitation and whose image signal-to-noise ratio meets the preset conditions from the multi-view image sequence, and to form an effective image subset from the multiple grating deformation images;

[0042] The analysis module is used to obtain displacement vector data of each monitoring point on the surface of the monitored object by performing phase analysis and three-dimensional reconstruction on the effective image subset, and to map the displacement vector data into displacement time history data of each monitoring point in physical space.

[0043] The generation module is used to perform numerical processing on the displacement time history data to generate acceleration time histories corresponding to each monitoring point, and to extract visual vibration features from the acceleration time histories to characterize the vibration source characteristics.

[0044] The solution module is used to perform inverse solution of the visual vibration characteristics by combining the propagation model of vibration waves in the structural medium, so as to determine the vibration parameters of the vibration source. The vibration parameters include the spatial location of the vibration source, the duration of action, and the frequency components. The inverse solution includes wave direction estimation and propagation path deconvolution.

[0045] Thirdly, this application provides an electronic device, comprising:

[0046] Memory, used to store computer programs;

[0047] A processor is configured to execute the computer program to implement the steps of the vision-based vibration source testing and identification method as described in the first aspect above.

[0048] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the vibration source testing and identification method based on a vision system as described in the first aspect above.

[0049] The vibration source testing and identification method based on a vision system provided in this application ensures comprehensive and synchronous vibration-related image data by acquiring synchronous multi-view image sequences of vibration events of monitored objects in the rail transit environment; it selects grating deformation images that are synchronized with impact excitation and meet the signal-to-noise ratio standard to form an effective subset, ensuring the data quality of subsequent analysis; it obtains the displacement vector data of the monitoring point through phase analysis and three-dimensional reconstruction and maps it to the physical space displacement time history, accurately capturing the spatial displacement information of the monitoring point; it performs numerical processing on the displacement time history data to generate acceleration time history and extracts visual vibration features to obtain key parameters characterizing the vibration source; and it combines the vibration wave propagation model to solve the visual vibration features in reverse, realizing the accurate identification of the spatial location, action time history, and frequency components of the vibration source.

[0050] Furthermore, a vibration wave propagation model is established based on the structural medium properties of the monitored object. Using this model, the direction of arrival is estimated based on visual vibration characteristics to determine the spatial location of the vibration source. Then, a deconvolution operation is performed between this spatial location and the vibration wave propagation model to obtain the vibration source's duration and frequency components. This method refines the inverse solution process for vibration source parameters, further improving the accuracy and reliability of identifying the vibration source's spatial location, duration, and frequency components, and ensuring the logical consistency and accuracy of the solution for key vibration source parameters. Attached Figure Description

[0051] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 A schematic flowchart of a vibration source testing and identification method based on a vision system provided in this application embodiment;

[0053] Figure 2 A schematic flowchart of another vibration source testing and identification method based on a vision system provided in this application embodiment;

[0054] Figure 3 This is a schematic diagram of a vibration source testing and identification system based on a vision system, provided as an embodiment of this application. Detailed Implementation

[0055] In vibration monitoring of rail transit, existing vibration source identification methods based on vision systems have significant shortcomings: the acquired images often lack synchronization with vibration events and are easily affected by environmental interference, resulting in poor image quality; there is a lack of strict standards when screening valid data; at the same time, the calculation of spatial displacement of monitoring points during data processing is not accurate enough, and the propagation law of vibration waves in the structure is not fully considered, ultimately leading to large deviations in the identification results of key information such as the location of the vibration source, vibration time, and frequency, which makes it difficult to meet the high-precision requirements of actual monitoring.

[0056] To address this issue, this application proposes a vibration source testing and identification method based on a vision system. The core of this method involves simultaneously acquiring vibration-related images from multiple perspectives, first selecting valid images with good synchronization and meeting quality standards, then accurately calculating the displacement and acceleration information of the monitoring points to extract vibration source features, and finally combining this with a vibration wave propagation model to deduce the vibration source parameters. This method ensures data quality through rigorous image selection and improves the accuracy of feature extraction and parameter solving by combining precise spatial displacement calculation with vibration wave propagation laws. It fundamentally solves the problem of insufficient accuracy in vibration source identification in existing technologies, enabling more reliable determination of key vibration source information and providing strong support for ensuring the structural safety of rail transit.

[0057] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0058] The core of this application is to provide a vibration source testing and identification method based on a vision system, and a flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes:

[0059] S101. For vibration events of monitored objects in the rail transit environment, acquire multi-view image sequences synchronized with the vibration events.

[0060] Among them, multi-view image sequences refer to continuous image data taken from different observation angles, which can comprehensively capture the appearance changes of the monitored object during vibration. The monitored objects are areas in the rail transit environment that require key attention to vibration impact, including key sections on the track that are susceptible to vibration, such as turnouts, curves, and vibration-sensitive locations in overpasses, such as the ground floor or floor slabs of buildings.

[0061] In one specific implementation, multiple high-definition industrial cameras are first reasonably arranged around the monitored object based on its distribution range and structural characteristics. The number of cameras arranged must meet the requirements of multi-view coverage and avoid monitoring blind spots. After the deployment is completed, all cameras are time-synchronized to ensure that the shooting time accuracy of each camera is consistent.

[0062] Meanwhile, vibration sensing triggering devices are set up in the monitoring area. When key sections along the track, such as turnouts and curves, vibrate due to train passage, or when sensitive locations of covered buildings, such as ground floor slabs and upper floors, vibrate, the vibration sensing triggering devices will send signals to simultaneously activate all calibrated high-definition industrial cameras, continuously capturing images of the vibration occurrence and development process, and finally obtaining a set of continuous image sequences that perfectly match the vibration event in time and cover the key parts of the monitored object in terms of perspective.

[0063] S102. Select multiple grating deformation images from the multi-view image sequence that are synchronized with the impact excitation and whose image signal-to-noise ratio meets the preset conditions, and form an effective image subset from the multiple grating deformation images.

[0064] Among them, impact excitation refers to the sudden force that causes vibration of the monitored object in a rail transit scenario, such as the impact force generated when a train passes through a key section of the track. Image signal-to-noise ratio (SNR) is the ratio of effective grating fringe information to interfering noise in an image; a higher ratio indicates more prominent useful information and better image quality. Grating deformation images are images showing the morphological changes of the grating fringes projected onto the surface of the monitored object due to vibration, directly reflecting the displacement changes caused by vibration. The effective image subset is a collection of grating deformation images that, after multiple rounds of screening, meet the requirements of synchronization and quality and can be used for subsequent vibration analysis.

[0065] In one specific implementation, this step first uses the time information synchronized with the vibration event in the multi-view image sequence to lock the target acquisition time period synchronized with the impact excitation, then extracts the image frames of each view within the time period, and sequentially filters out unqualified images by evaluating the clarity of grating stripes and filtering the intensity of brightness and darkness changes in non-striped areas, and finally determines the remaining images as grating deformation images and integrates them to form an effective image subset.

[0066] As an example, the acquisition time period is determined to be 10:00:00-10:00:30, based on the vibration start time and duration recorded by the vibration triggering device. This time period is completely synchronized with the vibration event, ensuring that the extracted images can accurately correspond to the occurrence and development of the vibration.

[0067] Secondly, all image frames captured by the five image acquisition devices during the acquisition period were extracted. Each device captured 50 frames per second, and each device acquired 1,500 images within 30 seconds. A total of 7,500 image frames from various perspectives were extracted from the five devices. These image frames completely covered the vibration process of the track curve section and the ground floor of the commercial building.

[0068] Next, an intuitive method for evaluating the sharpness of grating stripes was used to determine whether the edges of the grating stripes in each image frame were blurred, whether there were breaks, and whether the stripe details were distinguishable. After checking each image frame one by one, a total of 850 frames out of 7500 images were rejected due to blurred stripes and loss of details.

[0069] Then, for the remaining 6650 image frames that met the clarity standard, the brightness changes in the non-striped areas were evaluated to observe whether there were any sudden changes in brightness or drastic fluctuations in these areas. After checking frame by frame, a total of 350 frames out of the 6650 images were removed due to drastic changes in brightness in the non-striped areas.

[0070] Finally, the remaining 6300 image frames after two rounds of elimination and screening were identified as raster distorted images. Since this example only involves one acquisition time period, these raster distorted images were directly integrated to form an effective image subset.

[0071] The above example is only one example of this application. In practical applications, the duration of the data collection period can be flexibly adjusted according to the actual duration of the vibration event. This application does not limit this.

[0072] This application precisely locks in the acquisition time period synchronized with vibration, and uses an intuitive and reliable method of evaluating clarity and brightness changes to screen images. This effectively eliminates low-quality and interference-affected invalid data, ensuring the reliability and usability of the effective image subset, and providing high-quality data support for subsequent acquisition of displacement data at monitoring points and identification of vibration source parameters.

[0073] S103. By performing phase analysis and three-dimensional reconstruction on the effective image subset, displacement vector data of each monitoring point on the surface of the monitored object is obtained, and the displacement vector data is mapped to displacement time history data of each monitoring point in physical space.

[0074] Phase analysis is a technique for parsing and processing the fringe phase information in a grating deformation image, used to extract the phase change corresponding to the vibration at the monitoring point.

[0075] S103 specifically includes:

[0076] S1031. Perform phase analysis on the grating deformation images in the effective image subset to obtain the phase data of each monitoring point at each acquisition time.

[0077] Phase resolution is the core operation of phase analysis, which extracts the phase information of grating fringes in an image using specific techniques. Phase data is quantitative data characterizing changes in the position of grating fringes. Vibration at the monitoring point causes the surface grating fringes to shift, resulting in a corresponding change in the phase data, which can indirectly reflect the movement of the monitoring point.

[0078] S1032. Based on the phase data, calculate the displacement of each monitoring point on the image plane at each acquisition time.

[0079] The displacement of the image plane is the change in the two-dimensional position of the monitoring point in the image. This displacement can be used to preliminarily determine the movement range of the monitoring point and is the basis for subsequent three-dimensional spatial displacement calculations.

[0080] S1033. Combining the displacement from multiple perspectives with the spatial parameters of the visual acquisition system, reconstruct the three-dimensional spatial position of each monitoring point at each acquisition time.

[0081] Among them, spatial parameters are installation and performance-related parameters of the image acquisition equipment, including the spatial installation coordinates of the acquisition equipment, focal length, lens distortion parameters, etc. These parameters are used to establish the correspondence between the image plane and the physical space to ensure the accuracy of three-dimensional position reconstruction.

[0082] S1034 calculates the displacement vector data of each monitoring point relative to the fixed reference system based on the three-dimensional spatial position.

[0083] The fixed reference frame is a static, unchanging reference spatial coordinate system selected within the monitoring area. It is typically established with the fixed structure surrounding the monitored object as the origin and is used to accurately calculate the actual movement of the monitoring point. Displacement vector data includes the magnitude and spatial direction of the displacement, enabling comprehensive quantification of the vibration displacement of the monitoring point.

[0084] S1035. Arrange the displacement vector data in the order of acquisition time to generate displacement time history data for each monitoring point.

[0085] Among them, displacement time history data integrates displacement information in the time dimension. Through this data, the displacement change trend of the monitoring point at different times during the entire vibration event can be clearly understood.

[0086] In one specific implementation, phase analysis is first performed on the grating deformation image in the effective image subset to obtain the phase data of each monitoring point at each time. Then, the image plane displacement of the monitoring point is calculated based on the phase data. The three-dimensional spatial position is reconstructed by combining the multi-view displacement and the spatial parameters of the visual acquisition system. The displacement vector data is calculated based on the three-dimensional spatial position and the fixed reference frame. Finally, the displacement time history data is generated by sorting by time, thus fully realizing the conversion of image information into quantized displacement data.

[0087] As an example, phase analysis is first performed using the phase-shifting method in step S1031. This method extracts the phase data of each monitoring point at the corresponding acquisition time by analyzing the phase shift pattern of the stripes in the grating deformation image. In specific operation, for each frame of the grating deformation image in the effective image subset, a number of preset monitoring points are selected on the surface of the monitoring object. For example, 200 monitoring points are evenly selected on the surface of the track curve section, and 150 monitoring points are selected on the ground floor wall of the commercial building. The phase value of each monitoring point at the current acquisition time is analyzed by the phase-shifting method, and finally the phase data of all monitoring points at each acquisition time is obtained. In practical applications, the number of monitoring points can be adjusted according to the monitoring accuracy requirements, and this application does not limit this.

[0088] Next, in step S1032, the image plane displacement is calculated using the phase data difference at different acquisition times. The core is to realize the quantitative calculation based on the correspondence between phase change and image plane displacement. The commonly used algorithm formula is as follows (1):

[0089] (1)

[0090] In the formula, u represents the displacement of the monitoring point on the image plane, and the unit is pixels; is the phase difference between two adjacent acquisition times of the monitoring point, in radians; p is the period of the grating stripes on the image plane, in pixels. In this example, the grating stripe period p is set to 10 pixels.

[0091] Suppose that the phase value of a certain track monitoring point at time t1 is In radians, the phase value at time t2 is In radians, first calculate the phase difference. Radius, then substitute the value into equation (1) to calculate the plane displacement of the image, that is The image plane displacement of the monitoring point from time t1 to t2 is 10 pixels. Using this method, the image plane displacement of all monitoring points at each acquisition time can be calculated one by one.

[0092] Next, in step S1033, the spatial parameters of the visual acquisition system are pre-calibrated. The calibration process requires a standard calibration board, which is placed at different locations within the monitoring area. After acquiring multiple sets of calibration images, the parameter values ​​are calculated, including the spatial installation coordinates, lens focal length, and lens distortion parameters of each of the five image acquisition devices. In this example, the focal length of all acquisition devices is uniformly set to 50mm. Then, combining the planar displacement of the image at the same monitoring point acquired by each device, the mapping relationship between the image and physical space is established using the principle of multi-view triangulation. The planar displacement of the image is converted into three-dimensional coordinates in physical space, completing the reconstruction of the three-dimensional spatial position of each monitoring point at the corresponding acquisition time.

[0093] For example, monitoring point M on the ground floor wall of a building corresponds to 3 image acquisition devices with different viewing angles:

[0094] Camera A is installed on the left exterior wall of the building, with spatial installation coordinates of (10.0, 5.0, 3.0) meters;

[0095] Camera B is mounted on a bracket next to the track in front of the commercial building, with spatial installation coordinates of (15.0, 8.0, 3.0) meters;

[0096] Camera C is installed on the monitoring column on the right side of the commercial building, with spatial installation coordinates of (20.0, 5.0, 3.0) meters.

[0097] All three cameras have a focal length of 50mm. It is known from step S1032 that, at time t2, the planar displacement of the image acquired by camera A at monitoring point M is 8 pixels in the x-axis direction and 6 pixels in the y-axis direction, the planar displacement of the image acquired by camera B is 10 pixels in the x-axis direction and 5 pixels in the y-axis direction, and the planar displacement of the image acquired by camera C is 9 pixels in the x-axis direction and 7 pixels in the y-axis direction.

[0098] Then, combining the principles of triangulation, the spatial installation coordinates, focal length, and other parameters of the three cameras, along with their corresponding image plane displacement data, are substituted into the spatial mapping model. By calculating the correspondence between image coordinates and physical space coordinates under different viewpoints, the measurement bias of a single viewpoint is eliminated. Specifically, firstly, the image plane displacement (8, 6) pixels of camera A is converted into a direction vector in physical space based on its parameters. Then, combined with the direction vectors of cameras B and C, the intersection of the three direction vectors is found. This intersection point is the actual position of monitoring point M in physical space. Finally, its three-dimensional spatial position coordinates at time t2 are reconstructed as (2.5, 3.1, 4.8) meters, clearly showing the specific position of the monitoring point in three-dimensional space. The above example is only one example of this application. In practical applications, the specific values ​​of camera installation position and image plane displacement can be adjusted according to the site layout and monitoring situation. This application does not limit this.

[0099] Finally, in step S1035, the displacement vector data of each monitoring point at all acquisition times are sequentially arranged and integrated according to the acquisition time from earliest to latest, generating displacement time history data for each monitoring point. In this example, the acquisition device captures 50 frames per second, the vibration event lasts for 30 seconds, and each monitoring point corresponds to 600 acquisition times. The displacement vector data of the 600 acquisition times of the monitoring point are arranged in chronological order to form a complete displacement change sequence. Through this sequence, the magnitude and direction of displacement changes at each moment of the monitoring point from the start to the end of the vibration can be clearly seen. The specific values ​​of the spatial parameters of the above visual acquisition system are only examples. In actual applications, they can be flexibly calibrated according to the model of the acquisition device and the installation location.

[0100] In another specific implementation, phase analysis can use Fourier transform instead of phase shifting. This method does not require multiple phase-shifted images. It separates the fundamental frequency component of the stripes by performing Fourier transform on a single frame of grating deformation image, and then obtains the phase data of each monitoring point through inverse Fourier transform. It is suitable for scenarios of rapid acquisition of vibration events. In specific operation, the frequency filtering parameters of Fourier transform can be adjusted according to the image resolution and stripe density to ensure accurate extraction of phase data.

[0101] This application transforms grating deformation information in effective images into precisely quantified displacement data through a series of operations such as phase analysis, three-dimensional reconstruction, and data integration. This clearly presents the spatiotemporal variation characteristics of vibration displacement at the monitoring point, providing reliable digital data support for subsequent acceleration time history generation and vibration source characteristic extraction.

[0102] S104. Perform numerical processing on the displacement time history data to generate acceleration time histories corresponding to each monitoring point, and extract visual vibration features from the acceleration time histories to characterize the vibration source characteristics.

[0103] Among them, the acceleration time history is the acceleration data of the monitoring points arranged in chronological order, which can reflect the variation law of the intensity of vibration at the monitoring points. The stronger the vibration, the greater the acceleration value.

[0104] S104 specifically includes:

[0105] S1041. Differentiate the displacement time history data of each monitoring point to generate the acceleration time history of each monitoring point.

[0106] S1042. Extract waveform feature parameters from the acceleration time history, wherein the waveform feature parameters include the arrival time and waveform shape of the vibration at each monitoring point.

[0107] Among them, waveform morphology refers to the contour characteristics of the acceleration time history curve, such as pulse shape, sine shape, etc., and the waveform morphology caused by different vibration sources is significantly different.

[0108] S1043. Perform spectral transformation on the acceleration time history to extract spectral feature parameters.

[0109] S1044. Combine the waveform feature parameters with the spectral feature parameters to form the visual vibration feature.

[0110] Visual vibration features are a comprehensive set of features that integrate waveform and spectral characteristic parameters, and can fully characterize the core characteristics of the vibration source.

[0111] In one specific implementation, this step first performs differential processing on the displacement time history data to obtain the acceleration time history, then extracts waveform feature parameters from the acceleration time history, extracts spectral feature parameters through spectral transformation, and finally combines the two types of parameters to form visual vibration features that can characterize the vibration source characteristics, thus fully realizing the transformation from displacement data to vibration source features.

[0112] As an example, the central difference method is first used for differentiation in step S1041. This method calculates the displacement difference between two adjacent moments before and after a certain moment, and obtains the acceleration by combining the time interval. The core algorithm formula is shown in equation (2):

[0113] (2)

[0114] In the formula, The representative monitoring point at time The acceleration; , , These represent the monitoring points at different times. , , The displacement; This represents the time interval between two adjacent acquisition moments. In this example, the acquisition frame rate is 50 frames per second, therefore... =0.02 seconds.

[0115] Taking monitoring point M at the bottom of the building as an example, in its displacement time history data, time... Displacement Rice, at any time Displacement Rice, at any time Displacement Meters, substitute into equation (2) to calculate Acceleration at any moment: If at any time Displacement Meters, calculation Acceleration at any moment: By calculating the acceleration time history data of monitoring point M using this method, complete acceleration time history data can be obtained.

[0116] Next, waveform feature parameters are extracted in step S1042: for the vibration arrival time, the amplitude threshold is set to 50. By traversing the acceleration time history data of monitoring point M, it was found that at time... The acceleration reached 60 for the first time. The vibration arrival time at this monitoring point is determined to be [time value missing]. Regarding the waveform morphology, observing the acceleration time history curve reveals a pulse-shaped profile with a rapid rise followed by a slow fall, and the peak values ​​are concentrated within a single time period; therefore, the waveform morphology is determined to be pulse-type. Simultaneously, the acceleration time history curve at a monitoring point N within the track curve segment exhibits periodic sinusoidal fluctuations, with the vibration arrival time being... This completes the extraction of waveform characteristic parameters for all monitoring points.

[0117] Next, in step S1043, a Fast Fourier Transform (FFT) is used to perform a spectral transformation on the acceleration time history. This method can efficiently decompose the vibration signal in the time dimension into sinusoidal components of different frequencies, thereby obtaining the frequency composition of the signal. In specific operation, a FFT is performed on the acceleration time history data of monitoring point M. The acceleration time history data of M contains 1500 data points, and a spectrum diagram is obtained. The amplitude corresponding to each frequency is read from the spectrum diagram. It is found that the frequency with the highest amplitude is 20 Hz, and the frequency distribution is mainly concentrated in the 10-30 Hz range. Therefore, the main frequency of 20 Hz and the frequency distribution range of 10-30 Hz are extracted as the spectral characteristic parameters of this monitoring point. The spectrum diagram of monitoring point N shows that the main frequency is 15 Hz and the frequency distribution range is 5-25 Hz. Its spectral characteristic parameters are extracted in the same way.

[0118] Finally, in step S1044, the waveform characteristic parameters and spectral characteristic parameters of each monitoring point are combined: the visual vibration characteristics of monitoring point M are: vibration arrival time 10:00:00.10, waveform morphology pulse type, main frequency 20 Hz, frequency distribution range 10-30 Hz; the visual vibration characteristics of monitoring point N are: vibration arrival time 10:00:00.08, waveform morphology sinusoidal type, main frequency 15 Hz, frequency distribution range 5-25 Hz. Through this combination, the visual vibration characteristics of each monitoring point can comprehensively reflect the time, morphology, and frequency characteristics of the vibration, providing a complete basis for subsequent analysis of the vibration source. The specific algorithm parameters for the above spectral transformation can be adjusted according to the amount of data; this application does not limit this.

[0119] In another specific implementation, the differential processing can adopt the Simpson method, which obtains the acceleration by fitting the displacement data of three adjacent time points into a quadratic curve and then taking the derivative. This method can reduce the error caused by data noise and is suitable for scenarios with large fluctuations in displacement data. The spectrum transformation can use wavelet transform instead of fast Fourier transform. Wavelet transform can simultaneously reflect the time and frequency characteristics of the signal and is suitable for analyzing non-stationary vibration signals. The appropriate transformation method can be selected according to the stability of the vibration signal.

[0120] This application transforms displacement data into acceleration data, comprehensively extracts waveform and spectral features, and combines them to form visual vibration features, accurately capturing the core characteristics of the vibration source and providing comprehensive and reliable feature support for the subsequent reverse solution of vibration source parameters.

[0121] S105. Combining the propagation model of vibration waves in the structural medium, the visual vibration characteristics are solved in reverse to determine the vibration parameters of the vibration source. The vibration parameters include the spatial location of the vibration source, the duration of action, and the frequency components. The inverse solution includes wave direction estimation and propagation path deconvolution.

[0122] S105 specifically includes:

[0123] S1051. Based on the structural medium properties of the monitored object, establish a vibration wave propagation model from the vibration source to each monitoring point.

[0124] Among them, the structural medium properties are the physical characteristics of the materials constituting the monitored object, including the medium's elastic modulus, density, Poisson's ratio, etc. These properties directly determine the propagation speed and attenuation law of vibration waves within them. The core of the vibration wave propagation model is to establish a quantitative relationship between the vibration wave propagation time, propagation distance, and medium properties, ensuring an accurate description of the vibration propagation process from the source to the monitoring point.

[0125] S1052. Based on the visual vibration characteristics, the direction of arrival is estimated using the vibration wave propagation model to determine the main propagation direction of the vibration wave, and the spatial location of the vibration source is estimated by combining the three-dimensional spatial location of each monitoring point.

[0126] The main propagation direction of the vibration wave is the dominant direction in which vibration energy diffuses from the vibration source to the surrounding area. This direction can be used to initially determine the approximate location of the vibration source. The spatial location of the vibration source is the specific three-dimensional coordinates of the vibration occurrence and is one of the core parameters for vibration source identification.

[0127] S1052 specifically includes:

[0128] The vibration arrival time information of each monitoring point is obtained from the visual vibration features, and the vibration arrival time difference between every two monitoring points is calculated. Based on the wave velocity in the vibration wave propagation model and the vibration arrival time difference, the distance difference of the vibration wave propagation path between every two monitoring points is calculated. Combining the three-dimensional spatial position of each monitoring point and the distance difference of the vibration wave propagation path, the geometric relationship is solved to determine the propagation direction of the vibration wave and the spatial position of the vibration source.

[0129] Among them, the distance difference of the vibration wave propagation path is the difference in the propagation path length from the vibration source to the two monitoring points. It is obtained by the product of the wave speed and the time difference of arrival, and is the key geometric basis for solving the location of the vibration source.

[0130] S1053. Using the visual vibration characteristics as the system output, and based on the spatial location of the vibration source and the vibration wave propagation model, perform a deconvolution operation on the transmission path to solve for the action time history and frequency components of the vibration source.

[0131] The core of the transmission path deconvolution operation is to eliminate the distortion effect of the vibration wave during propagation and restore the original vibration state of the vibration source.

[0132] S1053 specifically includes:

[0133] Based on the spatial location of the vibration source and the vibration wave propagation model, the transmission path response of the vibration wave from the spatial location to each monitoring point is determined. According to the transmission path response, a system response function is constructed, which describes the relationship between the unit vibration input emitted from the spatial location and the expected visual vibration characteristics generated at each monitoring point. The visual vibration characteristics actually obtained from all monitoring points are used as the known overall system output. Based on the system response function, a deconvolution operation is performed on the overall system output to calculate the vibration input sequence of the vibration source. From the vibration input sequence, the action time history of the vibration source and its corresponding frequency components are extracted.

[0134] The propagation path response is a function describing the changes in parameters such as amplitude and phase during the propagation of the vibration wave from the vibration source to the monitoring point, reflecting the modifying effect of the propagation path on the vibration signal. The vibration input sequence is the original vibration amplitude data of the vibration source at each moment, and is the core component of the vibration source's operating time history.

[0135] In one specific implementation, such as Figure 2 As shown, a vibration wave propagation model is first established based on the structural medium properties of the monitored object. Then, the arrival time information in the visual vibration characteristics is used to determine the spatial location of the vibration source by estimating the direction of arrival. Finally, the visual vibration characteristics are used as the system output. The transmission path is deconvolved with the vibration source location and the propagation model to solve for the action time history and frequency components of the vibration source, thus realizing the complete reverse solution of the vibration source parameters.

[0136] As an example, a vibration wave propagation model is first established in step S1051: the structural medium of the track curve section in the monitored object is a steel rail, with a corresponding elastic modulus of 206 GPa and a density of 7850 g / L. The structural medium of the commercial building is reinforced concrete, with a corresponding elastic modulus of 30 GPa and a density of 2500 kJ / m³. By consulting the structural mechanics handbook, the vibration wave velocities of the two media were obtained as follows: the longitudinal wave velocity in the rail was 5900 m / s, and the longitudinal wave velocity in the concrete was 3200 m / s. Based on the spatial distribution of the monitoring points and the potential vibration source area, a propagation model was established, as shown in equation (3):

[0137] (3)

[0138] In the formula, d is the propagation distance of the vibration wave, v is the wave velocity of the corresponding medium, and t is the propagation time.

[0139] This model calculates the propagation distance by multiplying wave velocity by time, accurately describing the propagation law of vibration waves in two types of media. The above example is merely one example of this application. In practical applications, the structural medium properties can be obtained through on-site sampling and testing, and the wave velocity parameter can be adjusted according to the actual medium condition. This application does not impose any limitations on this.

[0140] Next, the direction of arrival and the spatial location of the vibration source are estimated in step S1052. The specific operation is as follows:

[0141] The first step is to extract the vibration arrival time of three key monitoring points from the visual vibration characteristics, namely the arrival time of monitoring point A (10.0, 8.0, 0.5) meters in the track curve section. The arrival time of monitoring point B at (15.0, 8.0, 0.5) meters is measured in seconds. The arrival time of the monitoring point C (12.0, 12.0, 3.0) meters after the roof is built. Seconds. Calculating the arrival time difference between any two monitoring points, the time difference between monitoring point A and monitoring point B is: The time difference between monitoring point A and monitoring point C is The time difference between monitoring point B and monitoring point C is Second.

[0142] The second step is to calculate the distance difference of each vibration wave propagation path based on the wave velocity in the propagation model:

[0143] Monitoring points A and B are located on the rail medium, corresponding to wave velocity =5900m / s, substituting into equation (3) yields the distance difference. rice;

[0144] Monitoring points A and C are located in the rail and concrete media respectively. The weighted average of the wave velocities of the two media is taken. m / s, substituting into equation (3) yields the distance difference. rice;

[0145] The same applies to monitoring points B and C; wave velocity... m / s, substituting into equation (3) yields the distance difference. rice.

[0146] The third step involves solving the geometric relationship by combining the three-dimensional coordinates of the monitoring points and the distance difference: A local coordinate system is established with monitoring point A as the origin. A hyperbola equation is constructed based on the distance difference condition. The trajectory of a point whose distance difference to two fixed points is constant is a hyperbola. The three monitoring points correspond to two sets of hyperbolas, and the intersection of the hyperbolas is the spatial location of the vibration source. Through geometric calculations, the three-dimensional coordinates of the vibration source are obtained as (8.0, 8.0, 0.5) meters. Simultaneously, based on the direction of the lines connecting the vibration source to each monitoring point, the main propagation direction of the vibration wave is determined to be radiating outwards from (8.0, 8.0, 0.5) meters, consistent with the direction of the track extension.

[0147] Next, in step S1053, a deconvolution operation of the transmission path is performed to solve for the time history and frequency components of the vibration source. The specific operation is as follows:

[0148] The first step, based on the spatial location of the vibration source (8.0, 8.0, 0.5 meters) and the propagation model, is to calculate the propagation distance from the vibration source to each monitoring point. The propagation distance from the vibration source to monitoring point A is... The propagation distance from meter to monitoring point B is [distance missing]. The propagation distance from meter to monitoring point C is Meters. Combining this with the attenuation coefficients of the corresponding media—0.001 dB per meter for rail and 0.002 dB per meter for concrete—the transmission path response function for each monitoring point is obtained. The transmission path response function is shown in equation (4):

[0149] (4)

[0150] In the formula, d is the propagation distance from the vibration source to the monitoring point. The attenuation coefficient of the structural medium at the monitoring point characterizes the rate of energy loss during the propagation of the vibration wave. Specifically, in this example, the attenuation coefficient of the rail is... Its value is 0.001 dB per meter, and the concrete attenuation coefficient is... Its value is 0.002 decibels per meter. This is a unit impulse function used to represent the instantaneous propagation characteristics of vibration waves. This is the time it takes for the vibration wave to travel from the vibration source to the monitoring point.

[0151] In this example, substituting the relevant information of each monitoring point into equation (4), the transmission path response function of each monitoring point is as follows:

[0152] The transmission path response function of monitoring point A is ;

[0153] The transmission path response function of monitoring point B is ;

[0154] The transmission path response function of monitoring point C is .

[0155] The second step is to construct the system response function by combining the transmission path response functions of the three monitoring points to obtain the overall system response function. For example, by combining the transmission path response functions of monitoring points A, B, and C, the overall system response function is obtained, as shown in equation (5):

[0156] (5)

[0157] Equation (5) describes the combination of visual vibration characteristics generated at three monitoring points after a unit vibration input is emitted from the vibration source.

[0158] The third step is to determine the overall system output. The overall system output is the set of vibration characteristics actually collected from all monitoring points. The visual vibration characteristics of each monitoring point need to be integrated according to a unified time dimension to form a matrix-based overall system output. For example, the visual vibration characteristics of monitoring points A, B, and C are integrated to form the overall system output, as shown in equation (6):

[0159] (6)

[0160] In the formula, For the overall system output, , , These are the acceleration time history features corresponding to monitoring points A, B, and C, respectively. Each feature contains complete time series and frequency information, providing a known output data basis for deconvolution operations. It should be noted that the core information of visual vibration features comes from acceleration time history, and is presented in the form of acceleration time history during integration, i.e., the acceleration time history feature in equation (6). This is to meet the requirements of subsequent deconvolution operations for continuous time series data.

[0161] The fourth step is to perform deconvolution operation, using the Wiener filtering deconvolution algorithm, and solve the vibration input sequence of the vibration source through equation (7). Equation (7) is shown below:

[0162] (7)

[0163] In the formula, for The conjugate function, This is a regularization parameter with a value of 0.001, used to prevent computational divergence. This algorithm eliminates the effects of attenuation and delay during propagation, restoring the time sequence of the original vibration amplitude of the vibration source.

[0164] The fifth step is to analyze the time history and frequency components of the vibration input sequence. The effective vibration period was extracted and found to start at 10:00:00:00 and end at 10:00:00:30, lasting 30 seconds, which is the duration of the vibration source. Spectral analysis revealed that the main frequencies were concentrated in the 15-25 Hz range, with the highest amplitude at 20 Hz, which is the main frequency component of the vibration source. The secondary frequencies included 18 Hz and 22 Hz.

[0165] The regularization parameters of the above deconvolution algorithm are only examples. In practical applications, they can be adjusted according to the signal and noise levels. This application does not limit this.

[0166] In another specific implementation, the finite element method can be used to establish the vibration wave propagation model. By dividing the structure of the monitored object into meshes, the reflection and refraction laws of vibration waves at different medium interfaces can be simulated, making the model more consistent with the actual propagation of complex structures. The deconvolution operation of the transmission path can be performed using a blind deconvolution algorithm, which does not require prior knowledge of the precise expression of the transmission path response. The vibration source input can be deduced by analyzing the statistical characteristics of the system output, which is suitable for scenarios with complex structural medium properties and where it is difficult to accurately model the transmission path.

[0167] This application establishes a propagation model by combining the characteristics of the structural medium, and achieves inverse solution by integrating wave direction estimation and propagation path deconvolution, accurately determining the spatial location, action time history and frequency components of the vibration source, providing reliable technical support for the accurate identification and control of vibration sources in rail transit.

[0168] S106. The vibration parameters are compared with the collected reference vibration data. Based on the comparison results, the propagation model is optimized by feedback, and the optimized vibration parameters are output.

[0169] The reference vibration data is baseline data with known and accurate vibration information, collected in advance using standard monitoring equipment under the same monitoring scenario in rail transit. This data includes the spatial location, duration, and frequency components of typical vibration sources, and is used to verify and calibrate the vibration parameters obtained in this identification. Feedback optimization refers to the process of adjusting the key parameters of the vibration wave propagation model in reverse, based on the differences between the vibration parameters and the reference vibration data, to make the vibration parameters output by the model closer to the actual values.

[0170] In one specific implementation, reference vibration data under different train operating conditions is first collected and archived in advance using standard vibration monitoring instruments in typical scenarios such as curved sections and sensitive locations of covered buildings on the same rail transit line.

[0171] Next, the vibration parameters obtained in step S105 are compared item by item with the archived reference vibration data, and the deviation between the two is calculated. If the deviation exceeds the preset reasonable range, the key parameters such as wave velocity and attenuation coefficient in the propagation model are adjusted according to the magnitude of the deviation; if the deviation is within the reasonable range, no adjustment of the model is required.

[0172] After model optimization, the vibration parameters are recalculated using the optimized propagation model, and the optimized vibration parameters are finally output to ensure the accuracy and reliability of the identification results.

[0173] Figure 3 This is a schematic diagram of a specific implementation of a vibration source testing and identification system based on a vision system provided in this application. (Refer to...) Figure 3 The system may include:

[0174] The acquisition module 31 is used to acquire a multi-view image sequence synchronized with the vibration event of the monitored object in the rail transit environment. The monitored object includes key sections along the track and sensitive locations of the covered buildings.

[0175] The filtering module 32 is used to filter out multiple grating deformation images that are synchronized with the impact excitation and whose image signal-to-noise ratio meets the preset conditions from the multi-view image sequence, and to form an effective image subset from the multiple grating deformation images;

[0176] Analysis module 33 is used to obtain displacement vector data of each monitoring point on the surface of the monitored object by performing phase analysis and three-dimensional reconstruction on the effective image subset, and to map the displacement vector data into displacement time history data of each monitoring point in physical space;

[0177] The generation module 34 is used to perform numerical processing on the displacement time history data to generate the acceleration time history corresponding to each monitoring point, and extract visual vibration features from the acceleration time history to characterize the vibration source characteristics.

[0178] The solver module 35 is used to combine the propagation model of vibration waves in the structural medium to perform inverse solution of the visual vibration characteristics in order to determine the vibration parameters of the vibration source. The vibration parameters include the spatial location of the vibration source, the duration of action, and the frequency components. The inverse solution includes wave direction estimation and propagation path deconvolution.

[0179] The vibration source testing and identification system based on the vision system in this application is used to implement the aforementioned vibration source testing and identification method based on the vision system. Therefore, the specific implementation of the vibration source testing and identification system based on the vision system can be found in the embodiment section of the vibration source testing and identification method based on the vision system above. The specific implementation can be referred to the description of the corresponding embodiment, and will not be repeated here.

[0180] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described vision-based vibration source testing and identification methods.

[0181] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described vibration source testing and identification methods based on a vision system.

[0182] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0183] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above embodiments of the vibration source testing and identification method based on a vision system.

[0184] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0185] The above provides a detailed description of the vibration source testing and identification method and system based on a vision system provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A vibration source testing and identification method based on a vision system, characterized in that, include: For vibration events of monitored objects in the rail transit environment, a multi-view image sequence synchronized with the vibration events is collected. The monitored objects include key sections along the track and sensitive locations of the superstructure. Multiple grating deformation images that are synchronized with the impact excitation and whose image signal-to-noise ratio meets the preset conditions are selected from the multi-view image sequence, and the multiple grating deformation images are combined into an effective image subset; The preset conditions include: filtering out unqualified images by evaluating the clarity of grating stripes and the intensity of brightness changes in non-striped areas, and determining the remaining images as grating deformation images; the grating deformation image is an image in which the shape of the grating stripes projected on the surface of the monitored object changes due to vibration. By performing phase analysis and three-dimensional reconstruction on the effective image subset, displacement vector data of each monitoring point on the surface of the monitored object is obtained, and the displacement vector data is mapped to displacement time history data of each monitoring point in physical space. The displacement time history data is numerically processed to generate acceleration time histories corresponding to each monitoring point, and visual vibration features for characterizing the vibration source are extracted from the acceleration time histories. The extraction of visual vibration features from the acceleration time history to characterize the vibration source includes: Waveform feature parameters are extracted from the acceleration time history, including the arrival time and waveform shape of the vibration at each monitoring point; The acceleration time history is subjected to spectral transformation to extract spectral feature parameters; The waveform feature parameters and the spectral feature parameters are combined to form the visual vibration feature; By combining the propagation model of vibration waves in the structural medium, the visual vibration characteristics are solved in reverse to determine the vibration parameters of the vibration source. The vibration parameters include the spatial location of the vibration source, the duration of action, and the frequency components. The inverse solution includes wave direction estimation and propagation path deconvolution.

2. The method according to claim 1, characterized in that, Also includes: The vibration parameters are compared with the collected reference vibration data. Based on the comparison results, the propagation model is optimized, and the optimized vibration parameters are output.

3. The method according to claim 1, characterized in that, The aforementioned propagation model of vibration waves in the structural medium is used to inversely solve the visual vibration characteristics to determine the vibration parameters of the vibration source, including: Based on the structural medium properties of the monitored object, a vibration wave propagation model is established to show the vibration wave propagation from the vibration source to each monitoring point. Based on the visual vibration characteristics, the direction of arrival is estimated using the vibration wave propagation model to determine the main propagation direction of the vibration wave, and the spatial location of the vibration source is estimated by combining the three-dimensional spatial location of each monitoring point. Using the visual vibration characteristics as the system output, and based on the spatial location of the vibration source and the vibration wave propagation model, a deconvolution operation of the transmission path is performed to solve for the action time history and frequency components of the vibration source.

4. The method according to claim 3, characterized in that, The system outputs the visual vibration features and performs a deconvolution operation on the transmission path based on the spatial location of the vibration source and the vibration wave propagation model to obtain the time history and frequency components of the vibration source, including: Based on the spatial location of the vibration source and the vibration wave propagation model, the transmission path response of the vibration wave from the spatial location to each monitoring point is determined. Based on the transmission path response, a system response function is constructed, which describes the relationship between the unit vibration input emitted from the spatial location and the expected visual vibration characteristics generated at each monitoring point. The visual vibration characteristics actually obtained from all monitoring points are used as the known overall system output; Based on the system response function, the overall system output is deconvolved to calculate the vibration input sequence of the vibration source. The duration of the vibration source and its corresponding frequency components can be extracted from the vibration input sequence.

5. The method according to claim 3, characterized in that, The step of estimating the direction of arrival (ROA) of the vibration wave based on the visual vibration characteristics using the vibration wave propagation model, determining the main propagation direction of the vibration wave, and estimating the spatial location of the vibration source by combining the three-dimensional spatial locations of each monitoring point includes: The vibration arrival time information of each monitoring point is obtained from the visual vibration characteristics, and the vibration arrival time difference between every two monitoring points is calculated. Based on the wave velocity in the vibration wave propagation model and the time difference of vibration arrival, calculate the distance difference of the vibration wave propagation path between every two monitoring points; By combining the three-dimensional spatial location of each monitoring point with the distance difference of the vibration wave propagation path, the geometric relationship is solved to determine the propagation direction of the vibration wave and the spatial location of the vibration source.

6. The method according to claim 1, characterized in that, The step of numerically processing the displacement time history data to generate the acceleration time history corresponding to each monitoring point includes: The displacement time history data of each monitoring point are differentiated to generate the acceleration time history of each monitoring point.

7. The method according to claim 1, characterized in that, The step of performing phase analysis and three-dimensional reconstruction on the effective image subset to obtain displacement vector data of each monitoring point on the surface of the monitored object, and mapping the displacement vector data to displacement time history data of each monitoring point in physical space, includes: Phase analysis is performed on the grating deformation images in the effective image subset to obtain the phase data of each monitoring point at each acquisition time; Based on the phase data, calculate the displacement of each monitoring point on the image plane at each acquisition time; By combining the displacement from multiple perspectives and the spatial parameters of the vision system, the three-dimensional spatial position of each monitoring point at each acquisition time is reconstructed. Based on the three-dimensional spatial position, calculate the displacement vector data of each monitoring point relative to the fixed reference system; The displacement vector data are arranged in chronological order of acquisition time to generate displacement time history data for each monitoring point.

8. A vibration source testing and identification system based on a vision system, characterized in that, include: The acquisition module is used to acquire multi-view image sequences synchronized with vibration events of monitored objects in the rail transit environment. The monitored objects include key sections along the track and sensitive locations of the superstructure. The filtering module is used to filter out multiple grating deformation images that are synchronized with the impact excitation and whose image signal-to-noise ratio meets the preset conditions from the multi-view image sequence, and to form an effective image subset from the multiple grating deformation images; The preset conditions include: filtering out unqualified images by evaluating the clarity of grating stripes and the intensity of brightness changes in non-striped areas, and determining the remaining images as grating deformation images; the grating deformation image is an image in which the shape of the grating stripes projected on the surface of the monitored object changes due to vibration. The analysis module is used to obtain displacement vector data of each monitoring point on the surface of the monitored object by performing phase analysis and three-dimensional reconstruction on the effective image subset, and to map the displacement vector data into displacement time history data of each monitoring point in physical space. The generation module is used to perform numerical processing on the displacement time history data to generate acceleration time histories corresponding to each monitoring point, and to extract visual vibration features from the acceleration time histories to characterize the vibration source characteristics. The extraction of visual vibration features from the acceleration time history to characterize the vibration source includes: Waveform feature parameters are extracted from the acceleration time history, including the arrival time and waveform shape of the vibration at each monitoring point; The acceleration time history is subjected to spectral transformation to extract spectral feature parameters; The waveform feature parameters and the spectral feature parameters are combined to form the visual vibration feature; The solution module is used to perform inverse solution of the visual vibration characteristics by combining the propagation model of vibration waves in the structural medium, so as to determine the vibration parameters of the vibration source. The vibration parameters include the spatial location of the vibration source, the duration of action, and the frequency components. The inverse solution includes wave direction estimation and propagation path deconvolution.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the vibration source testing and identification method based on a vision system as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the implementation of the vibration source testing and identification method based on a vision system as described in any one of claims 1 to 7.

Images