Bridge displacement monitoring method and system based on image recognition

Abstract

This invention discloses a bridge displacement monitoring method and system based on image recognition. The method involves deploying markers at key locations on the bridge structure, each marker including at least an infrared LED and an infrared reflective sticker. Multiple image sequences of the markers are acquired using an image acquisition device, and the light spot regions of these image sequences are fitted with a Gaussian distribution to extract the coordinates of the light spot center points. Anomaly detection is used to remove light spots whose center point coordinates do not conform to the Gaussian distribution characteristics, thus obtaining the target light spot center point coordinates. Based on the pixel position changes of the target light spot center point coordinates across different image frames, pixel displacement is calculated. This pixel displacement is converted into actual bridge deflection using an imaging geometric model, enabling the capture of the dynamic displacement process of the bridge beam and real-time output of monitoring results, facilitating timely monitoring of the bridge structure's health status by staff.

Description

Technical Field

[0001] This invention relates to the field of bridge displacement monitoring technology, and in particular to a bridge displacement monitoring method and system based on image recognition. Background Technology

[0002] As a core component of transportation infrastructure, the structural stability of bridges directly affects traffic safety and public property safety. Displacement and deflection monitoring are crucial means to assess the health of bridge structures and prevent structural damage and accidents. With the increasing service life of bridges and the continuous increase in traffic load, bridge beams are prone to deformation such as bending and displacement. If monitoring is not timely or the data is inaccurate, it may lead to serious consequences such as beam cracking, bearing damage, or even bridge collapse. Therefore, it is necessary to establish a high-precision, high-stability displacement monitoring system. Currently, the mainstream bridge displacement monitoring methods are mainly divided into two categories: contact and non-contact. Contact monitoring methods, such as strain gauges and displacement meters, require the monitoring equipment to be directly fixed to the bridge structure, which presents problems such as complex installation, some damage to the bridge structure, and high maintenance costs. Furthermore, in complex outdoor environments, they are easily affected by vibration, temperature, and humidity, making it difficult to guarantee monitoring accuracy. Among non-contact monitoring methods, GPS monitoring is greatly affected by satellite signal interference and cannot function properly in areas obstructed by bridges, making it difficult to meet sub-millimeter accuracy requirements. Traditional image recognition monitoring methods often use ordinary visible light markers, which are easily affected by changes in outdoor lighting and weather, resulting in blurred light spot extraction and a lack of effective abnormal light spot removal mechanisms. This leads to large coordinate extraction errors and makes it impossible to achieve accurate monitoring of dynamic displacement, thus reducing the accuracy of monitoring results. Summary of the Invention

[0003] The purpose of this invention is to solve the above problems by designing a bridge displacement monitoring method and system based on image recognition.

[0004] To achieve the above objectives, the technical solution of the present invention further includes the following steps in the above-mentioned bridge displacement monitoring method based on image recognition:

[0005] Markers are placed at key locations on the bridge structure, and the markers include at least infrared LED lights and infrared reflective stickers.

[0006] Multiple image sequences of the marker are acquired using an image acquisition device, and the light spot regions of the multiple image sequences are fitted with a Gaussian distribution to extract the coordinates of the center point of the light spot.

[0007] By detecting abnormal light spots, the center point coordinates of the target light spot are obtained by eliminating light spots whose center point coordinates do not conform to the Gaussian distribution characteristics.

[0008] The pixel displacement is calculated based on the pixel position changes of the target spot center point coordinates between different image frames;

[0009] The pixel displacement is converted into actual bridge deflection using an imaging geometry model.

[0010] Furthermore, in the aforementioned image recognition-based bridge displacement monitoring method, the placement of markers at key locations on the bridge structure, wherein the markers include at least infrared LED lights and infrared reflective stickers, comprising:

[0011] According to the bridge structural design drawings and displacement monitoring specifications, markers are set up at the monitoring points in the mid-span, support, beam end and web of the bridge beam, and the spacing of the markers is set according to the bridge span and the field of view of the image acquisition equipment.

[0012] The images were captured using an image acquisition device to confirm that each marker formed an independent bright area, with no overlapping, blurred, or offset light spots.

[0013] Furthermore, in the aforementioned bridge displacement monitoring method based on image recognition, the step of acquiring multiple image sequences of the marker using an image acquisition device, and performing Gaussian distribution fitting on the spot regions of the multiple image sequences to extract the coordinates of the spot center point includes:

[0014] The industrial-grade camera equipment was fixed on a stable bracket on the side of the bridge monitoring, the camera's intrinsic parameters were calibrated, and the camera shooting parameters were set according to the outdoor lighting conditions, using grayscale imaging mode.

[0015] The camera is activated to continuously and in real time capture images of the deployed markers, obtaining a sequence of grayscale images arranged in chronological order.

[0016] The pixel information of each frame in the grayscale image sequence is stored in the form of a two-dimensional array. The pixel grayscale values ​​are assigned according to industry standards, and basic noise reduction processing is performed on the pixel array to remove isolated grayscale abrupt changes caused by imaging noise, resulting in multiple image sequences.

[0017] A grayscale threshold is set, and threshold filtering is performed on multiple preprocessed image sequences to remove background pixels with grayscale values ​​less than the grayscale threshold, while retaining the bright pixels formed by the markers. The retained bright pixels are then segmented using a connected component analysis algorithm to divide the light spot region corresponding to each marker.

[0018] Furthermore, in the aforementioned bridge displacement monitoring method based on image recognition, the step of acquiring multiple image sequences of the marker using an image acquisition device, and performing Gaussian distribution fitting on the spot regions of the multiple image sequences to extract the coordinates of the spot center point includes:

[0019] For each segmented light spot region, a two-dimensional Gaussian distribution mathematical model is used for fitting to extract the coordinates of the light spot center point. The formula for the two-dimensional Gaussian distribution mathematical model is as follows:

[0020]

[0021] in, Represents the pixel grayscale value. Indicates the Gaussian peak amplitude. Indicates the coordinates of the center point of the light spot. This represents the standard deviation of a Gaussian distribution in the x and y directions. Background grayscale;

[0022] The least squares method is used to solve the parameters of the two-dimensional Gaussian distribution mathematical model, and the gray-level distribution characteristics of the light spot region are fitted to obtain the coordinates of the center point of each light spot. .

[0023] Furthermore, in the aforementioned image recognition-based bridge displacement monitoring method, the step of eliminating light spots whose center point coordinates do not conform to Gaussian distribution characteristics through light spot anomaly detection to obtain the target light spot center point coordinates includes:

[0024] The index values ​​of the center point coordinates of the light spot are set as the fitting residual, Gaussian peak amplitude, and standard deviation in the x and y directions under normal Gaussian distribution. The index values ​​of the center point coordinates of the light spot are compared with the threshold range.

[0025] If any indicator exceeds the effective threshold range, the spot is determined to be an abnormal spot. The coordinates of the center point of the abnormal spot are marked and deleted to obtain the coordinates of the center point of the target spot.

[0026] Furthermore, in the aforementioned bridge displacement monitoring method based on image recognition, the step of calculating the pixel displacement based on the pixel position change of the target spot center point coordinates between different image frames includes:

[0027] The first frame of the image sequence is set as the initial reference frame, and the initial coordinates of the center points of all target light spots in the initial reference frame are extracted. Establish a coordinate database according to the spot number to complete the initial position calibration;

[0028] Extract the real-time coordinates of the target spot center point in any subsequent time frame of the image sequence. Inter-frame coordinate matching is performed based on the spot number;

[0029] For each successfully matched target spot, calculate its pixel displacement in the x and y directions, using the following formula:

[0030]

[0031]

[0032] in, , These are the pixel displacements in the x and y directions, respectively. Positive values ​​indicate movement along the positive axis, while negative values ​​indicate movement along the negative axis. The calculated pixel displacement values ​​are checked for reasonableness, and abrupt pixel displacement values ​​caused by inter-frame matching errors are removed.

[0033] Furthermore, in the above-mentioned bridge displacement monitoring method based on image recognition, the step of converting the pixel displacement into actual bridge deflection through an imaging geometric model includes:

[0034] Using an imaging geometry model, the pixel displacement values ​​in the x and y directions are converted into preliminary bridge deflection of the bridge structure, as shown in the following model formula:

[0035]

[0036] in, This indicates the initial actual displacement of the bridge structure. Indicates the camera pixel size. Indicates the equivalent focal length of the lens;

[0037] If the camera is shooting from a non-perpendicular angle, the extrinsic parameter matrix is ​​obtained through camera extrinsic parameter calibration. Substituting the initial bridge deflection into the correction formula, the pixel displacement direction is corrected to obtain the actual bridge deflection in the direction of bridge beam deformation. The correction formula is as follows:

[0038]

[0039] in, This represents the actual bridge deflection, which is a vector value and includes both the magnitude and direction of the displacement.

[0040] Furthermore, in an image recognition-based bridge displacement monitoring system, the fiber optic-based telephone communication system includes the following modules:

[0041] The marker placement module is used to place markers at key locations on the bridge structure. The markers include at least infrared LED lights and infrared reflective stickers.

[0042] The spot coordinate extraction module is used to acquire multiple image sequences of the marker through an image acquisition device, and to perform Gaussian distribution fitting on the spot regions of the multiple image sequences to extract the coordinates of the center point of the spot.

[0043] The spot coordinate filtering module is used to remove spots whose center point coordinates do not conform to the Gaussian distribution characteristics by spot anomaly detection, and obtain the center point coordinates of the target spot.

[0044] The pixel displacement calculation module is used to calculate the pixel displacement based on the pixel position change between different image frames according to the coordinates of the center point of the target spot;

[0045] The bridge deflection calculation module is used to convert the pixel displacement into actual bridge deflection through an imaging geometry model.

[0046] Furthermore, in an image recognition-based bridge displacement monitoring system, the pixel displacement calculation module includes the following sub-modules:

[0047] The extraction submodule is used to set the first frame of the image sequence as the initial reference frame and extract the initial coordinates of the center points of all target spots in the initial reference frame. Establish a coordinate database according to the spot number to complete the initial position calibration;

[0048] The matching submodule is used to extract the real-time coordinates of the center point of the target spot in any subsequent time frame of the image sequence. Inter-frame coordinate matching is performed based on the spot number;

[0049] The calculation submodule is used to calculate the pixel displacement in the x and y directions for each successfully matched target spot. The calculation formula is as follows:

[0050]

[0051]

[0052] in, , These are the pixel displacements in the x and y directions, respectively. Positive values ​​indicate movement along the positive axis, while negative values ​​indicate movement along the negative axis. The calculated pixel displacement values ​​are checked for reasonableness, and abrupt pixel displacement values ​​caused by inter-frame matching errors are removed.

[0053] Furthermore, in an image recognition-based bridge displacement monitoring system, the bridge deflection calculation module includes the following sub-modules:

[0054] The conversion submodule is used to convert the pixel displacement values ​​in the x and y directions into preliminary bridge deflection of the bridge structure using the imaging geometry model. The model formula is as follows:

[0055]

[0056] in, This indicates the initial actual displacement of the bridge structure. Indicates the camera pixel size. Indicates the equivalent focal length of the lens;

[0057] The correction submodule is used to obtain the extrinsic parameter matrix through camera extrinsic parameter calibration if the camera is shooting from a non-perpendicular orthophoto angle. Substituting the initial bridge deflection into the correction formula, the pixel displacement direction is corrected to obtain the actual bridge deflection in the direction of bridge beam deformation. The correction formula is as follows:

[0058]

[0059] in, This represents the actual bridge deflection, which is a vector value and includes both the magnitude and direction of the displacement.

[0060] Its beneficial effects are as follows: 1. High monitoring accuracy, achieving sub-pixel level coordinate extraction. This method uses a combination of infrared LED lights and infrared reflective stickers to deploy markers, effectively avoiding interference from outdoor lighting changes and weather, ensuring clear light spot imaging; by fitting a two-dimensional Gaussian distribution and using the least squares method to solve for the coordinates of the light spot center point, sub-pixel level extraction is achieved. Combined with an abnormal light spot detection mechanism, invalid coordinates with excessively large fitting residuals or those not conforming to the Gaussian distribution are eliminated, significantly reducing coordinate extraction errors and laying a precise foundation for subsequent displacement calculations. 2. Non-contact monitoring, causing no damage to the bridge structure and simple installation and maintenance. This method does not require directly fixing the monitoring equipment to the bridge structure; only markers need to be deployed at key locations and cameras fixed in stable areas. The installation process is simple and will not damage the bridge structure. The markers are waterproof and shockproof, and the camera parameters can be dynamically adjusted according to the environment, resulting in low maintenance costs and suitability for long-term monitoring of various types of bridges. 3. Strong anti-interference capability, adaptable to complex outdoor environments. By selecting infrared markers, processing image noise reduction, and removing abnormal light spots, the system effectively resists outdoor interference factors such as changes in lighting, vibration, and obstruction, ensuring the continuity and stability of monitoring data. Simultaneously, through camera intrinsic parameter calibration and extrinsic parameter matrix correction, the system resolves the conversion error problem caused by tilted-view shooting, further improving the accuracy of actual deflection calculation. 4. It enables dynamic real-time monitoring and is highly practical. This method captures the dynamic displacement process of bridge beams through continuous image acquisition, outputting monitoring results in real time, facilitating timely understanding of the bridge's structural health status by staff. The monitoring process is standardized and highly operable; the deployment spacing and shooting parameters can be flexibly adjusted according to the bridge span and monitoring needs, making it suitable for different types and working conditions of bridge displacement monitoring scenarios. Attached Figure Description

[0061] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0062] Figure 1 This is a schematic diagram of the first embodiment of a bridge displacement monitoring method based on image recognition according to the present invention;

[0063] Figure 2 This is a schematic diagram of the first embodiment of a bridge displacement monitoring method based on image recognition according to the present invention;

[0064] Figure 3 This is a schematic diagram of the first embodiment of a bridge displacement monitoring system based on image recognition according to the present invention. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0066] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0067] The present invention will now be described in detail with reference to the accompanying drawings, such as... Figure 1 As shown, a bridge displacement monitoring method based on image recognition includes the following steps:

[0068] Step 101: Install markers at key locations on the bridge structure. The markers should include at least infrared LED lights and infrared reflective stickers.

[0069] Specifically, in this embodiment, according to the bridge structural design drawings and displacement monitoring specifications, markers are set up at monitoring points in the mid-span, supports, beam ends, and web of the bridge beam, and the spacing between markers is set according to the bridge span and the field of view of the image acquisition equipment. The image acquisition equipment is used to take pictures to confirm that each marker forms an independent bright area without overlapping, blurring, or offset light spots.

[0070] Based on the bridge structural design drawings and displacement monitoring specifications, markers were installed at key monitoring points prone to displacement / deformation, such as the mid-span, supports, beam ends, and webs of the bridge beams. The spacing between markers was rationally set according to the bridge span and the field of view of the image acquisition equipment, ensuring that all markers were within the effective shooting range of the equipment and that adjacent markers did not overlap. At least two types of markers were used: infrared LED lights and infrared reflective stickers. Infrared LED lights were installed at monitoring points without natural light reflection, using waterproof and shockproof installation methods and equipped with stable power supply modules to ensure continuous illumination. Infrared reflective stickers were affixed to flat, rust-free, and oil-free bridge surfaces, and after affixing, air bubbles were removed to ensure tight adhesion to the surface. Both types of markers were installed without obstruction or tilting. Trial shooting was conducted using the image acquisition equipment to check the imaging effect of the markers, confirming that each marker could form an independent bright area without overlapping, blurry, or offset light spots. If any of these problems existed, the installation position or angle of the markers was adjusted promptly.

[0071] Step 102: Acquire multiple image sequences of the marker using an image acquisition device, and perform Gaussian distribution fitting on the spot regions of the multiple image sequences to extract the coordinates of the center point of the spot;

[0072] Specifically, in this embodiment, an industrial-grade camera is fixed on a stable support on the bridge monitoring side. Camera intrinsic parameters are calibrated, and camera shooting parameters are set according to outdoor lighting conditions, using a grayscale imaging mode. The camera is then activated to continuously and in real-time capture images of the deployed markers, acquiring a sequence of multiple grayscale images arranged chronologically. The pixel information of each frame in the grayscale image sequence is stored in a two-dimensional array, and pixel grayscale values ​​are assigned according to industry standards. Basic noise reduction processing is performed on the pixel array to remove isolated grayscale abrupt changes caused by imaging noise, resulting in multiple image sequences. A grayscale threshold is set, and the pre-processed image sequences are subjected to threshold filtering to remove background pixels with grayscale values ​​less than the threshold, retaining the bright area pixels formed by the markers. A connected component analysis algorithm is used to segment the retained bright pixels, dividing the light spot area corresponding to each marker.

[0073] For each segmented light spot region, a two-dimensional Gaussian distribution mathematical model is used for fitting to extract the coordinates of the light spot center point. The formula for the two-dimensional Gaussian distribution mathematical model is as follows:

[0074]

[0075] in, Represents the pixel grayscale value. Indicates the Gaussian peak amplitude. Indicates the coordinates of the center point of the light spot. This represents the standard deviation of a Gaussian distribution in the x and y directions. Background grayscale;

[0076] The least squares method is used to solve the parameters of the two-dimensional Gaussian distribution mathematical model, and the gray-level distribution characteristics of the light spot region are fitted to obtain the coordinates of the center point of each light spot. .

[0077] Industrial-grade cameras were mounted on stable supports on the bridge monitoring side. Camera intrinsic parameter calibration was completed, and core parameters such as pixel size S and equivalent focal length f were obtained. Based on outdoor lighting conditions, camera shooting parameters, including exposure time, gain, and frame rate, were set. Grayscale imaging mode was adopted to ensure no camera vibration or viewpoint shift during shooting. The camera was activated to continuously and in real-time capture images of the deployed markers, obtaining a sequence of multiple grayscale images arranged chronologically. The frame rate was set according to the time resolution requirements of bridge displacement monitoring to ensure the capture of the dynamic displacement process of the beam. The pixel information of each frame in the image sequence was stored in a two-dimensional array. Pixel grayscale values ​​were assigned values ​​from 0 to 255 according to industry standards, with 0 representing pure black and 255 representing pure white. Basic noise reduction processing was performed on the pixel array to remove isolated grayscale abrupt changes caused by imaging noise, ensuring the continuity of grayscale distribution in the light spot area. The grayscale threshold T is set by experimental calibration and dynamically adjusted according to outdoor ambient light and the luminous / reflective intensity of the sign. It is generally set to 80-150. The preprocessed pixel array is then filtered by thresholding to remove background pixels with grayscale values ​​less than T, retaining only the bright area pixels formed by the sign. The retained bright pixels are then segmented using a connected component analysis algorithm to divide the area into independent light spot regions corresponding to each sign, ensuring that there are no pixel overlaps between the light spot regions.

[0078] Step 103: Eliminate light spots whose center point coordinates do not conform to the Gaussian distribution characteristics by detecting light spot anomalies, and obtain the center point coordinates of the target light spot;

[0079] Specifically, in this embodiment, the index values ​​of the center point coordinates of the light spot are set as the fitting residual, Gaussian peak amplitude, and standard deviation in the x and y directions under normal Gaussian distribution. The index values ​​in the center point coordinates of the light spot are compared with the threshold range.

[0080] If any indicator exceeds the effective threshold range, the spot is determined to be an abnormal spot. The coordinates of the center point of the abnormal spot are marked and deleted to obtain the coordinates of the center point of the target spot.

[0081] Through numerous outdoor simulation experiments, the fitting residual and Gaussian peak amplitude under the normal Gaussian distribution of the light spot were calibrated. x / y direction standard deviation ratio The effective threshold range is determined, and upper / lower limit critical values ​​are set for each indicator as the basis for anomaly judgment. For each Gaussian fitting result of a light spot, the fitting residual is extracted one by one. value, The ratio is one of three core indicators, and the value of each indicator is compared with the calibrated effective threshold range. If any indicator exceeds the effective threshold range, the fitting residual increases significantly. value drops / rises sharply If the ratio deviates too much from 1, the light spot is determined to be an abnormal light spot, as its grayscale distribution no longer conforms to the characteristics of a two-dimensional Gaussian distribution. The coordinates of the center point of the light spot determined to be abnormal are marked as null. This type of invalid coordinate is removed from the set of all light spot center point coordinates. The remaining unmarked light spot center point coordinates, which meet all threshold requirements, are the target light spot center point coordinates that can be used for subsequent calculations.

[0082] Step 104: Calculate the pixel displacement based on the pixel position changes of the target spot center point coordinates between different image frames;

[0083] Specifically, in this embodiment, the first frame of the image sequence is set as the initial reference frame, and the initial coordinates of the center points of all target light spots in the initial reference frame are extracted. Establish a coordinate database according to the spot number to complete the initial position calibration;

[0084] Extract the real-time coordinates of the target spot center point in any subsequent time frame of the image sequence. Inter-frame coordinate matching is performed based on the spot number;

[0085] For each successfully matched target spot, calculate its pixel displacement in the x and y directions, using the following formula:

[0086]

[0087]

[0088] in, , These are the pixel displacements in the x and y directions, respectively. Positive values ​​indicate movement along the positive axis, while negative values ​​indicate movement along the negative axis. The calculated pixel displacement values ​​are checked for reasonableness, and abrupt pixel displacement values ​​caused by inter-frame matching errors are removed.

[0089] Step 105: Convert pixel displacement into actual bridge deflection using the imaging geometry model.

[0090] Specifically, in this embodiment, an imaging geometry model is used to convert the pixel displacement values ​​in the x and y directions into preliminary bridge deflection of the bridge structure. The model formula is as follows:

[0091]

[0092] in, This indicates the initial actual displacement of the bridge structure. Indicates the camera pixel size. Indicates the equivalent focal length of the lens;

[0093] If the camera is shooting from a non-perpendicular angle, the extrinsic parameter matrix is ​​obtained through camera extrinsic parameter calibration. Substituting the initial bridge deflection into the correction formula, the pixel displacement direction is corrected to obtain the actual bridge deflection in the direction of bridge beam deformation. The correction formula is as follows:

[0094]

[0095] in, This represents the actual bridge deflection, which is a vector value and includes both the magnitude and direction of the displacement.

[0096] Its beneficial effects are as follows: 1. High monitoring accuracy, achieving sub-pixel level coordinate extraction. This method uses a combination of infrared LED lights and infrared reflective stickers to deploy markers, effectively avoiding interference from outdoor lighting changes and weather, ensuring clear light spot imaging; by fitting a two-dimensional Gaussian distribution and using the least squares method to solve for the coordinates of the light spot center point, sub-pixel level extraction is achieved. Combined with an abnormal light spot detection mechanism, invalid coordinates with excessively large fitting residuals or those not conforming to the Gaussian distribution are eliminated, significantly reducing coordinate extraction errors and laying a precise foundation for subsequent displacement calculations. 2. Non-contact monitoring, causing no damage to the bridge structure and simple installation and maintenance. This method does not require directly fixing the monitoring equipment to the bridge structure; only markers need to be deployed at key locations and cameras fixed in stable areas. The installation process is simple and will not damage the bridge structure. The markers are waterproof and shockproof, and the camera parameters can be dynamically adjusted according to the environment, resulting in low maintenance costs and suitability for long-term monitoring of various types of bridges. 3. Strong anti-interference capability, adaptable to complex outdoor environments. By selecting infrared markers, processing image noise reduction, and removing abnormal light spots, the system effectively resists outdoor interference factors such as changes in lighting, vibration, and obstruction, ensuring the continuity and stability of monitoring data. Simultaneously, through camera intrinsic parameter calibration and extrinsic parameter matrix correction, the system resolves the conversion error problem caused by tilted-view shooting, further improving the accuracy of actual deflection calculation. 4. It enables dynamic real-time monitoring and is highly practical. This method captures the dynamic displacement process of bridge beams through continuous image acquisition, outputting monitoring results in real time, facilitating timely understanding of the bridge's structural health status by staff. The monitoring process is standardized and highly operable; the deployment spacing and shooting parameters can be flexibly adjusted according to the bridge span and monitoring needs, making it suitable for different types and working conditions of bridge displacement monitoring scenarios.

[0097] like Figure 2 As shown, a bridge displacement monitoring method based on image recognition involves acquiring multiple image sequences of a marker using an image acquisition device, fitting a Gaussian distribution to the spot regions of the multiple image sequences, and extracting the coordinates of the spot center point. The method includes the following steps:

[0098] Step 201: Fix the industrial-grade camera equipment on the stable bracket on the bridge monitoring side, complete the camera intrinsic parameter calibration, set the camera shooting parameters according to the outdoor ambient light conditions, and adopt grayscale imaging mode;

[0099] Step 202: Start the camera to continuously and in real time capture images of the deployed markers, and acquire a sequence of grayscale images arranged in chronological order.

[0100] Step 203: Store the pixel information of each frame in the grayscale image sequence in the form of a two-dimensional array, assign pixel grayscale values ​​according to industry standards, and perform basic noise reduction processing on the pixel array to remove isolated grayscale abrupt changes caused by imaging noise, thereby obtaining multiple image sequences.

[0101] Step 204: Set a grayscale threshold, perform threshold filtering on multiple preprocessed image sequences, remove background pixels with grayscale values ​​less than the grayscale threshold, retain the bright area pixels formed by the markers, and perform region segmentation on the retained bright pixels using a connected component analysis algorithm to divide the light spot region corresponding to each marker.

[0102] The above describes an embodiment of the bridge displacement monitoring method based on image recognition of the present invention. Please refer to [link / reference]. Figure 3 In a bridge displacement monitoring system based on image recognition, the fiber optic-based telephone communication system includes the following modules:

[0103] The marker placement module is used to place markers at key locations on the bridge structure. The markers include at least infrared LED lights and infrared reflective stickers.

[0104] The spot coordinate extraction module is used to acquire multiple image sequences of the marker through an image acquisition device, and to perform Gaussian distribution fitting on the spot regions of the multiple image sequences to extract the coordinates of the center point of the spot.

[0105] The spot coordinate filtering module is used to remove spots whose center point coordinates do not conform to the Gaussian distribution characteristics by spot anomaly detection, and obtain the center point coordinates of the target spot.

[0106] The pixel displacement calculation module is used to calculate the pixel displacement based on the pixel position change between different image frames according to the coordinates of the center point of the target spot.

[0107] The bridge deflection calculation module is used to convert pixel displacements into actual bridge deflection using an imaging geometry model.

[0108] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A bridge displacement monitoring method based on image recognition, characterized in that, The image recognition-based bridge displacement monitoring method includes the following steps: Markers are placed at key locations on the bridge structure, and the markers include at least infrared LED lights and infrared reflective stickers. Multiple image sequences of the marker are acquired using an image acquisition device, and the light spot regions of the multiple image sequences are fitted with a Gaussian distribution to extract the coordinates of the center point of the light spot. By detecting abnormal light spots, the center point coordinates of the target light spot are obtained by eliminating light spots whose center point coordinates do not conform to the Gaussian distribution characteristics. The pixel displacement is calculated based on the pixel position changes of the target spot center point coordinates between different image frames; The pixel displacement is converted into actual bridge deflection using an imaging geometry model.

2. The bridge displacement monitoring method based on image recognition as described in claim 1, characterized in that, The placement of markers at key locations on the bridge structure, wherein the markers include at least infrared LED lights and infrared reflective stickers, including: According to the bridge structural design drawings and displacement monitoring specifications, markers are set up at the monitoring points in the mid-span, support, beam end and web of the bridge beam, and the spacing of the markers is set according to the bridge span and the field of view of the image acquisition equipment. The images were captured using an image acquisition device to confirm that each marker formed an independent bright area, with no overlapping, blurred, or offset light spots.

3. The bridge displacement monitoring method based on image recognition as described in claim 1, characterized in that, The step of acquiring multiple image sequences of the marker using an image acquisition device, and performing Gaussian distribution fitting on the spot regions of the multiple image sequences to extract the coordinates of the spot center point includes: The industrial-grade camera equipment was fixed on a stable bracket on the side of the bridge monitoring, the camera's intrinsic parameters were calibrated, and the camera shooting parameters were set according to the outdoor lighting conditions, using grayscale imaging mode. The camera is activated to continuously and in real time capture images of the deployed markers, obtaining a sequence of grayscale images arranged in chronological order. The pixel information of each frame in the grayscale image sequence is stored in the form of a two-dimensional array. The pixel grayscale values ​​are assigned according to industry standards, and basic noise reduction processing is performed on the pixel array to remove isolated grayscale abrupt changes caused by imaging noise, resulting in multiple image sequences. A grayscale threshold is set, and threshold filtering is performed on multiple preprocessed image sequences to remove background pixels with grayscale values ​​less than the grayscale threshold, while retaining the bright pixels formed by the markers. The retained bright pixels are then segmented using a connected component analysis algorithm to divide the light spot region corresponding to each marker.

4. The bridge displacement monitoring method based on image recognition as described in claim 1, characterized in that, The step of acquiring multiple image sequences of the marker using an image acquisition device, and performing Gaussian distribution fitting on the spot regions of the multiple image sequences to extract the coordinates of the spot center point includes: For each segmented light spot region, a two-dimensional Gaussian distribution mathematical model is used for fitting to extract the coordinates of the light spot center point. The formula for the two-dimensional Gaussian distribution mathematical model is as follows: in, Represents the pixel grayscale value. Indicates the Gaussian peak amplitude. Indicates the coordinates of the center point of the light spot. This represents the standard deviation of a Gaussian distribution in the x and y directions. Background grayscale; The least squares method is used to solve the parameters of the two-dimensional Gaussian distribution mathematical model, and the gray-level distribution characteristics of the light spot region are fitted to obtain the coordinates of the center point of each light spot. .

5. The bridge displacement monitoring method based on image recognition as described in claim 1, characterized in that, The step of eliminating light spots whose center point coordinates do not conform to the Gaussian distribution characteristics through light spot anomaly detection to obtain the target light spot center point coordinates includes: The index values ​​of the center point coordinates of the light spot are set as the fitting residual, Gaussian peak amplitude, and standard deviation in the x and y directions under normal Gaussian distribution. The index values ​​in the center point coordinates of the light spot are compared with the threshold range. If any indicator exceeds the effective threshold range, the spot is determined to be an abnormal spot. The coordinates of the center point of the abnormal spot are marked and deleted to obtain the coordinates of the center point of the target spot.

6. The bridge displacement monitoring method based on image recognition as described in claim 1, characterized in that, The step of calculating the pixel displacement based on the pixel position change of the target spot center point coordinates between different image frames includes: The first frame of the image sequence is set as the initial reference frame, and the initial coordinates of the center points of all target light spots in the initial reference frame are extracted. Establish a coordinate database according to the spot number to complete the initial position calibration; Extract the real-time coordinates of the target spot center point in any subsequent time frame of the image sequence. Inter-frame coordinate matching is performed based on the spot number; For each successfully matched target spot, calculate its pixel displacement in the x and y directions, using the following formula: in, , These are the pixel displacements in the x and y directions, respectively. Positive values ​​indicate movement along the positive axis, while negative values ​​indicate movement along the negative axis. The calculated pixel displacement values ​​are checked for reasonableness, and abrupt pixel displacement values ​​caused by inter-frame matching errors are removed.

7. The bridge displacement monitoring method based on image recognition as described in claim 1, characterized in that, The process of converting the pixel displacement into actual bridge deflection using an imaging geometry model includes: Using an imaging geometry model, the pixel displacement values ​​in the x and y directions are converted into preliminary bridge deflection of the bridge structure, as shown in the following model formula: in, This indicates the initial actual displacement of the bridge structure. Indicates the camera pixel size. Indicates the equivalent focal length of the lens; If the camera is shooting from a non-perpendicular angle, the extrinsic parameter matrix is ​​obtained through camera extrinsic parameter calibration. Substituting the initial bridge deflection into the correction formula, the pixel displacement direction is corrected to obtain the actual bridge deflection in the direction of bridge beam deformation. The correction formula is as follows: in, This represents the actual bridge deflection, which is a vector value and includes both the magnitude and direction of the displacement.

8. A bridge displacement monitoring system based on image recognition, characterized in that, The image recognition-based bridge displacement monitoring system includes the following modules: The marker placement module is used to place markers at key locations on the bridge structure. The markers include at least infrared LED lights and infrared reflective stickers. The spot coordinate extraction module is used to acquire multiple image sequences of the marker through an image acquisition device, and to perform Gaussian distribution fitting on the spot regions of the multiple image sequences to extract the coordinates of the center point of the spot. The spot coordinate filtering module is used to remove spots whose center point coordinates do not conform to the Gaussian distribution characteristics by spot anomaly detection, so as to obtain the center point coordinates of the target spot. The pixel displacement calculation module is used to calculate the pixel displacement based on the pixel position change between different image frames according to the coordinates of the center point of the target spot; The bridge deflection calculation module is used to convert the pixel displacement into actual bridge deflection through an imaging geometry model.

9. A bridge displacement monitoring system based on image recognition as described in claim 8, characterized in that, The pixel displacement calculation module includes the following sub-modules: The extraction submodule is used to set the first frame of the image sequence as the initial reference frame and extract the initial coordinates of the center points of all target light spots in the initial reference frame. Establish a coordinate database according to the spot number to complete the initial position calibration; The matching submodule is used to extract the real-time coordinates of the center point of the target spot in any subsequent time frame of the image sequence. Inter-frame coordinate matching is performed based on the spot number; The calculation submodule is used to calculate the pixel displacement in the x and y directions for each successfully matched target spot. The calculation formula is as follows: in, , These are the pixel displacements in the x and y directions, respectively. Positive values ​​indicate movement along the positive axis, while negative values ​​indicate movement along the negative axis. The calculated pixel displacement values ​​are checked for reasonableness, and abrupt pixel displacement values ​​caused by inter-frame matching errors are removed.

10. A bridge displacement monitoring system based on image recognition as described in claim 8, characterized in that, The bridge deflection calculation module includes the following sub-modules: The conversion submodule is used to convert the pixel displacement values ​​in the x and y directions into preliminary bridge deflection of the bridge structure using the imaging geometry model. The model formula is as follows: in, This indicates the initial actual displacement of the bridge structure. Indicates the camera pixel size. Indicates the equivalent focal length of the lens; The correction submodule is used to obtain the extrinsic parameter matrix through camera extrinsic parameter calibration if the camera is shooting from a non-perpendicular orthophoto angle. Substituting the initial bridge deflection into the correction formula, the pixel displacement direction is corrected to obtain the actual bridge deflection in the direction of bridge beam deformation. The correction formula is as follows: in, This represents the actual bridge deflection, which is a vector value and includes both the magnitude and direction of the displacement.

Images