Bridge support displacement and typical disease intelligent monitoring method
By installing movable rulers and reference rulers or built-in rulers and pointers on bridge bearings, and combining them with a detection segmentation network model, near real-time monitoring of bridge bearing displacement and defects has been achieved. This solves the problems of inconsistent manual judgment, poor real-time performance, and limited sensor lifespan in existing technologies, and improves the accuracy and efficiency of monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- RAILWAY CONSTR RES INST OF CHINA ACAD OF RAILWAY SCI CO LTD
- Filing Date
- 2022-09-22
- Publication Date
- 2026-08-04
AI Technical Summary
Existing methods for detecting bridge bearing defects suffer from problems such as inconsistent manual judgment, poor real-time performance, limited sensor lifespan, restrictions on drone operation, and high data processing resource requirements, making it impossible to achieve efficient and accurate monitoring of bearing displacement and defects.
Install movable rulers and reference rulers or built-in rulers and pointers on bridge bearings, and combine them with a detection segmentation network model to monitor bearing displacement and rotation in real time through image recognition technology, so as to realize the visual observation of typical defects.
It achieves near real-time monitoring of bridge bearing displacement and defects, with an identification accuracy of 1mm and 0.0002rad, reducing on-site workload, improving the intelligence and informatization level of monitoring, and enabling rapid acquisition of bearing service status.
Smart Images
Figure CN117405698B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of bridge health monitoring technology, and in particular relates to an intelligent monitoring method for bridge bearing displacement and typical defects. Background Technology
[0002] Current methods for detecting bridge bearing defects generally rely on manual observation, long-term monitoring, or video recording. This presents the following problems:
[0003] (1) Most of the visual defects and routine inspections of bearings are based on manual periodic or irregular visual inspections or inspections using appropriate testing equipment (such as magnifying glasses, squares, feeler gauges, vernier calipers, etc.). This method is greatly affected by the professional quality and experience of the technicians. Different personnel have different judgments on defects and interpretations of data, and there is a lack of a unified evaluation standard. Secondly, bearings of special structure bridges are generally inspected once every 3 months, which makes it impossible to obtain the bearing status in real time and lacks real-time feedback. Thirdly, for high-speed railway bridges, manual inspections are generally carried out during the maintenance window from 0:00 to 4:00 in the morning. The workers are in poor physical condition, the working conditions are harsh, and the labor intensity is high.
[0004] (2) Long-term monitoring using sensors installed on bearings is often part of bridge health monitoring or is installed as needed for bearings that have already developed defects. The sensors are typically displacement gauges or inclinometers, which upload real-time bearing displacement or rotation values to the application system via 4G / 5G or other methods. Combined with beam temperature and external load magnitude, a large amount of data is analyzed to assess the bearing condition. This method provides relatively accurate measurements, but the sensors are greatly affected by the service environment, have a limited lifespan, require periodic sensor replacement, and cannot perform multi-feature identification of surface defects in the bearings, thus presenting certain limitations.
[0005] (3) Using drones to collect images of bridge defects, perform image recognition on the data, and determine the type and level of defects. Considering the spatial relationship of the bridge structure, drones cannot get close to the supports, making it difficult to obtain information on the apparent defects of the supports; the operation of the drone system is limited by multiple factors such as the skill level of the operator, workload, and function allocation, which affects the accuracy or comprehensiveness of the support information collection; the training set for image recognition consists of various apparent defects of the supports (such as cracks), which requires a lot of annotation work, and the subsequent data processing requires high computing resources. Defect information cannot be fed back in a timely manner, and the displacement and rotation characteristics of the supports cannot be obtained, making this method inefficient for large-scale field application. Summary of the Invention
[0006] The purpose of this invention is to overcome the shortcomings of the existing technology and to propose an intelligent monitoring method for bridge bearing displacement and typical defects.
[0007] To achieve the above objectives, this invention proposes an intelligent monitoring method for bridge bearing displacement and typical defects, applicable to bridge bearings without built-in scales and pointers. The method includes:
[0008] Install a movable ruler and a reference ruler on the upper bearing plate and the pad stone of the bridge bearing to be monitored, respectively;
[0009] Target points are set on the upper bearing plate of the bridge bearing to be monitored;
[0010] Using the movable ruler and the reference ruler as monitoring targets, images of the target point, the movable ruler, and the reference ruler are collected.
[0011] Based on the collected images, the displacement and rotation data of the bridge supports under monitoring are obtained using a pre-established and trained detection and segmentation network model.
[0012] Based on the acquired support displacement and support rotation data, and simultaneously perform real-time visual observation of the supports, the monitoring of typical defects can be achieved.
[0013] As an improvement to the above method, the movable ruler and the reference ruler have the same length, both being dis. real .
[0014] As an improvement to the above method, the step of setting a target point on the upper bearing plate of the bridge support to be monitored specifically includes:
[0015] Left target point T1 and right target point T2 are set on the upper bearing plate of the bridge bearing to be monitored. The two target points are located on the same horizontal line, and the vertical distance from the upper bearing plate center line to the target point is D / 2, where D is the length of the lower bearing plate.
[0016] As an improvement to the above method, the detection and segmentation network model is based on a single-stage detection network. It improves the detection accuracy of targets by adding a small target detection layer or a fusion factor, and completes target detection and semantic segmentation simultaneously by adding a segmentation network.
[0017] As an improvement to the above method, the step of obtaining the bearing displacement and bearing rotation data of the bridge to be monitored based on the acquired images and a pre-established and trained detection and segmentation network model specifically includes:
[0018] The acquired image is input into a pre-established and trained detection and segmentation network model to obtain a segmented image P that includes the range of the movable ruler and the reference ruler, and the center pixel coordinates (X) of the left target point T1. t1 ,Y t1 ), center pixel coordinates (X) of right target point T2 t2 ,Y t2 );
[0019] Based on the segmented image P, the contour information of the active ruler and the reference ruler is extracted. The points on the contour are classified based on the position information to obtain the contour fitting line. Among them, the fitting lines of the upper and lower edges of the reference ruler are the horizontal lines of the view.
[0020] Based on the edge fitting of the movable ruler and the reference ruler, a straight line is obtained to obtain the 0 mark point C of the movable ruler. m Reference ruler 0 mark C b and the pixel distance corresponding to the reference ruler length (dis) px By calculating pixel distance dis px dis with actual length real The proportional relationship is used to obtain the proportional relationship for the conversion from pixel to actual distance;
[0021] Calculate point C on the 0 mark of the movable ruler. m The perpendicular intersection point M with the reference ruler is automatically obtained relative to the 0 mark point C of the reference ruler. b The left and right pixel displacement values are then converted into the actual displacement Dis of the support through the aforementioned proportional relationship. mv ;
[0022] The support rotation angle R is obtained from the following formula:
[0023]
[0024] As an improvement to the above method, the method further includes: for bridge bearings made of rubber, monitoring the shear deformation of the bridge bearing to be monitored is achieved based on target point localization; specifically, it includes: arranging a target point at the top and bottom of the centerline of the bridge bearing, and performing target detection on the collected target point images based on a pre-established and trained detection and segmentation network model to achieve real-time monitoring of shear deformation.
[0025] As an improvement to the above method, the typical defects include cracks and weld failures in steel components, wear or extrusion of sliding plates, shearing, missing or loose bolts, corrosion, voids in supports, and surface cracks.
[0026] On the other hand, the present invention also proposes an intelligent monitoring method for bridge bearing displacement and typical defects, for bridge bearings with built-in scales and pointers, the method comprising:
[0027] Target points are set on the upper bearing plate of the bridge bearing to be monitored;
[0028] Using a ruler and pointer as monitoring targets, images including target points, rulers, and pointers are acquired.
[0029] Based on the collected images, the displacement and rotation data of the bridge supports under monitoring are obtained using a pre-established and trained detection and segmentation network model.
[0030] Based on the acquired support displacement and support rotation data, and simultaneously perform real-time visual observation of the supports, the monitoring of typical defects can be achieved.
[0031] As an improvement to the above method, the step of setting a target point on the upper bearing plate of the bridge support to be monitored specifically includes:
[0032] Left target point T1 and right target point T2 are set on the upper bearing plate of the bridge bearing to be monitored. The two target points are located on the same horizontal line, and the vertical distance from the upper bearing plate center line to the target point is D / 2, where D is the length of the lower bearing plate.
[0033] As an improvement to the above method, the step of obtaining the bearing displacement and bearing rotation data of the bridge to be monitored based on the acquired images and a pre-established and trained detection and segmentation network model specifically includes:
[0034] The acquired image is input into a pre-established and trained detection and segmentation network model to obtain the segmented image P within the scale range and the center pixel coordinates (X) of the left target point T1. t1 ,Y t1 ), center pixel coordinates (X) of right target point T2 t2 ,Y t2 ) and the center point C of the pointer p ;
[0035] Based on the segmented image P of the scale, the scale contour information is extracted, and the contour points are classified based on the position information. The scale edge is fitted with a straight line according to the classified contour points to obtain the contour fitting line. The upper and lower edge fitting lines are the horizontal straight lines of the view.
[0036] By fitting a straight line to the profile of the scale, the 0 mark point C of the scale is obtained. r and the pixel distance corresponding to the scale length dis px By calculating pixel distance dis px Compared with the known actual length dis real The proportional relationship is used to obtain the proportional relationship of pixel to actual distance conversion;
[0037] Calculate the center point C of the pointer p The perpendicular intersection point M with the scale is used to obtain the distance between M and the 0 mark C on the scale. r The left and right pixel displacement values are converted into the actual displacement value Dis of the support according to the above proportional relationship. mv ;
[0038] The support rotation angle R is obtained from the following formula:
[0039]
[0040] The detection and segmentation network model is based on a single-stage detection network. It improves the detection accuracy of targets by adding a small target detection layer or a fusion factor, and completes target detection and semantic segmentation simultaneously by adding a segmentation network.
[0041] Compared with the prior art, the advantages of the present invention are:
[0042] 1. The method of the present invention realizes real-time monitoring of the displacement of movable bearings (including multi-directional movable, lateral movable and longitudinal movable bearings) and visual observation of defects through infrared high-definition network monitoring cameras, quickly obtains the service status of bearings, serves the operation and maintenance of bridge bearings, reduces the workload of on-site technicians, and has the technical characteristics of intelligence and information.
[0043] 2. The method of the present invention combines the scale and pointer built into the bridge bearing, or the method of setting a reference scale on the bearing pad stone and setting a movable scale on the bearing upper plate, to perform bridge bearing displacement identification in near real time based on the detection segmentation network, with an identification accuracy of 1mm.
[0044] 3. The method of the present invention sets two target points at both ends of the upper bearing plate of the bridge bearing to achieve near real-time accurate identification of the bearing rotation angle, with an identification accuracy of 0.0002 rad;
[0045] 4. The method of the present invention, combined with the set camera, enables real-time monitoring of apparent defects in the support;
[0046] 5. By using the method of the present invention, the deterioration pattern of the bearing can be evaluated through monitoring over a certain period of time. When the bearing has serious defects, it can be monitored in real time. This method is conducive to quickly obtaining the service status of the bearing, reducing the workload of on-site technicians, and has high identification accuracy, which can achieve the maximum input-output ratio. Attached Figure Description
[0047] Figure 1 This is a flowchart of the intelligent monitoring method for bridge bearing displacement and typical defects of the present invention;
[0048] Figure 2 This is a schematic diagram of the installation of the target point, movable ruler, and reference ruler in Embodiment 2 of the present invention;
[0049] Figure 3 This is a schematic diagram of the target point, pointer, and scale positions in Embodiment 1 of the present invention;
[0050] Figure 4 This is a schematic diagram of the rubber support target installation in Embodiments 1 and 2 of the present invention. Detailed Implementation
[0051] This invention establishes a method for monitoring bridge bearing displacement and observing typical defects based on image recognition technology, enabling real-time monitoring of bearing displacement and visual observation of apparent defects. The method includes: setting target points on the upper bearing plate of the bridge bearing to be monitored; acquiring images including the target points, scale, and pointer as monitoring targets; constructing a single-stage detection and segmentation network to simultaneously achieve target detection and semantic segmentation; further obtaining the bearing displacement and bearing rotation angle based on the acquired images and the detection and segmentation network model; monitoring the shear deformation of the bridge bearing based on target point localization; and simultaneously performing real-time visual observation of the bearing based on the acquired bearing displacement, bearing rotation angle, and shear deformation monitoring to achieve the monitoring of typical defects.
[0052] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and embodiments.
[0053] Example 1
[0054] Embodiment 1 of the present invention proposes an intelligent monitoring method for bridge bearing displacement and typical defects, for bridge bearings without built-in scales and pointers. The method includes:
[0055] Install a movable ruler and a reference ruler on the upper bearing plate and the pad stone of the bridge bearing to be monitored, respectively;
[0056] Target points are set on the upper bearing plate of the bridge bearing to be monitored;
[0057] Using the movable ruler and the reference ruler as monitoring targets, images of the target point, the movable ruler, and the reference ruler are collected.
[0058] A single-stage detection and segmentation network is constructed to simultaneously achieve object detection and semantic segmentation.
[0059] Based on the acquired images, the displacement and rotation angle of the bridge supports under monitoring are further obtained using a detection and segmentation network model.
[0060] Monitoring of shear deformation of bridge bearings is achieved by locating target points.
[0061] Based on the acquired support displacement, support rotation angle, and shear deformation monitoring, and simultaneously performing real-time visual observation of the supports, typical defects can be monitored.
[0062] Typical defects include, but are not limited to, cracks and weld failures in steel components, wear or extrusion of the sliding plate, shearing or loosening of bolts, corrosion, bearing voids, and surface cracks. Other defects may include excessive shear deformation, localized cracking, voids in the bearing surface, cracks in steel components, fractures and corrosion deformation, extrusion of rubber plates and sliding plates, and deformation and breakage of anchor bolts. Specifically, according to the type of bridge bearing, typical defects of plate rubber bearings include bearing voids, rubber cracks, exposed steel plates, uneven bulging of rubber, and excessive shear; typical defects of pot bearings include wear or extrusion of the sliding plate, corrosion of steel components, extrusion of the bearing rubber plate, immobility of movable bearings, excessive bearing displacement, and bearing non-rotation or excessive rotation angle; typical defects of spherical bearings include damage to steel components, failure of anchors and positioning components (such as shearing of anchor bolts), excessive displacement, immobility of movable bearings, excessive rotation angle, poor contact between the upper and lower bearing plates and the beam bottom and bearing pad, and defects in the bearing pad area.
[0063] Scale Selection and Installation
[0064] To accurately obtain the bearing displacement, a scale is needed as a marking fixture. In this embodiment, a center-dividing scale with an accuracy of 1mm is selected as the monitoring scale. A movable scale and a reference scale are respectively installed at the center of the upper bearing plate and the center of the bearing pad. Figure 2 As shown. Furthermore, by monitoring the relative displacement between the 0 mark of the movable ruler and the 0 mark of the reference ruler, the direction and value of the support displacement can be accurately obtained. The specific steps are as follows:
[0065] 1. Given that the lengths of both the movable ruler and the reference ruler are dis real The movable ruler and the reference ruler are selected as monitoring targets. Target points T1 and T2 are preset on the upper support plate, ensuring that T1 and T2 are on the same horizontal line. The length D of the lower support plate is...
[0066] Given that the vertical distances from the left and right target points to the center line of the upper support plate are D / 2 respectively.
[0067] 2. Install a camera on the scale side of the bridge bearing. The camera placement requirements are as follows.
[0068] 3. Collect 1,000-3,000 images of bridge bearings (movable ruler and reference ruler) taken at different times of day and night, annotate the images with rulers, and generate sample set A.
[0069] 4. Construct a single-stage detection and segmentation network. Single-stage detection and segmentation networks can be such as YOLOv5, FPN, etc., but are not limited to these. Add a small target detection layer to the YOLOv5 network, or add a fusion factor to the FPN-based detection network to improve target detection accuracy; add a convolutional segmentation head to the YOLOv5 network, or add a pixel stream segmentation network to the FPN-based detection network to complete semantic segmentation.
[0070] 5. Train a detection and segmentation network model based on sample set A to segment the active ruler and the reference ruler. It is recommended that 80% of the sample data A1 be used to train the model and 20% of the sample data A2 be used to test the model performance.
[0071] 6. The image is processed by a trained detection and segmentation network model to calculate the segmented image P of the measuring range of the movable ruler and the reference ruler.
[0072] 7. Based on the segmented image P, extract the contour information of the moving ruler and the reference ruler. Classify the points on the contour based on the position information to obtain their contour fitting lines. Among them, the fitting lines for the upper and lower edges of the reference ruler are the horizontal lines of the viewpoint.
[0073] 8. Based on the edge fitting lines of the movable ruler and the reference ruler, obtain the 0 mark point C of the movable ruler. m Reference ruler 0 mark C b And the pixel distance corresponding to the length of the reference ruler (dis) px By calculating pixel distance dis px dis with actual length real The proportional relationship is used to convert pixels to actual distances.
[0074] 9. Calculate the 0 mark C on the movable ruler. m The perpendicular intersection point M with the reference ruler is automatically obtained relative to the 0 mark point C of the reference ruler. b The left and right pixel displacement values are then converted into the actual displacement data of the support, Dis, through the aforementioned proportional relationship. mv .
[0075] 10. Monitoring of support rotation angle requires preset target points, as shown in the figure. Target point detection and scale segmentation model can be calculated synchronously, and displacement and rotation angle values can be measured in near real-time (second level).
[0076]
[0077] 11. By analyzing the actual displacement value Dis of the support mv Real-time monitoring of the bearing rotation angle R is used to understand the service status of the bearing and to respond promptly.
[0078] 12. For the shear deformation of rubber bearings, target points are also arranged, and shear deformation is monitored by locating the target points.
[0079] Target placement method:
[0080] Target points are placed at the top and bottom of the centerline of the rubber bearing. The method is analogous to bearing rotation detection. Real-time monitoring of shear deformation can be achieved by using a single-stage target detection network on the rubber bearing target points. Figure 4 As shown.
[0081] Camera selection and installation
[0082] (1) No special requirements for the camera. The standard configuration is a high-definition PTZ camera with more than 2 million pixels and infrared fill light. The horizontal range is 360° and the vertical range is -15° to 90°. The network interface supported is RJ45 network port, which is adaptive to 10M / 100M network data and supports 4G / 5G transmission, etc.
[0083] (2) In order to collect images of markers that can be used for monitoring, the deployment of cameras should meet the following requirements:
[0084] 1. The central axis of the camera lens should be as collinear as possible with the central axis of the reference scale.
[0085] 2. The main content of the shot is filled by markers that form the frame.
[0086] 3. The camera is installed stably and reliably.
[0087] 4. Ensure the camera captures detailed information about the markers.
[0088] • Diseases identified through camera-based visual inspection
[0089] Based on a camera, the following surface features of the support can be identified through visual inspection:
[0090] —Cracks in steel components
[0091] — Steel component detachment
[0092] —Skateboard wear or extrusion
[0093] — Bolts that are broken, missing, or loose
[0094] ——rust
[0095] —Support detached
[0096] —Support position movement
[0097] —Exposed steel plate
[0098] Surface cracks
[0099] • When dust covers are installed around the support
[0100] At this time, the space for camera deployment is limited, and it is difficult to face the reference ruler or measuring ruler directly. In this case, a small high-definition camera can be selected and deployed in an adjacent position. Based on the Sobel operator, Hough transform, and homography transform, the tilted image is corrected and preprocessed. The corrected image is then semantically segmented and target detected according to the above methods to achieve displacement and rotation monitoring.
[0101] Example 2
[0102] Embodiment 2 of the present invention provides an intelligent monitoring method for bridge bearing displacement and typical defects, for bridge bearings with built-in scales and pointers, the method comprising:
[0103] Target points are set on the upper bearing plate of the bridge bearing to be monitored;
[0104] Using a ruler and pointer as monitoring targets, images including target points, rulers, and pointers are acquired.
[0105] A single-stage detection and segmentation network is constructed to simultaneously achieve object detection and semantic segmentation.
[0106] Based on the acquired images, the displacement and rotation angle of the bridge supports under monitoring are further obtained using a detection and segmentation network model.
[0107] Monitoring of shear deformation of bridge bearings is achieved by locating target points.
[0108] Based on the acquired support displacement, support rotation angle, and shear deformation monitoring, and simultaneously performing real-time visual observation of the supports, typical defects can be monitored.
[0109] Typical defects include, but are not limited to, cracks and weld failures in steel components, wear or extrusion of the sliding plate, shearing or loosening of bolts, corrosion, bearing voids, and surface cracks. Other defects may include excessive shear deformation, localized cracking, voids in the bearing surface, cracks in steel components, fractures and corrosion deformation, extrusion of rubber plates and sliding plates, and deformation and breakage of anchor bolts. Specifically, according to the type of bridge bearing, typical defects of plate rubber bearings include bearing voids, rubber cracks, exposed steel plates, uneven bulging of rubber, and excessive shear; typical defects of pot bearings include wear or extrusion of the sliding plate, corrosion of steel components, extrusion of the bearing rubber plate, immobility of movable bearings, excessive bearing displacement, and bearing non-rotation or excessive rotation angle; typical defects of spherical bearings include damage to steel components, failure of anchors and positioning components (such as shearing of anchor bolts), excessive displacement, immobility of movable bearings, excessive rotation angle, poor contact between the upper and lower bearing plates and the beam bottom and bearing pad, and defects in the bearing pad area.
[0110] Scale Selection and Installation
[0111] To accurately obtain bearing displacement, a scale is required as a marking fixture. This embodiment directly utilizes the pointer and scale that come with the bearing from the factory. The scale range is mainly determined based on the design displacement levels of the multi-directional (DX) and longitudinal (ZX) movable bearings: longitudinal design displacement levels are ±30mm, ±50mm, ±100mm, ±150mm, ±200mm, ±250mm, and ±300mm; transverse design displacement levels for multi-directional (DX) and lateral (HX) movable bearings are divided into four levels: ±10mm, ±20mm, ±30mm, and ±40mm. The scale range should not be less than the bearing design displacement level.
[0112] To accurately measure support displacement, a 1mm precision center-division scale was selected as the monitoring scale. The direction and value of the support displacement can then be precisely obtained by monitoring the relative displacement of the pointer to the 0 mark on the scale. The specific steps are as follows:
[0113] 1. Preset target points T1 and T2 on the upper plate of the support, ensuring that T1 and T2 are on the same horizontal line. The length D of the lower plate of the support is known. The vertical distances from the left and right target points to the center line of the upper plate of the support are D / 2 respectively.
[0114] 2. Select a preset target point as the corner monitoring marker, and select the scale and pointer built into the support itself as the displacement monitoring marker. The scale length is dis. real It is a known quantity. For example... Figure 3 .
[0115] 3. Install a camera on one side of the target point and scale of the bridge support. To ensure high-quality image acquisition of the marker and complete and usable image information, the center axis of the lens should be as collinear as possible with the center axis of the scale, and the camera should be stable and able to capture detailed information of the marker.
[0116] 4. Collect 1,000-3,000 images of bridge bearings (including preset target points and scale pointers) taken at different time points of day and night, and label the target points, scales and pointers in the images to generate sample set A.
[0117] 5. Construct a single-stage detection and segmentation network. Examples include YOLOv5 and FPN, but not limited to these. Add a small target detection layer to the YOLOv5 network, or add a fusion factor to the FPN-based detection network to improve target detection accuracy; add a convolutional segmentation head to the YOLOv5 network, or add a pixel stream segmentation network to the FPN-based detection network to complete semantic segmentation.
[0118] 6. Based on sample set A, train a detection and segmentation network model to simultaneously detect marker targets and pointers, as well as segment the scale. It is recommended that 80% of the sample data A1 be used for model training, and 20% of the sample data A2 be used for model performance testing.
[0119] 7. After the image is processed by the trained detection and segmentation network model, the segmented image P of the marker scale range, the left target point T1 and its center pixel coordinates (X) are calculated. t1 ,Y t1 ), right target point T2 and its center pixel coordinates (X) t2 ,Y t2 ) and the center point C of the pointer p .
[0120] 8. Based on the segmented image P of the scale, extract the scale contour information, classify the contour points based on the position information, and perform straight line fitting on the scale edge according to the classified contour points to obtain the fitted straight line of each edge. Among them, the fitted straight line of the upper and lower edges is the horizontal straight line of the view.
[0121] 9. Based on the profile fitting of the ruler, obtain the 0 mark point C of the ruler. r and the pixel distance corresponding to the scale length dis px By calculating pixel distance dis px dis with actual length real The proportional relationship is used to convert pixels to actual distances.
[0122] 10. Calculate the center point C of the pointer. p The perpendicular intersection point M with the scale is automatically obtained relative to the 0 mark point C on the scale. r The left and right pixel displacement values are then converted into the actual displacement value Dis of the support through the aforementioned proportional relationship. mv .
[0123] 11. Based on the pixel coordinates of target points T1 and T2, the support rotation angle R can be obtained.
[0124]
[0125] Arrangement ideas
[0126] The tooling is retained. After being put into operation, for special bearings of important bridges, real-time monitoring can be carried out using the network conditions of the bridge health monitoring system to replace the traditional bearing displacement monitoring method. For other types of bridge bearings of concern, after being put into operation, monitoring can be carried out for one year (through a complete summer and winter) to obtain the initial state and development pattern of the bearings when they are working normally. In the later stage, monitoring can be carried out every 5 years / 10 years to evaluate the deterioration pattern of the bearings. When the bearings have serious defects, they can be monitored in real time, thus achieving the maximum input-output ratio.
[0127] The core key points and protection points of this article include:
[0128] (1) By combining the scale and pointer of the bridge bearing itself or by setting a fixed scale on the bearing pad stone and a movable scale on the bearing upper plate, the bridge bearing displacement is identified in near real time based on the detection segmentation network, with an identification accuracy of 1mm.
[0129] (2) By setting two target points at both ends of the upper bearing plate of the bridge bearing, the accurate identification of the bearing rotation angle can be realized in near real time, with an identification accuracy of 0.0002 rad.
[0130] (3) By setting two target points at the upper and lower ends of the center line of the rubber support, the accurate identification of shear deformation can be achieved.
[0131] (4) Real-time monitoring of surface defects of the support is achieved by setting up a camera.
[0132] (5) The bridge bearing inspection mechanism is upgraded and rebuilt based on this monitoring technology.
[0133] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art should understand that modifications or equivalent substitutions to the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A method for intelligent monitoring of bridge bearing displacement and typical defects, used for bridge bearings without built-in scales and pointers, the method comprising: Install a movable ruler and a reference ruler on the upper bearing plate and the pad stone of the bridge bearing to be monitored, respectively; Target points are arranged on the upper seat plate of the bridge support to be monitored; specifically, left and right target points are arranged on the upper seat plate of the bridge support to be monitored , the two target points are located on the same horizontal line, and the vertical distance to the center line of the upper seat plate of the support is , wherein is the length of the lower seat plate of the support . Using the movable ruler and the reference ruler as monitoring targets, images of the target point, the movable ruler, and the reference ruler are collected. Based on the acquired images, the displacement and rotation data of the bridge supports under monitoring are obtained using a pre-established and trained detection and segmentation network model; specifically including: The acquired images are input into a pre-established and trained detection and segmentation network model to obtain segmented images covering the measurement range of both the movable ruler and the reference ruler. Left target point center pixel coordinates Right target center pixel coordinates ; Based on image segmentation The contour information of the active ruler and the reference ruler is extracted, and the points on the contour are classified based on the position information to obtain the contour fitting line. Among them, the fitting line of the upper and lower edges of the reference ruler is the horizontal line of the view. The zero mark of the movable ruler is obtained by fitting a straight line along the edges of the movable ruler and the reference ruler. 0 mark on the reference ruler and the pixel distance corresponding to the reference ruler length By calculating pixel distance Compared to the actual length The proportional relationship is used to obtain the proportional relationship for the conversion from pixel to actual distance; Calculate the 0 mark on the movable ruler Perpendicular intersection with the reference ruler Automatically obtain Relative to the 0 mark of the reference ruler The left and right pixel displacement values are then converted into the actual displacement of the support through the aforementioned proportional relationship. ; The support rotation angle is obtained from the following formula. : ； The detection and segmentation network model is based on a single-stage detection network. It improves the detection accuracy of targets by adding a small target detection layer or a fusion factor, and completes target detection and semantic segmentation simultaneously by adding a segmentation network. Based on the acquired support displacement and support rotation data, and simultaneously perform real-time visual observation of the supports, the monitoring of typical defects can be achieved.
2. The intelligent monitoring method for bridge bearing displacement and typical defects according to claim 1, characterized in that, The movable ruler and the reference ruler are of the same length. .
3. The intelligent monitoring method for bridge bearing displacement and typical defects according to claim 1, characterized in that, The method further includes: for bridge bearings made of rubber, monitoring the shear deformation of the bridge bearing to be monitored is achieved based on target point localization; specifically, it includes: arranging a target point at the top and bottom of the centerline of the bridge bearing, and performing target detection on the collected target point images based on a pre-established and trained detection and segmentation network model to achieve real-time monitoring of shear deformation.
4. The intelligent monitoring method for bridge bearing displacement and typical defects according to claim 1, characterized in that, Typical defects include cracks and weld failures in steel components, wear or extrusion of sliding plates, shearing, missing or loose bolts, corrosion, voids in supports, and surface cracks.
5. A method for intelligent monitoring of bridge bearing displacement and typical defects, used for bridge bearings with built-in scales and pointers, the method comprising: Target points are set on the upper bearing plate of the bridge bearing to be monitored; Specifically, it includes: Left target points were set on the upper bearing plate of the bridge bearing to be monitored. and right target The two target points are located on the same horizontal line, and the vertical distance to the center line of the upper support plate is the same. ,in The length of the lower support plate of the support; Using a ruler and pointer as monitoring targets, images including target points, rulers, and pointers are acquired. Based on the acquired images, the displacement and rotation data of the bridge supports under monitoring are obtained using a pre-established and trained detection and segmentation network model; specifically including: The acquired images are input into a pre-established and trained detection and segmentation network model to obtain segmented images within the scale range. Left target point center pixel coordinates Right target center pixel coordinates and the center point of the pointer ; Scale-based image segmentation Extract the ruler contour information, classify the contour points based on the position information, and perform straight line fitting on the ruler edge according to the classified contour points to obtain the contour fitting line. Among them, the upper and lower edge fitting lines are the horizontal lines of the view. By fitting a straight line to the profile of the scale, the 0 mark of the scale is obtained. and the pixel distance corresponding to the ruler length By calculating pixel distance Compared with the known actual length The proportional relationship is used to obtain the proportional relationship of pixel to actual distance conversion; Calculate the center point of the pointer Perpendicular intersection with the scale , get Compared to the 0 mark on the scale The left and right pixel displacement values are converted into the actual displacement value of the support according to the above proportional relationship. ; The support rotation angle is obtained from the following formula. : ； The detection and segmentation network model is based on a single-stage detection network. It improves the detection accuracy of targets by adding a small target detection layer or a fusion factor, and completes target detection and semantic segmentation simultaneously by adding a segmentation network. Based on the acquired support displacement and support rotation data, and simultaneously perform real-time visual observation of the supports, the monitoring of typical defects can be achieved.