Intelligent pre-pressing detection system and method based on vision measurement

Abstract

The invention provides an intelligent pre-pressing detection system and method based on vision measurement. The intelligent pre-pressing detection system based on vision measurement comprises a pre-pressing loading module, a data acquisition module and an analysis and control module, according to the invention, a reaction frame in the pre-pressing loading module uses a stable foundation beside a to-be-pressed support structure as a fulcrum, a certain load is applied to the to-be-pressed support structure through a pressure applying part in the reaction frame, and the top of the to-be-pressed support structure is loaded while inelastic deformation is eliminated; the data acquisition module can automatically acquire bracket displacement deformation image data in the whole prepressing loading process through a combined target mounted on a bracket structure to be pressed, and transmits the bracket displacement deformation image data into the analysis and control module for deformation data analysis and prepressing control; in the actual loading process, the change rate of the to-be-pressed support structure is judged, when the change rate exceeds the preset safety threshold value, loading can be stopped in time, an alarm is given, and loading safety is guaranteed.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of bridge construction, and particularly relates to an intelligent pre-press detection system and method based on visual measurement. BACKGROUND

[0002] The pre-press construction before the bridge hanging basket cantilever pouring is a key process for ensuring the safety and linear precision of bridge construction. Its core function is to simulate the real pouring state to eliminate the non-elastic deformation and measure the elastic deformation value, so as to provide accurate pre-lifting data for the formwork elevation of subsequent segments. The pre-press quality is directly related to the smoothness of the bridge line, and is also a comprehensive test of the design, manufacture and installation quality of the hanging basket.

[0003] Traditionally, precast blocks are usually used to simulate the main beam load pre-press, and in recent years, counterforce frames have gradually been used for pre-press. However, the deformation measurement during the pre-press process completely relies on manual measurement by measurement personnel, the counterforce frame loading and unloading control relies on manual operation of hydraulic jacks, the continuity of elastic deformation measurement data is poor, and the pre-press safety guarantee is low. SUMMARY

[0004] In view of the deficiencies in the prior art, the application provides an intelligent pre-press detection system and method based on visual measurement, which is based on remote visual measurement technology, has automatic measurement, elastic deformation full-process measurement, counterforce frame pre-press automatic control functions, and can realize automatic and safe pre-press of the hanging basket cantilever pouring bridge.

[0005] The technical scheme of the application is as follows: an intelligent pre-press detection system based on visual measurement, comprising:

[0006] a pre-press loading module, a data acquisition module and an analysis and control module;

[0007] The pre-press loading module comprises a counterforce frame arranged above a to-be-pressed support structure, the counterforce frame is fixedly arranged on a stable foundation beside the to-be-pressed support structure, and a pressure applying component for applying pre-press force to the to-be-pressed support structure is arranged on the counterforce frame.

[0008] The data acquisition module comprises a plurality of combined targets arranged on the to-be-pressed support structure and a visual camera for acquiring image information of the combined targets.

[0009] The analysis and control module is in communication connection with the visual camera, is used for receiving image information and processing through a visual measurement algorithm, and obtains the spatial displacement value of the combined target.

[0010] The analysis and control module is in control connection with the pressure applying component, and is used for controlling the pressure applying component to load or unload the to-be-pressed support structure according to a preset loading program.

[0011] Further, the combination target comprises a prism target and an infrared target, the prism target is used for total station calibration.

[0012] The visual camera comprises an infrared camera, and the infrared target is used for image acquisition of the infrared camera.

[0013] Further, the pressure applying component is a jack or an oil cylinder.

[0014] The intelligent pre-pressing detection method based on visual measurement comprises the following steps:

[0015] Pre-pressing position component installation: fixing the counterforce frame on a stable foundation beside the to-be-pressed support structure, installing a combination target at a designed measurement point of the to-be-pressed support structure, and setting the combination target towards a planned setting position of a visual camera;

[0016] Acquisition device installation: setting the visual camera at a preset position, and adjusting the visual camera to clearly capture images of the combination target;

[0017] Dynamic calibration and measurement: capturing images of the combination target by the visual camera in real time, dynamically solving a three-dimensional displacement field of each monitoring point based on a visual measurement algorithm, and obtaining a deformation result;

[0018] Graded loading and intelligent monitoring: controlling the pressure applying component to perform step-by-step loading on the to-be-pressed support structure; during and after each loading process, continuously obtaining the deformation result;

[0019] During the loading process, the change trend of the deformation result is judged in real time:

[0020] If the change rate of the deformation result exceeds a preset safety threshold, it is determined that the curve is suddenly changed, the pressure applying component is immediately controlled to stop loading or perform unloading, and an alarm signal is sent;

[0021] If the change rate of the deformation result gradually decreases and is lower than a preset stability threshold within a preset time period, it is determined that the deformation is convergent and stable, a signal of completing the current loading is then sent, and the next loading process is entered.

[0022] Further, the visual measurement algorithm comprises the following steps:

[0023] Feature points of the combination target in the image are extracted, and pixel coordinates thereof are obtained;

[0024] Based on camera calibration parameters, preliminary three-dimensional coordinate solving is performed;

[0025] A physical constraint model of the to-be-pressed support structure is introduced as a filtering condition to optimize the solved three-dimensional coordinates, and the physical constraint model comprises a relative position relationship constraint between support nodes.

[0026] Output the optimized three-dimensional coordinates and calculate the spatial displacement value by comparing the coordinate values at different times.

[0027] Further, the visual camera captures images of the combined targets in real time using a multi-target synchronous measurement method, which includes the following steps:

[0028] In the first frame of image obtained by the system, all visible combined targets are recognized and segmented, and each recognized combined target is assigned a unique identity identifier;

[0029] In each subsequent frame of image, all combined targets are re-identified and located, and the detection results of the current frame are matched with the existing identity identifiers based on the appearance features, motion trajectory prediction or location proximity principle of the combined targets;

[0030] For successfully matched combined targets, their current position information is used to update the corresponding motion trajectory; for failed matched combined targets, they are initialized as new trajectories or marked as temporarily lost according to the preset rules.

[0031] Further, in the dynamic calibration and measurement step, the captured target images are processed according to the following steps to eliminate the centroid positioning error introduced by the imaging size difference of the targets:

[0032] The binarized target blob image is subjected to morphological closing operation processing, and the size of the structure element used is determined according to the expected imaging size and maximum imaging deviation of the target;

[0033] The parameters of the morphological processing are adjusted until the discrete spot regions belonging to the same target are fused into a single connected domain, and the pixel area difference of the connected domains formed by different targets is controlled within a preset tolerance range;

[0034] Based on the processed and area-normalized connected domains, the centroid coordinates of each target are calculated.

[0035] Further, the visual measurement algorithm further includes a calibration algorithm, which includes the following steps:

[0036] To cope with the slow creep and instantaneous vibration of the support during loading, a dynamic filtering method based on time series analysis is used to distinguish between the true displacement of the structure and the instantaneous jitter, wherein the parameters of the filtering method are adaptively adjusted according to the expected motion model of the combined targets;

[0037] To compensate for the thermal expansion and contraction of the support structure caused by uneven sunlight, a system error model related to the temperature field distribution is established, and real-time data read by temperature sensors arranged in the monitoring area are used to compensate the three-dimensional coordinates obtained by visual measurement in real time.

[0038] Further, the intelligent pre-press detection method based on visual measurement further comprises a report generation step:

[0039] During the whole process of hierarchical loading and intelligent monitoring, the load value and time stamp corresponding to each deformation result are recorded synchronously;

[0040] Based on the recorded time sequence data, a curve reflecting the deformation of the pre-press support structure and the load-time relationship is automatically generated.

[0041] According to the preset report template and engineering specification, the deformation-time-load curve, maximum displacement value, and stability state judgment result are integrated to automatically output the pre-press detection result report.

[0042] Further, the step-by-step loading includes at least one initial loading stage, one intermediate loading stage, one full load stage, and one overload stage, and the load value of the overload stage is greater than that of the full load stage.

[0043] The beneficial effects of the present application are:

[0044] (1) In the present application, the counterforce frame in the pre-press loading module uses the stable foundation beside the pre-press support structure as a fulcrum, and applies a certain load to the pre-press support structure through the pressing component in the counterforce frame to eliminate non-elastic deformation and load the top of the pre-press support structure;

[0045] (2) The data acquisition module can automatically collect pre-press loading process support displacement deformation image data by installing a combined target on the pre-press support structure, and transmit the data to the analysis and control module for deformation data analysis and pre-press control;

[0046] (3) In the actual loading process, the change rate of the pre-press support structure is judged, and when it exceeds the preset safety threshold, the loading can be stopped in time and an alarm is given to ensure the safety of the loading. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 The figure is a structural schematic diagram of the intelligent pre-press detection system based on visual measurement in the present application.

[0048] Figure 2 The figure is a structural schematic diagram of one embodiment of the pre-press loading module in the present application.

[0049] Figure 3 The figure is a schematic diagram of the visual camera focusing target in the present application.

[0050] Figure 4 Flow chart of the intelligent pre-press detection method based on visual measurement in the present application. DETAILED DESCRIPTION

[0051] Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. The description of the exemplary embodiments is merely illustrative in nature and is in no way intended to limit the application, its application, or its uses, to the specific embodiments described. The application can be embodied in a multitude of different forms and should not be construed as limited to the embodiments set forth herein. These embodiments are described so that this disclosure will be thorough and complete, and fully convey the scope of the application to those skilled in the art. It should be noted that the relative arrangement of components and steps, the numerical expressions, and the numerical values set forth in these embodiments are to be interpreted as merely exemplary, and not as a limitation of the application unless otherwise specifically indicated.

[0052] The terms "first", "second", and similar terms in the present application do not denote any order, number, or importance, but are only used to distinguish different parts. The terms "comprise", "include", and similar terms mean that the elements before the term encompass the elements listed after the term, and do not exclude the possibility of also encompassing other elements. "Up", "down", "left", "right", and the like are only used to indicate relative positional relationships, and when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0053] As shown in Figure 1 and 2 A visual measurement-based intelligent pre-press detection system is disclosed, comprising a pre-press loading module 300, a data acquisition module 200, and an analysis and control module 100. The pre-press loading module 300 comprises a counterforce frame 91 arranged above the support structure to be pressed, the counterforce frame 91 being fixedly arranged on a stable foundation beside the support structure to be pressed 93, and a pressure applying component 92 being arranged on the counterforce frame 91 for applying pre-press to the support structure to be pressed 93. The data acquisition module 200 comprises a plurality of combined targets arranged on the support structure to be pressed 93 and a visual camera for acquiring image information of the combined targets. The analysis and control module 100 is in communication connection with the visual camera, for receiving image information and processing through a visual measurement algorithm to obtain the spatial displacement value of the combined targets. The analysis and control module 100 is in control connection with the pressure applying component 92, for controlling the pressure applying component 92 to load or unload the support structure to be pressed 93 according to a pre-set loading program.

[0054] In the above embodiment, the counter-force frame 91 in the preloading module 300 uses the stable foundation beside the to-be-pressed support structure 93 as a fulcrum, applies a certain load to the to-be-pressed support structure 93 through the pressing component 92 in the counter-force frame 91, eliminates the non-elastic deformation, and loads the top of the to-be-pressed support structure 93; the data acquisition module 200 can automatically acquire the support displacement deformation image data in the whole preloading process by installing a combined target on the to-be-pressed support structure 93, and transmit the data to the analysis and control module 100 for deformation data analysis and preloading control.

[0055] In some embodiments, the combined target includes a prism target and an infrared target, the prism target is used for total station calibration; specifically, the prism target is an L-shaped prism target, when the air temperature or air state changes greatly, the position of the L-shaped prism target can be measured by the total station, the change of the external parameter of the visual camera is corrected, and the accuracy of the visual measurement technology in long-time measurement in actual engineering application can be significantly improved; the visual camera includes an infrared camera, and the infrared target is used for image acquisition of the infrared camera; the infrared target is a self-luminous target with narrow beam emission capability, the target displacement is captured in real time by a high-performance infrared camera array, and the embedded industrial computer dynamically calculates the three-dimensional displacement field of the monitoring point based on the camera calibration parameter and the space coordinate conversion algorithm, generates a millimeter-level precision deformation result by acquiring more than one frame of image data per second; more specifically, the bottom of the combined target is supported by a triangular support.

[0056] In some embodiments, the principle of focus adjustment of the visual camera is to make the target in the center of the frame, but not too close to the edge of the frame, as shown in Figure 3 , so that the distance between the target edge and the left, right, top and bottom edges of the visual camera frame is less than about 1 / 6 of the length of the frame edge; when there is only one row of targets in a certain direction, the targets are in the central position in this direction; a gimbal is arranged on the visual camera to adjust the angle of the visual camera.

[0057] In some embodiments, as shown in Figure 2 , the counter-force frame 91 is fixedly arranged on one side of the concrete main beam 95, and a jack or an oil cylinder is arranged in the counter-force frame 91 as the pressing component 92 to press the to-be-pressed support structure 93 below the counter-force frame 91 by mechanical energy; specifically, the counter-force frame 91 is fixedly arranged on one side of the concrete main beam 95 through the pre-embedded anchor 94 pre-embedded in the concrete main beam 95.

[0058] In some embodiments, the pressing component 92 is a jack or an oil cylinder; specifically, the pressing component 92 is a numerical control jack, which can remotely communicate with the control component such as an industrial computer to receive a control signal to realize the loading and unloading functions.

[0059] In some embodiments, as shown in Figure 4As shown, an intelligent pre-press detection method based on visual measurement is disclosed, comprising the following steps:

[0060] Pre-press position component installation, the counterforce frame 91 is fixedly arranged on the stable foundation beside the to-be-pressed support structure 93, the combined target is installed at the designed measurement point of the to-be-pressed support structure 93, and the combined target is arranged towards the preset position of the visual camera;

[0061] Acquisition device installation, the visual camera is arranged at the preset position, and the visual camera is adjusted to clearly capture the image of the combined target;

[0062] Dynamic calibration and measurement, the image of the combined target is captured in real time through the visual camera, the three-dimensional displacement field of each monitoring point is dynamically solved based on the visual measurement algorithm, and the deformation result is obtained;

[0063] Graded loading and intelligent monitoring, the to-be-pressed support structure is loaded by the pressing component in stages; during each loading process and after the loading is completed, the deformation result is continuously obtained;

[0064] During the loading process, the change trend of the deformation result is judged in real time:

[0065] If the change rate of the deformation result exceeds the preset safety threshold, it is determined that the curve is suddenly changed, the pressing component is immediately controlled to stop loading or perform unloading, and an alarm signal is sent;

[0066] If the change rate of the deformation result gradually decreases and is lower than the preset stability threshold within the preset time period, it is determined that the deformation converges stably, then a signal that the current loading is completed is sent, and the next loading process is entered.

[0067] The intelligent pre-press detection method based on visual measurement in the above embodiment can judge the change rate of the to-be-pressed support structure during the actual loading process, stop loading in time and alarm when the change rate exceeds the preset safety threshold, and ensure the loading safety.

[0068] In some embodiments, the visual measurement algorithm comprises the following steps:

[0069] Feature points of the combined target in the image are extracted, and pixel coordinates thereof are obtained;

[0070] Based on the camera calibration parameters, preliminary three-dimensional coordinate solving is performed;

[0071] The physical constraint model of the to-be-pressed support structure is introduced as a filtering condition to optimize the solved three-dimensional coordinates, and the physical constraint model comprises the relative position relationship constraint between support nodes;

[0072] The optimized three-dimensional coordinates are output, and the spatial displacement value is calculated by comparing the coordinate values at different times.

[0073] As a more specific embodiment of the above embodiment, the visual measurement algorithm comprises the following steps:

[0074] Image feature extraction: the analysis and control module receives one or more frames of images from the visual camera, pre-processes the images, including grayscale, Gaussian filter denoising, to enhance image quality, automatically identifies the preset feature points of each combined target in the image using a feature point detection algorithm based on Blob detection, and accurately calculates the sub-pixel level coordinates (u, v) of these feature points in the image pixel coordinate system;

[0075] Preliminary three-dimensional coordinate calculation: call the camera parameters obtained in advance through high-precision camera calibration, including camera intrinsic parameters and extrinsic parameters, the intrinsic parameters include focal length f x ,f y , principal point coordinates c x ,c y , distortion coefficients k1, k2, p1, p2, etc., and the extrinsic parameters include the position T and rotation R of the camera relative to the world coordinate system; for a monocular camera system, use the PnP algorithm to calculate the preliminary three-dimensional coordinates P initial =(X i ,Y i ,Z i ) of the feature points in combination with the known initial coordinates of the target in the world coordinate system; for a multi-camera system, use triangulation or forward intersection algorithm to convert the pixel coordinates of the same feature point under different viewing angles into preliminary three-dimensional coordinates; in this embodiment, a monocular camera system is used;

[0076] Coordinate optimization based on physical constraint model: a simplified mechanical model describing the structure of the support to be pressed is pre-stored or constructed in real time in the algorithm, which defines a series of relative position relationship constraints between nodes, the most typical one being the rigid rod assumption, that is, the length of the support rod connecting two nodes is considered to remain unchanged in a short time; for any rod defined as rigid in the model, connecting nodes i and j, the theoretical rod length L ij is known, and the optimization problem is constructed as follows: find a set of optimal three-dimensional coordinates P optimized , so that the sum of the square errors between the calculated rod length and the theoretical rod length is minimized; the objective function is:

[0077] M inimize ∑[||P i -P j ||-L ij ] 2

[0078] Where P i and P j are the optimized node coordinates, and ||P i -P j|| is the optimized rod calculation length;

[0079] The preliminary three-dimensional coordinates P initial as the initial value of optimization, substitute into the above objective function; adopt the nonlinear least square optimization algorithm to solve iteratively, and finally output a set of more accurate three-dimensional coordinates P optimized that satisfy the physical constraints.

[0080] Displacement calculation and output: compare the optimized coordinates P optimized (t) of the current frame with the coordinates P optimized (t-1) of the previous frame or the reference frame; for each target point, calculate its displacement components (ΔX, ΔY, ΔZ) in the X, Y, and Z directions and the total displacement; output these spatial displacement values and pass them to the subsequent control logic unit for generating the time-history-deformation curve and intelligent loading decision.

[0081] In some embodiments, the visual camera captures images of the combined targets in real time using a multi-target synchronous measurement method, which includes the following steps:

[0082] In the first frame of image obtained by the system, all visible combined targets are recognized and segmented, and each recognized combined target is assigned a unique identity identifier;

[0083] In each subsequent frame of image, all combined targets are re-recognized and located, and the detection results of the current frame are matched with the existing identity identifiers based on the appearance features, motion trajectory prediction, or location proximity principle of the combined targets;

[0084] For successfully matched combined targets, their current position information is used to update the corresponding motion trajectory; for failed matched combined targets, they are initialized as new trajectories or marked as temporarily lost according to the preset rules.

[0085] As a more specific embodiment of the above embodiment, the multi-target synchronous measurement method includes the following steps:

[0086] Initial frame target recognition and identity identifier assignment: use a pre-trained target detection neural network or a feature-based image segmentation algorithm to recognize all visible combined targets in the image and accurately segment each target from the image background; extract the unique feature information of each segmented target region; if the target is a coded target, directly decode its built-in ID information; if it is a non-coded target, extract its appearance features and geometric features. Then, the system assigns a unique identity identifier to each target and creates a new motion trajectory for each target using the ID as an index. The trajectory will record its timestamp.

[0087] Subsequent frame data association: for each subsequent frame, repeat the target detection and localization, obtain the detection result set D t = {d1, d2,..., dm}; next, match D t with the existing trajectory set T_t-1 = {T1, T2,..., Tn}; the matching is based on the integrated judgment of one or more of the following principles:

[0088] Motion trajectory prediction: for each existing trajectory T i , use Kalman filter or particle filter algorithm to predict its expected position P predicted in the current frame t according to its motion state (position, velocity) of the previous frames; then, calculate the Euclidean distance between each detection result d j and P predicted , forming a motion association cost matrix; the smaller the distance, the higher the association possibility;

[0089] Appearance feature matching: calculate the similarity between the appearance feature of the current frame detection result d j and the historical appearance feature stored in the existing trajectory T i , forming an appearance association cost matrix; this method is particularly effective for solving the ID exchange problem caused by target crossing and temporary occlusion and then reappearing;

[0090] Position proximity: as an auxiliary means, simply calculate the distance between the detection result d j and the trajectory T i in the position P t-1 of the previous frame, which is very effective in the case of slow target motion or high frame rate;

[0091] Finally, use the Hungarian algorithm or greedy algorithm to comprehensively consider the above cost matrices to assign the most reasonable existing identity identifier to the detection result of the current frame, achieving global optimal or local optimal matching;

[0092] Trajectory update and management: according to the above matching structure, the system adopts different trajectory management strategies:

[0093] Successful matching: for the successfully associated <detection result d j , trajectory T i > pair, use the current position information of d j to update the trajectory T i ; at the same time, update the state of its Kalman filter and reset the life cycle counter of the trajectory;

[0094] Matching failure, new trajectory initialization: for a detection result detected in the current frame but failed to match any existing trajectory, it is determined as a newly appeared target; the system assigns it a new, unused identity identifier and creates a new trajectory for it;

[0095] Matching failure, trajectory temporarily lost: for an existing trajectory T i If no detection result is found in the current frame to match, it is determined that the target may be temporarily occluded; the system does not immediately delete the trajectory, but marks it as "temporarily lost" and continues to predict its position using the Kalman filter, while searching in the vicinity of its predicted position in subsequent frames; if N consecutive frames are not successfully re-matched, it is determined that the target has permanently disappeared, and the trajectory is terminated.

[0096] The embodiment realizes stable and synchronous tracking of a large number of targets in a complex construction environment through data association and trajectory management mechanism. It effectively solves the ID jump or loss problem caused by target occlusion, similar appearance and rapid motion, etc., ensures the generation of continuous, accurate and unique displacement and time course correlation data for each monitoring point, and provides a reliable data basis for subsequent structure safety evaluation. The method greatly improves the efficiency and reliability of automatic monitoring and reduces manual intervention.

[0097] In some embodiments, in the dynamic calibration and measurement step, the target image collected is processed according to the following steps to eliminate the centroid positioning error introduced by the size difference of the target imaging:

[0098] The binarized target blob image is subjected to morphological closing operation processing, and the size of the structural element used is determined according to the expected imaging size and maximum imaging deviation of the target;

[0099] The parameters of morphological processing are adjusted until the discrete spot regions belonging to the same target are fused into a single connected domain, and the pixel area difference of the connected domains formed by different targets is controlled within a preset tolerance range;

[0100] Based on the processed and area-normalized connected domains, the centroid coordinates of each target are calculated.

[0101] As a more specific embodiment of the above embodiment, it includes the following steps:

[0102] Image preprocessing and binarization: after the collected target image is grayed and Gaussian filtered, adaptive threshold algorithm is used for binarization processing to separate the target blob and the background;

[0103] Adaptive morphological closing operation: Morphological closing operation is performed using a circular structuring element, whose initial radius r is set according to the expected imaging radius R0 of the target in the image, for example, r = 0.5 × R0; the radius of the structuring element is increased iteratively until the following condition is met:

[0104] Region fusion: Discrete light spots formed by reflection from the same target are fused into a single connected region;

[0105] Area balancing: Calculate the area of ​​all connected components such that the ratio of their standard deviation to their mean is less than a set threshold, which is set to 15% as an example;

[0106] Centroid Calculation and Output: For the morphologically processed normalized connected components, calculate their pixel-level geometric centroid coordinates and use these coordinates as the final positioning result of the target for subsequent spatial coordinate calculation.

[0107] In this embodiment, adaptive morphological operations effectively eliminate the differences in aperture size caused by imaging angle, ensuring that the centroid of each target is located at a consistent scale, significantly reducing system errors and improving the overall accuracy and reliability of three-dimensional displacement measurement.

[0108] In some embodiments, the visual measurement algorithm further includes a calibration algorithm, which includes the following steps:

[0109] To address the slow creep and instantaneous vibration of the support during loading, a dynamic filtering method based on time series analysis is adopted to distinguish between the actual displacement of the structure and the instantaneous jitter. The parameters of the filtering method are adaptively adjusted according to the expected motion model of the combined target.

[0110] To compensate for the thermal expansion and contraction of the support structure caused by uneven sunlight, a systematic error model related to the temperature field distribution was established, and the three-dimensional coordinates obtained by visual measurement were compensated in real time by reading real-time data from temperature sensors deployed in the monitoring area.

[0111] As a more specific implementation of the above embodiments, the calibration algorithm includes the following steps:

[0112] Dynamic displacement extraction based on adaptive filtering: For mixed displacement signals containing slow creep and instantaneous vibration during loading, a state-space model is constructed. The state variables of this model include the target's actual displacement, velocity, and possible low-frequency drift. Adaptive Kalman filtering or recursive least squares algorithm is used for processing. The process noise covariance matrix Q and observation noise covariance matrix R of the filter are not fixed values, but are adjusted online according to the expected motion model of the combined target. For example, during the loading holding phase, when the expected displacement changes slowly, the noise estimation related to velocity is automatically reduced, making the filter more confident in the model prediction, thereby strongly suppressing high-frequency vibration noise. At the instant of load change, the noise estimation is increased accordingly, making the filter more confident in sensor observations, so as to quickly respond to the actual displacement. Finally, the filter outputs smooth, true structural displacement time-series data after removing instantaneous jitter.

[0113] Real-time system error compensation based on temperature field modeling: Multiple high-precision temperature sensors are deployed at key locations within the monitoring area to form a distributed temperature monitoring network, such as the top, bottom, sun-facing side, and shaded side of the support frame; a system error compensation model is established:

[0114] ΔL corrected =ΔL measured -∑[α i ×(T i -T ref )×L i ]

[0115] Where, ΔL measured For displacement measured directly by vision, α i T represents the coefficient of thermal expansion of the material in different parts. i For the real-time temperature of each sensor, T ref For reference temperature, L i The characteristic length under the influence of this temperature; the analysis and control module receives data from the temperature sensor network in real time, calculates the theoretical deformation value caused by non-uniform thermal expansion based on the above model, and then subtracts it from the original visual measurement data, thereby outputting the temperature-compensated three-dimensional coordinates.

[0116] This implementation significantly improves the accuracy and reliability of visual measurement data in complex engineering environments through adaptive dynamic filtering and temperature compensation via multi-sensor fusion. It can effectively separate physical deformation from environmental interference, making the final displacement data more realistically reflect the structural response caused by the load, and providing a clean data foundation for safety judgment.

[0117] In some embodiments, the vision-based intelligent prestressing detection method further includes a report generation step:

[0118] Throughout the entire process of graded loading and intelligent monitoring, the load value and timestamp corresponding to each deformation result are recorded synchronously.

[0119] Based on the recorded time-series data, a curve reflecting the relationship between the deformation of the support structure under pressure and the load and time is automatically generated, showing the change of deformation with time and load.

[0120] Based on the preset report template and engineering specifications, the system integrates the curves of deformation over time and load, the maximum displacement value, and the results of stability assessment, and automatically outputs a preload test result report.

[0121] In some embodiments, progressive loading includes at least one initial loading level, an intermediate loading level, a full load level, and an overload level, wherein the load value of the overload level is greater than the load value of the full load level.

[0122] As an example, the initial loading level is 10% of the maximum construction load, the intermediate loading level is 50% of the maximum construction load, the full load level is 100% of the maximum construction load, and the overload level is 110% of the maximum construction load.

[0123] The various embodiments of the present invention have now been described in detail. To avoid obscuring the concept of the invention, some details known in the art have not been described. Those skilled in the art will fully understand how to implement the technical solutions disclosed herein based on the above description.

[0124] The above embodiments only illustrate some implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A vision-based intelligent pre-stressing detection system, characterized in that, include: Preload module, data acquisition module, analysis and control module; The preload module includes a reaction frame disposed above the support structure to be pressed. The reaction frame is fixedly disposed on a stable foundation next to the support structure to be pressed. The reaction frame is provided with a pressure-applying component for applying preload to the support structure to be pressed. The data acquisition module includes multiple combined targets disposed on the structure to be pressed and a visual camera for acquiring image information of the combined targets; The analysis and control module is communicatively connected to the vision camera and is used to receive image information and process it through a visual measurement algorithm to obtain the spatial displacement value of the combined target. The analysis and control module is connected to the pressure application component and is used to control the pressure application component to load or unload the support structure to be pressed according to a preset loading program.

2. The intelligent pre-stressing detection system based on vision measurement according to claim 1, characterized in that: The combined target includes a prism target and an infrared target, and the prism target is used for total station calibration. The visual camera includes an infrared camera, and the infrared target is used for image acquisition by the infrared camera.

3. The intelligent pre-stressing detection system based on vision measurement according to claim 1, characterized in that: The pressure-applying component is a jack or a hydraulic cylinder.

4. A visual measurement-based intelligent pre-stress detection method, characterized in that, Includes the following steps: The pre-compression position component is installed by fixing the reaction frame on a stable foundation next to the support structure to be compressed, installing the combined target at the design measurement point of the support structure to be compressed, and aligning the combined target with the planned setting position of the vision camera. The acquisition equipment is installed, and a vision camera is set at a preset position. The vision camera is adjusted so that it can clearly capture the image of the combined target. Dynamic calibration and measurement involves capturing images of the combined target in real time using the vision camera and dynamically calculating the three-dimensional displacement field of each monitoring point based on a vision measurement algorithm to obtain deformation results. The system employs graded loading and intelligent monitoring to control the pressure-applying components to apply loads to the support structure under pressure in stages; and continuously acquires the deformation results during and after each loading stage. During the loading process, the trend of the deformation result is judged in real time: If the rate of change of the deformation result exceeds the preset safety threshold, it is determined to be a curve abrupt change, and the pressure-applying component is immediately controlled to stop loading or perform unloading, and an alarm signal is issued. If the rate of change of the deformation result gradually decreases and falls below the preset stability threshold within a preset time period, it is determined that the deformation has converged and stabilized. Subsequently, a signal indicating completion of the current loading stage is issued, and the next loading stage begins.

5. The intelligent pre-stress detection method based on vision measurement according to claim 4, characterized in that: The visual measurement algorithm includes the following steps: Extract feature points of the combined target in the image and obtain their pixel coordinates; Based on the camera calibration parameters, preliminary three-dimensional coordinate calculations are performed; The physical constraint model of the support structure to be compressed is introduced as a filtering condition to optimize the calculated three-dimensional coordinates. The physical constraint model includes the relative positional relationship constraints between the support nodes. Output the optimized 3D coordinates and calculate the spatial displacement value by comparing the coordinate values ​​at different times.

6. The intelligent pre-stress detection method based on vision measurement according to claim 4, characterized in that: The vision camera captures images of the combined targets in real time using a multi-target synchronous measurement method, which includes the following steps: In the first frame image acquired by the system, all visible combined targets are identified and segmented, and a unique identifier is assigned to each identified combined target. In each subsequent frame, all combined targets are re-identified and located, and the detection results of the current frame are matched with the trajectories of existing identifiers based on the appearance features, motion trajectory prediction, or location proximity of the combined targets. For successfully matched target combinations, their current position information is used to update the corresponding motion trajectory; for unmatched target combinations, they are initialized to a new trajectory or marked as temporarily lost according to preset rules.

7. The intelligent pre-stress detection method based on vision measurement according to claim 4, characterized in that: In the dynamic calibration and measurement steps, the acquired target images are processed according to the following steps to eliminate the centroid positioning error introduced by the difference in target imaging size: Morphological closing operations are performed on the binarized target light cluster image. The size of the structuring element used is determined based on the expected imaging size and maximum imaging deviation of the target. The parameters of the morphological processing are adjusted until discrete spot regions belonging to the same target are merged into a single connected region, and the pixel area difference of the connected regions formed by different targets is controlled within a preset tolerance range. Based on the processed, area-normalized connected domains, the centroid coordinates of each target are calculated.

8. The intelligent pre-stress detection method based on vision measurement according to claim 4, characterized in that: The visual measurement algorithm also includes a calibration algorithm, which includes the following steps: To address the slow creep and instantaneous vibration of the support during loading, a dynamic filtering method based on time series analysis is adopted to distinguish between the actual displacement of the structure and the instantaneous jitter. The parameters of the filtering method are adaptively adjusted according to the expected motion model of the combined target. To compensate for the thermal expansion and contraction of the support structure caused by uneven sunlight, a systematic error model related to the temperature field distribution was established, and the three-dimensional coordinates obtained by visual measurement were compensated in real time by reading real-time data from temperature sensors deployed in the monitoring area.

9. The intelligent pre-stress detection method based on vision measurement according to claim 4, characterized in that, It also includes the report generation step: Throughout the entire process of graded loading and intelligent monitoring, the load value and timestamp corresponding to each deformation result are recorded synchronously. Based on the recorded time-series data, a curve reflecting the relationship between the deformation of the support structure under pressure and the load and time is automatically generated, showing the change of deformation with time and load. Based on the preset report template and engineering specifications, the system integrates the curves of deformation over time and load, maximum displacement values, and stability judgment results to automatically output a preload test result report.

10. The intelligent pre-stress detection method based on vision measurement according to claim 4, characterized in that, The progressive loading includes at least one initial loading level, one intermediate loading level, one full load level, and one overload level, wherein the load value of the overload level is greater than the load value of the full load level.

Images