A bridge construction intelligent monitoring system fusing machine vision and inclination sensing

Abstract

The application discloses a kind of fusion machine vision and bridge construction intelligent monitoring system of inclination sensing, including multi-source sensing synchronous acquisition module, visual deformation detection module, inclination data correction module, multi-source data fusion evaluation module, construction risk grading early warning module and construction linkage control module;The present application combines the advantages of machine vision non-contact, wide coverage and inclination sensing continuous stability, strong anti-interference, through confidence dynamic weighted fusion to realize complementary advantages, significantly improve the all-weather adaptability of monitoring system under complex construction environment.Sub-pixel visual positioning improves the resolution of visual detection, and the second-order temperature drift correction eliminates the error of inclination sensing system, which ensures the accuracy of deformation detection from single-source data level and fusion level, and meets the high-precision monitoring requirements of bridge construction.

Description

Technical Field

[0001] This invention relates to the field of bridge construction safety monitoring technology, specifically to an intelligent bridge construction monitoring system that integrates machine vision and tilt sensing. Background Technology

[0002] As bridge engineering develops towards longer spans, higher piers, and more complex structures, monitoring structural deformation during construction directly determines construction safety and the quality of the completed bridge. Current bridge construction monitoring technologies suffer from several technical problems: First, single sensing methods lack adaptability. Pure visual monitoring is easily affected by lighting, dust, and water mist at the construction site, resulting in significant accuracy degradation in low-visibility scenarios. Pure tilt sensing is susceptible to zero-point drift due to changes in ambient temperature, leading to significant cumulative errors over long-term monitoring. Second, multi-source data lacks unified time-base synchronization. The time misalignment between visual and tilt data prevents the fusion results from accurately reflecting the instantaneous structural deformation state, easily leading to misjudgments. Third, monitoring is disconnected from the construction process. Most systems only record data and perform post-construction analysis, failing to dynamically adjust construction parameters based on real-time deformation. Delayed risk warnings can easily lead to the accumulation of construction deviations. Fourth, risk assessment dimensions are limited, often relying solely on deformation amplitude as a basis for judgment, neglecting the impact of deformation rate on structural safety, resulting in insufficient rationality in early warning grading. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent monitoring system for bridge construction that integrates machine vision and tilt sensing, so as to solve the problems existing in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a bridge construction intelligent monitoring system integrating machine vision and tilt sensing, comprising a multi-source sensor synchronous acquisition module, a visual deformation detection module, a tilt data correction module, a multi-source data fusion evaluation module, a construction risk classification and early warning module, and a construction linkage control module;

[0005] The output of the multi-source sensor synchronous acquisition module is connected to the input of the visual deformation detection module and the input of the tilt data correction module, respectively, and outputs a synchronous trigger signal with a unified time base, and synchronously outputs bridge construction visual image data and structural tilt data with timestamp labels.

[0006] The output of the visual deformation detection module is connected to the first input of the multi-source data fusion evaluation module. It performs feature matching and sub-pixel localization on the component marker points in the visual image and outputs the physical deformation data of the component.

[0007] The output of the tilt angle data correction module is connected to the second input of the multi-source data fusion evaluation module. It performs temperature drift compensation and filtering on the original tilt angle data and outputs the corrected tilt angle data and converted deflection data.

[0008] The output of the multi-source data fusion evaluation module is connected to the input of the construction risk classification and early warning module. It performs confidence-weighted fusion of visual deformation variables and tilt angle converted deformation variables, and outputs the overall structural deformation evaluation result.

[0009] The output of the construction risk classification and early warning module is connected to the input of the construction linkage control module. Based on the deformation assessment results, the construction risk degree is calculated and the risk level is matched, and the corresponding early warning signal is output.

[0010] The output of the construction linkage control module is connected to an external bridge construction control system, and generates and outputs construction parameter adjustment instructions based on the early warning signal.

[0011] Furthermore, the multi-source sensing synchronous acquisition module includes a visual acquisition unit, a tilt sensing unit, a synchronous triggering unit, and a timestamp marking unit;

[0012] The first output terminal of the synchronization triggering unit is connected to the triggering terminal of the visual acquisition unit, and the second output terminal is connected to the triggering terminal of the tilt angle sensing unit, outputting a pulse trigger signal with the same frequency and phase; the output terminal of the visual acquisition unit is connected to the first input terminal of the timestamp annotation unit, outputting the acquired visual image data; the output terminal of the tilt angle sensing unit is connected to the second input terminal of the timestamp annotation unit, outputting the acquired raw tilt angle data; the timestamp annotation unit adds a timestamp label with a unified time base to each set of input data, completing the one-to-one correspondence between the two types of data according to the time sequence.

[0013] Furthermore, the visual deformation detection module includes a marker point recognition unit, a sub-pixel optimization unit, and a deformation calculation unit;

[0014] The output of the marker recognition unit is connected to the input of the sub-pixel optimization unit. It performs edge extraction and shape matching on the artificial markers in the reference frame and the current frame image, and outputs the pixel-level coordinates of the markers.

[0015] The output of the subpixel optimization unit is connected to the input of the deformation calculation unit. Based on the gray-level gradient least squares method, the subpixel offset correction is performed on the pixel-level coordinates, and the subpixel coordinates of the marker point are output.

[0016] The deformation calculation unit, in conjunction with pre-stored camera pixel equivalent coefficients, obtains the actual physical deformation of the component using a sub-pixel deformation calculation formula. The sub-pixel deformation calculation formula is as follows:

[0017]

[0018] Where ΔL is the actual deformation of the component corresponding to the marker point, (x0, y0) are the sub-pixel coordinates of the marker point in the reference frame image, (x s ys ) represents the subpixel coordinates of the marker point in the current frame image, and k is the image pixel equivalent coefficient, that is, the actual physical space size corresponding to a single pixel.

[0019] Furthermore, the tilt data correction module includes a temperature acquisition unit, a temperature compensation unit, a filtering unit, and a deflection conversion unit;

[0020] The output terminal of the temperature acquisition unit is connected to the first input terminal of the temperature compensation unit, and outputs the real-time temperature data of the environment where the sensor is located.

[0021] The second input terminal of the temperature compensation unit receives the original tilt angle data, performs temperature drift compensation on the original tilt angle measurement value using the temperature drift correction formula, and outputs the temperature-corrected tilt angle data.

[0022] The output of the temperature compensation unit is connected to the input of the filtering unit. After random noise is suppressed by sliding window filtering, a smooth tilt angle sequence is output.

[0023] The output of the filtering unit is connected to the input of the deflection conversion unit, and the tilt angle data is converted into the deflection deformation at the corresponding position based on the calculated span of the component.

[0024] The temperature drift correction formula is as follows:

[0025] θ c =θ m -α·ΔT-β·ΔT 2

[0026] Where, θ c The tilt angle value after temperature correction, θ m The original measurement value of the tilt sensor is given, ΔT is the difference between the current ambient temperature and the calibration reference temperature, α is the first-order temperature drift coefficient, and β is the second-order temperature drift coefficient.

[0027] Furthermore, the construction risk classification and early warning module includes a threshold storage unit, a risk degree calculation unit, and a level early warning output unit;

[0028] The output of the threshold storage unit is connected to the first input of the risk calculation unit, and pre-stores and outputs the allowable deformation threshold and allowable deformation rate threshold corresponding to different construction conditions.

[0029] The second input terminal of the risk calculation unit receives the fused deformation assessment result and calculates the real-time construction risk degree using the construction risk degree calculation formula.

[0030] The output of the risk degree calculation unit is connected to the input of the grade warning output unit. The grade warning output unit matches the corresponding risk level according to the risk degree value and outputs a graded warning signal.

[0031] The formula for calculating the construction risk level is as follows:

[0032]

[0033] Where R represents the construction risk level, and D... f v represents the actual deformation after fusion, [D] represents the allowable deformation threshold under the corresponding construction condition, and v D Let ω1 be the deformation rate per unit time, [v] be the allowable deformation rate threshold under the corresponding working condition, ω1 be the deformation amplitude weighting coefficient, and ω2 be the deformation rate weighting coefficient, and satisfy ω1+ω2=1.

[0034] Furthermore, the multi-source data fusion evaluation module includes a visual confidence calculation unit, a tilt confidence calculation unit, a weight dynamic allocation unit, and a weighted fusion calculation unit;

[0035] The visual confidence calculation unit receives the illumination uniformity parameter of the visual image as input and outputs the confidence value of the visual detection data; the tilt angle confidence calculation unit receives the ambient temperature fluctuation amplitude parameter as input and outputs the confidence value of the tilt angle detection data; the weight dynamic allocation unit receives the two types of confidence values ​​respectively and calculates the visual data fusion weight and the tilt angle data fusion weight according to the confidence ratio; the weighted fusion calculation unit performs a weighted summation of the visual deformation variable and the tilt angle converted deformation variable according to the allocated weights and outputs the final fusion deformation evaluation result.

[0036] Furthermore, the construction linkage control module includes an early warning signal analysis unit, a construction parameter matching unit, and a control command output unit;

[0037] The input of the early warning signal analysis unit is connected to the graded early warning signal, and the early warning level is analyzed and the corresponding construction conditions are analyzed. The output of the construction parameter matching unit is connected to the input of the control command output unit, and the pre-stored construction parameter adjustment strategy is retrieved according to the early warning level and construction conditions. The control command output unit converts the adjustment strategy into standardized control commands and outputs them to the control system of the tensioning equipment, jacking equipment or formwork control equipment corresponding to bridge construction.

[0038] Furthermore, the marker recognition unit first extracts a set of closed contours in the image through an edge detection algorithm, calculates the roundness feature value of each contour, filters out candidate contours that meet the circular marker features, and then completes the pixel-level coordinate positioning of the marker point through normalized cross-correlation template matching.

[0039] Furthermore, the risk level warning output unit divides the risk level into four levels: normal state, attention state, warning state, and alarm state. Different levels correspond to different output forms, including data annotation prompts, audible and visual warning signals, and construction suspension control commands.

[0040] A monitoring method for an intelligent monitoring system for bridge construction that integrates machine vision and tilt sensing includes the following steps:

[0041] Step 1: Complete camera calibration to obtain pixel equivalent coefficients, complete tilt sensor temperature calibration to obtain drift coefficients, input deformation tolerance thresholds and weight parameters for each construction condition, and enter standby mode after system initialization is complete.

[0042] Step 2: The synchronous triggering unit outputs a synchronous pulse signal to trigger the visual acquisition unit and the tilt sensing unit to synchronously acquire data. The timestamp annotation unit adds a unified timestamp to the two types of data and then transmits them to the visual deformation detection module and the tilt data correction module, respectively.

[0043] Step 3: The marker recognition unit completes the pixel-level positioning of the marker points in the image, and the sub-pixel coordinates are obtained after correction by the sub-pixel optimization unit. The deformation calculation unit calculates the actual deformation of the component through the formula and outputs it.

[0044] Step 4: The temperature compensation unit performs temperature drift correction on the original tilt angle by combining real-time temperature data. After filtering, the deflection conversion unit converts it into the deflection deformation at the corresponding position and outputs it.

[0045] Step 5: Calculate the confidence scores of the visual data and the tilt angle data respectively, dynamically allocate the fusion weights, and perform weighted fusion of the two types of deformation variables to obtain the overall structural deformation assessment results;

[0046] Step 6: Retrieve the allowable threshold for the corresponding working condition, calculate the construction risk level using the formula, and output the corresponding early warning signal after matching the risk level;

[0047] Step 7: Generate construction parameter adjustment instructions based on the early warning signal and output them to the construction control system. At the same time, store all the monitoring data for subsequent data analysis and threshold optimization.

[0048] Compared with the prior art, the beneficial effects of the present invention are:

[0049] This invention integrates the advantages of machine vision (non-contact, wide coverage) with tilt sensing (continuous, stable, and strong anti-interference), achieving complementary strengths through dynamic weighted fusion of confidence levels, significantly improving the all-weather adaptability of monitoring systems in complex construction environments. Sub-pixel visual positioning enhances visual detection resolution, and second-order temperature drift correction eliminates tilt sensing system errors, ensuring the accuracy of deformation detection from both single-source data and fusion levels, meeting the high-precision monitoring requirements of bridge construction. A unified time-base synchronous acquisition mechanism eliminates time misalignment of multi-source data, hierarchical risk assessment enables rapid anomaly identification, and construction linkage control directly connects to construction equipment, shortening the risk response cycle and avoiding the accumulation of construction deviations. The modular architecture is adaptable to various bridge construction scenarios, thresholds and strategy libraries are flexibly configurable, support the access of more types of sensor data, and can expand monitoring points and linkage devices according to project needs. Attached Figure Description

[0050] Figure 1 This is a system module diagram of the present invention;

[0051] Figure 2 This is a flowchart of the monitoring method of the present invention. Detailed Implementation

[0052] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0053] Please see Figure 1-2 This invention provides an intelligent monitoring system for bridge construction that integrates machine vision and tilt sensing. The system includes a multi-source sensor synchronous acquisition module, a visual deformation detection module, a tilt data correction module, a multi-source data fusion evaluation module, a construction risk grading and early warning module, and a construction linkage control module. The multi-source sensor synchronous acquisition module serves as the system's data entry point, providing multi-source raw data with a consistent time reference for subsequent processing. The visual deformation detection module and the tilt data correction module preprocess and correct errors in the two types of sensor data, respectively, outputting deformation data of the same dimension. The multi-source data fusion evaluation module performs confidence-weighted fusion of the two types of data, outputting reliable deformation evaluation results. The construction risk grading and early warning module quantifies and grades risks based on the fusion results. The construction linkage control module transforms monitoring results into construction actions, forming a complete closed loop of monitoring, early warning, and control.

[0054] The multi-source sensor synchronous acquisition module includes a visual acquisition unit, a tilt sensing unit, a synchronous triggering unit, and a timestamp annotation unit. The synchronous triggering unit generates a fixed-frequency square wave pulse signal based on the system master clock. The two output signals maintain the same frequency and phase, and are respectively connected to the external trigger interfaces of the visual acquisition unit and the tilt sensing unit, ensuring that both types of units start data acquisition at the same time. After receiving the trigger signal, the visual acquisition unit completes the acquisition and output of one frame of the construction scene image; after receiving the trigger signal, the tilt sensing unit completes the sampling and output of one tilt angle data. The timestamp annotation unit connects to the same system clock as the triggering unit, and adds a timestamp label to each frame of data output. The timestamp accuracy matches the acquisition frequency, ensuring that each set of visual images and tilt angle data can be matched one-to-one through the timestamp, eliminating fusion errors caused by time misalignment from the data source. After the data is timestamped, the visual image data is transmitted to the visual deformation detection module, and the raw tilt angle data and the synchronously acquired temperature data are transmitted together to the tilt angle data correction module.

[0055] The visual deformation detection module includes a marker recognition unit, a sub-pixel optimization unit, and a deformation calculation unit. The marker recognition unit first converts the input image to grayscale, transforming a color image into a single-channel grayscale image to reduce computational load. Then, it smooths the image using Gaussian filtering to suppress image noise caused by dust and uneven lighting at the construction site. Subsequently, it extracts edge information from the image using an edge detection operator and extracts all closed contours using a contour tracking algorithm. It calculates the area, perimeter, and circularity of each contour, and filters candidate marker contours based on a preset marker feature threshold. To further eliminate interference from similar shapes in the environment, the unit retrieves a pre-stored standard marker template and performs template matching on the candidate contours using a normalized cross-correlation algorithm. It calculates the matching degree, and contours with a matching degree higher than a set threshold are determined as valid markers, while simultaneously outputting the pixel-level center coordinates of the markers.

[0056] The sub-pixel optimization unit centers on the pixel-level coordinates of the marker point, selects the grayscale values ​​of surrounding pixels, and calculates the distribution of grayscale gradients in the x and y directions. Based on the continuity of grayscale distribution at the sub-pixel level, a least-squares optimization equation is constructed to solve for the sub-pixel offset of the pixel-level coordinates. This offset is then superimposed onto the pixel-level coordinates to obtain the marker point center coordinates with sub-pixel precision. Compared to pure pixel-level localization, sub-pixel optimization can improve localization accuracy by an order of magnitude, significantly enhancing the resolution of visual deformation detection.

[0057] The deformation calculation unit retrieves the pixel equivalent coefficient obtained during the system calibration phase. This coefficient, determined jointly by camera calibration and on-site scale calibration, represents the actual physical spatial size corresponding to a single pixel in the image. The unit calculates the difference between the sub-pixel coordinates of the marker point in the current frame and the reference frame to obtain the displacement at the pixel level. This displacement is then multiplied by the pixel equivalent coefficient to obtain the actual physical deformation of the component corresponding to the marker point. The calculation process uses the sub-pixel deformation calculation formula, comprehensively considering the displacement components in the x and y directions to obtain the composite deformation in the plane. After the calculation is completed, the deformation data, along with the corresponding timestamp, is output to the multi-source data fusion evaluation module.

[0058] The tilt angle data correction module includes a temperature acquisition unit, a temperature compensation unit, a filtering unit, and a deflection conversion unit. The temperature acquisition unit samples synchronously with the tilt angle sensing unit to obtain the ambient temperature data at the tilt angle sensor's installation location, maintaining the same timestamp as the original tilt angle data to ensure the time correspondence of temperature compensation. The temperature compensation unit uses a second-order polynomial temperature drift model to correct the original tilt angle data. The bridge construction environment experiences large diurnal temperature variations, which can cause deformation of the tilt angle sensor's internal structure, resulting in zero-point drift and sensitivity changes. First-order linear compensation can only eliminate part of the drift, while the second-order polynomial model can better fit the nonlinear drift characteristics, improving correction accuracy. The temperature drift correction formula is based on the original measured value, subtracting the first and second-order temperature drift terms to obtain the temperature-corrected tilt angle value. The first and second-order drift coefficients are determined through temperature calibration experiments during the system initialization phase. The calibration process covers the temperature range that may occur in the construction environment, ensuring the effectiveness of correction across the entire temperature range.

[0059] The filtering unit receives temperature-corrected tilt angle data and smooths it using a sliding window filtering algorithm. The window length can be adjusted according to the construction vibration characteristics. The data within the window are arranged in chronological order, and the average value of all data within the window is taken as the filtered output value for the current moment. As new data is input, the window gradually slides forward, continuously outputting the smoothed tilt angle sequence. Sliding window filtering can effectively suppress high-frequency vibration noise caused by mechanical operations and personnel movement at the construction site, preserve the true trend of tilt angle changes, and avoid misjudgment of deformation due to noise interference.

[0060] The deflection conversion unit receives smoothed inclination angle data and converts the angle values ​​into linear displacement values ​​based on the geometric parameters of the measured component and geometric relationships. For beam components, the deflection deformation at the measuring point can be calculated based on the distance from the measuring point to the support and the inclination angle value; for pier components, the horizontal offset of the pier top can be calculated based on the measuring point height and the inclination angle value. The converted deformation variable and the deformation variable obtained by visual inspection are in the same physical dimension, providing a unified dimensional basis for subsequent data fusion. The converted deflection deformation variable, along with a timestamp, is output to the multi-source data fusion evaluation module.

[0061] The multi-source data fusion evaluation module includes a visual confidence calculation unit, a tilt angle confidence calculation unit, a dynamic weight allocation unit, and a weighted fusion calculation unit. The visual confidence calculation unit performs illumination analysis on the current frame image, calculating the overall grayscale mean, grayscale variance, and local illumination gradient to comprehensively evaluate the lighting suitability of the current scene. When the illumination intensity is moderate, uniformly distributed, and without significant local overexposure or underexposure, the visual detection results are highly reliable, and the confidence value is high. When strong direct light, backlighting, or dust obstruction cause a decrease in image contrast, the visual detection error increases, and the confidence value decreases proportionally. The confidence value ranges from 0 to 1, with higher values ​​indicating stronger data reliability.

[0062] The tilt confidence calculation unit assesses the reliability of tilt data based on the amplitude of ambient temperature fluctuations per unit time. When temperature changes are gradual and fluctuations are small, the temperature drift correction effect is good, the tilt data is stable, and the confidence value is high. However, when there are sudden temperature changes, such as sudden rainfall or direct sunlight on the sensor, the temperature drift correction residual increases, the data reliability decreases, and the confidence value decreases proportionally. The confidence value range is also from 0 to 1.

[0063] The dynamic weight allocation unit receives the confidence scores of the two data types and calculates the corresponding fusion weights according to the confidence score ratio. Data sources with higher confidence scores are assigned larger weights, making the fusion result more biased towards reliable data; when the confidence score of a certain data type is too low, its weight is reduced accordingly to avoid unreliable data causing excessive interference to the fusion result. The sum of the two weights remains at 1 to ensure that the physical meaning of the fusion result remains unchanged.

[0064] The weighted fusion calculation unit multiplies the visual deformation variable and the tilt angle-converted deformation variable by their respective weights, then sums them to obtain the final fused deformation assessment result. The dynamic weighting mechanism allows the system to automatically adjust the contribution ratio of the two types of sensors according to the site environment. In good lighting conditions, it leverages the high-resolution advantage of visual detection more, while in poor lighting conditions, it relies more on the stability of the tilt angle sensor, achieving complementary advantages between the two types of sensors and improving the robustness of the monitoring system in complex environments. The fused deformation assessment result is output to the construction risk classification and early warning module.

[0065] The construction risk classification and early warning module includes a threshold storage unit, a risk degree calculation unit, and a level early warning output unit. The threshold storage unit stores corresponding allowable deformation thresholds and allowable deformation rate thresholds according to different construction conditions and stages. For cantilever construction, thresholds can be set separately for different beam segments such as block 0, standard section, and closure section; for high pier construction, thresholds can be set according to different pier heights. The system can automatically switch the corresponding thresholds based on construction progress information, or administrators can manually adjust them to ensure the thresholds match the current construction status.

[0066] The risk calculation unit retrieves the allowable threshold corresponding to the current working condition and calculates the ratio of the current deformation to the allowable threshold, as well as the ratio of the current deformation rate to the allowable rate threshold. The deformation rate is calculated by dividing the difference in deformation between adjacent time points by the time interval, reflecting the trend of deformation change. The construction risk calculation formula weights and sums the deformation amplitude ratio and deformation rate ratio according to preset weights to obtain a comprehensive risk value. The weights of deformation amplitude and deformation rate can be adjusted according to construction characteristics; for slowly loading construction conditions, the amplitude weight can be increased, and for dynamic construction conditions, the rate weight can be increased. The higher the risk value, the higher the construction safety risk.

[0067] The graded early warning output unit has four preset risk level ranges, corresponding to normal, attention, warning, and alarm states, respectively. When the risk level is in the lowest range, it is considered a normal state, and the system only records data without triggering any additional prompts. When the risk level enters the attention range, it is considered a attention state, and the system annotates the corresponding measuring points with data and provides color prompts on the monitoring interface to remind management personnel to pay closer attention. When the risk level enters the warning range, it is considered a warning state, and the system triggers an audible and visual warning signal while simultaneously sending a warning notification to management personnel. When the risk level enters the alarm range, it is considered an alarm state, and in addition to triggering a strong warning, the system can also output a pause command to the construction control system to prevent further escalation of the risk. The graded early warning mechanism enables differentiated risk management, avoiding over-warning or under-warning.

[0068] The construction linkage control module includes an early warning signal analysis unit, a construction parameter matching unit, and a control command output unit. The early warning signal analysis unit analyzes the input early warning signals and extracts key information such as risk level, deformation direction, deformation location, and current construction conditions to provide a basis for parameter matching.

[0069] The construction parameter matching unit has a built-in construction parameter adjustment strategy library, which is categorized and stored according to construction conditions and risk levels. For cantilever casting construction, in the early warning state, the tension of the front slings of the hanging basket can be adjusted to counteract downward deflection; in the alarm state, concrete pouring can be suspended until the deformation stabilizes before resuming construction. For high pier formwork construction, in the early warning state, the tightness of the formwork tie rods can be adjusted to correct the verticality deviation of the formwork; in the alarm state, the formwork lifting operation can be suspended. The adjustment parameters in the strategy library can be preset based on engineering experience and simulation calculation results, and can also be optimized and updated during construction based on actual results.

[0070] The control command output unit converts the matched adjustment parameters into standardized control commands recognizable by the construction equipment. The command format is compatible with the communication protocol of the construction equipment control system and can be transmitted to the corresponding equipment controller via industrial bus, Ethernet, or other means. After the command is output, the unit receives feedback signals from the equipment to confirm the command execution status, forming a closed-loop control system. The construction linkage mechanism directly translates monitoring results into construction actions, significantly shortening risk response time, avoiding delays and errors caused by manual intervention, and improving the automation level of construction management.

[0071] Example 1: Application of construction monitoring for cantilever continuous beams

[0072] This embodiment is applied to the cantilever casting construction process of long-span continuous beams. The monitoring object is the vertical deformation of the formwork and the beam end, realizing real-time monitoring of deformation and linkage adjustment of tensioning parameters during the construction process.

[0073] During the system initialization phase, on-site camera calibration was completed to obtain the pixel equivalent coefficient of the monitoring area at the beam end; temperature calibration of the tilt sensor was completed to obtain the first and second order drift coefficients for the corresponding temperature range; allowable deformation thresholds, deformation rate thresholds, and risk weight parameters for each standard section of the cantilever casting were entered, and the adjustment strategy library for the hanging basket tensioning equipment was configured. Circular artificial markers were set up at the front end of the beam segment, and the tilt sensor was installed on the top surface of the beam end, at the same cross-sectional position as the marker.

[0074] During construction, the synchronous triggering unit outputs synchronous pulses at a set frequency, simultaneously triggering image acquisition and tilt angle sampling. The timestamp annotation unit adds a uniform timestamp to each frame of data. After the visual image is transmitted to the visual deformation detection module, it first undergoes grayscale conversion and filtering preprocessing. The marker recognition unit extracts the image contour and filters out candidate markers, determines the pixel-level coordinates of the markers through template matching, and then obtains precise sub-pixel coordinates through sub-pixel optimization. The deformation calculation unit compares the coordinates with the reference frame to calculate the vertical deformation at the beam end.

[0075] The original tilt angle data and synchronous temperature data are transmitted to the tilt angle data correction module. The temperature compensation unit compensates for the original tilt angle using a second-order temperature drift correction formula, eliminating drift errors caused by changes in ambient temperature. The filtering unit uses a sliding window filter to suppress high-frequency noise caused by construction vibration, resulting in a smooth tilt angle sequence. The deflection conversion unit calculates the tilt angle value into a vertical deflection value at the beam end based on the span of the beam segment.

[0076] The multi-source data fusion evaluation module calculates the confidence levels of the two types of data separately: when the construction site has good lighting and no dust obstruction, the visual data has a higher confidence level and is assigned a larger fusion weight; when there is insufficient lighting in the evening or a lot of dust, the visual confidence level decreases, and the weight of the tilt angle data increases accordingly. After weighted fusion, the comprehensive deformation result of the beam end is obtained and output to the risk warning module.

[0077] The risk calculation unit retrieves the allowable threshold corresponding to the current beam segment, calculates the proportion of deformation amplitude and deformation rate, and obtains the comprehensive risk level through the construction risk calculation formula. When the beam end deflection gradually increases during pouring and the risk level enters the monitoring range, the system displays a warning on the monitoring interface. When the risk level enters the warning range, the system triggers an audible and visual warning, and simultaneously links with the construction parameter matching unit to retrieve the corresponding level of hanging basket tensioning adjustment strategy, generating a tension force adjustment command and outputting it to the hanging basket tensioning control system to actively tension and offset some of the deflection deformation. After adjustment, the system continuously monitors deformation changes. If the risk level falls back to the normal range, normal monitoring resumes; if the risk level continues to rise to the alarm range, a pouring pause command is output, and construction resumes only after the cause is identified and measures are taken.

[0078] Once the construction of each beam segment is completed, the system automatically stores all the monitoring data for that segment. Managers can analyze deformation patterns through historical data to optimize the thresholds and adjustment strategies for subsequent beam segments, thereby continuously improving the monitoring effect.

[0079] Example 2: Application of Verticality Monitoring for High Pier Column Construction Formwork

[0080] This embodiment is applied to the formwork construction process of high bridge piers. The monitoring objects are the verticality deviation and horizontal deformation of the formwork, so as to realize real-time monitoring of the formwork posture and linkage control of formwork adjustment.

[0081] During system initialization, camera calibration is completed to obtain the pixel equivalent coefficient of the template monitoring area, and tilt sensor temperature calibration is completed to obtain the drift coefficient. Allowable thresholds for template verticality and offset rate at different pier heights are entered, and the adjustment strategy library for the template control equipment is configured. Manual markers are placed at the four corners of the template, and tilt sensors are installed vertically along the template to detect the template tilt angle.

[0082] After the template is lifted into place, the system collects images and tilt angle data under the reference state as reference values. During construction, the synchronous trigger unit outputs synchronous pulses at a set frequency, synchronously collecting template images and tilt angle data. The visual deformation detection module identifies four marker points on the template, obtains the precise coordinates of each marker point after sub-pixel optimization, calculates the relative displacement between the top and bottom marker points of the template, and obtains the verticality deviation and horizontal deformation of the template.

[0083] The tilt angle data correction module performs temperature compensation and filtering on the original tilt angle data to obtain the corrected template tilt angle, which is then converted into the horizontal offset of the pier top. The multi-source data fusion evaluation module calculates the confidence scores of the two types of data based on illumination conditions and temperature fluctuations, and dynamically allocates weights to obtain the fusion evaluation result of the template horizontal deformation.

[0084] When formwork shifts during concrete pouring, the risk calculation unit calculates the combined risk level of the shift amplitude and rate. When the risk level is in a "watchful" state, the system prompts construction personnel to observe carefully. When the risk level enters an "alarm" state, the system triggers an alarm and activates the formwork control equipment, generating adjustment commands for the formwork tie rods. The verticality of the formwork is corrected by adjusting the length of the tie rods on both sides. During the adjustment process, the system monitors deformation changes in real time until the formwork deviation returns to the allowable range. When the risk level enters an "alarm" state, the system pauses concrete pouring and resumes construction only after the formwork is properly adjusted and safety is confirmed.

[0085] After the construction of each pier column is completed, the system stores the monitoring data of that column, which can be used to analyze the formwork offset pattern, optimize the construction process and adjust parameters of subsequent columns, and ensure the verticality accuracy and structural safety of the high pier construction.

[0086] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A bridge construction intelligent monitoring system integrating machine vision and tilt sensing, characterized in that: It includes a multi-source sensor synchronous acquisition module, a visual deformation detection module, a tilt angle data correction module, a multi-source data fusion evaluation module, a construction risk classification and early warning module, and a construction linkage control module; The output of the multi-source sensor synchronous acquisition module is connected to the input of the visual deformation detection module and the input of the tilt data correction module, respectively, and outputs a synchronous trigger signal with a unified time base, and synchronously outputs bridge construction visual image data and structural tilt data with timestamp labels. The output of the visual deformation detection module is connected to the first input of the multi-source data fusion evaluation module. It performs feature matching and sub-pixel localization on the component marker points in the visual image and outputs the physical deformation data of the component. The output of the tilt angle data correction module is connected to the second input of the multi-source data fusion evaluation module. It performs temperature drift compensation and filtering on the original tilt angle data and outputs the corrected tilt angle data and converted deflection data. The output of the multi-source data fusion evaluation module is connected to the input of the construction risk classification and early warning module. It performs confidence-weighted fusion of visual deformation variables and tilt angle converted deformation variables, and outputs the overall structural deformation evaluation result. The output of the construction risk classification and early warning module is connected to the input of the construction linkage control module. Based on the deformation assessment results, the construction risk degree is calculated and the risk level is matched, and the corresponding early warning signal is output. The output of the construction linkage control module is connected to an external bridge construction control system, and generates and outputs construction parameter adjustment instructions based on the early warning signal.

2. The intelligent monitoring system for bridge construction integrating machine vision and tilt sensing according to claim 1, characterized in that: The multi-source sensing synchronous acquisition module includes a visual acquisition unit, a tilt sensing unit, a synchronous triggering unit, and a timestamp marking unit. The first output terminal of the synchronous triggering unit is connected to the triggering terminal of the visual acquisition unit, and the second output terminal is connected to the triggering terminal of the tilt sensing unit, outputting a pulse triggering signal with the same frequency and phase. The output of the visual acquisition unit is connected to the first input of the timestamp annotation unit, and outputs the acquired visual image data; the output of the tilt sensing unit is connected to the second input of the timestamp annotation unit, and outputs the acquired raw tilt angle data; the timestamp annotation unit adds a timestamp label with a unified time base to each set of input data, and completes the one-to-one correspondence between the two types of data according to the time sequence.

3. The intelligent monitoring system for bridge construction integrating machine vision and tilt sensing according to claim 1, characterized in that: The visual deformation detection module includes a marker point recognition unit, a sub-pixel optimization unit, and a deformation calculation unit; The output of the marker recognition unit is connected to the input of the sub-pixel optimization unit. It performs edge extraction and shape matching on the artificial markers in the reference frame and the current frame image, and outputs the pixel-level coordinates of the markers. The output of the subpixel optimization unit is connected to the input of the deformation calculation unit. Based on the gray-level gradient least squares method, the subpixel offset correction is performed on the pixel-level coordinates, and the subpixel coordinates of the marker point are output. The deformation calculation unit, in conjunction with pre-stored camera pixel equivalent coefficients, obtains the actual physical deformation of the component using a sub-pixel deformation calculation formula. The sub-pixel deformation calculation formula is as follows: Where ΔL is the actual deformation of the component corresponding to the marker point, (x0, y0) are the sub-pixel coordinates of the marker point in the reference frame image, (x s y s ) represents the subpixel coordinates of the marker point in the current frame image, and k is the image pixel equivalent coefficient, i.e., the actual physical space size corresponding to a single pixel.

4. The intelligent monitoring system for bridge construction integrating machine vision and tilt sensing according to claim 1, characterized in that: The tilt data correction module includes a temperature acquisition unit, a temperature compensation unit, a filtering unit, and a deflection conversion unit. The output terminal of the temperature acquisition unit is connected to the first input terminal of the temperature compensation unit, and outputs the real-time temperature data of the environment where the sensor is located. The second input terminal of the temperature compensation unit receives the original tilt angle data, performs temperature drift compensation on the original tilt angle measurement value using the temperature drift correction formula, and outputs the temperature-corrected tilt angle data. The output of the temperature compensation unit is connected to the input of the filtering unit. After random noise is suppressed by sliding window filtering, a smooth tilt angle sequence is output. The output of the filtering unit is connected to the input of the deflection conversion unit, and the tilt angle data is converted into the deflection deformation at the corresponding position based on the calculated span of the component. The temperature drift correction formula is as follows: i c =θ m -a·ΔT-b·ΔT 2 Where, θ c The tilt angle value after temperature correction, θ m The original measurement value of the tilt sensor is given, ΔT is the difference between the current ambient temperature and the calibration reference temperature, α is the first-order temperature drift coefficient, and β is the second-order temperature drift coefficient.

5. The intelligent monitoring system for bridge construction integrating machine vision and tilt sensing according to claim 1, characterized in that: The construction risk classification and early warning module includes a threshold storage unit, a risk degree calculation unit, and a classification and early warning output unit. The output of the threshold storage unit is connected to the first input of the risk calculation unit, and pre-stores and outputs the allowable deformation threshold and allowable deformation rate threshold corresponding to different construction conditions. The second input terminal of the risk calculation unit receives the fused deformation assessment result and calculates the real-time construction risk degree using the construction risk degree calculation formula. The output of the risk degree calculation unit is connected to the input of the grade warning output unit. The grade warning output unit matches the corresponding risk level according to the risk degree value and outputs a graded warning signal. The formula for calculating the construction risk level is as follows: Where R represents the construction risk level, and D... f v represents the actual deformation after fusion, [D] represents the allowable deformation threshold under the corresponding construction condition, and v D Let ω1 be the deformation rate per unit time, [v] be the allowable deformation rate threshold under the corresponding working condition, ω1 be the deformation amplitude weighting coefficient, and ω2 be the deformation rate weighting coefficient, and satisfy ω1+ω2=1.

6. The intelligent monitoring system for bridge construction integrating machine vision and tilt sensing according to claim 1, characterized in that: The multi-source data fusion evaluation module includes a visual confidence calculation unit, a tilt confidence calculation unit, a weight dynamic allocation unit, and a weighted fusion calculation unit; The visual confidence calculation unit receives the illumination uniformity parameter of the visual image as input and outputs the confidence value of the visual detection data; the tilt confidence calculation unit receives the ambient temperature fluctuation amplitude parameter as input and outputs the confidence value of the tilt detection data; the weight dynamic allocation unit receives the two types of confidence values ​​respectively and calculates the visual data fusion weight and the tilt data fusion weight according to the confidence ratio. The weighted fusion calculation unit performs a weighted summation of the visual deformation variable and the tilt angle conversion deformation variable according to the assigned weights, and outputs the final fusion deformation evaluation result.

7. The intelligent monitoring system for bridge construction integrating machine vision and tilt sensing according to claim 1, characterized in that: The construction linkage control module includes an early warning signal analysis unit, a construction parameter matching unit, and a control command output unit. The input of the early warning signal analysis unit is connected to the graded early warning signal, and the early warning level is analyzed and the corresponding construction conditions are analyzed. The output of the construction parameter matching unit is connected to the input of the control command output unit, and the pre-stored construction parameter adjustment strategy is retrieved according to the early warning level and construction conditions. The control command output unit converts the adjustment strategy into standardized control commands and outputs them to the control system of the tensioning equipment, jacking equipment or formwork control equipment corresponding to bridge construction.

8. The intelligent monitoring system for bridge construction integrating machine vision and tilt sensing according to claim 3, characterized in that: The marker recognition unit first extracts a set of closed contours in the image using an edge detection algorithm, calculates the roundness feature value of each contour, filters out candidate contours that meet the circular marker features, and then completes the pixel-level coordinate positioning of the markers through normalized cross-correlation template matching.

9. The intelligent monitoring system for bridge construction integrating machine vision and tilt sensing according to claim 5, characterized in that: The risk level output unit divides the risk level into four levels: normal state, attention state, warning state, and alarm state. Different levels correspond to different output forms, including data annotation prompts, audible and visual warning signals, and construction suspension control commands.

10. A monitoring method for a bridge construction intelligent monitoring system integrating machine vision and tilt sensing according to any one of claims 1-9, characterized in that: Includes the following steps: Step 1: Complete camera calibration to obtain pixel equivalent coefficients, complete tilt sensor temperature calibration to obtain drift coefficients, input deformation tolerance thresholds and weight parameters for each construction condition, and enter standby mode after system initialization is complete. Step 2: The synchronous triggering unit outputs a synchronous pulse signal to trigger the visual acquisition unit and the tilt sensing unit to synchronously acquire data. The timestamp annotation unit adds a unified timestamp to the two types of data and then transmits them to the visual deformation detection module and the tilt data correction module, respectively. Step 3: The marker recognition unit completes the pixel-level positioning of the marker points in the image, and the sub-pixel coordinates are obtained after correction by the sub-pixel optimization unit. The deformation calculation unit calculates the actual deformation of the component through the formula and outputs it. Step 4: The temperature compensation unit performs temperature drift correction on the original tilt angle by combining real-time temperature data. After filtering, the deflection conversion unit converts it into the deflection deformation at the corresponding position and outputs it. Step 5: Calculate the confidence scores of the visual data and the tilt angle data respectively, dynamically allocate the fusion weights, and perform weighted fusion of the two types of deformation variables to obtain the overall structural deformation assessment results; Step 6: Retrieve the allowable threshold for the corresponding working condition, calculate the construction risk level using the formula, and output the corresponding early warning signal after matching the risk level; Step 7: Generate construction parameter adjustment instructions based on the early warning signal and output them to the construction control system. At the same time, store all the monitoring data for subsequent data analysis and threshold optimization.

Images