Bridge component risk management method and system

Abstract

This invention proposes a risk management method and system for bridge components. The system's management method includes: acquiring real-time images of the corresponding bridge segment through an image camera device and performing image preprocessing; extracting, identifying, and segmenting features from the input bridge segment analysis images, eliminating irrelevant information, and classifying them; performing feature analysis on the component feature set and calculating the initial deformation value; performing matching calculations on bridge segment images under different conditions; extracting, identifying, and segmenting features from the registered images and calculating the registration deformation value; taking the arithmetic mean of the initial deformation value and the registration deformation value as the evaluation deformation value; and judging whether the evaluation deformation value is within the safe range based on a preset safety value, issuing an early warning if it exceeds the limit. This management method and system can simplify data to the greatest extent, improve the computational efficiency and analytical reliability of detection, and reduce the influence of environmental and perspective factors on the acquired images.

Description

Technical Field

[0001] This invention belongs to the field of bridge construction technology, and specifically refers to a method and system for risk management of bridge components. Background Technology

[0002] During bridge construction, bridge supports and Bailey beams, as the main components of bridge erection, play a crucial role in ensuring their stability. Traditional technologies for inspecting bridge components largely rely on manual verification, which is highly subjective and cannot be performed in real time. Furthermore, the complex construction environment can reduce the accuracy of inspections. Visual differences from different angles and image acquisition devices can lead to unreliable judgments and make it impossible to obtain a reliable deformation measurement. Existing computer vision methods for calculating the deformation of bridge supports and Bailey beams generally use images of the bridge in its correct installation state or its original images as a reference to extract and analyze the displacement points and displacement amounts in the live images. This method requires ensuring the accuracy of the original images or reference bridge images, and the reliability of comparative analysis is also affected by the angle and size of the acquired live images. Therefore, high requirements are placed on the acquisition of live images.

[0003] In view of this, the control method of the present invention performs feature analysis on real-time images and directly calculates the specific state and stability of the support or Bailey beam based on the real-time images, thereby improving the calculation efficiency and accuracy of detection, realizing real-time detection of bridge components, and providing more scientific data for taking subsequent measures. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for risk management of bridge components to solve the problems mentioned in the background art.

[0005] The technical solution of this invention is implemented as follows:

[0006] A method for risk management of bridge components includes the following steps:

[0007] S1. Obtain real-time images of the corresponding bridge segment through an image camera device, and perform image preprocessing to obtain analysis images;

[0008] S2. Perform image processing on the analysis image in S1, including feature extraction on the input bridge segment analysis image, feature recognition on the bridge segment according to the model framework of each component, select the segmented image of the corresponding component, eliminate irrelevant information, obtain feature sets of multiple target control units, and classify the multiple feature sets into the corresponding target control units respectively.

[0009] S3. Perform feature analysis on the feature set of each target control unit and calculate the initial deformation value;

[0010] S4. Perform multiple matching calculations on the bridge segment analysis images under different conditions, collect the same scene images under different conditions as registration images, process the registration images according to step S2, and calculate several registration deformation values ​​according to step S3. Take the initial deformation value and the arithmetic mean of several registration deformation values ​​as the evaluation deformation value of the target control unit.

[0011] S5: Determine whether the deformation value is within the safe range based on the preset safety value; if it exceeds the range, issue a warning.

[0012] Preferably, the establishment of the component model framework includes the establishment and testing of training sets for each component, collecting several independent images and installation images of different components, randomly allocating the images to the training set and test set in an 8:2 ratio, establishing the component model and testing the model, and defining the component model structure based on the component model.

[0013] More preferably, the image preprocessing in S1 includes scaling down and enlarging the real-world image proportionally to the actual size of the component, and adjusting the angle of the real-world image through the principle of image perspective cropping and program compilation.

[0014] More preferably, the feature analysis in S3 includes line detection of the component based on the Canny edge detection algorithm and principal component analysis, and the detected edge point set of the feature set is p1, p2, ..., p n Treating its two-dimensional coordinates as two variables, after subtracting their respective means, we get:

[0015]

[0016] in

[0017] Calculate the covariance matrix of variables x and y:

[0018]

[0019] Calculate the eigenvalues ​​λ1, λ2 and the corresponding eigenvectors u1, u2 of matrix F, i.e., Fu1 = λ1u1, Fu2 = λ2u2;

[0020] Comparing eigenvalues ​​λ1 and λ2, assuming λ1 > λ2, the eigenvector u1 is the principal component of the data. Using the eigenvector u1 as the direction of the line, the geometric center of the data points is determined. Is it a point on a straight line?

[0021] More preferably, the feature analysis in S3 further includes fitting feature curves to the boundary feature points of non-linear structures and calculating the initial deformation. The boundary point set of the feature set is {(x i ,y iThe fitted characteristic curve matrix expression is: (i = 1, 2, ..., N)

[0022]

[0023] Wherein, the coefficient vector C = [X T X] -1 [X T [Y], when X is a non-singular matrix, the coefficient vector C = X -1 Y;

[0024] The equation for the fitted curve is: y = f(x) = c0 + c1x + c2x 2 +c3x 3 +c4x 4 +c5x 5 +c6x 6 The initial deformation value of the component is obtained by solving the equation.

[0025] More preferably, the feature analysis in S3 also includes correction detection, which determines whether the component installation is vertical or horizontal based on the slope of the feature vector u1 in the standard coordinate system.

[0026] More preferably, S2 also includes acquiring overall images of the bridge section for training and testing, discovering the installation patterns of components, extracting and recognizing features from the analyzed images, matching the number of target control units, and determining whether the component assembly is correct.

[0027] More preferably, the image camera device is installed on the intelligent adjustment device. The image camera device controls the intelligent adjustment device to start lifting and rotating movements according to the perspective principle, changing the position and shooting angle of the camera device to capture the bridge section from multiple angles.

[0028] A bridge component risk management system includes a camera device, an intelligent adjustment device, and a risk management center. The camera device is mounted on the intelligent adjustment device to adjust its height and viewing angle. The camera device inputs the analysis images into the risk management center via a data cable for risk assessment. The risk management center includes an image preprocessing module, a feature analysis module, an image registration module, and an early warning module. The image preprocessing module extracts, identifies, and segments features from the input analysis images, eliminates irrelevant information, and obtains multiple component feature sets. The feature analysis module performs deformation analysis and calculation on the component feature sets. The image registration module matches and calculates the deformation values ​​of the analysis images under different conditions and calibrates them. The early warning module determines whether the calibrated deformation values ​​are within a safe range and issues an early warning for deformations exceeding the range.

[0029] Preferably, the intelligent adjustment device includes a three-claw base, an electric telescopic rod, and an electric steering shaft. The electric telescopic rod is fixed on the three-claw base, and the electric steering shaft is installed on the top of the electric telescopic rod. The camera device is fixed on the electric steering shaft. The camera device includes an image acquisition module, an image adjustment module, and a drive module. The image acquisition module is used to capture the center of the bridge section and take pictures. The image adjustment module is used to preprocess the real-time image to generate an analysis image. The drive module is used to send action commands to the electric telescopic rod and the electric steering shaft.

[0030] Compared with the prior art, the bridge component risk management system and method of the present invention have the following beneficial effects:

[0031] 1. After preprocessing the on-site images, the generated analysis images are processed by feature extraction. The trained component model is then used for component recognition and bounding box segmentation. Irrelevant information in the images is eliminated and the images are classified to reduce interference factors in the analysis. At the same time, useful real information is restored, the detectability of relevant information is enhanced, and the data is simplified to the maximum extent, thereby improving the reliability of feature extraction, image segmentation, matching and recognition.

[0032] 2. Based on the Canny edge detection algorithm and principal component analysis, straight lines are detected on the edges of components, non-straight lines are fitted, the deformation value of the components is calculated, and it is determined whether the bridge components exceed the maximum allowable deformation value. The calculation is small and the calculation efficiency of the detection is improved.

[0033] 3. By collecting images of bridge components in the same scene under different conditions and matching the initial deformation values ​​multiple times, the initial deformation values ​​are corrected to obtain evaluation deformation values ​​with smaller errors. Therefore, the deformation calculation results are highly reliable and reduce the influence of environmental and viewing angle factors on the collected images.

[0034] 4. By installing the camera device through intelligent adjustment equipment, not only can the stability and level of the camera device be guaranteed, but the position and angle of the camera device can also be automatically adjusted according to the preset shooting angle to obtain high-quality component images that meet the analysis requirements. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a schematic diagram of the assembly structure of the camera device and intelligent adjustment device disclosed in this invention;

[0037] Figure 2This is a flowchart of the bridge component risk management method disclosed in this invention;

[0038] Figure 3 This is a structural schematic diagram of the bridge component risk management system disclosed in this invention.

[0039] In the picture: 1. Three-claw base; 2. Electric telescopic rod; 3. Electric steering shaft; 4. Camera device. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] like Figure 2 The bridge component risk management method shown is mainly used for the deformation and installation status analysis and evaluation of bridge supports or Bailey beams, and includes the following steps:

[0042] S1. Obtain on-site real-time images of the cross-section and longitudinal section of the corresponding bridge segment through an image camera device. In order to obtain complete images of the cross-section and longitudinal section of each component of the Bailey beam or construction support, an intelligent adjustment device is used to assist the camera device in fine-tuning the image acquisition angle during the acquisition operation of the cross-section and longitudinal section of the bridge segment. During acquisition, after the camera device captures the center of the target bridge segment in the capture frame, it automatically and slightly adjusts the height and angle of the camera device according to the camera perspective principle and the preset optimal shooting angle of the cross-section or longitudinal section to obtain cross-section or longitudinal section images of the bridge segment components suitable for image analysis.

[0043] Meanwhile, the camera device preprocesses the live images, scaling them down proportionally to the actual dimensions of the components to obtain images of bridge components of the same size and specifications. This mainly involves acquiring images of support structures or Bailey beams. Through image perspective cropping principles and program compilation, the angles of the support structures or Bailey beams in the live images are adjusted to eliminate the influence of the shooting angle and output analysis images to improve the uniformity and accuracy of feature analysis.

[0044] S2. Perform image processing on the analysis images in S1. Select one or two analysis images with good clarity and lighting for each bridge segment and perform image processing. Based on the edge detection principle, extract features from the input bridge segment analysis images. Use the trained component model framework to identify features of the bridge segments, automatically lock and select the corresponding components, and segment and extract component images. Eliminate irrelevant information in the component images except for the component itself to obtain the segmented images of the corresponding components, which are the feature sets of the target control units. Process multiple bridge segments in this way to obtain feature sets of multiple components, and automatically classify and categorize the corresponding components into the corresponding target control units according to the type of the constructed model.

[0045] S3. Perform feature analysis on the feature sets of each component of each target control unit, calculate the initial deformation value, and perform correction to determine whether the installation of each component is skewed.

[0046] First, based on the Canny edge detection algorithm and principal component analysis, straight line detection of the components is performed, and the detected edge point set of the feature set is p1, p2, ..., p n Treating its two-dimensional coordinates as two variables, after subtracting their respective means, we get:

[0047] x = [x1, x2, ..., x n ] T -μ x

[0048] y = [y1, y2, ..., y n ] T -μ y

[0049] in

[0050] Calculate the covariance matrix of variables x and y:

[0051]

[0052] Calculate the eigenvalues ​​λ1, λ2 and the corresponding eigenvectors u1, u2 of matrix F, i.e., Fu1 = λ1u1, Fu2 = λ2u2;

[0053] Comparing eigenvalues ​​λ1 and λ2, assuming λ1 > λ2, the eigenvector u1 is the principal component of the data. Using the eigenvector u1 as the direction of the line, the geometric center of the data points is determined. Is it a point on a straight line? If the edge point is located on the feature vector u1, it means that the set of edge points is a set of straight line points, and the target analysis component has not deformed on the corresponding cross section;

[0054] For undeformed components, the standard coordinate system is used to determine whether the feature vector u1 is horizontal or vertical, so as to determine whether the corresponding component is installed crookedly. The standard coordinate system is established with the horizontal direction as the horizontal axis and the vertical direction as the vertical axis.

[0055] Feature curves are fitted to the boundary feature points of the deformable component to obtain a fitted numerical curve, which is then used to calculate the deformation. The boundary point set of the image is obtained as {(x i ,y i ), i = 1, 2, ..., N}, and use the minimum mean square error ε as the standard to find the curve y = f(x) that approximates the set of fitted boundary points. That is, when the "distance" between each point on the curve and the set of boundary points is the smallest, the mean square error ε is the smallest. Finally, the fitted curve y = f(x) is used to concisely describe the original curve.

[0056] therefore,

[0057] The curve f(x) has the form: y = f(x) = c0 + c1x + c2x 2 +…+c M x M

[0058] The matrix expression is:

[0059]

[0060] Solve for the coefficient vector C = [X] of the minimum mean square error. T X] -1 [X T [Y], when X is a non-singular matrix, C = X -1 Y

[0061] By inverting the matrix and solving the system of linear equations containing multiple unknowns, the coefficient matrix is ​​obtained, and the equation of the fitted curve is determined as follows:

[0062] y = f(x) = c0 + c1x + c2x 2 +c3x 3 +c4x 4 +c5x 5 +c6x 6

[0063] It should be noted that the boundary point set used is calculated from the component image segmented by the region growing method. The maximum value of y in the curve equation y=f(x) is the initial deformation value.

[0064] S4. In order to eliminate visual differences from different perspectives or errors between images collected under different weather conditions, multiple matching calculations are performed on the bridge segment analysis images under different conditions. This includes collecting and selecting real-time images of the same scene under different lighting conditions and different perspectives, at least two of which are preprocessed as registration images. The initial deformation value is further corrected. The registration images are processed in the same way according to step S2, and several registration deformation values ​​are calculated according to step S3. The arithmetic mean of the initial deformation value and several registration deformation values ​​is taken as the evaluation deformation value of the target control unit. Similarly, the components in each registration image are corrected and detected again.

[0065] S5: Evaluate and judge whether the deformation value is within the safe range based on the preset safety value. If it exceeds the safe range, an early warning will be issued. At the same time, based on the results of the analysis image and the registration image correction detection, if the correction result of any image is negative, an early warning will be issued.

[0066] As a further technical solution, the establishment of the component model framework includes the establishment and testing of training sets for each component. At least 10,000 independent images and installation images of different components are collected. All images are randomly allocated to the training set and the test set in an 8:2 ratio. The training set is used to build the component model and classify them according to the different characteristics of each bridge component so as to jump to different target control units for feature analysis. The test is used to evaluate the predictive ability of the model. The component model is established and the component model structure is defined using OPEN CV to identify and analyze the components in the images.

[0067] As a further technical solution, S2 also includes acquiring overall bridge images or overall images of individual bridge segments for training and testing, discovering the installation phase and installation pattern of each component, and constructing corresponding bridge segment models. When performing feature recognition, the overall bridge model or corresponding bridge segment model is used to extract and recognize the features of the analysis image of the corresponding bridge segment. The feature type and feature quantity in the analysis image are matched with the corresponding bridge segment of the same length in terms of quantity and type. If the match is successful, it is judged that the component is assembled correctly. If the match is unsuccessful, it indicates that a certain type of component is missing.

[0068] like Figure 1 and Figure 3The bridge component risk management system shown includes a camera device, an intelligent adjustment device, and a risk management center. The camera device is mounted on the intelligent adjustment device to adjust its height and viewing angle. The camera device inputs the analysis images into the risk management center via a data cable for risk assessment. The risk management center includes an image preprocessing module, a feature analysis module, an image registration module, and an early warning module. The image preprocessing module extracts, recognizes, and segments features from the input analysis images, eliminates irrelevant information, and obtains multiple component feature sets. The feature analysis module performs deformation analysis and calculation on the component feature sets. The image registration module matches and calculates the deformation values ​​of the analysis images under different conditions and calibrates them. The early warning module determines whether the calibrated deformation values ​​are within a safe range and issues an early warning for deformations exceeding the range.

[0069] As a further technical solution, refer again... Figure 1 The intelligent adjustment device includes a three-claw base 1, an electric telescopic rod 2, and an electric steering shaft 3. The electric telescopic rod 2 is a three-section telescopic rod fixed on the three-claw base 1. The electric steering shaft 3 is installed on the top of the electric telescopic rod 2. The camera device is fixed on the electric steering shaft 3. The camera device includes a control module, an image acquisition module, an image adjustment module, and a drive module. The control module receives and feeds back signals from each module and sends corresponding control signals. The image acquisition module is used to capture the center of the bridge section and take pictures to obtain real-time images. The image adjustment module is used to preprocess the real-time images in real time to generate analysis images. The drive module is used to send action commands to the controller. The control module receives and processes the feedback signals from the drive module and the image acquisition module according to the preset shooting angle and controls their actions.

[0070] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. 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 member risk management method, characterized by, Comprise: S1, obtain the live image of the corresponding bridge section by the image camera device, and perform image preprocessing to obtain the analysis image, the image preprocessing including corresponding equal proportion contraction and enlargement according to the actual size of the component, and adjusting the angle of the live image through image perspective clipping principle and program compilation; S2, image processing is performed on the analysis image in S1, including feature extraction on the input bridge section analysis image, feature recognition of the analysis image according to the component model framework, frame selection of the corresponding component segmentation image, elimination of irrelevant information, obtaining of the feature set of multiple target control units, and classification of the multiple feature sets into the corresponding target control units respectively; wherein, the establishment of the component model framework includes the establishment and test of the component training set, the collection of a plurality of independent images and installation images of different components, the random allocation of the images to the training set and the test set in the ratio of 8:2, the establishment of the component model and the model test, and the definition of the component model structure according to the component model; S3, performing feature analysis on the feature set of each target management unit, calculating an initial deformation value, the feature analysis including performing straight line detection on the component based on a Canny edge detection algorithm and a principal component analysis method, and detecting edge point sets p1, p2,..., p n , and the two-dimensional coordinates are regarded as two variables, and after subtracting the respective means, the following is obtained: x = [x1, x2,..., x n ] T -μ x y = [y1, y2,..., y n ] T -μ y wherein Calculate the covariance matrix of variables x and y: Calculate the eigenvalues λ1, λ2 of the matrix F and the corresponding eigenvectors u1, u2, i.e. Fu1 = λ1u1, Fu2 = λ2u2; Comparing the eigenvalues λ1, λ2, assuming λ1> λ2, the eigenvector u1 is the principal component of the data, and the geometric center of the edge point set is determined as a straight line direction with the eigenvector u1 whether it is a point on a straight line; S4, multiple matching calculations are performed on the bridge section analysis images under different conditions, the images of the same scene under different conditions are collected as registration images, the registration images are processed according to step S2, and a plurality of registration deformation values are obtained according to step S3, and the arithmetic mean of the initial deformation value and the plurality of registration deformation values is taken as the evaluation deformation value of the target control unit; S5: according to the preset safety value, it is judged whether the evaluation deformation value is within the safety range, and if it is exceeded, a warning is given.

2. The bridge member risk management method according to claim 1, wherein: The feature analysis in S3 also includes non-linear boundary feature points for feature curve fitting, and calculating the initial deformation amount, the boundary point set of the feature set is {(x i ,y i ), i = 1, 2, …, N}, and the expression of the fitted feature curve matrix is: where the coefficient vector C = [X T Y] -1 X T Y], and when X is non-singular, the coefficient vector C = X -1 Y; The equation of the fitted curve is determined as: y = f(x) = c0+ c1x + c2x 2 +c3x 3 +c4x 4 +c5x 5 +c6x 6 The initial deformation value of the component is obtained by solving y.

3. The bridge member risk management method according to claim 1, wherein: The feature analysis in S3 also includes correction detection, which judges whether the component installation is vertical or horizontal according to the slope of the feature vector u1 in the standard coordinate system.

4. The bridge member risk management method according to claim 2, wherein: S2 also includes collecting the overall image of the bridge section for training and testing, discovering the component installation rule, matching the number of target control units after feature extraction and feature recognition of the analysis image, and judging whether the component assembly is correct.

5. The bridge member risk management method according to any one of claims 1 to 4, characterized by: The image camera device is arranged on the intelligent adjusting device, and the image camera device controls the intelligent adjusting device to start lifting action and rotating action according to the perspective principle, and adjusts the position and collection angle of the camera device.

Images