Global space-time displacement monitoring method for long-thin ratio structure based on line segment template matching

Abstract

The application discloses a kind of large slenderness ratio structure global space-time displacement monitoring methods based on line sub-template matching, comprising: S1. the sequence image including the structure and its movement range of measured structure is collected;S2. determine the structure movement area and structure image area of any one frame image in sequence image;S3. to any one frame image, along the length direction of measured structure, the structure movement area image space is divided into several pixel column units with pixel column as unit;S4. template area and searched area are selected from each column pixel column unit, and the displacement value of each pixel column unit in this frame is calculated by using template area and searched area to carry out template matching processing to pixel column unit;S5. according to steps S3-S4 analogy, the displacement value of all pixel column units on structure length at different time is calculated frame by frame.The application can realize frame by frame each column uniform distribution measuring point setting, effectively avoids the occurrence of mismatching, improves calculation efficiency and monitoring accuracy.

Description

Technical Field

[0001] This invention relates to the field of large structure monitoring, specifically to a method for monitoring the spatiotemporal displacement of large-slenderness structures based on line-segment template matching. Background Technology

[0002] Bridges and other large structures are composed of components such as main beams, arches, and cables, whose geometric dimensions in one direction are much larger than those in the other two directions; they are also known as structures with a high slenderness ratio. Bridge health monitoring can record and analyze the technical condition of in-service bridges in real time, and is an important guarantee for maintaining the durability and safety of bridge structures, maintaining and extending their service life, and improving their disaster prevention capabilities and structural toughness.

[0003] Deploying high-precision, densely packed measuring points on the surface of bridge structures can effectively obtain their full-area deformation and perform damage identification and location. However, deploying precise, dense measuring points on important bridge components is a core technical challenge for health diagnosis. Current bridge health monitoring systems have a limited number of measuring points with large spacing, allowing monitoring of displacement, strain, and tilt angles only on typical sections of important and special components, such as mid-span beams, supports, cables, and main towers. This fails to achieve precise, dense measurement of all bridge components. Furthermore, traditional monitoring methods face difficulties in selecting displacement monitoring benchmarks or reference points, and the number of measuring points is limited and affected by power supply constraints. While machine vision-based bridge monitoring technology has been widely applied with the rapid development of machine intelligence, current machine vision-based corner point and feature extraction and matching for dynamic structural displacement monitoring suffers from high computational redundancy and mismatch rates, still presenting challenges in measuring the spatiotemporal displacement field of structures.

[0004] Therefore, a spatiotemporal displacement monitoring method for large slenderness ratio structures based on line-segment template matching is needed to solve the above problems. Summary of the Invention

[0005] In view of this, the purpose of this invention is to overcome the defects in the prior art and provide a method for monitoring the spatiotemporal displacement of large-slenderness structures based on line-segment template matching. This method can realize the arrangement of uniformly distributed measuring points in a single row along the surface of the structure, avoid the possibility of mismatch, and improve the calculation efficiency and monitoring accuracy.

[0006] The present invention provides a method for monitoring the full-domain spatiotemporal displacement of a large aspect ratio structure based on line-segment template matching, comprising the following steps:

[0007] S1. Acquire a sequence of images including the structure under test and its range of motion;

[0008] S2. Based on prior information about structural geometry and motion, determine the structural motion region and structural image region of any frame in the image sequence;

[0009] S3. For any frame of image, divide the image space of the structure motion region along the length direction of the measured structure in units of pixel columns to obtain several pixel column units;

[0010] S4. Select the template region and the searched region from each column of pixel units, and use the template region and the searched region to perform template matching processing on the pixel unit. Using the position of each column structure in the first frame image as the initial position, calculate the displacement value of each pixel unit in the frame.

[0011] S5. Following steps S3 to S4, calculate the displacement values ​​of all pixel column units along the structural length at different times frame by frame.

[0012] Furthermore, the pixel column unit includes information about the structure under test and information about the surrounding environment of the structure under test.

[0013] Furthermore, the template area is selected according to the following method: along the direction perpendicular to the length of the structure being measured, a template area of ​​length N is selected. o i It also includes the pixel set of the structure under test as the template region.

[0014] Furthermore, the search area is selected according to the following method: along the direction perpendicular to the length of the measured structure, the set of pixels occupied by the expected motion area of ​​the measured structure is taken as the search area.

[0015] Furthermore, the length of the searched region is N. s i And N s i Not less than N o i .

[0016] Furthermore, the pixel column unit comprises a column of pixels.

[0017] Furthermore, the pixel column unit includes n columns of pixels, and the n columns of pixels are n neighboring pixels; where n is greater than 1.

[0018] Furthermore, it also includes: S6. Comparing the displacement values ​​of several pixel column units with the deflection values ​​of the theoretical deflection curve equation to obtain the maximum difference value at a single point, and determining whether the maximum difference value at a single point is within the set threshold range.

[0019] Furthermore, the deflection value w(x) of the theoretical deflection curve equation is determined according to the following formula:

[0020]

[0021] Where a is the distance between the equivalent concentrated load and the first support end of the structure under test, b is the distance between the equivalent concentrated load and the second support end of the structure under test, F is the concentrated force generated by the equivalent concentrated load, l is the length of the structure under test, and EI z denoted as , where is the bending stiffness of the cross section of the structure under test, and x is the distance between the location point on the structure under test and the first support end.

[0022] The beneficial effects of this invention are as follows: This invention discloses a method for monitoring the full-domain spatiotemporal displacement of large-slenderness structures based on line-segmentation template matching. By concretizing the prior information of machine vision displacement measurement of large bridge structures into the geometric features of large slenderness ratio of components and the basic assumptions of small deformation in engineering mechanics, a machine vision deformation measurement model is introduced, a single-column vertically distinguishable lightweight feature is constructed, and a line-segmentation template matching method is proposed. This forms a simple and lightweight high-precision full-domain spatiotemporal displacement measurement method for structures, providing technical support for efficient and low-consumption health monitoring of large structures. Attached Figure Description

[0023] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0024] Figure 1 This is a schematic diagram of the monitoring method of the present invention;

[0025] Figure 2 This is a schematic diagram of the extraction of non-uniformly distributed SIFT feature points from the natural texture of concrete components according to the present invention.

[0026] Figure 3 This is a schematic diagram of the ideal uniformly distributed monitoring points of the present invention;

[0027] Figure 4 This is a schematic diagram of the single-row uniformly distributed measuring point arrangement of the present invention;

[0028] Figure 5 This is a schematic diagram of the storage of the beam structure of the present invention in image space;

[0029] Figure 6 This is a schematic diagram illustrating the principle of the line-splitting template matching model of the present invention;

[0030] Figure 7 (a) is a schematic diagram of the single-line template matching principle of the present invention;

[0031] Figure 7 (b) is a schematic diagram of the three-line template matching principle of the present invention;

[0032] Figure 7 (c) is a schematic diagram of the five-line template matching principle of the present invention;

[0033] Figure 8 This is a schematic diagram of the cross-sectional dimensions of the composite box girder structure of the present invention;

[0034] Figure 9 This is a schematic diagram showing the geometric dimensions of the test beam, the load arrangement, and the verification method of the present invention.

[0035] Figure 10 This is a schematic diagram of the present invention, showing the installation of a strip of reflective material along the beam length.

[0036] Figure 11 This is a schematic diagram of the actual reflective strip arrangement of the present invention;

[0037] Figure 12 This is a front view of the simply supported composite beam of the present invention.

[0038] Figure 13 This is a schematic diagram of the template boundary marking in the line-division template matching method of the present invention. Detailed Implementation

[0039] The present invention will be further described below with reference to the accompanying drawings, as shown in the figures:

[0040] The present invention provides a method for monitoring the full-domain spatiotemporal displacement of a large aspect ratio structure based on line-segment template matching, comprising the following steps:

[0041] S1. Acquire a sequence of images including the structure under test and its range of motion; wherein, in order to ensure better acquisition results when acquiring the sequence of images, a high-contrast reflective strip is pasted along the length direction of the surface of the structure under test and supplementary illumination is set to enhance the identifiability and distinguishability of the structure surface.

[0042] S2. Based on prior information about structural geometry and motion, determine the structural motion region and structural image region of any frame in the image sequence;

[0043] S3. For any frame of image, divide the image space of the structure motion region along the length direction of the measured structure in units of pixel columns to obtain several pixel column units;

[0044] S4. Select the template region and the searched region from each column of pixel units, and use the template region and the searched region to perform template matching processing on the pixel unit. Using the position of each column structure in the first frame image as the initial position, calculate the displacement value of each pixel unit in the frame.

[0045] S5. Following steps S3 to S4, calculate the displacement values ​​of all pixel column units along the structural length at different times frame by frame.

[0046] In this embodiment, feature extraction and matching are the main methods for machine vision motion tracking, which can invert the motion patterns of objects in the real world recorded in sequence images. By tracking the natural texture features of the structure surface, the feature extraction and matching methods are applied to the extraction and measurement of the global displacement field of the structure. However, the feature points extracted using the natural texture of the object surface are non-uniformly distributed, such as... Figure 2 As shown, the SIFT points extracted from the concrete structure surface are due to the non-uniform spatial distribution of the gray-level gradient on the structure surface. The accuracy of calculating the structural displacement and strain fields using non-uniformly distributed feature points is lower than that using uniformly distributed feature points. Figure 3 The uniformly distributed high-quality feature points shown can be achieved by pre-laying artificial markers or spraying speckle on the structural surface to achieve a uniformly distributed feature point or corner point.

[0047] When measuring the spatiotemporal deformation of large structures using machine vision, densely covering the surface with distinguishable manually coded markers is labor-intensive, difficult to implement, and has a limited number of markers. In contrast, the digital image correlation (DIC) method for measuring the spatiotemporal deformation of large structures suffers from complex speckle fabrication and time-consuming computational iterations. The DIC method originates from characterizing the correlation between two sets of signals; the correlation coefficient characterizes the approximation of two digital images of the same size and is applied to motion tracking. The most commonly used methods for defining the image correlation coefficient are normalized least squares and the correlation coefficient method.

[0048] Large structural engineering projects such as bridges and wind turbine blades are composed of long and slender components, where one dimension of a component is much larger than the other two. Typical components include beams, columns, rods, cables, and arches. Taking beam structures as an example, based on the fundamental assumption of small deformation in structural mechanics, the axial deformation of the structure can be ignored; only the deflection of the beam structure needs to be monitored. Figure 3 The densely packed measuring point layout shown can be simplified as follows: Figure 4 The diagram shows a single-row uniformly distributed measuring point arrangement.

[0049] A single-row, evenly distributed measuring point arrangement can be achieved by setting artificial markers on the structural surface, but the number of measuring points is limited and the distinction between each measuring point is poor. Therefore, it is necessary to... Figure 4 The measurement point layout shown is introduced into the machine vision measurement model for the deformation of structures with large aspect ratios, and is used as prior information to construct a new feature paradigm, preparing for the extraction of the deformation of the entire structure.

[0050] In this embodiment, displacement monitoring of large slenderness ratio components generally only monitors displacement perpendicular to the direction of maximum dimension. For example, for the main beam of a bridge, only deflection deformation needs to be monitored, and the deformation of the structure along the beam length direction is not monitored. This invention proposes a full-domain spatiotemporal deflection field monitoring method based on the line-splitting template matching method for monitoring the dynamic deflection field displacement deformation of large slenderness ratio structures such as beams, arches, and cables.

[0051] In image space, grayscale values ​​are distinguishable along the vertical direction of the beam but not along the beam length. Based on this characteristic, this invention proposes a novel template matching construction method.

[0052] like Figure 5 The beam shown is sampled and stored in the digital image space as illustrated (the length of the image space is M, and the width is N). The beam structure occupies M pixel columns in the figure. l In this model, any column of pixels contains both environmental and structural information, and exhibits good distinguishability along the direction perpendicular to the beam length in sub-image i, allowing for the establishment of a template matching model; for example... Figure 6 As shown, in any sub-image i that is a 1D image, a template matching method is constructed, with a length of N... o i The structural pixel set is selected as template I o i Based on the prior information about the motion of this section of the beam structure, the possible motion region of this column of structure is taken as the search region I of the i-th column. s i Its length is N s i N can be determined based on the actual working conditions. o i and N s i The specific value to be taken.

[0053] Following the above method, the continuous motion tracking of the cantilever beam structure is decomposed into M... l The 1D template tracking problem, specifically the structural motion tracking problem for each column, can be solved using the template matching method described above. This invention refers to this model as the Slice Template Match Model (STMM). Using the STMM, displacement can be calculated column by column along the beam length. The matching results between each line segment do not affect each other, avoiding the possibility of mismatches between column images. Simultaneously, the displacement measurements along the beam length are continuous, satisfying the requirement for continuous and consistent displacement of adjacent points on the structure, facilitating result inspection and analysis.

[0054] As can be seen from the above template construction method, the line-splitting template matching method does not set artificial marks on the structural surface, realizing pixel-by-pixel "full-field" dynamic deflection displacement measurement along the beam length, which greatly increases the number of measuring points and effectively ensures the uniform distribution of measuring points. It breaks through the requirement that the pixel subset of the template area must have obvious distinguishability in both horizontal and vertical directions, reducing this condition to grayscale distinguishability in only the deflection direction. Undoubtedly, this condition can be well satisfied in the measurement of structures with large slenderness ratios, solving the problem of applying the template matching method to the full-field displacement of large slender structures.

[0055] In this embodiment, to reduce the impact of mismatches and noise, the n-neighborhood of the i-th column pixel, comprising n columns of sub-images, can be considered as a whole, establishing an n-line template matching method or a multiple slice template matching (MSTM) method. Figure 7 It is the single-line, three-line, and five-line subdivision template I corresponding to the i-th column pixel. o i and the searched region I S i .

[0056] In this embodiment, the monitoring method of the present invention can be tested, verified, and evaluated using a simply supported composite beam, such as... Figure 8 As shown, the cross-section of the test box girder is a composite box girder with a length of 5.77m, a height of 40mm, and a width of 160mm, and the wall thickness of the box girder is 2mm.

[0057] like Figure 9 As shown, the experiment achieved dynamic loading by moving an 8kg counterweight truck on the composite beam. During the loading process, the truck moved at a constant speed from the right simply supported end to the left simply supported end. Under the action of the counterweight truck, the beam structure satisfied the small deformation assumption. The deflection curve equation of the simply supported beam under a concentrated force F can be obtained from the differential equation of the deflection curve. The deflection value w(x) of the theoretical deflection curve equation is determined according to the following formula:

[0058]

[0059] Where a is the distance between the equivalent concentrated load and the first support end of the structure under test, b is the distance between the equivalent concentrated load and the second support end of the structure under test, F is the concentrated force generated by the equivalent concentrated load, l is the length of the structure under test, and EI z denoted as , where is the bending stiffness of the cross section of the structure under test, and x is the distance between the location point on the structure under test and the first support end.

[0060] The test beam has 3770 pixels distributed along its length, and the calculated object resolution of the measurement system is 1.53 mm / px. According to the analysis of the theory of mechanics of materials, the deflection of the simply supported beam is the largest when the concentrated load is applied to the mid-span of the structure. The deflection curve of the beam structure at any time can be calculated using the line-division template matching method.

[0061] In machine vision measurement, low-precision measurements can utilize the inherent features of objects, such as natural textures. High-precision measurements require enhancing structural features through artificial markers with distinct characteristics. These artificial markers effectively improve the recognizability of the tracked target in the image. The reflective brightness of the tiny glass microbead coating on the surface of the directional reflective strip is far greater than that of the diffuse white marker, improving target recognizability and efficiency. This invention, based on the needs of full-domain spatiotemporal deflection displacement measurement, designs a 1.5cm wide strip of directional reflective strip along the beam length direction. The arrangement of the reflective strip along the beam length direction is illustrated in the diagram. Figure 10 As shown, Figure 11 The image shows a partial view of the structure after the reflective strips were installed. It can be seen that the reflective strips significantly improve the surface reflectivity and distinguishability. Each column of reflective strips is used as a template for tracking and positioning.

[0062] According to the theory, method, and steps of template matching, a discriminative region needs to be selected as the mask from any column of pixels in any frame j of the image. j i (Template area) and Region of Interest (ROI) j i (Searched area) Figure 12 The 140th frame image was acquired, showing the beam structure, vehicle load, and laser ranging position. The structure and vehicle morphology at different times were acquired and measured in the manner described above.

[0063] For any given frame of image, select the Mask column by column within the image region containing the beam structure. j i Region of Interest (ROI) j i The selection method refers to the boundary shown in the figure. Figure 13 The red dots in the image represent the top and bottom boundaries of each column of pixel template mask.

[0064] The dynamic deflection curve measurement results of the beam structure are displacement values ​​of a large number of measuring points uniformly distributed along the beam length at different times. In the experiment of this invention, 3770 displacement measuring points are uniformly distributed along the beam length in each frame of the image, and there are 94250 displacement measuring points per second. It is difficult to verify and evaluate the accuracy of the dynamic displacement measurement results of a large number of measuring points.

[0065] This invention employs two methods: theoretical verification and single-point time history curve verification. The theoretical verification method verifies the displacement of all 3770 measuring points on the beam structure when the vehicle travels to a specific position at a certain moment. The result is verified by comparing the displacement with the calculation result of the beam structure deflection curve equation (2). This method is a single-moment full-space domain static verification method. The second method verifies the difference between the dynamic time history curve measured by machine vision and the laser rangefinder at a single measuring point. This method is a single-point full-time domain dynamic verification method.

[0066] The theoretical deflection curves at different positions of the structure were calculated using equation (2) for verification. When the experimental vehicle reached the mid-span, the theoretical deflection curve and the deflection obtained by the STMM method showed good consistency overall, with a maximum single-point measurement error of 0.72 mm. The STMM method can measure the box girder structure deflection curve fitting diagram at any time, i.e., when the vehicle reaches different positions.

[0067] In the single-point time-domain dynamic verification method, the measurement results of the laser rangefinder at the mid-span position are compared with the corresponding displacement dynamic time history curve measured by the STMM method. The laser measurement time history curve and the measurement results of the line-division template matching method have good consistency.

[0068] In static verification, the overall error evaluation method for all measuring points can be represented by the root mean square error (RMSE); the correlation between the STMM at the mid-span measuring point and the two sets of dynamic time history curves obtained by the laser rangefinder can be represented by the correlation coefficient ρ and the coefficient of determination R. 2 The results showed that the correlation between the two sets of curves measured by the STMM method proposed in this invention is much greater than that of the prior art, and the calculation accuracy is greatly improved.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for monitoring the full-domain spatiotemporal displacement of a large-slenderness structure based on line-segment template matching, characterized in that: Includes the following steps: S1. Acquire a sequence of images including the structure under test and its range of motion; S2. Based on prior information about structural geometry and motion, determine the structural motion region and structural image region of any frame in the image sequence; S3. For any frame of image, divide the image space of the structure motion region along the length direction of the measured structure in units of pixel columns to obtain several pixel column units; S4. Select the template region and the searched region from each column of pixel units. Perform template matching processing on the pixel units using the template region and the searched region. Taking the structural positions of each column in the first frame image as the initial positions, calculate the displacement value of each pixel unit in that frame. Select the template region according to the following method: along the direction perpendicular to the length of the measured structure, place a template region of length N... o i The template region includes the pixel set of the structure under test; the search region is selected according to the following method: along the direction perpendicular to the length of the structure under test, the pixel set occupied by the expected motion region of the structure under test is taken as the search region; the length of the search region is N. s i And N s i Not less than N o i ; S5. Following steps S3 to S4, calculate the displacement values ​​of all pixel column units along the structural length at different times frame by frame.

2. The method for monitoring the full-domain spatiotemporal displacement of a large aspect ratio structure based on line-splitting template matching according to claim 1, characterized in that: The pixel column unit includes information about the structure under test and information about the surrounding environment of the structure under test.

3. The method for monitoring the full-domain spatiotemporal displacement of a large aspect ratio structure based on line-splitting template matching according to claim 1, characterized in that: The pixel column unit comprises a column of pixels.

4. The method for monitoring the full-domain spatiotemporal displacement of a large aspect ratio structure based on line-splitting template matching according to claim 1, characterized in that: The pixel column unit includes n columns of pixels, and the n columns of pixels are n neighboring pixels; where n is greater than 1.

5. The method for monitoring the full-domain spatiotemporal displacement of a large-slenderness structure based on line-splitting template matching according to claim 1, characterized in that: Also includes: S6. Compare the displacement values ​​of several pixel column units with the deflection values ​​of the theoretical deflection curve equation to obtain the maximum difference value at a single point, and determine whether the maximum difference value at a single point is within the set threshold range.

6. The method for monitoring the full-domain spatiotemporal displacement of a large-slenderness structure based on line-splitting template matching according to claim 1, characterized in that: The deflection value w(x) of the theoretical deflection curve equation is determined using the following formula: Where a is the distance between the equivalent concentrated load and the first support end of the structure under test, b is the distance between the equivalent concentrated load and the second support end of the structure under test, F is the concentrated force generated by the equivalent concentrated load, l is the length of the structure under test, and EI z denoted as , where is the bending stiffness of the cross section of the structure under test, and x is the distance between the location point on the structure under test and the first support end.

Images