Visual-based monitoring pole finite element automatic modeling and simulation method and device

Abstract

The application discloses a kind of visual-based monitoring pole finite element automatic modeling and simulation method and device, it is related to computer vision and earthquake engineering simulation field, the method includes: obtaining field image and scale calibration information screening and scale calibration;Initial closed contour is generated using visual model and is obtained by morphological correction to get pole piece binary mask;Based on mask and scale relationship, center line and width curve are extracted, and rod type topology, section change type and accessory equipment position are identified;Image coordinates are converted into physical coordinates to generate structured engineering parameters;Finite element simulation project file is automatically generated;After consistency check, dynamic analysis is executed and response time history is exported.The application seamlessly connects visual recognition and finite element simulation, realizes the automatic conversion from field image to executable structure model, guarantees the traceability of image evidence, engineering simplification and simulation result, and greatly improves the efficiency of urban pole type facility seismic response analysis.

Description

Technical Field

[0001] This application relates to the fields of computer vision and earthquake engineering simulation, and in particular to a vision-based finite element automatic modeling and simulation method and device for monitoring poles. Background Technology

[0002] Streetlight poles, surveillance poles, traffic signal poles, and cantilever poles equipped with cameras, lights, electrical boxes, and other auxiliary equipment are widely deployed in urban roads, parks, and transportation infrastructure. These pole-type facilities are numerous and densely distributed, and possess power supply and communication capabilities. If their structural simulation models can be quickly developed, it can provide fundamental data for seismic response analysis, structural safety assessment, monitoring data interpretation, and subsequent training of seismic motion inversion models.

[0003] Existing structural simulations of pole-type facilities typically rely on manual measurements, manual drawing of node elements, manual setting of cross-sectional and mass parameters, and manual writing of analysis scripts. For a large number of dispersed monitoring poles with significant structural differences, manual modeling is not only inefficient but also prone to inconsistencies in pole topology, cross-sectional dimensions, auxiliary equipment mass, foundation constraints, response record locations, and seismic input interfaces, thereby affecting the traceability of subsequent dynamic analysis results.

[0004] Conventional object detection, semantic segmentation, or image description methods can identify pole regions or auxiliary equipment categories in images, but their outputs typically remain at the image level as bounding boxes, masks, or semantic labels, making it difficult to directly convert them into node coordinates, element connections, cross-sectional parameters, material parameters, lumped masses, constraint conditions, and recorder positions required for finite element analysis. Therefore, an image-to-finite element model conversion method is needed for dynamic simulation of monitoring poles, transforming on-site image evidence into structured engineering parameters, and further forming a verifiable, executable, and batch-generated finite element simulation project. Summary of the Invention

[0005] The purpose of this application is to provide a vision-based automatic finite element modeling and simulation method and device for monitoring poles, which can solve the problems of low efficiency of manual modeling and difficulty in directly converting conventional vision methods into finite element analysis parameters in the prior art.

[0006] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a vision-based finite element automatic modeling and simulation method for monitoring poles, including: S1, acquire on-site images and scale calibration information containing the target monitoring pole, determine whether the target monitoring pole can be modeled, and determine the scale relationship between pixel coordinates and structural physical coordinates; S2, using a visual language model to assist in generating the initial closed contour of the target monitoring pole, and obtaining a pole binary mask for structural modeling through morphological processing and connected component filtering; S3. Based on the binary mask of the rod and the scale relationship, extract the center line and width curve of the vertical main rod and the horizontal cantilever, and identify the rod topology, physical size, cross-sectional change type, cross-sectional shape and location of auxiliary equipment; S4. Based on the rod topology, physical dimensions, cross-sectional change type, cross-sectional shape, and location of auxiliary equipment, and in conjunction with the scale relationship, the image coordinates are converted into structural physical coordinates to generate structured engineering parameters that include the locations of nodes, elements, cross-sections, materials, mass, constraints, and response records. S5, Read the structured engineering parameters and automatically generate the model files and analysis scripts required for the finite element simulation project; S6. Perform a consistency check on the finite element simulation project. After the check passes, perform dynamic analysis and export the response time history data of the target response location.

[0007] Optionally, in step S1, determining whether the target monitoring pole is modelable includes: outputting the existence of a modelable main pole and the confidence level of the pole through an inference model; when a main pole exists and the confidence level is not lower than a preset gating threshold, and there is a usable scale source, it is determined to be modelable. The scale calibration information includes ArUco markers, known-size components, camera parameters, engineering prior dimensions, or multi-view geometric constraints. When an ArUco marker is detected, the pixel scale is calculated based on the marker corner coordinates, and the marker number, marker quantity, marker side length, scale confidence level, and usage status are recorded as the scale relationship.

[0008] Optionally, in step S2, generating the initial closed contour of the target monitoring pole using a visual language model includes: The visual language model is constrained to draw a pure red closed target outline on the same canvas; The binary mask for structural modeling, obtained through morphological processing and connected component filtering, includes: Extract red pixels in HSV space, perform closing operations using structuring elements to fill closed contours; use dynamic connection kernels to connect local broken boundaries, retain the largest connected foreground region, and obtain a continuous binary mask; judge the mask quality based on the foreground area ratio and the number of connected regions, and if the quality does not meet the preset conditions, return a correction prompt or transfer to manual review.

[0009] Optionally, in step S3, extracting the centerline and width curve of the vertical main rod and the horizontal cantilever includes: The main connected domain is selected from the binary mask of the rod as the target rod region, and the calibration plate region and auxiliary equipment protrusions are excluded from the width sampling. The width curve is sampled along the centerline of the member. The first RANSAC residual threshold is used for the centerline of the vertical member, and the second RANSAC residual threshold is used for the cantilever. The horizontal cantilever is identified according to the horizontal extension constraint to avoid misjudging the auxiliary equipment as the load-bearing cantilever.

[0010] Optionally, in step S3, identifying the type of cross-sectional change includes: Construct a segmented engineering diameter model to represent constant cross-section, segmented variable cross-section, and continuous tapered variable cross-section; The cross-section variation model is selected based on length-weighted diameter change, monotonicity ratio, and fitting error; when the relative diameter change is less than a preset threshold, it is determined to be an equal cross-section; when the width curve is represented by discrete segments, it is determined to be a piecewise variable cross-section; when the diameter changes continuously and meets the requirements of monotonicity and fitting error, it is determined to be a continuous conical variable cross-section. For cross-sectional shape, identify circular or polygonal cross-sections. When a polygonal hollow cross-section is identified, convert it into an equivalent circular tube cross-section with matching area and bending moment of inertia.

[0011] Optionally, in S4, the structured engineering parameters further include quality control metadata regarding scale source, mask quality, and cross-section recognition quality; The location of the auxiliary equipment is mapped to a concentrated mass and its position of action, and the mass of the auxiliary equipment is applied to the corresponding installation node or the nearest load-bearing member node.

[0012] Optionally, in step 5, the model files and analysis scripts required for the finite element simulation project are automatically generated, including: Establish three-dimensional six-degree-of-freedom nodes based on the pole base, segment point, pole top, cantilever end, and equipment installation point; Beam-column elements are created based on adjacent nodes, and cross sections are assigned accordingly. The distributed mass of the members is calculated based on the material density, cross-sectional area and element length. The element mass is distributed to the end nodes, and the mass of the auxiliary equipment is applied as concentrated mass to the corresponding nodes. Generates a finite element project file containing materials, sections, nodes, elements, masses, constraints, seismic inputs, analysis steps, main execution script, and response recorder.

[0013] Optionally, in step 6, the consistency check includes: checking whether the unit system is consistent, whether the node connectivity is complete, whether the size and cross-sectional parameters are positive, whether the mass is repeatedly applied, whether the foundation constraints are reasonable, whether the seismic input exists, whether the recorder covers the target response location, and whether the short-term trial run can output the response normally.

[0014] Optionally, in step 6, performing dynamic analysis includes: Gravity load is applied in stages; The transient integral algorithm is used for dynamic analysis. When the algorithm fails to converge, the line search algorithm, Broyden algorithm, or the time step is reduced to a preset ratio of the original step size is tried in turn.

[0015] Secondly, this application provides a vision-based finite element automatic modeling and simulation device for monitoring poles, comprising: The scale calibration module is used to acquire on-site images and scale calibration information containing the target monitoring pole, determine whether the target monitoring pole can be modeled, and determine the scale relationship between pixel coordinates and structural physical coordinates; The mask correction module is used to generate the initial closed contour of the target monitoring pole with the help of a visual language model, and obtains a binary mask of the pole for structural modeling through morphological processing and connected component filtering. The topology recognition module is used to extract the centerline and width curve of the vertical main rod and the horizontal cantilever based on the binary mask of the rod and the scale relationship, and to identify the rod topology, physical dimensions, cross-sectional change type, cross-sectional shape and location of auxiliary equipment; The parameter generation module is used to convert image coordinates into structural physical coordinates based on the rod topology, physical dimensions, cross-sectional change type, cross-sectional shape, and location of auxiliary equipment, combined with the scale relationship, and generate structured engineering parameters including the locations of nodes, elements, cross-sections, materials, masses, constraints, and response records. The project generation module is used to read the structured engineering parameters and automatically generate the model files and analysis scripts required for the finite element simulation project. The simulation module is used to perform a consistency check on the finite element simulation project. After the check passes, dynamic analysis is performed to export the response time history data of the target response location.

[0016] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides a vision-based finite element automatic modeling and simulation method and device for monitoring poles, which has the following significant effects: This invention connects visual recognition, morphological geometry extraction, and finite element simulation project generation into a continuous process, enabling the transformation of structural evidence in field images into executable structural models, significantly reducing the workload of manual modeling and manual document conversion.

[0017] This invention outputs structured engineering parameters, clearly recording the dimensions of members, cross-sectional assumptions, mass of auxiliary equipment, source of dimensions, and quality control status, thus enabling a traceable relationship between image recognition errors, engineering simplification, and simulation results.

[0018] This invention provides specialized processing for the horizontal cantilever, tapered pole segment, polygonal cross section, and mass of auxiliary equipment of monitoring poles, effectively avoiding the problem that general target detection or semantic segmentation results cannot meet the requirements of dynamic simulation parameters.

[0019] This invention uses "finite element simulation project" as the overarching technical object, and is not limited to a specific software platform. It can be exported as an input file for OpenSees Tcl project or other finite element software.

[0020] This invention can automatically generate response time histories and datasets under seismic action in batches, providing an efficient data foundation for subsequent structural response analysis, seismic inversion model training, and urban pole-mounted facility monitoring applications. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart illustrating a vision-based finite element automatic modeling and simulation method for a monitoring pole, provided in an embodiment of this application; Figure 2 This is a schematic diagram of the image-to-structure modeling and finite element response simulation workflow provided in one embodiment of this application; Figure 3 A schematic diagram of the morphological geometry extraction and finite element model generation process provided in an embodiment of this application; Figure 4 This is a schematic diagram of the image recognition, masking, topology, and cross-sectional results of a field monitoring pole provided in an embodiment of this application. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] In one exemplary embodiment, such as Figure 1As shown, a vision-based finite element automatic modeling and simulation method for monitoring poles is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method includes the following steps: S1, acquire on-site images and scale calibration information containing the target monitoring pole, determine whether the target monitoring pole can be modeled, and determine the scale relationship between pixel coordinates and structural physical coordinates.

[0026] The determination of whether the target monitoring pole can be modeled includes: outputting the existence of a modelable main pole and the confidence level of the pole through the inference model; when a main pole exists and the confidence level is not lower than a preset gating threshold, and there is a usable scale source, it is determined to be modelable. The scale calibration information includes ArUco markers, known-size components, camera parameters, engineering prior dimensions, or multi-view geometric constraints. When an ArUco marker is detected, the pixel scale is calculated based on the marker corner coordinates, and the marker number, marker quantity, marker side length, scale confidence level, and usage status are recorded as the scale relationship.

[0027] In this embodiment, target poles include upright monitoring poles, L-shaped monitoring poles with horizontal cantilevered arms, streetlight poles, traffic signal poles, and vertical main poles, cantilevered arms, cameras, lighting fixtures, and electrical boxes belonging to the same load-bearing structure. The inference model outputs the existence of modelable main poles and the confidence level of the poles; the current local gating condition is the existence of a main pole and a confidence level of not less than 0.45. When the gating condition is met and a usable scale is available, automatic processing begins; if the subject is identifiable but the scale or local boundaries need to be confirmed, manual verification is required; if the subject is missing, severely obscured, or a continuous pole area cannot be formed, automatic modeling is rejected. The confidence level is output by the inference model along with the screening results, and the local program no longer fits its internal weights.

[0028] Available scales refer to at least one standard that can stably convert pixel lengths into physical lengths. The preferred standard is the ArUco square marker board with a unique code and known physical side lengths; the current implementation supports the DICT_4X4_50 and DICT_5X5_50 dictionaries. The side length of the marked pixel is the average of the four sides. A single-marker scale is the ratio of the known side length to the pixel side length, and multiple-marker scales are weighted by the effective marked area. Scale usage status can be recorded as primary scale, multi-marker fusion, consistency check only, rejection due to occlusion or missing corners, or the use of a user-confirmed engineering prior scale. The purpose of calculating the scale is to convert mask width, pole height, cantilever length, and device position from pixel coordinates to structural physical coordinates. Quality control metadata is not a pre-existing external database, but rather a JSON or equivalent record saved with each sample, including scale source, marker number, screening conclusion, mask parameters, section selection, and manual revision status. The conclusions of this step are: the image modelability status, the reasons for rejection or manual review, the pixel-physical scale relationship and its source record; the above results serve as inputs for target contour generation in step 2 and physical size conversion in step 3.

[0029] S2, using a visual language model to generate the initial closed contour of the target monitoring pole, and obtaining a binary mask for structural modeling of the pole through morphological processing and connected component filtering.

[0030] The initial closed contour of the target monitoring pole is generated using a visual language model, including: The visual language model is constrained to draw a pure red closed target outline on the same canvas; The binary mask for structural modeling, obtained through morphological processing and connected component filtering, includes: Extract red pixels in HSV space, perform closing operations using structuring elements to fill closed contours; use dynamic connection kernels to connect local broken boundaries, retain the largest connected foreground region, and obtain a continuous binary mask; judge the mask quality based on the foreground area ratio and the number of connected regions, and if the quality does not meet the preset conditions, return a correction prompt or transfer to manual review.

[0031] Specifically, the input to this step is the original site image screened in step 1, the target object constraints, and scale metadata. The output is a target contour map with the same canvas size as the original image and the contour generation status. The image generation model is constrained to draw only pure red #FF0000 closed lines on the outer boundary of the same target pole structure, retaining the original image and ArUco markers, without outputting natural language coordinate points, and without replacing the contour with large-area smearing. Contour objects may include vertical main poles, horizontal cantilever arms, and cameras, lights, and electrical boxes belonging to the same load-bearing structure.

[0032] The local parser extracts red pixels in the HSV space with a hue no higher than 10 or no lower than 170, and a saturation and brightness no lower than 120. It performs two closing operations using a 5×5 structuring element and fills a closed contour with an area no less than 16 pixels and the larger of 0.00002 times the image area. Then, a dynamic connectivity kernel of 5 to 11 pixels is used to connect local break boundaries, preserving the largest 8-connected foreground region to obtain a continuous binary mask. A quality warning is triggered when the foreground area ratio of the mask is lower than 0.003 or higher than 0.12. If multiple regions, excluding the main connected component, have an area exceeding 5% of the main connected component, a message indicating contour breakage or background infiltration is displayed, and the mask is regenerated or manually reviewed.

[0033] S3. Based on the binary mask of the rod and the scale relationship, extract the center line and width curve of the vertical main rod and the horizontal cantilever, and identify the rod topology, physical size, cross-sectional change type, cross-sectional shape and location of auxiliary equipment.

[0034] The extraction of the centerline and width curve of the vertical main rod and the horizontal cantilever includes: The main connected domain is selected from the binary mask of the rod as the target rod region, and the calibration plate region and auxiliary equipment protrusions are excluded from the width sampling. The width curve is sampled along the centerline of the member. The first RANSAC residual threshold is used for the centerline of the vertical member, and the second RANSAC residual threshold is used for the cantilever. The horizontal cantilever is identified according to the horizontal extension constraint to avoid misjudging the auxiliary equipment as the load-bearing cantilever.

[0035] Among them, identifying the types of cross-sectional changes includes: Construct a segmented engineering diameter model to represent constant cross-section, segmented variable cross-section, and continuous tapered variable cross-section; The cross-section variation model is selected based on length-weighted diameter change, monotonicity ratio, and fitting error; when the relative diameter change is less than a preset threshold, it is determined to be an equal cross-section; when the width curve is represented by discrete segments, it is determined to be a piecewise variable cross-section; when the diameter changes continuously and meets the requirements of monotonicity and fitting error, it is determined to be a continuous conical variable cross-section. For cross-sectional shape, identify circular or polygonal cross-sections. When a polygonal hollow cross-section is identified, convert it into an equivalent circular tube cross-section with matching area and bending moment of inertia.

[0036] Specifically, this step takes the binary mask of the rod obtained in step (2) and the scale metadata from step (1) as input, and outputs the vertical main rod, horizontal cantilever, centerline, endpoints, connection relationships, width curves, and physical dimensions. The system first enlarges the region of a single ArUco marker by 2.2 times, and the joint convex hull formed by multiple markers by 1.35 times before excluding it from the width sampling. The centerline of the vertical rod adopts a 3.5-pixel RANSAC residual threshold, allowing vertical discontinuities of no more than 8 pixels, and the stable segment covers at least 20 rows, representing the width taken as the 60th percentile of the effective samples; the cantilever adopts a 3.0-pixel residual threshold, the effective horizontal segment covers at least 10 columns, the horizontal span is at least 1.25 times the corresponding vertical change, representing the width taken as the 30th percentile. The purpose of this processing is not to obtain the region code again, but to transform the mask into topological and geometric quantities that can be used for engineering modeling.

[0037] After obtaining the centerline, width curve, and physical length of each observation segment, the system further determines the cross-sectional dimension changes of the member along the height direction. The cross-sectional variation model is used to describe the variation of the outer diameter with height, which belongs to two different dimensions from the cross-sectional shape determination of the next segment. This implementation uses the segmented engineering diameter model shown in the following formula to represent constant cross-section, segmented variable cross-section, and continuous tapered variable cross-section, and obtains the inner diameter from the outer diameter and wall thickness: ; In the formula, and These represent the outer diameter and inner diameter at height z, respectively. This represents the measured vertical height of the pole, with z measured from the bottom of the pole upwards. and These represent the outer diameters of the base and top of the rod, respectively. Indicates the first The outer diameter of each segment This indicates the estimated wall thickness. The system selects a constant cross section, a segmented variable cross section, or a continuous conical model based on the length-weighted diameter variation, monotonicity ratio, and fitting error. When the relative error of the constant cross section is less than 0.06, the constant cross section is selected first. Segmented candidates are merged according to the diameter difference and the short segment threshold. The penalized errors of the conical and segmented models are compared only when the conditions of relative variation, monotonicity, and fitting error are met.

[0038] The cross-section variation type describes whether the outer diameter changes along the rod height, while the cross-section shape describes the cross-sectional profile at any height. These two can be combined, for example, a uniform cross-section circular tube, a segmented variable-diameter circular tube, or a continuous conical octagonal hollow rod. The system can identify polygonal cross-sections such as circles and octagons; for other regular thin-walled cross-sections, they can be equivalent to circular tubes or corresponding beam cross-sections based on area and primary bending stiffness. The identification results of auxiliary equipment are converted into concentrated masses and their application locations; the mass of cameras, lighting fixtures, or electrical boxes is applied to the mounting node or the nearest load-bearing rod node. The conclusion of this step is: the topology of straight or L-shaped rods, the physical dimensions of the main rod and cantilever, the cross-section variation type, the cross-sectional shape, and the location and mass mapping information of auxiliary equipment are formed by a binary mask, and this information is used as input for generating structured engineering parameters.

[0039] S4. Based on the rod topology, physical dimensions, cross-sectional change type, cross-sectional shape, and location of auxiliary equipment, and in conjunction with the scale relationship, the image coordinates are converted into structural physical coordinates to generate structured engineering parameters that include the locations of nodes, elements, cross-sections, materials, mass, constraints, and response records.

[0040] Among them, the structured engineering parameters also include quality control metadata for scale source, mask quality, and cross-section recognition quality; The location of the auxiliary equipment is mapped to a concentrated mass and its position of action, and the mass of the auxiliary equipment is applied to the corresponding installation node or the nearest load-bearing member node.

[0041] Specifically, this step takes the topology, width, cross-section, and auxiliary equipment location from step 3, and the scale from step 1 as input. First, the pixel coordinates are converted to millimeter physical coordinates according to the scale, and then written into the structural object using a unified field. The structured engineering parameters include node coordinates, element connections, segment lengths, outer diameters, inner diameters, wall thicknesses, cross-section types, material properties, auxiliary equipment mass, foundation constraints, response record locations, seismic motion input interfaces, scale sources, mask quality, and cross-section recognition quality. This structural object serves as a unified interface between the vision module and the finite element generation module.

[0042] The finite element generation module generates the model file item by item according to the nodes and connections in the structural object: It establishes three-dimensional six-degree-of-freedom nodes based on the pole base, segment point, pole top, cantilever end, and equipment mounting point; it creates beam-column elements based on adjacent nodes and assigns cross-sections; it calculates the mass of the members based on material density, cross-sectional area, and element length, and adds half of the mass of each element to the two end nodes; it applies the mass of the camera or lighting fixture as a concentrated mass to the corresponding node; the foundation includes fixed constraints at the pole base, and rotational springs can also be set if rotational stiffness is given separately. Beam-column elements belong to the element definition part of the finite element model.

[0043] In one OpenSees implementation, the finite element project includes material files, section files, node files, element files, mass files, constraint files, seismic input files, analysis step files, main execution script, response recorder, structural parameters, and a run summary. The current generator uses a three-dimensional six-DOF model, P-Delta coordinate transformation, forceBeamColumn elements, and 5-point Lobatto integration; the circular fiber section is partitioned into 96 circumferential and 4 radial sections. This implementation is merely an example of an executable platform; the structural engineering parameters can also be converted to input for other finite element software or a custom solver.

[0044] S5, read the structured engineering parameters and automatically generate the model files and analysis scripts required for the finite element simulation project.

[0045] This includes automatically generating the model files and analysis scripts required for finite element simulation projects, including: Establish three-dimensional six-degree-of-freedom nodes based on the pole base, segment point, pole top, cantilever end, and equipment installation point; Beam-column elements are created based on adjacent nodes, and cross sections are assigned accordingly. The distributed mass of the members is calculated based on the material density, cross-sectional area and element length. The element mass is distributed to the end nodes, and the mass of the auxiliary equipment is applied as concentrated mass to the corresponding nodes. Generates a finite element project file containing materials, sections, nodes, elements, masses, constraints, seismic inputs, analysis steps, main execution script, and response recorder.

[0046] S6. Perform a consistency check on the finite element simulation project. After the check passes, perform dynamic analysis and export the response time history data of the target response location.

[0047] The execution dynamics analysis includes: Gravity load is applied in stages; The transient integral algorithm is used for dynamic analysis. When the algorithm fails to converge, the line search algorithm, Broyden algorithm, or the time step is reduced to a preset ratio of the original step size is tried in turn.

[0048] Before performing dynamic analysis, the system checks the unit system, node and element connectivity, positive values ​​of dimensions and section parameters, whether the mass is applied repeatedly, foundation constraints, seismic input, recorder target nodes, and short-time history test run output. Gravity loads are applied in 10 steps; if the conventional Newton algorithm fails to converge, the system switches to NewtonLineSearch. The system extracts the first two modes to calculate Rayleigh damping.

[0049] After the consistency check passes, the ground motion is input along the X, Y, and Z directions using UniformExcitation mode, with transient integration using Newmark parameters of 0.5 and 0.25. If convergence fails, NewtonLineSearch, Broyden, and reducing the time step to one-tenth of the original step size are tried sequentially. The system outputs the displacement, velocity, or acceleration response time history of the pole top, cantilever end, or camera node, and simultaneously saves the original image, contour map, mask, geometric results, structural parameters, finite element input, and simulation summary.

[0050] See Figure 2 , Figure 2 This is a schematic diagram of the workflow from image to structure modeling and finite element response simulation, showing the dependencies between user constraints, visual analysis, engineering parameter organization, model generation, consistency checks, and dynamic simulation, and indicating the roles of scale, mass, constraints, seismic input, and response record location in the process.

[0051] See Figure 3 , Figure 3 This is a schematic diagram of the morphological geometry extraction and finite element model generation process. It shows the process of obtaining a binary mask from the original image and target contour, further recovering the center line, vertical rod and cantilever topology, sampling the width along the center line, selecting the cross-sectional variation model, and mapping the results to nodes, elements, cross sections, materials, mass and constraints.

[0052] See Figure 4 , Figure 4 This is a schematic diagram of the image recognition, masking, topology, and cross-sectional results of the on-site monitoring pole. Each group of sub-images sequentially represents the original on-site image, the outer contour of the target, the binary mask obtained from contour analysis, the center line and endpoints of the main pole and cantilever, as well as the width sampling or cross-sectional judgment results, used to illustrate the intermediate outputs of the same image at each processing stage.

[0053] Based on the same inventive concept, this application also provides a vision-based automatic finite element modeling and simulation device for implementing the aforementioned vision-based automatic finite element modeling and simulation method for monitoring poles. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more vision-based automatic finite element modeling and simulation device embodiments provided below can be found in the limitations of the vision-based automatic finite element modeling and simulation method for monitoring poles described above, and will not be repeated here.

[0054] In one exemplary embodiment, a vision-based finite element automatic modeling and simulation device for monitoring poles is provided, comprising: The scale calibration module is used to acquire on-site images and scale calibration information containing the target monitoring pole, determine whether the target monitoring pole can be modeled, and determine the scale relationship between pixel coordinates and structural physical coordinates; The mask correction module is used to generate the initial closed contour of the target monitoring pole with the help of a visual language model, and obtains a binary mask of the pole for structural modeling through morphological processing and connected component filtering. The topology recognition module is used to extract the centerline and width curve of the vertical main rod and the horizontal cantilever based on the binary mask of the rod and the scale relationship, and to identify the rod topology, physical dimensions, cross-sectional change type, cross-sectional shape and location of auxiliary equipment; The parameter generation module is used to convert image coordinates into structural physical coordinates based on the rod topology, physical dimensions, cross-sectional change type, cross-sectional shape, and location of auxiliary equipment, combined with the scale relationship, and generate structured engineering parameters including the locations of nodes, elements, cross-sections, materials, masses, constraints, and response records. The project generation module is used to read the structured engineering parameters and automatically generate the model files and analysis scripts required for the finite element simulation project. The simulation module is used to perform a consistency check on the finite element simulation project. After the check passes, dynamic analysis is performed to export the response time history data of the target response location.

[0055] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0056] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A vision-based finite element automatic modeling and simulation method for monitoring poles, characterized in that, The vision-based finite element automatic modeling and simulation method for monitoring poles includes: S1, acquire on-site images and scale calibration information containing the target monitoring pole, determine whether the target monitoring pole can be modeled, and determine the scale relationship between pixel coordinates and structural physical coordinates; S2, using a visual language model to assist in generating the initial closed contour of the target monitoring pole, and obtaining a pole binary mask for structural modeling through morphological processing and connected component filtering; S3. Based on the binary mask of the rod and the scale relationship, extract the center line and width curve of the vertical main rod and the horizontal cantilever, and identify the rod topology, physical size, cross-sectional change type, cross-sectional shape and location of auxiliary equipment; S4. Based on the rod topology, physical dimensions, cross-sectional change type, cross-sectional shape, and location of auxiliary equipment, and in conjunction with the scale relationship, the image coordinates are converted into structural physical coordinates to generate structured engineering parameters that include the locations of nodes, elements, cross-sections, materials, mass, constraints, and response records. S5, Read the structured engineering parameters and automatically generate the model files and analysis scripts required for the finite element simulation project; S6. Perform a consistency check on the finite element simulation project. After the check passes, perform dynamic analysis and export the response time history data of the target response location.

2. The vision-based finite element automatic modeling and simulation method for monitoring poles according to claim 1, characterized in that, In step S1, determining whether the target monitoring pole can be modeled includes: outputting the existence of a modelable main pole and the confidence level of the pole through an inference model; when a main pole exists and the confidence level is not lower than a preset gating threshold, and there is a usable scale source, it is determined to be modelable. The scale calibration information includes ArUco markers, known-size components, camera parameters, engineering prior dimensions, or multi-view geometric constraints. When an ArUco marker is detected, the pixel scale is calculated based on the marker corner coordinates, and the marker number, marker quantity, marker side length, scale confidence level, and usage status are recorded as the scale relationship.

3. The vision-based finite element automatic modeling and simulation method for monitoring poles according to claim 1, characterized in that, In step S2, the initial closed contour of the target monitoring pole is generated using a visual language model, including: The visual language model is constrained to draw a pure red closed target outline on the same canvas; The binary mask for structural modeling, obtained through morphological processing and connected component filtering, includes: Extract red pixels in HSV space, perform closing operations using structuring elements to fill closed contours; use dynamic connection kernels to connect local broken boundaries, retain the largest connected foreground region, and obtain a continuous binary mask; judge the mask quality based on the foreground area ratio and the number of connected regions, and if the quality does not meet the preset conditions, return a correction prompt or transfer to manual review.

4. The vision-based finite element automatic modeling and simulation method for monitoring poles according to claim 1, characterized in that, In step S3, the centerline and width curve of the vertical main rod and the horizontal cantilever are extracted, including: The main connected domain is selected from the binary mask of the rod as the target rod region, and the calibration plate region and auxiliary equipment protrusions are excluded from the width sampling. The width curve is sampled along the centerline of the member. The first RANSAC residual threshold is used for the centerline of the vertical member, and the second RANSAC residual threshold is used for the cantilever. The horizontal cantilever is identified according to the horizontal extension constraint to avoid misjudging the auxiliary equipment as the load-bearing cantilever.

5. The vision-based finite element automatic modeling and simulation method for monitoring poles according to claim 1, characterized in that, In S3, identifying the type of cross-sectional change includes: Construct a segmented engineering diameter model to represent constant cross-section, segmented variable cross-section, and continuous tapered variable cross-section; The cross-section variation model is selected based on length-weighted diameter change, monotonicity ratio, and fitting error; when the relative diameter change is less than a preset threshold, it is determined to be an equal cross-section; when the width curve is represented by discrete segments, it is determined to be a piecewise variable cross-section; when the diameter changes continuously and meets the requirements of monotonicity and fitting error, it is determined to be a continuous conical variable cross-section. For cross-sectional shape, identify circular or polygonal cross-sections. When a polygonal hollow cross-section is identified, convert it into an equivalent circular tube cross-section with matching area and bending moment of inertia.

6. The vision-based finite element automatic modeling and simulation method for monitoring poles according to claim 1, characterized in that, In S4, the structured engineering parameters also include quality control metadata for scale source, mask quality, and cross-section recognition quality; The location of the auxiliary equipment is mapped to a concentrated mass and its position of action, and the mass of the auxiliary equipment is applied to the corresponding installation node or the nearest load-bearing member node.

7. The vision-based finite element automatic modeling and simulation method for monitoring poles according to claim 1, characterized in that, In step 5, the model files and analysis scripts required for the finite element simulation project are automatically generated, including: Establish three-dimensional six-degree-of-freedom nodes based on the pole base, segment point, pole top, cantilever end, and equipment installation point; Beam-column elements are created based on adjacent nodes, and cross sections are assigned accordingly. The distributed mass of the members is calculated based on the material density, cross-sectional area and element length. The element mass is distributed to the end nodes, and the mass of the auxiliary equipment is applied as concentrated mass to the corresponding nodes. Generates a finite element project file containing materials, sections, nodes, elements, masses, constraints, seismic inputs, analysis steps, main execution script, and response recorder.

8. The vision-based finite element automatic modeling and simulation method for monitoring poles according to claim 1, characterized in that, In step 6, the consistency check includes: checking whether the unit system is consistent, whether the node connectivity is complete, whether the size and cross-sectional parameters are positive, whether the mass is applied repeatedly, whether the foundation constraints are reasonable, whether the seismic input exists, whether the recorder covers the target response location, and whether the short-term trial run can output the response normally.

9. The vision-based finite element automatic modeling and simulation method for monitoring poles according to claim 1, characterized in that, In step 6, dynamic analysis is performed, including: Gravity load is applied in stages; The transient integral algorithm is used for dynamic analysis. When the algorithm fails to converge, the line search algorithm, Broyden algorithm, or the time step is reduced to a preset ratio of the original step size is tried in turn.

10. A vision-based finite element automatic modeling and simulation device for monitoring poles, characterized in that, The vision-based finite element automatic modeling and simulation device for monitoring poles includes: The scale calibration module is used to acquire on-site images and scale calibration information containing the target monitoring pole, determine whether the target monitoring pole can be modeled, and determine the scale relationship between pixel coordinates and structural physical coordinates; The mask correction module is used to generate the initial closed contour of the target monitoring pole with the help of a visual language model, and obtains a binary mask of the pole for structural modeling through morphological processing and connected component filtering. The topology recognition module is used to extract the centerline and width curve of the vertical main rod and the horizontal cantilever based on the binary mask of the rod and the scale relationship, and to identify the rod topology, physical dimensions, cross-sectional change type, cross-sectional shape and location of auxiliary equipment; The parameter generation module is used to convert image coordinates into structural physical coordinates based on the rod topology, physical dimensions, cross-sectional change type, cross-sectional shape, and location of auxiliary equipment, combined with the scale relationship, and generate structured engineering parameters including the locations of nodes, elements, cross-sections, materials, masses, constraints, and response records. The project generation module is used to read the structured engineering parameters and automatically generate the model files and analysis scripts required for the finite element simulation project. The simulation module is used to perform a consistency check on the finite element simulation project. After the check passes, dynamic analysis is performed to export the response time history data of the target response location.

Images