A flexible support deflection rapid evaluation method and system based on unmanned aerial vehicle inspection images

Abstract

This invention belongs to the field of power facility inspection and discloses a method and system for rapid assessment of flexible support deflection based on UAV inspection images. The method involves automatically acquiring multi-angle inspection images and pose data using a UAV; accurately extracting the support contour from the images using a trained semantic segmentation model and filtering out background interference; and reconstructing a 3D point cloud model of the support's current state using structure-of-motion (SOG) technology based on the contour image and camera parameters. This model is then registered with the design 3D model to unify the coordinate space. In the registered space, the vertical distance from each point of the reconstructed model to the baseline is calculated to obtain accurate deflection values, a distribution curve is plotted, and the deflection is automatically compared to the deflection limit for safety assessment, generating a visual assessment report. This invention eliminates the need for markers or sensors on the support; the acquired images and flight control data, through deep learning and 3D vision technology, determine the support deflection, providing data support for structural safety early warning.

Description

Technical Field

[0001] This invention belongs to the field of power facility inspection, specifically relating to a method and system for rapid assessment of flexible support deflection based on UAV inspection images. Background Technology

[0002] The deflection of flexible supports is a key parameter for assessing structural safety, operational status, and load-bearing capacity. Excessive deflection can lead to insufficient electrical clearance, abnormal structural stress, material fatigue, and even serious accidents such as structural collapse. Therefore, regular and accurate monitoring of the deflection of flexible supports is crucial.

[0003] Traditional deflection measurement methods have the following limitations: Manual measurement relies on operators using physical methods such as total stations, theodolites, or string lines. This method is not only inefficient but also requires inspectors to work close to or at heights, posing significant safety risks in high-pressure, high-altitude, and complex terrain environments. Furthermore, it is difficult to conduct rapid surveys of large-scale facilities. Contact sensor methods involve real-time monitoring using devices such as inclinometers, displacement gauges, or distributed fiber optic sensors mounted on supports. While this method enables continuous measurement, the installation, wiring, and maintenance costs of the sensors are high, and it is unsuitable for large-scale deployment and retrofitting of existing, numerous facilities. Traditional photogrammetry uses cameras at fixed ground positions to capture images and analyzes the displacement of specific marker points through image processing. However, limited by a fixed, single, or finite viewpoint, it is difficult to fully capture the overall deformation information of the structure in three-dimensional space; its automation level is typically low, the processing flow is cumbersome, and it is susceptible to environmental factors such as lighting changes and background interference.

[0004] In recent years, drone inspection technology has been widely used in power, infrastructure and other fields due to its flexibility, efficiency and safety. Currently, most drone image-based inspection technologies focus on identifying defects in discrete components such as insulator damage, loose bolts, and broken conductor strands. However, quantitative and rapid assessment methods for deflection, a core parameter characterizing the overall mechanical state of flexible supports, remain immature. Existing image-based deflection assessment schemes typically rely on pre-attaching manual markers (targets) to the structure under test, or on deploying and measuring high-precision ground control points on-site. These preliminary steps not only increase the complexity and time cost of the operation, but also fail to fully explore and utilize the inherent value of the multi-view image sequences and their accompanying high-precision position and attitude data naturally provided by drone inspection. Summary of the Invention

[0005] The purpose of this invention is to solve the problem that the existing technology requires preset markings for rapid assessment of flexible support deflection and cannot fully utilize UAV inspection data to achieve automated and rapid assessment of flexible support deflection. The invention provides a method and system for rapid assessment of flexible support deflection based on UAV inspection images.

[0006] To achieve the above objectives, the present invention employs the following technical solution: The present invention proposes a method for rapid assessment of the deflection of flexible supports based on UAV inspection images, comprising the following steps: Obtain pixel position data from multi-angle inspection images of the target flexible support, and process the pixel position data to obtain the support outline image; Based on the scaffold contour image and camera parameters, a dense 3D point cloud model is obtained by combining feature matching, sparse reconstruction and dense matching processing. Register the dense 3D point cloud model with the preset benchmark point cloud model to unify the coordinate space; In the registered space, the distance from the preset reference point cloud model to the baseline is obtained to obtain the accurate deflection value, thus realizing the rapid assessment of the flexible support deflection.

[0007] Preferably, the step of acquiring pixel position data of the multi-angle inspection image of the target flexible support specifically involves: The drone carries a camera and flies along the flight path, triggering shooting at fixed time intervals or distance intervals, simultaneously recording the GNSS position and IMU attitude of each image, and storing the camera's intrinsic parameters; among them, an automatic inspection route is planned in the drone's ground control station according to the span, height and spatial shape of the flexible support. Image pixel coordinates With camera coordinate system The mapping is described by the camera model, represented by the pinhole model:

[0008] in, For the camera intrinsic parameter matrix, ; , These are the rotation matrix and translation vector from the world coordinate system to the camera coordinate system, respectively. The coordinates of a three-dimensional point in the world coordinate system. f For camera focal length, These are the coordinates of the camera's principal point.

[0009] Preferably, the step of processing the pixel position data to obtain the bracket outline image specifically involves: A fully convolutional neural network with an encoder-decoder structure is used as the semantic segmentation model. The loss function is a weighted sum of cross-entropy loss and Dice loss. It is expressed as follows:

[0010] in, For pixels The true label, 0 for background, 1 for support; For pixels The probability of a stent being prescribed; , All are weighting coefficients; The inspection image is input into the trained semantic segmentation model and the output is a binary mask image, which is the outline image of the support structure; where white pixels represent the support structure area and black pixels represent the background.

[0011] Preferably, the process of obtaining a dense 3D point cloud model based on the scaffold contour image and camera parameters, combined with feature matching, sparse reconstruction, and dense matching, specifically involves: Feature points are extracted from multiple masked images, and feature matching is performed. Based on epipolar geometry constraints, the matching results are constrained, and the essential matrix is ​​solved using the five-point method. : ;in, x 1. x 2 represents the normalized coordinates of the matching points in the two images. This is the camera intrinsic parameter matrix; Camera parameters are obtained based on the essential matrix; the 3D coordinates of the matching points are calculated through triangulation; and the camera parameters and 3D points are optimized using bundle adjustment.

[0012] Based on the optimization results, a multi-view stereo vision algorithm is used to generate a dense 3D point cloud through epipolar search and photometric consistency calculation, which is the dense 3D point cloud model.

[0013] in, For projection functions; Let be a rotation matrix. It is a translation vector. For the first The world coordinates of a three-dimensional point For the first The first image The pixel coordinates of each feature point For the first point cloud A three-dimensional point.

[0014] Preferably, the registration of the dense 3D point cloud model with a preset reference point cloud model to unify the coordinate space specifically involves: Obtain the baseline point cloud model:

[0015] Minimize the distance between the dense 3D point cloud model and the baseline point cloud model: Solving for the optimal rotation matrix using SVD R With translation vector t Iterate until convergence; Dense 3D point cloud model Through transformation Mapping to a reference coordinate system unifies the coordinate space; in, for P The Middle The point is at Q The nearest neighbor index in the middle, A vector containing three-dimensional coordinates is m The spatial position of a point in a reference coordinate system. For the first point cloud A three-dimensional point.

[0016] Preferably, in the registered space, by obtaining the distance from the preset reference point cloud model to the reference line, a precise deflection value is obtained, thereby realizing rapid assessment of the flexible support deflection. Specifically: In the baseline point cloud model, select two suspension points and connect them to form the reference baseline. ; After registration, the point cloud Every point in Calculate its distance from the baseline. The vertical distance is used as the deflection value:

[0017] Where × represents the cross product of vectors, and Indicates the two suspension points; Mid-span lowest point identification: The point cloud is divided into bins along the support structure. The point with the minimum Z-value in each longitudinal interval is selected, and then the global minimum point is selected as the mid-span lowest point. Its deflection value is... .

[0018] Preferably, the deflection value is compared with a set allowable deflection threshold. ,like If the deflection does not exceed the limit, it meets the safety criterion; if If the deflection exceeds the limit, it is determined that the safety criterion is violated.

[0019] This invention proposes a rapid deflection assessment system for flexible supports based on UAV inspection images, comprising: The contour extraction module is used to acquire pixel position data of multi-angle inspection images of the target flexible support, and to process the pixel position data to obtain the support contour image. The 3D reconstruction module is used to obtain a dense 3D point cloud model based on the bracket contour image and camera parameters, combined with feature matching, sparse reconstruction and dense matching processing. The model registration module is used to register the dense 3D point cloud model with the preset reference point cloud model to unify the coordinate space. The deflection analysis module is used to obtain accurate deflection values ​​in the registered space by acquiring the distance from the preset reference point cloud model to the reference line, thereby enabling rapid assessment of the deflection of the flexible support.

[0020] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the method for rapid assessment of flexible support deflection based on UAV inspection images.

[0021] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method for rapid assessment of flexible support deflection based on UAV inspection images.

[0022] Compared with the prior art, the present invention has the following beneficial effects: This invention proposes a rapid deflection assessment method for flexible supports based on UAV inspection images. First, multi-angle inspection images and pose data are acquired. A semantic segmentation model is used to accurately extract the support contour from the images, filtering out background interference. Based on these contour images and camera parameters, combined with feature matching, sparse reconstruction, and dense matching processing, a dense 3D point cloud model of the support's current state is reconstructed. This model is then registered with a pre-set 3D model to unify the coordinate space. In the registered space, the precise deflection value is obtained by calculating the vertical distance from each point of the reconstructed model to the baseline, and a distribution curve can be plotted. The system automatically compares the deflection to the deflection limit for safety judgment and generates a visual assessment report. This method eliminates the need for markers or sensors on the support. Utilizing images and flight control data acquired through conventional UAV inspections, and employing deep learning and 3D vision technology, it efficiently and accurately calculates the support deflection, providing data support for structural safety early warning and realizing a solution for automated and rapid deflection assessment of flexible supports.

[0023] This invention proposes a rapid deflection assessment system for flexible supports based on UAV inspection images. By dividing the system into a contour extraction module, a 3D reconstruction module, a model registration module, and a deflection analysis module, accurate deflection values ​​are obtained, enabling rapid assessment of the flexible support deflection. The modular approach ensures that each module is independent, facilitating unified management of all modules. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a flowchart of the method for rapid assessment of flexible support deflection based on UAV inspection images according to the present invention.

[0026] Figure 2 This is a detailed flowchart of the method for rapid assessment of flexible support deflection based on UAV inspection images according to the present invention.

[0027] Figure 3 This is a diagram of the flexible support deflection rapid assessment system based on UAV inspection images of the present invention.

[0028] Figure 4 This is a schematic diagram of the structure of an electronic device according to the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0030] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0031] The present invention will now be described in further detail with reference to the accompanying drawings: Example 1 This invention provides a rapid deflection assessment method for flexible supports based on UAV inspection images. It eliminates the need for markers or sensors on the support, utilizing images and flight control data acquired through conventional UAV inspections. Through deep learning and 3D vision technology, it efficiently and accurately calculates the support deflection, providing data support for structural safety early warning. The method first automatically collects multi-angle inspection images and pose data using a UAV. A trained deep learning semantic segmentation model is then used to accurately extract the support contour from the images, filtering out background interference. Based on these contour images and camera parameters, a 3D point cloud model of the support's current state is reconstructed using motion reconstruction technology. This model is then registered with a 3D model of the design or baseline state to unify the coordinate space. In the registered space, the precise deflection value is obtained by calculating the vertical distance from each point on the reconstructed model to the baseline, and a distribution curve can be plotted. The system automatically compares the deflection with the deflection limit for safety assessment and generates a visual assessment report, such as... Figure 1 As shown, it includes the following steps: S1. Obtain pixel position data of multi-angle inspection images of the target flexible support, and process the pixel position data to obtain the support outline image; The acquisition of pixel position data from multi-angle inspection images of the target flexible support specifically involves: The drone carries a camera and flies along the flight path, triggering shooting at fixed time intervals or distance intervals, simultaneously recording the GNSS position and IMU attitude of each image, and storing the camera's intrinsic parameters; among them, an automatic inspection route is planned in the drone's ground control station according to the span, height and spatial shape of the flexible support. Image pixel coordinates With camera coordinate system The mapping is described by the camera model, represented by the pinhole model:

[0032] in, For the camera intrinsic parameter matrix, ; , These are the rotation matrix and translation vector from the world coordinate system to the camera coordinate system, respectively. The coordinates of a three-dimensional point in the world coordinate system. f For camera focal length, These are the coordinates of the camera's principal point.

[0033] The process of processing pixel position data to obtain the bracket outline image specifically involves: A fully convolutional neural network with an encoder-decoder structure is used as the semantic segmentation model. The loss function is a weighted sum of cross-entropy loss and Dice loss. It is expressed as follows:

[0034] in, For pixels The true label, 0 for background, 1 for support; For pixels The probability of a stent being prescribed; , All are weighting coefficients; The inspection image is input into the trained semantic segmentation model and the output is a binary mask image, which is the outline image of the support structure; where white pixels represent the support structure area and black pixels represent the background.

[0035] S2. Based on the bracket contour image and camera parameters, combined with feature matching, sparse reconstruction and dense matching processing, a dense 3D point cloud model is obtained. The process of obtaining a dense 3D point cloud model based on the scaffold contour image and camera parameters, combined with feature matching, sparse reconstruction, and dense matching, is as follows: Feature points are extracted from multiple masked images, and feature matching is performed. Based on epipolar geometry constraints, the matching results are constrained, and the essential matrix E is solved using the five-point method. ;in, x 1. x 2 represents the normalized coordinates of the matching points in the two images; Camera parameters are obtained based on the essential matrix; the 3D coordinates of the matching points are calculated through triangulation; and the camera parameters and 3D points are optimized using bundle adjustment.

[0036] Based on the optimization results, a multi-view stereo vision algorithm is used to generate a dense 3D point cloud through epipolar search and photometric consistency calculation, which is the dense 3D point cloud model.

[0037] in, For projection functions; Let be a rotation matrix. It is a translation vector. For the first The world coordinates of a three-dimensional point For the first The first image The pixel coordinates of each feature point For the first point cloud A three-dimensional point.

[0038] S3. Register the dense 3D point cloud model with the preset reference point cloud model to unify the coordinate space; The process of registering the dense 3D point cloud model with a preset reference point cloud model to unify the coordinate space specifically involves: Obtain the baseline point cloud model:

[0039] Minimize the distance between the dense 3D point cloud model and the baseline point cloud model: Solving for the optimal rotation matrix using SVD R With translation vector t Iterate until convergence; Dense 3D point cloud model Through transformation Mapping to a reference coordinate system unifies the coordinate space; in, for P The Middle The point is at Q The nearest neighbor index in the middle, A vector containing three-dimensional coordinates is m The spatial position of a point in a reference coordinate system. For the first point cloud A three-dimensional point.

[0040] S4. In the registered space, by obtaining the distance from the preset reference point cloud model to the baseline, the accurate deflection value is obtained, realizing rapid assessment of the flexible support deflection.

[0041] In the registered space, by obtaining the distance from the preset reference point cloud model to the reference line, a precise deflection value is obtained, enabling rapid assessment of the flexible support deflection. Specifically: In the baseline point cloud model, select two suspension points and connect them to form the reference baseline. ; After registration, the point cloud Every point in Calculate its distance from the baseline. The vertical distance is used as the deflection value:

[0042] Where × represents the cross product of vectors, and Indicates the two suspension points; Mid-span lowest point identification: The point cloud is divided into bins along the support structure. The point with the minimum Z-value in each longitudinal interval is selected, and then the global minimum point is selected as the mid-span lowest point. Its deflection value is... .

[0043] Compare the deflection value with the set allowable deflection threshold. ,like If the deflection does not exceed the limit, it meets the safety criterion; if If the deflection exceeds the limit, it is determined that the safety criterion is violated.

[0044] The following is combined Figure 2 The method is described in detail below: S1, Image Acquisition. Plan the drone's automated inspection route to ensure complete coverage of the target flexible support from multiple angles. The drone is equipped with a high-resolution visible light camera to capture images, simultaneously recording POS data such as GNSS position and IMU attitude for each frame.

[0045] S1.1, Flight Path Planning: Based on the span, height, and spatial shape of the flexible support structure, plan an automated inspection flight path in the UAV ground control station. The flight path should ensure that the UAV covers the support structure from multiple perspectives to ensure the integrity of subsequent 3D reconstruction.

[0046] S1.2, Data Acquisition: The UAV, equipped with a high-resolution visible light camera, flies along the flight path, triggering images at fixed time intervals or distance intervals, simultaneously recording the GNSS position (latitude, longitude, and altitude) and IMU attitude (roll angle) of each image. φ Pitch angle θ Yaw angle ψ POS data such as focal length, and store camera intrinsic parameters (focal length) f Principal point coordinates Distortion coefficient k 1, k 2, p 1, p 2).

[0047] S1.3, coordinate mapping relationship, image pixel coordinates With camera coordinate system The mapping is described by the camera model, represented by the pinhole model:

[0048] in, For the camera intrinsic parameter matrix,

[0049] in, , These are the rotation matrix and translation vector from the world coordinate system to the camera coordinate system, respectively, calculated from the POS data; These are the coordinates of a three-dimensional point in the world coordinate system.

[0050] S2, Contour Extraction. A deep learning semantic segmentation model for flexible scaffolds is constructed and trained. The inspection image is input into this model, which automatically and accurately segments the main contour of the scaffold, outputting a binarized mask image to filter out background interference.

[0051] S2.1, Model Construction: A fully convolutional neural network with an encoder-decoder structure is used as the semantic segmentation model. The encoder uses pre-trained networks such as ResNet to extract features, and the decoder recovers details through upsampling and skip connections.

[0052] S2.2, the loss function, uses a weighted sum of cross-entropy loss and Dice loss to improve edge segmentation accuracy:

[0053] in For pixels The actual labels (0 for background, 1 for support); For pixels The probability of a stent being prescribed; , The weighting coefficient has a value of [value]. =0.5、 =0.5.

[0054] S2.3, Training and Inference: The model is trained using a labeled flexible support image dataset. After the inspection image is input into the model, a binary mask image is output, in which white pixels (value 1) represent the support area and black pixels (value 0) represent the background.

[0055] S3, 3D Reconstruction. Based on multi-view contour mask images, camera intrinsics, and POS data, motion reconstruction technology is employed. First, feature matching and sparse reconstruction are performed to generate a 3D sparse point cloud of key points on the scaffold; then, dense matching is performed to generate a dense 3D point cloud model describing the complete geometry of the scaffold.

[0056] S3.1, Feature Matching and Sparse Reconstruction: ORB or SIFT feature points are extracted from multiple masked images, and FLANN is used for feature matching. Based on epipolar geometry constraints, the essential matrix is ​​solved using the five-point method. :

[0057] in, x 1. x 2 represents the normalized coordinates of the matching points in the two images.

[0058] The 3D coordinates of the matching points are calculated using triangulation, and the camera parameters and 3D points are optimized using bundle adjustment.

[0059] in, This is the projection function.

[0060] S3.2, Dense Reconstruction: Employing a multi-view stereo vision algorithm, dense 3D point clouds are generated through epipolar search and photometric consistency calculation.

[0061] Statistical filtering and radius filtering are applied to the point cloud to remove outliers.

[0062] S4, Reference Model Registration. Obtain the 3D CAD model of the flexible support in its design state or the reference 3D model established through the initial inspection as a reference. Use the iterative nearest point algorithm to spatially register the current state reconstructed model obtained in step S3 with the reference model to unify the coordinate system.

[0063] S4.1, Baseline Model Acquisition: This adopts a 3D CAD model from the design phase or a baseline point cloud model established through the initial inspection.

[0064] S4.2, Iterative nearest point registration, the objective function is to minimize the current point cloud. P With reference point cloud Q Distance between:

[0065] in, for P The Middle i The point is at Q The nearest neighbor index in the database.

[0066] Solving the optimal rotation matrix using SVD R With translation vector t Iterate until convergence.

[0067] S4.3, Coordinate System 1, will reconstruct the point cloud. Through transformation Mapped to the reference coordinate system.

[0068] S5, Deflection Calculation and Analysis. In the registered space, the line connecting the suspension points of the reference model is defined as the reference baseline. Along the direction of gravity (or perpendicular to the baseline), calculate the vertical distance from each point on the current reconstructed model (especially the lowest point at mid-span) to the corresponding point on the reference model or the baseline; this distance is the deflection value. Calculate the deflection at a series of points and plot the deflection distribution curve along the length of the support.

[0069] S5.1, Baseline Definition: In the baseline model, select two suspension points A and B at both ends, and the line connecting them is the reference baseline. .

[0070] S5.2, Deflection calculation for registered point clouds Every point in Calculate its distance from the baseline. The vertical distance is used as the deflection value:

[0071] Where × represents the cross product of vectors.

[0072] Mid-span lowest point identification: The point cloud is divided into bins along the support structure. The point with the minimum Z-value in each longitudinal interval is selected, and then the global minimum point is selected as the mid-span lowest point. Its deflection value is... .

[0073] S5.3 Deflection distribution curve: Along the length of the support (with A as the origin and AB as the X-axis), calculate the average deflection using equally spaced sampling points and plot the deflection-position curve.

[0074] S6, Safety Assessment and Report Generation. The system compares the calculated deflection with design allowable values ​​or historical data to automatically determine if limits are exceeded and generates a visual report.

[0075] S6.1, Safety Criterion: Setting the Allowable Deflection Threshold The judgment condition is: .

[0076] S6.2, Visual Report, automatically generates charts including deflection distribution curves, maximum deflection location markers, 3D model comparison renderings, and highlighted over-limit areas; outputs structured reports containing information such as time, location, deflection extremes, and safety status.

[0077] Example 2 This invention proposes a rapid deflection assessment system for flexible supports based on UAV inspection images, such as... Figure 3 As shown, it includes a contour extraction module, a 3D reconstruction module, a model registration module, and a deflection analysis module; The contour extraction module is used to obtain pixel position data of multi-angle inspection images of the target flexible support, and to process the pixel position data to obtain the support contour image. The 3D reconstruction module is used to obtain a dense 3D point cloud model based on the bracket contour image and camera parameters, combined with feature matching, sparse reconstruction and dense matching processing. The model registration module is used to register the dense 3D point cloud model with the preset reference point cloud model to unify the coordinate space; The deflection analysis module is used to obtain accurate deflection values ​​in the registered space by acquiring the distance from the preset reference point cloud model to the baseline, thereby enabling rapid assessment of the flexible support deflection.

[0078] Example 3 Please see Figure 4As shown, the present invention also provides an electronic device 100 for a rapid assessment method of flexible support deflection based on UAV inspection images; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.

[0079] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the rapid assessment method for flexible support deflection based on UAV inspection images described in Embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device 100 (such as audio data), etc. In addition, the memory 101 may include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other non-volatile solid-state storage device.

[0080] The at least one processor 102 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor. The processor 102 is the control center of the electronic device 100, connecting various parts of the electronic device 100 via various interfaces and lines.

[0081] The memory 101 in the electronic device 100 stores multiple instructions to implement a method for rapid assessment of the deflection of a flexible support based on UAV inspection images, and the processor 102 can execute the multiple instructions to achieve the following: Obtain pixel position data from multi-angle inspection images of the target flexible support, and process the pixel position data to obtain the support outline image; Based on the scaffold contour image and camera parameters, a dense 3D point cloud model is obtained by combining feature matching, sparse reconstruction and dense matching processing. Register the dense 3D point cloud model with the preset benchmark point cloud model to unify the coordinate space; In the registered space, the distance from the preset reference point cloud model to the baseline is obtained to obtain the accurate deflection value, thus realizing the rapid assessment of the flexible support deflection.

[0082] Example 4 If the modules / units integrated in the electronic device 100 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0083] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0084] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0085] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0086] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for rapid assessment of flexible support deflection based on UAV inspection images, characterized in that, Includes the following steps: Obtain pixel position data from multi-angle inspection images of the target flexible support, and process the pixel position data to obtain the support outline image; Based on the scaffold contour image and camera parameters, a dense 3D point cloud model is obtained by combining feature matching, sparse reconstruction and dense matching processing. Register the dense 3D point cloud model with the preset benchmark point cloud model to unify the coordinate space; In the registered space, the distance from the preset reference point cloud model to the baseline is obtained to obtain the accurate deflection value, thus realizing the rapid assessment of the flexible support deflection.

2. The method for rapid assessment of flexible support deflection based on UAV inspection images according to claim 1, characterized in that, The acquisition of pixel position data from multi-angle inspection images of the target flexible support specifically involves: The drone carries a camera and flies along the flight path, triggering shooting at fixed time intervals or distance intervals, simultaneously recording the GNSS position and IMU attitude of each image, and storing the camera's intrinsic parameters; among them, an automatic inspection route is planned in the drone's ground control station according to the span, height and spatial shape of the flexible support. Image pixel coordinates With camera coordinate system The mapping is described by the camera model, represented by the pinhole model: in, For the camera intrinsic parameter matrix, ; , These are the rotation matrix and translation vector from the world coordinate system to the camera coordinate system, respectively. The coordinates of a three-dimensional point in the world coordinate system. f For camera focal length, These are the coordinates of the camera's principal point.

3. The method for rapid assessment of flexible support deflection based on UAV inspection images according to claim 1, characterized in that, The process of processing pixel position data to obtain the bracket outline image specifically involves: A fully convolutional neural network with an encoder-decoder structure is used as the semantic segmentation model. The loss function is a weighted sum of cross-entropy loss and Dice loss. It is expressed as follows: in, For pixels The true label, 0 for background, 1 for support; For pixels The probability of a stent being prescribed; , All are weighting coefficients; The inspection image is input into the trained semantic segmentation model and the output is a binary mask image, which is the outline image of the support structure; where white pixels represent the support structure area and black pixels represent the background.

4. The method for rapid assessment of flexible support deflection based on UAV inspection images according to claim 1, characterized in that, The process of obtaining a dense 3D point cloud model based on the scaffold contour image and camera parameters, combined with feature matching, sparse reconstruction, and dense matching, is as follows: Feature points are extracted from multiple masked images, and feature matching is performed. Based on epipolar geometry constraints, the matching results are constrained, and the essential matrix is ​​solved using the five-point method. : ; in, x 1. x 2 represents the normalized coordinates of the matching points in the two images. This is the camera intrinsic parameter matrix; Camera parameters are obtained based on the essential matrix; the 3D coordinates of the matching points are calculated through triangulation; and the camera parameters and 3D points are optimized using bundle adjustment. Based on the optimization results, a multi-view stereo vision algorithm is used to generate a dense 3D point cloud through epipolar search and photometric consistency calculation, which is the dense 3D point cloud model. in, For projection functions; Let be a rotation matrix. It is a translation vector. For the first The world coordinates of a three-dimensional point For the first The first image The pixel coordinates of each feature point For the first point cloud A three-dimensional point.

5. The method for rapid assessment of flexible support deflection based on UAV inspection images according to claim 1, characterized in that, The process of registering the dense 3D point cloud model with a preset reference point cloud model to unify the coordinate space specifically involves: Obtain the baseline point cloud model: Minimize the distance between the dense 3D point cloud model and the baseline point cloud model: Solving for the optimal rotation matrix using SVD R With translation vector t Iterate until convergence; Dense 3D point cloud model Through transformation Mapping to a reference coordinate system unifies the coordinate space; in, for P The Middle The point is at Q The nearest neighbor index in the middle, A vector containing three-dimensional coordinates is m The spatial position of a point in a reference coordinate system. For the first point cloud A three-dimensional point.

6. The method for rapid assessment of flexible support deflection based on UAV inspection images according to claim 1, characterized in that, In the registered space, by obtaining the distance from the preset reference point cloud model to the reference line, a precise deflection value is obtained, enabling rapid assessment of the flexible support deflection. Specifically: In the baseline point cloud model, select two suspension points and connect them to form the reference baseline. ; After registration, the point cloud Every point in Calculate its distance from the baseline. The vertical distance is used as the deflection value: Where × represents the cross product of vectors, and Indicates the two suspension points; Mid-span lowest point identification: The point cloud is divided into bins along the support structure. The point with the minimum Z-value in each longitudinal interval is selected, and then the global minimum point is selected as the mid-span lowest point. Its deflection value is... .

7. The method for rapid assessment of flexible support deflection based on UAV inspection images according to claim 6, characterized in that, Compare the deflection value with the set allowable deflection threshold. ,like If the deflection does not exceed the limit, it meets the safety criterion; if If the deflection exceeds the limit, it is determined that the safety criterion is violated.

8. A rapid deflection assessment system for flexible supports based on UAV inspection images, characterized in that, include: The contour extraction module is used to acquire pixel position data of multi-angle inspection images of the target flexible support, and to process the pixel position data to obtain the support contour image. The 3D reconstruction module is used to obtain a dense 3D point cloud model based on the bracket contour image and camera parameters, combined with feature matching, sparse reconstruction and dense matching processing. The model registration module is used to register the dense 3D point cloud model with the preset reference point cloud model to unify the coordinate space. The deflection analysis module is used to obtain accurate deflection values ​​in the registered space by acquiring the distance from the preset reference point cloud model to the reference line, thereby enabling rapid assessment of the deflection of the flexible support.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for rapid assessment of flexible support deflection based on UAV inspection images as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for rapid assessment of flexible support deflection based on UAV inspection images as described in any one of claims 1 to 7.

Images