Risk assessment methods, devices, equipment and storage media for bridge supports

Abstract

This application provides a method, apparatus, device, and storage medium for risk assessment of bridge piers. When a bridge pier comprises multiple individual piers, first point cloud data, visible light images, and location data are acquired for each individual pier. Then, a three-dimensional model of each individual pier is constructed based on this data. Next, a defect severity factor, deterioration trend factor, and detection confidence factor are determined for each individual pier based on the three-dimensional model. A first risk level for each individual pier is then determined based on these factors. Finally, a second risk level for the bridge pier is determined based on the first risk levels of all individual piers. This method effectively solves the problem that relying solely on two-dimensional image data for risk assessment of bridge piers leads to a single assessment dimension and potentially inaccurate results. By using multi-dimensional data, a more comprehensive risk assessment of bridge piers can be achieved.

Description

Technical Field

[0001] This application relates to the field of bridge risk assessment technology, and in particular to a method, apparatus, equipment and storage medium for risk assessment of bridge supports. Background Technology

[0002] Bridge piers, as load-bearing components and supporting structures of bridges, are responsible for supporting the weight of the superstructure and resisting various external forces. Their safety is directly related to the overall load-bearing capacity and traffic safety of the bridge. Once a bridge pier malfunctions, the normal use of the entire bridge will be severely affected.

[0003] In existing technologies, drone inspections are commonly used to monitor the risks of bridge structures. This involves using drones equipped with cameras to capture two-dimensional images of bridge supports, then analyzing and detecting the collected images to extract defect information and assess the safety status of the bridge supports.

[0004] However, this method of risk assessment of bridge supports that relies solely on two-dimensional image data has a limited scope and is prone to resulting in inaccurate risk assessment results for bridge supports. Summary of the Invention

[0005] This application provides a method, apparatus, equipment, and storage medium for risk assessment of bridge piers, in order to solve the problem of inaccurate risk assessment of bridge piers.

[0006] In a first aspect, embodiments of this application provide a risk assessment method for bridge piers, wherein the bridge piers comprise: multiple single piers, including:

[0007] Acquire the first point cloud data, visible light image, and positioning data corresponding to each of the single pillars;

[0008] Based on the first point cloud data, visible light image and positioning data corresponding to each single pillar, construct a three-dimensional model of the pillar corresponding to each single pillar;

[0009] Based on the three-dimensional model of each pillar, determine the defect severity factor, degradation trend factor, and detection confidence factor for each pillar.

[0010] The first risk level of each single pillar is determined based on the defect severity factor, degradation trend factor, and detection confidence factor corresponding to each single pillar.

[0011] The second risk level of the bridge support is determined based on the first risk level of each individual support.

[0012] Optionally, the step of constructing a 3D model of each pillar based on the first point cloud data, visible light image, and positioning data corresponding to each pillar includes:

[0013] Point cloud preprocessing is performed on the first point cloud data corresponding to each single pillar to obtain the second point cloud data corresponding to each single pillar;

[0014] Based on the positioning data of each single pillar, point cloud registration processing is performed on the second point cloud data corresponding to each single pillar to obtain the third point cloud data corresponding to each single pillar.

[0015] Based on the visible light image corresponding to each single pillar and the third point cloud data corresponding to each single pillar, a three-dimensional model of the pillar corresponding to each single pillar is constructed.

[0016] Optionally, determining the defect severity factor, degradation trend factor, and detection confidence factor for each individual support pillar based on its corresponding 3D model includes:

[0017] The 3D model of each single pillar is unfolded to obtain a 2D panoramic view of each single pillar.

[0018] Based on the two-dimensional panoramic view of each pillar, the defect severity factor, degradation trend factor, and detection confidence factor corresponding to each pillar are determined.

[0019] Optionally, determining the first risk level of each single pillar based on the defect severity factor, degradation trend factor, and detection confidence factor corresponding to each single pillar includes:

[0020] For any single pillar, a first risk value for the single pillar is determined based on the defect severity factor, degradation trend factor, and detection confidence factor corresponding to the single pillar.

[0021] Obtain a single-pillar risk level correspondence table, which includes: multiple first risk value ranges, and a first level corresponding to each first risk value range;

[0022] From the plurality of first risk value ranges, determine the second risk value range corresponding to the first risk value;

[0023] The first level corresponding to the second risk value range is determined as the first risk level of the single pillar.

[0024] Optionally, determining the second risk level of the bridge support based on the first risk level of each of the individual supports includes:

[0025] The first risk level of each single pillar is compared to obtain the third risk level, and the level of the third risk level is greater than the level of the other first risk levels.

[0026] Based on the third risk level, determine the baseline value of the level limit and the superimposed value of the level limit;

[0027] The first risk value corresponding to each single pillar is weighted and summed to obtain the first value.

[0028] Multiply the first value by the superimposed value of the grade limit, and then add the grade limit benchmark value to obtain the second value;

[0029] The second value is compared with the preset upper limit value to obtain the second risk value, and the second risk level is determined based on the second risk value.

[0030] Optionally, determining the second risk level based on the second risk value includes:

[0031] Obtain a bridge support risk level correspondence table, which includes: multiple third risk value ranges, and a second level corresponding to each third risk value range;

[0032] From the plurality of third risk value ranges, determine the fourth risk value range corresponding to the second risk value;

[0033] The second level corresponding to the fourth risk value range is determined as the second risk level.

[0034] Optionally, the method further includes:

[0035] When the second risk level is level one, a first alarm message is generated, which is used to instruct the immediate handling and repair of the bridge support.

[0036] When the second risk level is level two, a second alarm message is generated. The second alarm message is used to instruct the bridge support to be monitored and / or maintained according to the first preset cycle.

[0037] If the second risk level is classified as Level 3 risk, a risk assessment report will be generated.

[0038] If the second risk level is level four, a health assessment report is generated.

[0039] Secondly, embodiments of this application provide a risk assessment device for bridge supports, wherein the bridge supports include: multiple single supports, including:

[0040] The acquisition module is used to acquire the first point cloud data, visible light image and positioning data corresponding to each of the single pillars;

[0041] The construction module is used to construct a 3D model of each single pillar based on the first point cloud data, visible light image and positioning data corresponding to each single pillar.

[0042] The determination module is used to determine the defect severity factor, degradation trend factor, and detection confidence factor for each single support based on the three-dimensional model of the support corresponding to each single support.

[0043] The determining module is further configured to determine the first risk level of each single pillar based on the defect severity factor, deterioration trend factor, and detection confidence factor corresponding to each single pillar;

[0044] The determining module is further configured to determine the second risk level of the bridge support based on the first risk level of each of the single supports.

[0045] Optionally, the device further includes: a processing module;

[0046] The processing module is used to perform point cloud preprocessing on the first point cloud data corresponding to each single pillar to obtain the second point cloud data corresponding to each single pillar.

[0047] The processing module is further configured to perform point cloud registration processing on the second point cloud data corresponding to each single pillar based on the positioning data of each single pillar, so as to obtain the third point cloud data corresponding to each single pillar.

[0048] The construction module is specifically used to construct a three-dimensional model of each single pillar based on the visible light image corresponding to each single pillar and the third point cloud data corresponding to each single pillar.

[0049] Optionally, the processing module is further configured to unfold the three-dimensional model of each single pillar to obtain a two-dimensional panoramic view of each single pillar.

[0050] The determining module is specifically used to determine the defect severity factor, deterioration trend factor, and detection confidence factor for each single support pillar based on the two-dimensional panoramic image of the support pillar corresponding to each single support pillar.

[0051] Optionally, the determining module is further configured to determine a first risk value for any single pillar based on the defect severity factor, degradation trend factor, and detection confidence factor corresponding to the single pillar.

[0052] The acquisition module is further configured to acquire a single-pillar risk level correspondence table, which includes: multiple first risk value ranges, and a first level corresponding to each first risk value range;

[0053] The determining module is further configured to determine a second risk value range corresponding to the first risk value from the plurality of first risk value ranges;

[0054] The determining module is specifically used to determine the first level corresponding to the second risk value range as the first risk level of the single pillar.

[0055] Optionally, the processing module is further configured to perform a level comparison process on the first risk level of each single pillar to obtain a third risk level, wherein the level of the third risk level is greater than the level of the other first risk levels.

[0056] The determining module is further configured to determine the baseline value of the level limit and the superimposed value of the level limit based on the third risk level;

[0057] The processing module is also used to perform a weighted summation of the first risk value corresponding to each single pillar to obtain a first value;

[0058] The determining module is further configured to multiply the first value by the level limit superposition value and then add the level limit benchmark value to obtain the second value.

[0059] The determining module is specifically used to compare the second value with the preset upper limit value to obtain a second risk value, and to determine the second risk level based on the second risk value.

[0060] Optionally, the acquisition module is further configured to acquire a bridge support risk level correspondence table, the bridge support risk level correspondence table including: multiple third risk value ranges, and a second level corresponding to each third risk value range;

[0061] The determining module is further configured to determine a fourth risk value range corresponding to the second risk value from the plurality of third risk value ranges;

[0062] The determining module is specifically used to determine the second level corresponding to the fourth risk value range as the second risk level.

[0063] Optionally, the apparatus further includes: a generation module;

[0064] The generation module is used to generate a first alarm message when the second risk level is level one risk. The first alarm message is used to indicate that the bridge support should be dealt with and repaired immediately.

[0065] The generation module is further configured to generate a second alarm message when the second risk level is level two risk. The second alarm message is used to instruct the bridge support to be monitored and / or maintained according to the first preset cycle.

[0066] The generation module is also used to generate a risk assessment report when the second risk level is level three risk;

[0067] The generation module is also used to generate a health assessment report when the second risk level is level four.

[0068] Thirdly, embodiments of this application provide a risk assessment device for bridge supports, including: a memory and a processor;

[0069] The memory stores computer-executed instructions;

[0070] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0071] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0072] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0073] The bridge support risk assessment method, apparatus, equipment, and storage medium provided in this application first acquire point cloud data, visible light images, and positioning data for each of multiple individual supports. Then, a three-dimensional model of each individual support is constructed based on this data. Next, a defect severity factor, deterioration trend factor, and detection confidence factor are determined for each individual support based on the three-dimensional model. Then, a first risk level for each individual support is determined based on these factors. Finally, a second risk level for the bridge support is determined based on the first risk levels of all individual supports. This method, by acquiring multi-dimensional data and introducing defect severity factors, deterioration trend factors, and detection confidence factors, effectively solves the problem of a single assessment dimension caused by relying solely on two-dimensional image data. It achieves multi-dimensional risk assessment of bridge supports, improves the accuracy of bridge support risk assessment results, and better meets the actual needs of bridge safety inspection. Attached Figure Description

[0074] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0075] Figure 1 A schematic diagram illustrating a scenario for the risk assessment method for bridge supports provided in this application;

[0076] Figure 2 Flowchart of the risk assessment method for bridge supports provided in this application Figure 1 ;

[0077] Figure 3 Flowchart of the risk assessment method for bridge supports provided in this application Figure 2 ;

[0078] Figure 4 A schematic diagram showing the unfolded view of the three-dimensional model of the support column provided in this application, oriented towards the two-dimensional panoramic view of the support column;

[0079] Figure 5 Flowchart of the risk assessment method for bridge supports provided in this application Figure 3 ;

[0080] Figure 6 A structural schematic diagram of the risk assessment device for bridge supports provided in this application;

[0081] Figure 7 A schematic diagram of the risk assessment device for bridge supports provided in this application.

[0082] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0083] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0084] With the rapid development of urban transportation networks, bridges, as the load-bearing structures of key infrastructure, are directly related to public safety and traffic efficiency. Bridge pillars, as the core load-bearing components of the bridge structure, are subjected to multiple factors such as traffic loads, environmental erosion (such as acid rain and salt spray), temperature changes, and material aging over a long period of time, making them prone to defects such as cracks, spalling, exposed reinforcement, and corrosion.

[0085] In existing technologies, drone inspections are commonly used to monitor the risks of bridge structures. This involves using drones equipped with cameras to capture two-dimensional images of bridge supports, then analyzing and detecting the collected images to extract defect information and assess the safety status of the bridge supports.

[0086] However, this method of risk assessment of bridge supports that relies solely on two-dimensional image data has a limited scope and is prone to resulting in inaccurate risk assessment results for bridge supports.

[0087] To address the aforementioned issues, this application proposes a risk assessment method for bridge supports. Figure 1 A schematic diagram illustrating a scenario for the risk assessment method for bridge supports provided in this application. (Example) Figure 1 As shown, bridge 1 includes 11 single pillars. When the preset detection cycle is reached, the drone 2 is controlled to fly along the preset circular route corresponding to each single pillar. The drone arrives at the preset waypoints in sequence and collects point cloud data, visible light images and positioning data corresponding to each single pillar when it arrives at each waypoint. Then, the above data is analyzed and processed to obtain the risk level of each single pillar. Finally, the risk levels of all single pillars are combined to determine the overall risk level of the bridge pillars.

[0088] First, for multiple individual support pillars, the first point cloud data, visible light image, and location data for each pillar are acquired. Next, a 3D model of each pillar is constructed based on this data. Then, based on the 3D model, the defect severity factor, deterioration trend factor, and detection confidence factor for each pillar are determined. Furthermore, the first risk level for each pillar is determined based on these factors. Finally, the second risk level of all individual pillars is determined based on their respective first risk levels. This method, by acquiring multi-dimensional data and introducing defect severity, deterioration trend, and detection confidence factors, effectively solves the problem of limited assessment dimensions caused by traditional methods relying solely on two-dimensional image data. It achieves a comprehensive, multi-dimensional assessment of bridge pillar risk, thereby improving the accuracy of bridge pillar risk assessment results and better meeting the needs of bridge safety inspection.

[0089] Figure 2 A flowchart illustrating the risk assessment method for bridge supports provided in this application embodiment. Figure 1The implementing entity of this application could, for example, be a drone inspection and risk assessment system. Figure 2 As shown in this embodiment, the risk assessment method for bridge supports includes multiple single supports, including:

[0090] S101. Obtain the first point cloud data, visible light image and positioning data corresponding to each single pillar.

[0091] The first point cloud data includes multiple spatial point clouds, which can represent the three-dimensional geometry of a single pillar.

[0092] Visible light images consist of multiple pixels, which can reveal the color and texture information of a single pillar surface, such as surface defects like cracks, peeling, corrosion, and exposed reinforcement.

[0093] Location data refers to the spatial position of a single support column within the overall bridge structure.

[0094] Understandably, point cloud data can present the three-dimensional geometry of a single support column, reflecting its spatial morphology; visible light images can display the color and texture information of the column's surface; and location data can reflect the spatial position of the column within the overall bridge structure. Therefore, by acquiring the point cloud data, visible light images, and location data corresponding to each single support column, a basis for risk assessment of the overall bridge supports can be provided.

[0095] Different acquisition factors have different acquisition methods. When the acquisition factor is the first point cloud data, the acquisition method may be, for example, through a sensor device (such as LiDAR) installed on the drone. When the acquisition factor is a visible light image, the acquisition method may be, for example, through a sensor device (such as a visible light camera) installed on the drone. When the acquisition factor is positioning data, the acquisition method may be, for example, through a positioning module installed on the drone.

[0096] The timing of acquiring the first point cloud data, visible light image, and positioning data corresponding to each single support pillar can be, for example, according to a preset period or in real time. This application does not impose any special restrictions on this.

[0097] S102. Based on the first point cloud data, visible light image and positioning data corresponding to each single pillar, construct a three-dimensional model of the pillar corresponding to each single pillar.

[0098] Understandably, firstly, the first point cloud data can only present the three-dimensional geometry of a single pillar, lacking the color and texture information of the single pillar surface and the position information of the single pillar;

[0099] Secondly, visible light images can only display the color and texture information of a single pillar surface, and cannot reflect the three-dimensional geometry and position information of a single pillar.

[0100] Furthermore, the positioning data can only determine the spatial location of a single pillar, but cannot characterize its three-dimensional geometry and surface condition.

[0101] Therefore, by comprehensively considering the first point cloud data, visible light image and positioning data corresponding to a single pillar, a three-dimensional model of the pillar is constructed. This model can not only present the three-dimensional geometry of the single pillar, but also show its surface color and texture, and clearly define its position in the bridge.

[0102] Optionally, since the first point cloud data and the visible light image were acquired through different sensor devices, they cannot be directly matched in spatial coordinates, resulting in spatial inconsistency. Therefore, spatial alignment processing is also required for the first point cloud data and the visible light image, including:

[0103] The first step is to obtain calibration parameters, which include at least one of the following: the camera focal length of the visible light camera, the pixel size of the visible light camera, and the positional relationship between the LiDAR and the visible light camera.

[0104] In this context, the focal length of a visible light camera refers to the distance between the optical center of the camera lens and the photosensitive surface of the image sensor.

[0105] The pixel size of a visible light camera refers to the physical size of a single photosensitive pixel on the camera's image sensor.

[0106] The positional relationship between the lidar and the visible light camera includes, but is not limited to, the relative translational or rotational relationship between the lidar and the visible light camera.

[0107] The calibration parameters can be obtained in this step by means of, for example, from the local database of the UAV inspection and risk assessment system, from the cloud database of the UAV inspection and risk assessment system, or through user input. This application does not impose any special restrictions on this.

[0108] The second step involves spatially aligning the first point cloud data and the visible light image based on calibration parameters and positioning data for any single pillar.

[0109] The purpose of this step is to achieve spatial matching between the lidar point cloud and the visible light image.

[0110] Understandably, since the first point cloud data and the visible light image were acquired by different sensor devices and are in independent coordinate systems, they cannot be matched by relying solely on the two.

[0111] The calibration parameters can help spatially align the first point cloud data with the visible light image, and the positioning data can reflect the position of a single support column in the overall bridge structure.

[0112] Therefore, when processing data for any single pillar, the spatial alignment of the first point cloud data with the visible light image can be achieved by using calibration parameters and positioning data, with the visible light image as a reference.

[0113] S103. Based on the three-dimensional model of each support pillar, determine the defect severity factor, deterioration trend factor, and detection confidence factor for each support pillar.

[0114] The defect severity factor characterizes the overall destructive extent of surface and internal defects in a single support column. A higher defect severity factor indicates a more severe overall destructive extent of surface and internal defects in the single support column; conversely, a lower defect severity factor indicates a less severe overall destructive extent of surface and internal defects in the single support column.

[0115] The degradation trend factor can characterize the degradation rate and trend of surface and internal defects in a single support. The larger the value of the degradation trend factor, the faster the degradation rate and the more significant the degradation trend of the surface and internal defects in the single support; conversely, the smaller the value of the degradation trend factor, the slower the degradation rate and the more gradual the degradation trend of the surface and internal defects in the single support.

[0116] The detection confidence factor characterizes the reliability of the detection results for surface and internal defects of a single support column. A higher detection confidence factor indicates more reliable detection results; conversely, a lower detection confidence factor indicates less reliable detection results.

[0117] Understandably, a 3D model of a single pillar can not only present the 3D structure of the single pillar, but also show its surface color and texture, and clearly define its spatial position in the bridge.

[0118] Therefore, by considering the three-dimensional model of the pillar corresponding to a single pillar, the defect severity factor, deterioration trend factor, and detection confidence factor corresponding to the single pillar can be quantified, providing a reliable basis for subsequent assessment of the risk level of the single pillar.

[0119] S104. Determine the first risk level of each single pillar based on the defect severity factor, deterioration trend factor, and detection confidence factor corresponding to each single pillar.

[0120] The purpose of this step is to assess and determine the risk level of each individual pillar based on its defect severity factor, degradation trend factor, and detection confidence factor.

[0121] Understandably, the defect severity factor can characterize the overall destructive degree of surface and internal defects of a single pillar, which has a direct impact on structural safety; the deterioration development trend factor can characterize the deterioration development speed and trend of surface and internal defects of a single pillar, and can predict the development and changes of defects in a single pillar in the future; the detection confidence factor can characterize the reliability of the detection results of surface and internal defects of a single pillar, and can ensure the accuracy of risk assessment.

[0122] Therefore, by comprehensively considering the defect severity factor, deterioration trend factor, and detection confidence factor corresponding to a single pillar, the risk level of a single pillar can be determined more accurately.

[0123] S105. Determine the second risk level of the bridge support based on the first risk level of each individual support.

[0124] The purpose of this step is to comprehensively determine the overall risk level of the entire bridge support based on the risk levels of each individual support, thus achieving the transition from the safety status of local components to the overall structural safety assessment.

[0125] Understandably, the first risk level of a single strut only reflects the risk level of that individual strut, while the safety and stability of a bridge as a whole structure depends on the combined condition of all its individual struts. Therefore, by determining the second risk level of a bridge strut, the overall risk level of the entire bridge can be determined.

[0126] The risk assessment method for bridge supports provided in this application first acquires the first point cloud data, visible light image, and positioning data corresponding to each individual support, and constructs a three-dimensional model of each individual support based on this data. Then, based on this three-dimensional model, it determines the defect severity factor, deterioration trend factor, and detection confidence factor for the corresponding individual support, thereby determining the first risk level of each individual support. Finally, it combines the risk levels of all individual supports to determine the second risk level of the entire bridge support system. This method effectively solves the problem that relying solely on two-dimensional image data for risk assessment of bridge supports leads to a single assessment dimension and potentially inaccurate results. By using multi-dimensional data, a more comprehensive risk assessment of bridge supports can be achieved.

[0127] Figure 3 Flowchart of the risk assessment method for bridge supports provided in this application Figure 2 ,like Figure 3 As shown, in this embodiment... Figure 2 Based on the examples, the risk assessment method for bridge supports is described in detail, and the method includes:

[0128] S201. Obtain the first point cloud data, visible light image and positioning data corresponding to each single pillar.

[0129] The explanation of step S201 is the same as that in the above embodiments, and will not be repeated here.

[0130] S202. Perform point cloud preprocessing on the first point cloud data corresponding to each single pillar to obtain the second point cloud data corresponding to each single pillar.

[0131] Understandably, the raw first point cloud data contains issues such as noisy points, high data redundancy, and inconsistent coordinate systems, which would negatively impact the quality of subsequent 3D pillar models if used directly. Therefore, point cloud preprocessing can be performed on the first point cloud data corresponding to a single pillar to obtain clean, concise, and coordinate-consistent second point cloud data.

[0132] Optionally, this application provides a possible implementation method, including:

[0133] The first step is to denoise the first point cloud data of any single pillar to obtain the denoised first point cloud data.

[0134] Since the original first point cloud data usually contains noise points caused by environmental factors, equipment vibration and other factors, these noise points will affect the subsequent model construction. Therefore, the first point cloud data needs to be denoised.

[0135] Understandably, the first step is to calculate the average distance and standard deviation from each spatial point cloud to its neighboring points. Then, the average distance plus three times the standard deviation is used as a judgment threshold. Points with an average distance greater than this threshold are identified as noise points and removed. At the same time, the total number of noise points removed is controlled to not exceed the preset value of the total number of points in the current single-pillar point cloud data, such as 5%.

[0136] The second step is to perform point cloud downsampling on the first point cloud data after denoising to obtain the first point cloud data after downsampling.

[0137] In particular, since the original first point cloud data is large in volume and has a high point density, it will lead to low processing efficiency and large computational load in subsequent processing. Therefore, it is necessary to downsample the point cloud to reduce the amount of data while preserving structural features.

[0138] Understandably, the first step is to determine the 3D spatial range based on the initial point cloud data. Then, this 3D spatial range is divided according to a preset voxel grid size (e.g., 0.5cm × 0.5cm × 0.5cm), resulting in multiple voxel grids. Next, for each voxel grid, only a preset number of representative points (e.g., one) are retained. This representative point can be the average coordinates of all points within the voxel or the voxel's center point, and the remaining points within that voxel are deleted. After traversing all voxels, the remaining representative points constitute the downsampled spatial point cloud.

[0139] The third step is to perform coordinate normalization on the first point cloud data after downsampling to obtain the second point cloud data corresponding to a single pillar.

[0140] Since the point cloud data of a single pillar is collected by the UAV at different waypoints of the preset circular route corresponding to the single pillar, the coordinate systems of the spatial point clouds in the first point cloud data are not uniform and cannot be directly spliced ​​into a complete pillar model. Therefore, the point cloud data needs to be normalized.

[0141] Understandably, a local coordinate system is first established with the bottom center point of the single pillar as the origin (for example, with the X-axis as the horizontal radial direction, the Y-axis as the horizontal circumferential direction, and the Z-axis as the pillar height direction). Then, all point cloud coordinates are uniformly transformed into this local coordinate system to eliminate positional deviations between different point clouds and achieve coordinate normalization.

[0142] S203. Based on the positioning data of each single pillar, perform point cloud registration processing on the second point cloud data corresponding to each single pillar to obtain the third point cloud data corresponding to each single pillar.

[0143] Point cloud registration processing includes, but is not limited to: coarse registration, fine registration, and global registration.

[0144] The purpose of this step is to perform point cloud registration processing on the second point cloud data corresponding to a single pillar, eliminate position and attitude deviations between different point clouds, and obtain complete, unified, and misaligned third point cloud data.

[0145] In this step, for any single pillar, the first point cloud data can be registered in the order of coarse registration, fine registration, and finally global registration.

[0146] For coarse registration, for any single support pillar, firstly, the geometric feature parameters of the single support pillar are obtained, including the center of the bottom cross-section and the center of the top cross-section. Secondly, based on the positioning data and geometric feature parameters of the single support pillar, an initial transformation matrix is ​​constructed. This initial transformation matrix is ​​used to describe the rotation and translation relationships between the point clouds acquired from multiple waypoints of the single support pillar, providing a basic transformation basis for subsequent point cloud fusion. Subsequently, the initial transformation matrix is ​​used to transform and fuse the first point cloud data of the single support pillar to obtain the coarsely registered point cloud data. Throughout the fusion process, the registration error is controlled within a first preset distance (e.g., 5 cm), for example, the error is less than or equal to 5 cm, to ensure the accuracy of the initial point cloud fusion.

[0147] For fine registration, for any single pillar, the constraint parameters of the single pillar are determined. The constraint parameters include the point cloud surface normal vector and the distance between point clouds. The point cloud surface normal vector is used to constrain the orientation consistency of the point cloud surface, and the distance between point clouds is used to measure the matching degree between point clouds. Then, based on the constraint parameters of the single pillar, the initial transformation matrix obtained from coarse registration is subjected to iterative optimization processing under preset iterative conditions (such as preset iterative conditions including the number of iterations being greater than or equal to 50 and the convergence threshold being 0.001) to obtain the transformation matrix after fine registration, thereby further reducing the registration error.

[0148] For global registration, for any single pillar, a multi-view point cloud transformation matrix is ​​constructed based on the transformation matrix after fine registration of that single pillar. The multi-view point cloud transformation matrix is ​​used to record the overall pose relationship between point clouds acquired at each waypoint. Subsequently, the multi-view point cloud transformation matrix is ​​globally optimized based on a global energy function. The global energy function is used to comprehensively evaluate the overall registration error of all point clouds. By minimizing the overall registration error of all point clouds, a third point cloud data that is completely stitched together and spatially continuous is finally obtained.

[0149] S204. Based on the visible light image corresponding to each single pillar and the third point cloud data corresponding to each single pillar, construct a three-dimensional model of the pillar corresponding to each single pillar.

[0150] Understandably, visible light images can provide color and texture information about the surface of a pillar, but it is difficult to obtain a complete 3D geometry of a single pillar based solely on this image. Third-point cloud data, however, can provide the 3D geometry of a single pillar. Therefore, by combining visible light images and third-point cloud data of a single pillar, the limitations of relying solely on either image or data can be overcome, and a complete 3D model of the pillar can be generated.

[0151] Optionally, this application provides a possible implementation method, including:

[0152] The first step is to use the KD-tree algorithm to traverse the third point cloud data for any single pillar, and obtain the traversed third point cloud data.

[0153] In point cloud data processing, the KD-tree algorithm can efficiently traverse the point cloud data, identify and remove duplicate points whose distance is less than or equal to a second preset distance (e.g., the second preset distance is 0.1cm). The second preset distance is less than the first preset distance.

[0154] Understandably, the presence of duplicate spatial point clouds in the third-level point cloud data increases computational complexity, impacts model accuracy, and, since multiple sets of point clouds are discrete, they cannot be directly used for reconstruction. Therefore, it is necessary to remove duplicate points from the third-level point cloud data and integrate multiple discrete point clouds to obtain complete and clean pillar point cloud data.

[0155] The second step involves using the Poisson reconstruction algorithm to perform gridded reconstruction on the traversed third point cloud data, resulting in a triangular mesh model and the corresponding triangular facets.

[0156] Among them, the Poisson reconstruction algorithm can convert discrete point cloud data into a continuous, smooth triangular mesh model, which can restore the three-dimensional geometry of a single pillar.

[0157] The triangular mesh model is a continuous three-dimensional structure formed by reconstructing discrete spatial point clouds through meshing, which can restore the three-dimensional geometry of a single pillar.

[0158] Triangular facets are the basic units that make up a triangular mesh model. They are formed by connecting three adjacent point cloud vertices to form a triangular plane. Each triangular facet corresponds to a small area on the surface of a single pillar. All triangular facets are pieced together to form a complete three-dimensional surface structure of the pillar.

[0159] The purpose of this step is to convert discrete third-point cloud data into a continuous triangular mesh model, restoring the three-dimensional geometry of a single pillar.

[0160] Understandably, third-party point cloud data is a discrete spatial point cloud, which cannot intuitively represent the continuous surface structure of a single pillar, nor can it be used for subsequent texture mapping. By using the Poisson reconstruction algorithm to construct a triangular mesh model, the discrete spatial point cloud can be transformed into a continuous three-dimensional structure, providing a structural carrier for texture mapping and defect identification.

[0161] The third step is to construct a 3D model of the pillar based on the visible light image of the single pillar and the triangular facets of each triangular mesh model.

[0162] The purpose of this step is to add surface textures to the triangular faces of each triangular mesh model to generate a textured pillar 3D model.

[0163] Understandably, the triangular mesh model can only show the geometric shape of the support, lacking detailed information such as surface color and texture, and cannot intuitively identify surface defects.

[0164] Therefore, by using the visible light image of a single pillar to perform texture mapping on the triangular facets of each triangular mesh model, the pixel information of the visible light image can be assigned to the triangular facets, allowing the 3D model of the pillar to restore the real surface state of the single pillar.

[0165] S205. Unfold the 3D model of each single pillar to obtain a 2D panoramic view of each pillar.

[0166] The purpose of this step is to convert the three-dimensional model of the pillar into a two-dimensional planar image.

[0167] Understandably, while a 3D model can reflect the spatial form of a single pillar, directly observing and analyzing the pillar surface in 3D space is inconvenient, especially when inspecting defects, textures, or markings on the pillar surface. Therefore, the 3D model of each individual pillar can be unfolded separately to obtain a 2D panoramic view of each pillar.

[0168] For example, when a single pillar has a circular cross-section and is cylindrical in shape, this application provides a specific process for unfolding the three-dimensional model of the single pillar, such as... Figure 4 As shown, it includes:

[0169] Assume a single column has a cross-sectional radius of r and a height of H. At any point on the side of this cylindrical column... The three-dimensional parametric equations are:

[0170]

[0171] in, The inscribed angle (with the positive X-axis as the reference point) ), The coordinates are in the height direction. This represents the minimum value in the height direction of a single pillar. This represents the maximum value in the height direction of a single pillar.

[0172] Unfold the side of the cylinder along a generatrix of the Y-axis, flatten it onto a two-dimensional plane, and establish a two-dimensional coordinate system for the unfolded plane. ,in The axis corresponds to the circumferential direction. Corresponding to the height direction, origin Corresponding three-dimensional points Based on the above parametric equations, the mapping formula from the 3D model of the support pillar to the 2D panoramic image of the support pillar is:

[0173]

[0174] in, This formula can be used to map a 3D model of a support pillar into a 2D panoramic view of the pillar.

[0175] Conversely, the mapping formula from a 2D panoramic image of a support pillar to a 3D model of the support pillar is:

[0176]

[0177] S206. Based on the two-dimensional panoramic view of each support pillar, determine the defect severity factor, deterioration trend factor, and detection confidence factor for each support pillar.

[0178] Understandably, a two-dimensional panoramic image of a support pillar can present the defect distribution and morphological details of the entire surface of a single pillar. Therefore, in order to assess the safety status and risk level of a single pillar, the corresponding defect severity factor, degradation trend factor, and detection confidence factor can be determined.

[0179] Optionally, regarding the defect severity factor, this application provides a possible implementation method, including:

[0180] The first step is to obtain a two-dimensional panoramic view of any single pillar and determine the severity of cracks and associated defects based on the corresponding two-dimensional panoramic view of the single pillar.

[0181] Crack severity is used to indicate the extent of cracks in a single support column. A higher crack severity indicates more severe cracks in the column; conversely, a lower crack severity indicates less severe cracks.

[0182] The severity of associated defects is used to characterize the degree of mutual influence between different defects on the surface of a single column. A higher severity of associated defects indicates a stronger mutual influence between the defects on the surface of the single column; conversely, a lower severity of associated defects indicates a weaker mutual influence between the defects. For example, if the surface defects of a single column include exposed rebar corrosion, concrete spalling, and water seepage, a higher severity of associated defects means a stronger mutual influence between these defects.

[0183] The purpose of this step is to analyze and identify the two-dimensional panoramic image of any single pillar to obtain the severity of cracks and associated defects of that single pillar.

[0184] The second step is to perform a weighted summation of the severity of cracks and the severity of associated defects to obtain the defect severity factor.

[0185] The purpose of this step is to calculate a weighted sum of the severity of cracks and the severity of associated defects according to preset weights, so as to obtain a defect severity factor that can comprehensively reflect the overall defect status of a single support.

[0186] Optionally, the defect severity factor can be determined by the following expression.

[0187]

[0188] in, Indicates the defect severity factor; Indicates the severity of the crack; Indicates the severity of the associated defect; , where is a weighting coefficient for the severity of the crack, and is a constant; is the weighting coefficient for the severity of the associated defects, and is a constant.

[0189] Optionally, depending on the severity of the cracks, this application provides a possible implementation method, including:

[0190] The first step is to input the 2D panoramic image of any single pillar into the crack instance segmentation model to obtain the pixel-level crack instances output by the crack instance segmentation model.

[0191] Among them, pixel-level crack instances refer to the segmentation result image output by the crack instance segmentation model, which is the same size as the two-dimensional panoramic image of the pillar. Each pixel in the image is assigned a category label to indicate whether it is a crack, which can represent the location, shape and extent of the crack.

[0192] The purpose of this step is to input the two-dimensional panoramic image of the support pillar into the crack instance segmentation model to achieve pixel-level accurate identification and segmentation of the crack area.

[0193] The second step is to traverse and identify the pixel-level crack instances to obtain multiple individual cracks and the corresponding pixel-level segmentation results. Each pixel-level segmentation result contains the coordinates of the corresponding first pixel.

[0194] The pixel-level segmentation result refers to the set of pixels corresponding to a single independent crack identified from a pixel-level crack instance, which is used to distinguish different cracks.

[0195] The first pixel coordinate refers to the image coordinates of each pixel in the two-dimensional panoramic image of the support column, which is used to characterize the position of the crack in the two-dimensional panoramic image of the support column.

[0196] The purpose of this step is to traverse and identify the pixel-level crack instances obtained from the segmentation, split the overall crack into multiple independent cracks, and obtain the pixel coordinates corresponding to each crack to achieve the differentiation of individual cracks.

[0197] The third step is to perform curve fitting on the pixel-level segmentation results of multiple individual cracks to obtain the main crack curve and the corresponding second pixel coordinates.

[0198] The main crack curve refers to a smooth curve obtained by curve fitting the pixels of a single crack, used to characterize the overall extension trend and center direction of the crack. The number of main crack curves can be one or multiple.

[0199] The second pixel coordinates corresponding to the main crack curve refer to the pixel coordinates of each point on the fitted main crack curve in the two-dimensional panoramic view of the support column.

[0200] The purpose of this step is to perform curve fitting on each individual crack and extract the main crack curve and its pixel coordinates that can characterize the direction of the crack body.

[0201] The fourth step is to determine the three-dimensional point cloud coordinates of the main crack curve based on the second pixel coordinates corresponding to the main crack curve and the three-dimensional model of the single support column.

[0202] Among them, the three-dimensional point cloud coordinates of the main crack curve refer to the actual spatial coordinates of the curve in three-dimensional space after mapping the two-dimensional pixel coordinates of the main crack curve to the three-dimensional model of the support column.

[0203] The purpose of this step is to map the main crack curve on the two-dimensional plane to three-dimensional space, and combine it with the three-dimensional model of the support column to obtain the three-dimensional point cloud coordinates, thereby realizing the positioning conversion from two-dimensional to three-dimensional.

[0204] The fifth step is to determine the current length, current depth, and current width of the main crack curve based on the three-dimensional point cloud coordinates of the main crack curve.

[0205] The current length refers to the extension length of the main crack curve in three-dimensional space.

[0206] The current depth refers to the depth of the depression in the main crack curve in three-dimensional space.

[0207] The current width refers to the opening width of the main crack curve perpendicular to the extension direction.

[0208] Understandably, based on the three-dimensional point cloud coordinates of the main crack curve, the length, depth, and width of the main crack curve in the actual single-column structure are calculated, and the crack information on the image is transformed into actual physical dimensions that can be used for risk assessment.

[0209] The sixth step is to perform linear fitting on the three-dimensional point cloud coordinates of the main crack curve to obtain the crack fitting line.

[0210] Among them, the crack fitting line refers to the spatial line obtained by fitting the three-dimensional point cloud coordinates of the main crack curve through linear fitting algorithms such as the least squares method, which is used to characterize the spatial orientation and extension direction of the main crack curve.

[0211] The purpose of this step is to fit the discrete three-dimensional point cloud coordinates of the main crack curve into a straight line by performing linear fitting on the three-dimensional point cloud coordinates, thus obtaining the crack fitting line.

[0212] Step 7: Based on the 3D model of the single support column and the crack fitting line, determine the curve type, curve direction, and penetration thickness of the crack fitting line.

[0213] The types of curves include, but are not limited to: bending cracks and shear cracks. Bending cracks refer to cracks that occur in a single column under bending loads; shear cracks refer to cracks that occur in a single column under shear stress.

[0214] The curve orientation refers to the orientation and extension direction of the fitted straight line of the crack in real three-dimensional space.

[0215] Penetration thickness refers to the maximum depth to which the fitted line of the crack extends in three-dimensional space along the direction perpendicular to the cross-section of the support column.

[0216] Understandably, by using the three-dimensional model of a single support column as a reference benchmark, and combining it with the fitted crack fitting line that represents the overall extension trend of the main crack curve, the curve type and direction in three-dimensional space can be determined, as well as the penetration thickness of the curve.

[0217] Step 8: If the curve direction of the crack fitting line meets the preset curve direction, or if the penetration thickness of the crack fitting line meets the preset penetration thickness, determine the current reference value of the main crack curve as the first preset reference value.

[0218] The preset curve direction includes: the bending crack extending along the height direction of the support or the angle between the shear crack and the height direction of the support is within a preset angle range (such as 30°-60°).

[0219] The preset through-thickness can be, for example, 1 / 3 of the cross-sectional thickness of a single column.

[0220] The first preset reference value can be, for example, 1.

[0221] Understandably, when the curve direction of the crack fitting line meets the preset curve direction, or the penetration thickness of the crack fitting line meets the preset penetration thickness, it indicates that the main crack curve belongs to the stress-type crack. Therefore, the current reference value of the main crack curve is determined as the first preset reference value.

[0222] Step 9: If the curve direction of the crack fitting line does not meet the preset curve direction, and if the curve direction of the crack fitting line does not meet the preset curve direction, determine the current reference value of the main crack curve as the second preset reference value.

[0223] The second preset reference value is 0.5.

[0224] Understandably, when the curve direction of the crack fitting line does not meet the preset curve direction, and the penetration thickness of the crack fitting line does not meet the preset penetration thickness, it indicates that the main crack curve is a non-stressed crack. Therefore, the current reference value of the main crack curve can be determined as the second preset reference value.

[0225] Step 10: Determine the comprehensive crack reference value based on the current length, depth, and width of the main crack curve and the current reference value of the main crack curve.

[0226] Understandably, by combining the intuitive geometric parameters such as the current length, current depth, and current width of the main crack curve with the current reference value that reflects whether the crack is under stress, a comprehensive crack reference value can be obtained.

[0227] Optionally, this application may determine the comprehensive crack reference value using the following expression.

[0228]

[0229] in, This represents the overall crack reference value; Indicates the total number of main crack curves; This represents the curve of the i-th main crack. This indicates the maximum number of main shaft cracks allowed per pillar, which is a preset value; This indicates the maximum allowable crack length for a single support column, which is a preset value. This indicates the maximum allowable crack width for a single support column, which is a preset value. This indicates the maximum allowable crack depth for a single support column, which is a preset value. This indicates the current reference value for the main crack curve; This represents the current length of the i-th main crack curve; This represents the current width of the i-th main crack curve; This represents the current depth of the i-th main crack curve.

[0230] The eleventh step involves comparing the comprehensive crack reference value with the preset crack upper limit value to obtain the crack severity.

[0231] The severity of the crack is the minimum value between the comprehensive crack reference value and the preset crack upper limit value.

[0232] The preset crack limit value can be, for example, 1.

[0233] Understandably, by comparing the comprehensive crack reference value with the preset crack upper limit value, the severity of the crack can be determined based on the comparison result, thus enabling a quantitative assessment of the severity of cracks in a single support column.

[0234] Optionally, the severity of cracks can be determined by the following expression.

[0235]

[0236] in, Indicates the severity of the crack; This represents the overall crack reference value; This indicates the preset upper limit value for cracks, which is a preset value, such as 1.

[0237] Optionally, this application provides a method for constructing a crack instance segmentation model, including:

[0238] The first step is to acquire images of cracks in historical bridge supports to determine the training and test datasets.

[0239] The training dataset is used to train the crack instance segmentation model. The test dataset is used to evaluate the model's performance after training is complete.

[0240] Understandably, by dividing and labeling the collected images of cracks in historical bridge pillars, training and testing datasets can be separated, providing a data source for the construction, training, and validation of subsequent crack instance segmentation models.

[0241] The second step is to train the initial model based on the training model to obtain candidate segmentation models.

[0242] The initial model can be, for example, an object detection model. This model establishes a mapping between image features and target regions to locate crack targets in historical bridge support crack images and regress their bounding boxes. Based on this mapping, it achieves accurate segmentation and identification of the crack region.

[0243] The purpose of this step is to use historical bridge support crack images in the training dataset to train and optimize the parameters of the initial model, so that the initial model can learn and fit the morphology, distribution and boundary features of the bridge support cracks, and transform the original general blank model framework into a candidate segmentation model that is adapted to the bridge crack detection conditions.

[0244] Understandably, the historical bridge support crack images in the training dataset are first standardized and preprocessed, including image size unification, grayscale normalization, and noise filtering. At the same time, the crack regions in the historical bridge support crack images are labeled pixel by pixel to form crack mask labels. Finally, these are integrated to form a standardized training feature set containing image data and corresponding crack labels, ensuring that the data format fully matches the input requirements of the initial model and avoiding noise and labeling bias from affecting the training effect.

[0245] Subsequently, the functions required for model training are configured, the intersection-union ratio loss is selected as the loss function for model training, and the Adam optimizer is selected as the model optimizer. The intersection-union ratio loss is used to quantify the degree of overlap between the crack region predicted by the model and the labeled real region, and the Adam optimizer is used to optimize the model parameters in reverse based on the loss value.

[0246] Next, the standardized training feature set is input in batches into the initial object detection model. This model fits the visual features and boundary patterns of cracks based on the object detection algorithm and outputs preliminary predicted crack segmentation results. During training, the overlap between the predicted and labeled regions is calculated in real time using the intersection-union loss, and the Adam optimizer is used to adjust the model parameters to continuously improve the overlap and enhance segmentation accuracy. When the loss function converges to a preset threshold and the model segmentation accuracy reaches the standard, training stops, thus obtaining a candidate segmentation model with stable crack segmentation capabilities.

[0247] The third step is to test the candidate segmentation models based on the test dataset and obtain the test results.

[0248] The purpose of this step is to use the test dataset to comprehensively validate the performance of the trained candidate segmentation models.

[0249] Understandably, the historical bridge support crack images in the test dataset are first subjected to standardized preprocessing consistent with the training dataset. The images are sized, grayscale normalized, and noise filtered. At the same time, the crack regions in the historical bridge support crack images are labeled pixel by pixel to form crack mask labels. These are then integrated into a standardized test feature set that fully matches the input format of the candidate segmentation model, ensuring the standardization and consistency of the test data and avoiding data deviation from affecting the test results.

[0250] Subsequently, the functions required for model testing were configured, and the average intersection-over-union ratio (OCU) and average precision were selected as the core segmentation accuracy evaluation indicators, supplemented by pixel accuracy as the boundary fit evaluation indicator. The OCU is used to quantify the overall overlap between the crack region predicted by the model and the real region of the test data, the average precision is used to measure the accuracy of the model in locating the crack target, and the pixel accuracy is used to evaluate the fineness of the model's fit to the crack boundary.

[0251] Next, the preprocessed standardized test feature set is input in batches into the candidate segmentation model, which outputs predicted crack segmentation results based on the fitted crack visual features. Subsequently, segmentation accuracy and goodness of fit are calculated using preset evaluation metrics. The overlap between the predicted and real regions is calculated using the average intersection-over-union ratio (OCU), the accuracy of crack localization is measured by average accuracy, and the fineness of boundary fitting is evaluated by pixel accuracy. Finally, the evaluation results and accuracy distribution are integrated to form a complete quantitative test result, which is used to determine whether the candidate segmentation model's performance meets the preset standards, thus obtaining the test results of the candidate segmentation model.

[0252] The fourth step is to determine the candidate segmentation model as the crack instance segmentation model if the test result is passed.

[0253] The purpose of this step is to determine the candidate segmentation model as the crack instance segmentation model, provided that the test results are deemed satisfactory.

[0254] Understandably, the test result being passed reflects that the candidate segmentation model's performance metrics, such as segmentation accuracy, localization accuracy, and boundary fit, all meet the preset standards on an independent test dataset. Therefore, this candidate segmentation model can be identified as the crack instance segmentation model.

[0255] The fifth step is to retrain the candidate segmentation model if the test result is unsuccessful, until the test is successful.

[0256] Understandably, if the test result is that the test fails, the training dataset is called again to retrain the candidate segmentation model that does not meet the preset performance standard. The training process uses the previously preset loss function, optimization algorithm and training logic, and continuously adjusts the built-in parameters of the model. After each retraining, the model is tested again using the test dataset until the test result of the candidate segmentation model reaches the preset test pass standard.

[0257] Optionally, depending on the severity of the associated defect, this application provides a possible implementation method, including:

[0258] The first step is to input the two-dimensional panoramic image of any single pillar into the defect semantic segmentation model to obtain the pixel-level semantic segmentation instance of the defect output by the defect semantic segmentation model. The pixel-level semantic segmentation instance of the defect includes the pixel-level semantic segmentation result corresponding to at least one type of defect, such as exposed rebar corrosion defect, spalling defect, and water seepage defect.

[0259] Among them, exposed rebar corrosion defect refers to the phenomenon of exposed and corroded rebar on the surface of a single column.

[0260] Spalling defects refer to the phenomenon of concrete layers peeling off or breaking on the surface of a single column.

[0261] Water seepage defects refer to the phenomenon of water penetration and dampness on the surface of a single support column.

[0262] The pixel-level semantic segmentation result of exposed rebar corrosion defects refers to the pixels marked as exposed rebar corrosion defects by the defect semantic segmentation model in the two-dimensional panoramic image of the support column.

[0263] The pixel-level semantic segmentation result of the spalling defect refers to the pixel points marked as spalling defects by the defect semantic segmentation model in the two-dimensional panoramic image of the support pillar.

[0264] The pixel-level semantic segmentation result of seepage defects refers to the pixels marked as seepage defects by the defect semantic segmentation model in the two-dimensional panoramic image of the support column.

[0265] The purpose of this step is to input the two-dimensional panoramic image of the support pillar into the defect semantic segmentation model, identify and distinguish at least one type of structural defect in the two-dimensional panoramic image of the support pillar, such as exposed rebar corrosion, spalling, and water seepage, and output the pixel-level location of each type of defect.

[0266] The second step is to determine the total number of pixels of exposed rust defects based on the pixel-level semantic segmentation results of exposed rust defects, the total number of pixels of spalling defects based on the pixel-level semantic segmentation results of spalling defects, and the total number of pixels of water seepage defects based on the pixel-level semantic segmentation results of water seepage defects.

[0267] The total number of pixels representing exposed rebar corrosion defects is used to indicate the image area occupied by these defects in the 2D panoramic view of the support column.

[0268] The total number of pixels of the peeling defect is used to indicate the size of the image area occupied by the peeling defect in the 2D panoramic view of the pillar.

[0269] The total number of pixels of the seepage defect is used to indicate the size of the image area occupied by the seepage defect in the 2D panoramic view of the pillar.

[0270] The purpose of this step is to calculate the total number of pixels occupied by exposed rust defects, peeling defects, and water seepage defects, respectively, based on the pixel-level semantic segmentation results of each defect.

[0271] The third step is to determine the comprehensive defect reference value based on the total number of pixels of exposed rust defects, peeling defects, and water seepage defects.

[0272] Understandably, since different types of defects have varying degrees of impact on the safety of the support structure, by comprehensively considering the total number of pixels of exposed rebar corrosion defects, spalling defects, and water seepage defects, and combining the structural hazard weights of various defects, it is possible to achieve quantitative fusion of multiple types of defects, and finally obtain a value that can reflect the overall severity of defects in a single support.

[0273] Optionally, this application may determine the comprehensive defect reference value using the following expression.

[0274]

[0275] in, This represents a comprehensive defect reference value; The total number of pixels representing exposed steel reinforcement corrosion defects; Represents the total number of pixels with peeling defects; The total number of pixels with water seepage defects is represented; a is the severity factor of exposed rust defects, which is a preset value, such as 3.6; b is the severity factor of peeling defects, which is a preset value, such as 2.4; c is the severity factor of water seepage defects, which is a preset value, such as 1.2. This represents the pixel height of the 2D panoramic image of the pillar. This represents the pixel width of the 2D panoramic image of the pillar.

[0276] The fourth step is to compare the comprehensive defect reference value with the preset defect upper limit value to obtain the severity of the associated defects.

[0277] The severity of the associated defect is the minimum value between the comprehensive defect reference value and the preset defect upper limit value.

[0278] The preset defect limit value can be, for example, 1.

[0279] Understandably, by comparing the comprehensive defect reference value with the pre-set defect upper limit value, the minimum value can be determined from the two, and the minimum value can be determined as the defect severity.

[0280] Optionally, this application may determine the severity of associated defects using the following expressions, including:

[0281]

[0282] in, Indicates the severity of the associated defect; This represents a comprehensive defect reference value; This indicates the preset upper limit value for defects, which is a preset value, such as 1.

[0283] Optionally, for the degradation trend factor, this application provides a possible implementation method, including:

[0284] The first step is to obtain the historical change parameters of any single support pillar based on the current moment. The historical change parameters include: the historical change parameters of the main crack curve and the historical deterioration trend factor.

[0285] The historical variation parameters of the main crack curve include: the total number of historical main crack curves, the historical length, historical depth, and historical width of the main crack curve.

[0286] The second step is to determine the growth rate of the main crack curve's length, depth, and width based on the current length, depth, width, and historical change parameters of the main crack curve.

[0287] Optionally, this application can determine the length increase rate of the main crack curve using the following expression, including:

[0288]

[0289]

[0290] in, This represents the growth rate of the length of the i-th main crack curve; This represents the difference between the current time and the previous time. Indicates the curve of the i-th main crack. The length increment within the crack is determined based on the current length of the main crack curve and the corresponding historical length. Indicates the curve of the i-th main crack. The maximum allowable length increase within the range is a preset value.

[0291] Optionally, this application can determine the depth increase rate of the main crack curve using the following expression, including:

[0292]

[0293]

[0294] in, This represents the depth growth rate of the i-th main crack curve; This represents the difference between the current time and the previous time. Indicates the curve of the i-th main crack. The depth increment within the crack is determined based on the current depth of the main crack curve and the corresponding historical depth. Indicates the curve of the i-th main crack. The maximum allowable depth increase rate within the range is a preset value.

[0295] Optionally, this application can determine the width increase rate of the main crack curve using the following expression, including:

[0296]

[0297]

[0298] in, This represents the rate of increase in the width of the i-th main crack curve; This represents the difference between the current time and the previous time. Indicates the curve of the i-th main crack. The width increment within the crack is determined based on the current width of the main crack curve and the corresponding historical width. Indicates the curve of the i-th main crack. The maximum allowable width increase within the range is a preset value.

[0299] The third step involves multiplying the growth rates of the length, depth, and width of the main crack curve to obtain the three-dimensional expansion variation factor of the main crack curve.

[0300] Optionally, this application can determine the three-dimensional propagation variation factor of the main crack curve using the following expression, including:

[0301]

[0302] in, This represents the three-dimensional expansion variation factor of the i-th main crack curve; This represents the growth rate of the length of the i-th main crack curve; This represents the depth growth rate of the i-th main crack curve; This represents the rate of increase in the width of the i-th main crack curve.

[0303] The fourth step is to determine the crack size propagation rate based on the three-dimensional propagation variation factor of the main crack curve and the total number of main crack curves.

[0304] Optionally, this application can determine the crack size propagation rate using the following expression, including:

[0305]

[0306] in, Indicates the rate of crack size propagation; Indicates the total number of main crack curves; This represents the three-dimensional propagation variation factor of the main crack curve. Wherein, if ,but .

[0307] The fifth step is to determine the rate of increase in the number of cracks based on the total number of historical main crack curves and the total number of main crack curves.

[0308] Optionally, this application can determine the rate of increase in the number of cracks using the following expression, including:

[0309]

[0310]

[0311] in, This indicates the rate of increase in the number of cracks; This represents the difference between the current time and the previous time. This represents the difference between the total number of historical main crack curves and the total number of main crack curves. express The maximum allowable increase in the number of cracks is a preset value.

[0312] The sixth step is to determine reference values ​​for deterioration development based on the rate of crack size expansion and the rate of increase in the number of cracks.

[0313] Optionally, this application can determine the reference value for degradation development using the following expressions, including:

[0314]

[0315] in, This indicates a reference value for the deterioration process; Indicates the rate of crack growth; Indicates the rate of crack size propagation; This is the weighting coefficient for crack growth rate, and it is a preset value, for example, 0.2; This is a weighting coefficient for the crack size propagation rate, with a preset value, such as 0.8.

[0316] The seventh step is to compare the reference value for deterioration development with the preset upper limit value for deterioration development to obtain the current deterioration development trend factor.

[0317] Optionally, this application can determine the degradation trend factor using the following expression, including:

[0318]

[0319] in, Factors indicating the current trend of degradation; This indicates a reference value for the deterioration process; This indicates the preset upper limit of the degradation development reference value, which is a preset value, such as 1.

[0320] The eighth step is to perform a weighted summation of the current deterioration trend factor and the historical deterioration trend factor to obtain the deterioration trend factor.

[0321] Optionally, this application can determine the degradation trend factor using the following expression, including:

[0322]

[0323] in, Factors indicating a deterioration trend; This factor represents the historical deterioration trend; when calculated for the first time, it is assigned an initial value of 0.1. The weighting coefficients for factors representing historical deterioration trends are preset values, such as 0.4. Factors indicating the current trend of degradation; This represents the weighting coefficient of the factor indicating the current deterioration trend, which is a preset value, for example, 0.6. Among them, .

[0324] Optionally, this application provides a method for constructing a defect semantic segmentation model, including:

[0325] The first step is to acquire images of defects in historical bridge supports to determine the training and test datasets.

[0326] The training dataset is used to train the defect semantic segmentation model. The test dataset is used to evaluate the model's performance after training is complete.

[0327] Understandably, by dividing the collected and labeled images of defects in multiple sets of historical bridge supports, training and testing datasets can be separated, providing a data source for the subsequent construction, training, and validation of the defect semantic segmentation model.

[0328] The second step is to train the initial model based on the training model to obtain the candidate defect detection model.

[0329] The initial model could be, for example, an object detection model. This model extracts feature information from historical images of bridge support defects, locates, classifies, and segments target regions within those images, and identifies and segments defective areas of the bridge supports based on the learned features.

[0330] The purpose of this step is to use historical bridge support defect images in the training dataset to train and optimize the parameters of the initial model, so that the initial model can learn and fit the intrinsic features such as the shape, texture, and distribution of various defects on the surface of the bridge support, and transform the original general blank model framework into a candidate defect detection model that is adapted to the bridge defect detection conditions.

[0331] Understandably, the first step is to standardize and preprocess multiple sets of historical bridge support defect images in the training dataset. This involves unifying the size, normalizing the grayscale, and filtering the noise in the historical bridge support defect images. At the same time, the defect regions in the images are labeled to form defect label data. Finally, these are integrated to form a standardized training feature set containing defect images and corresponding labels. This ensures that the data format fully matches the input requirements of the initial model and avoids noise, blurring, and labeling bias from affecting the training effect.

[0332] Subsequently, the functions required for model training are configured, cross-entropy loss is selected as the loss function for model training, and Adam optimizer is selected as the model optimizer. Cross-entropy loss is used to quantify the degree of difference between the model prediction results and the labeled true results, and Adam optimizer is used to optimize the model parameters in reverse based on the loss value.

[0333] Next, the standardized training feature set is input in batches into the initial object detection model. The model learns the visual features and distribution patterns of defects based on the object detection algorithm and outputs preliminary defect detection and segmentation results. During training, the error between the predicted results and the true labels is calculated in real time using cross-entropy loss, and the Adam optimizer is used to adjust the model parameters to continuously reduce the error and improve detection accuracy. When the loss function converges to a preset threshold and the model's recognition accuracy reaches the standard, training stops, thus obtaining a candidate defect detection model with stable defect detection capabilities.

[0334] The third step is to test the candidate defect detection model based on the test dataset and obtain the test results.

[0335] The purpose of this step is to use the test dataset to comprehensively validate the performance of the trained candidate defect detection model.

[0336] Understandably, firstly, the test dataset is subjected to standardized preprocessing consistent with the training dataset for multiple sets of historical bridge support defect images. The historical bridge support defect images are subjected to size unification, grayscale normalization, and noise filtering, while retaining the corresponding defect labels. They are then integrated into a standardized test feature set that fully matches the input format of the candidate defect detection model, ensuring the standardization and consistency of the test data and avoiding data deviation from affecting the test results.

[0337] Subsequently, the functions required for model testing were configured, and mean intersection-over-union ratio (MIR), precision, and recall were selected as core evaluation metrics. MIR is used to quantify the degree of overlap between the defect regions segmented by the model and the real defect regions. Precision measures the accuracy of the defect results identified by the model, and recall measures the model's ability to detect real defects.

[0338] Next, the preprocessed standardized test feature sets are input in batches into the candidate defect detection model. The model outputs defect detection and segmentation results based on the learned feature patterns. Subsequently, the detection accuracy and segmentation effect are calculated using preset evaluation indicators. The mean intersection-over-union ratio (MIR) is used to evaluate the region overlap, and the accuracy and recall are used to comprehensively evaluate the model's recognition reliability. Finally, the evaluation results and accuracy distribution are integrated to form a complete quantitative test result, which is used to determine whether the candidate defect detection model's performance meets the preset standards, thus obtaining the test result of the candidate defect detection model.

[0339] The fourth step is to determine the candidate defect detection model as the crack instance segmentation model if the test result is passed.

[0340] The purpose of this step is to determine the candidate defect detection model as a defect semantic segmentation model, provided that the test results are deemed satisfactory.

[0341] Understandably, the test result being passed reflects that the candidate defect detection model's performance metrics, such as detection accuracy, segmentation overlap, and recognition robustness, all meet the preset standards on an independent test dataset. Therefore, this candidate defect detection model can be identified as a defect semantic segmentation model.

[0342] The fifth step is to retrain the candidate defect detection model if the test result is unsuccessful, until the test is successful.

[0343] Understandably, if the test result is that the test fails, the training dataset is called again to retrain the candidate defect detection model that does not meet the preset performance standard. The training process uses the previously preset loss function, optimization algorithm and training logic, and continuously adjusts the built-in parameters of the model. After each retraining, the model is tested again using the test dataset until the model's test result reaches the preset test pass standard.

[0344] Optionally, for detecting the confidence factor, this application provides a possible implementation method, including:

[0345] The first step is to obtain the crack segmentation accuracy of the crack instance segmentation model, the defect segmentation accuracy of the defect semantic segmentation model, and the device segmentation accuracy of the sensor equipment.

[0346] The crack segmentation accuracy of the crack instance segmentation model refers to its ability to accurately segment crack regions when identifying cracks. This accuracy is obtained through training and represents the model's accuracy in segmenting cracks on test data.

[0347] The defect segmentation accuracy of a defect semantic segmentation model refers to the degree of accuracy the model can achieve when identifying and segmenting defect regions on a single pillar or other structures. This accuracy is derived through training and represents the model's ability to accurately segment defect regions on test data.

[0348] Device segmentation accuracy of sensor equipment refers to the degree of accuracy that a sensor equipment can achieve when acquiring data. For example, the accuracy of LiDAR in acquiring point cloud data indicates the accuracy of LiDAR in acquiring data.

[0349] The second step is to determine the detection confidence factor based on the crack segmentation accuracy of the crack instance segmentation model, the defect segmentation accuracy of the defect semantic segmentation model, and the device segmentation accuracy of the sensor equipment.

[0350] Optionally, this application may determine the detection confidence factor using the following expression, including:

[0351]

[0352] in, Indicates the confidence factor for detection; The weighting coefficient corresponding to the crack separation accuracy is a preset value, such as 0.4; This indicates the crack segmentation accuracy of the crack instance segmentation model; This represents the weighting coefficient corresponding to the defect separation precision, which is a preset value, such as 0.2; This indicates the defect segmentation accuracy of the defect semantic segmentation model; This represents the weighting coefficient corresponding to the device separation accuracy, which is a preset value, such as 0.4; This indicates the device separation accuracy of the sensor equipment. Among them, .

[0353] S207. For any single pillar, determine the first risk value of the single pillar based on the defect severity factor, deterioration trend factor, and detection confidence factor corresponding to the single pillar.

[0354] The purpose of this step is to comprehensively calculate the defect severity factor, the deterioration trend factor, and the detection confidence factor to obtain a first risk value that can uniformly characterize the safety status of a single pillar.

[0355] Optionally, this application may determine the first risk value of a single pillar using the following expression, including:

[0356]

[0357] in, This represents the first risk value for a single pillar; The defect severity factor for a single pillar; Factors indicating the deterioration trend of a single pillar; This represents the confidence factor for a single-pillar detection. .

[0358] For example, suppose If the defect severity factor S=0.8, the deterioration trend factor T=0.6, and the detection confidence factor K=0.9 for a certain single pillar, then based on the above information, the first risk value of the single pillar can be determined to be 0.745.

[0359] S208. Obtain the single-pillar risk level correspondence table, which includes: multiple first risk value ranges, and the first level corresponding to each first risk value range.

[0360] The first risk value range can be divided into, for example, 0.8 or greater, 0.6 or greater and less than 0.8, 0.4 or greater and less than 0.6, and less than 0.4. The first level corresponding to each range is level one risk, level two risk, level three risk, and level four risk, with level one risk being the most severe level of risk.

[0361] The purpose of this step is to obtain a pre-defined table of single-pillar risk levels, providing a basis for mapping the first risk value to a risk level in the future.

[0362] This step can obtain the data from either the local database within the drone inspection and risk assessment system or from the cloud database within the same system. This application does not impose any specific restrictions on this.

[0363] S209. From multiple ranges of first risk values, determine the range of second risk values ​​corresponding to the first risk value.

[0364] The purpose of this step is to match the calculated first risk value to its corresponding risk value range.

[0365] S210. The first level corresponding to the second risk value range is determined as the first risk level of the single pillar.

[0366] The purpose of this step is to determine the first risk level of a single pillar based on the level corresponding to the second risk value range.

[0367] For example, suppose the first risk value range includes: greater than or equal to 0.8, greater than or equal to 0.6 and less than 0.8, greater than or equal to 0.4 and less than 0.6, and less than 0.4. The first risk levels corresponding to each range are, respectively, Level 1 risk, Level 2 risk, Level 3 risk, and Level 4 risk. Given that the first risk value of a certain single pillar is 0.745, based on the above information, we can first determine that the second risk value range corresponding to this single pillar is greater than or equal to 0.6 and less than 0.8; subsequently, we determine that the first risk level of this single pillar is Level 2 risk.

[0368] S211. Determine the second risk level of the bridge support based on the first risk level of each individual support.

[0369] Optionally, after determining the second risk level of the bridge piers, corresponding treatment methods can be generated based on different second risk levels, including:

[0370] The first method involves generating a first alarm message when the second risk level is classified as Level 1 risk. This first alarm message is used to instruct immediate action and repair of the bridge supports.

[0371] The first warning information includes, but is not limited to: risk details of each individual support (number, location, first risk value, first risk level), second risk level of the bridge support, second risk value of the bridge support, defect parameters of the individual support (such as crack length / width / depth and rate of increase), three-dimensional model of the individual support, and two-dimensional panoramic view of the individual support.

[0372] Understandably, a Level 1 risk indicates the presence of a single support pillar with Level 1 risk within the overall bridge structure, or the overlapping of multiple single supports with Level 2 or higher risk, significantly increasing the overall safety risk and placing it at a high level. If this high-risk condition is not addressed promptly, it could lead to structural damage to the bridge within a short period, resulting in irreparable losses. Therefore, a first warning message can be generated.

[0373] This step generates the first alarm message in various ways, such as by sending a text message to the relevant personnel's terminal device, by sending an email to the relevant responsible person's terminal device, or by sending an alarm message to the relevant responsible person through an application installed on the terminal device. This application does not impose any special restrictions on this.

[0374] The second method involves generating a second alarm message when the second risk level is Level 2. The second alarm message is used to instruct the bridge support to be monitored and / or maintained according to the first preset cycle.

[0375] The first preset cycle can be, for example, once a week or once a month. This application does not impose any special restrictions on this.

[0376] The second warning information includes, but is not limited to: risk details for each individual support pillar (number, location, first risk value, first risk level), second risk level of the bridge support pillar, second risk value of the bridge support pillar, defect parameters of the individual support pillar (such as crack length / width / depth and rate of increase), three-dimensional model of the individual support pillar, and two-dimensional panoramic view of the individual support pillar.

[0377] Understandably, Level 2 risk means that the highest risk level for the entire bridge support is Level 2, with no single support at Level 1 risk, but there are situations where local risks are relatively severe. In response to this situation, a second alarm message is generated, which can instruct relevant personnel to conduct monitoring and treatment of the bridge support according to the first preset cycle, and / or to carry out maintenance treatment of the bridge support according to the second preset cycle.

[0378] This step generates the second alarm message in various ways, such as by sending a text message to the relevant personnel's terminal device, by sending an email to the relevant responsible person's terminal device, or by sending an alarm message to the relevant responsible person through an application installed on the terminal device. This application does not impose any special restrictions on this.

[0379] The third approach involves generating a risk assessment report when the second risk level is classified as Level 3.

[0380] The risk assessment report includes, but is not limited to: risk details for each individual support pillar (number, location, first risk value, first risk level), second risk level of the bridge support pillar, second risk value of the bridge support pillar, handling recommendations, inspection plan, defect severity factor, deterioration trend factor, and detection confidence factor for each individual support pillar.

[0381] The risk assessment report can be in PDF or Word format, and this application does not impose any special restrictions on the format.

[0382] Understandably, a Level 3 risk rating means that the highest risk level for the entire bridge pier is Level 3. There are no individual piers with Level 2 or Level 1 risks; only minor, localized risks exist, and no emergency measures are required. Therefore, a risk assessment report can be generated to allow relevant personnel to obtain information on the risk status of the bridge piers.

[0383] The fourth scenario involves generating a health assessment report when the second risk level is level four.

[0384] The risk assessment report includes, but is not limited to: risk details for each individual support pillar (number, location, first risk value, first risk level), second risk level of the bridge support pillar, second risk value of the bridge support pillar, handling recommendations, inspection plan, defect severity factor, deterioration trend factor, and detection confidence factor for each individual support pillar.

[0385] The format of the health assessment report can be, for example, PDF or Word format; this application does not impose any special restrictions on this.

[0386] Understandably, a level four risk rating means that all bridge supports are at level four, with no medium- or high-risk supports, and the bridge as a whole is in a safe condition. Therefore, a health assessment report can be generated to allow relevant personnel to obtain information on the risk status of the bridge supports.

[0387] The risk assessment method for bridge supports provided in this application first acquires first point cloud data, visible light images, and positioning data for each individual support. Then, the first point cloud data is preprocessed to obtain second point cloud data, and the second point cloud data is registered based on the positioning data to obtain third point cloud data. Next, a three-dimensional model of each individual support is constructed based on the visible light image and the third point cloud data. The three-dimensional model is then unfolded to obtain a two-dimensional panoramic image of the support. Based on the two-dimensional panoramic image, a defect severity factor, a deterioration trend factor, and a detection confidence factor are determined for each individual support. For any individual support, the above three factors are multiplied to obtain a first risk value. A risk level correspondence table for individual supports is obtained, and the range of second risk values ​​corresponding to the first risk value is determined from the table, thereby determining the first risk level of the individual support. Finally, the second risk level of the bridge support is determined based on the first risk level of each individual support. This method, through multi-dimensional data acquisition and the introduction of three types of factors—defect severity, deterioration trend, and detection confidence—solves the problem of the single dimension in traditional two-dimensional image assessment, achieving multi-dimensional risk assessment of bridge supports and improving the accuracy of the assessment results.

[0388] Figure 5 Flowchart of the risk assessment method for bridge supports provided in this application Figure 3 ,like Figure 5 As shown, this embodiment, based on the above embodiments, provides a detailed explanation of the process for determining the second risk level of a bridge support based on the first risk level of each individual support. The method includes:

[0389] S301. The first risk level of each single pillar is compared to obtain the third risk level. The level of the third risk level is higher than the level of the other first risk levels.

[0390] Understandably, by comparing the risk levels of each individual pillar, the one with the most severe risk and the highest level is selected and defined as the third risk level, which represents the most dangerous situation in the current overall structure.

[0391] S302. Based on the third risk level, determine the baseline value of the level limit and the superimposed value of the level limit.

[0392] The baseline and superimposed threshold values ​​for different third-risk levels vary. For a level 1 risk, the baseline threshold is 0.8 and the superimposed threshold is 0.2; for a level 2 risk, the baseline threshold is 0.6 and the superimposed threshold is 0.4; for a level 3 risk, the baseline threshold is 0.4 and the superimposed threshold is 0.6; and for a level 4 risk, both the baseline and superimposed threshold values ​​are 0.

[0393] The purpose of this step is to match and determine the corresponding level limit benchmark value and level limit superposition value based on the most severe third risk level.

[0394] For example, if the third risk level is the same as the first risk level, then the baseline value for the level limit is 0.8, and the superposition value for the level limit is 0.2.

[0395] S303. Perform a weighted summation on the first risk value corresponding to each single pillar to obtain the first value.

[0396] Understandably, since the first risk value of a single pillar only represents its own risk level and cannot reflect the overall risk status of the bridge pillar structure, by weighted summing of all the first risk values, the influence of each pillar can be comprehensively considered to obtain the first value for quantifying the overall risk.

[0397] Optionally, this application may determine the first numerical value using the following expression, including:

[0398]

[0399] in, The first value is represented by n; the total number of single pillars is represented by n; and the number of the single pillar is represented by i, which indicates the i-th single pillar. This represents the weight of the i-th single pillar, using an equal-weight configuration. ; This represents the first risk value of the i-th single pillar.

[0400] For example, assuming there are 3 single pillars, and the first risk values ​​of these 3 single pillars are 0.6, 0.8, and 0.4 respectively, then based on the above information, we can determine that n is 3. for Subsequently, the first risk value corresponding to each single pillar is weighted and summed to obtain a first value of 0.6.

[0401] S304. Multiply the first value by the level limit superposition value, and then add the level limit base value to obtain the second value.

[0402] Understandably, by multiplying the first value by the superimposed level limit value and adding the level limit benchmark value, the calculation result can be made to fit the threshold range of the current overall risk level, so that the second value can reflect the actual risk level of the overall structure of the bridge support and ensure that the subsequent risk assessment results are reasonable and reliable.

[0403] Optionally, this application may determine the second numerical value using the following expressions, including:

[0404]

[0405] in, Indicates the second value. Indicates the baseline value for the grade limit; Indicates the superposition value of level limits; This represents the first numerical value.

[0406] For example, assuming the baseline value of the grade limit is 0.8, the superposition value of the grade limit is 0.2, and the first value is 0.6, then based on the above information, the second value can be obtained as 0.92.

[0407] S305. Compare the second value with the preset upper limit value to obtain the second risk value, and determine the second risk level based on the second risk value.

[0408] The preset upper limit value can be, for example, 1.

[0409] The second risk value is the minimum of the second value and the preset upper limit value.

[0410] Understandably, comparing the second value with the preset upper limit value can prevent the result from exceeding a reasonable range and ensure that the second risk value is within a valid range; then, based on the second risk value, the second risk level can be determined to determine the final safety risk level of the overall structure of the bridge support.

[0411] Optionally, this application provides a possible method for determining a second risk level based on a second risk value, including:

[0412] The first step is to obtain the bridge support risk level correspondence table, which includes: multiple third risk value ranges, and the second level corresponding to each third risk value range.

[0413] The third risk value range can be divided into, for example, 0.8 or greater, 0.6 or greater and less than 0.8, 0.4 or greater and less than 0.6, and less than 0.4. The first level corresponding to each range is, in order, level one risk, level two risk, level three risk, and level four risk, with level one risk being the most severe level of risk.

[0414] The purpose of this step is to obtain a pre-defined risk level correspondence table for bridge supports, providing a basis for subsequently mapping the second risk value to a risk level.

[0415] This step can obtain the data from either the local database within the drone inspection and risk assessment system or from the cloud database within the same system. This application does not impose any specific restrictions on this.

[0416] The second step is to determine the fourth risk value range corresponding to the second risk value from multiple third risk value ranges.

[0417] The purpose of this step is to find the actual range of the second risk value within a set range of third risk values, and to determine the corresponding range of the fourth risk value.

[0418] The third step is to determine the second level corresponding to the fourth risk value range as the second risk level.

[0419] Understandably, in the pre-established risk level correspondence table for bridge supports, the second level bound to the fourth risk value range can be directly used as the final second risk level of the overall structure of the bridge support. This can achieve the final mapping from risk range to risk level and complete the final determination of the risk level of the overall structure.

[0420] For example, suppose the third risk value range includes: greater than or equal to 0.8, greater than or equal to 0.6 and less than 0.8, greater than or equal to 0.4 and less than 0.6, and less than 0.4. The first level corresponding to each range is respectively Level 1 risk, Level 2 risk, Level 3 risk, and Level 4 risk. Given that the second risk value is 0.92, based on the above information, we can first determine that the fourth risk value range is greater than or equal to 0.8; subsequently, we determine the second risk level as Level 1 risk.

[0421] The risk assessment method for bridge supports provided in this application first compares the first risk level of each individual support to determine the highest third risk level. Next, based on the third risk level, a baseline value and a superimposed value for the risk level limit are determined. Then, the first risk values ​​corresponding to each individual support are weighted and summed to obtain a first value. The first value is then multiplied by the superimposed value for the risk level limit and added to the baseline value for the risk level limit to calculate a second value. Finally, the second value is compared with a preset upper limit value to obtain a second risk value, and the second risk level is determined accordingly. This method can highlight the impact of the highest-risk individual support on the overall bridge safety while also comprehensively considering the risk levels of all individual supports, thus enabling a comprehensive and reliable assessment of the overall risk of the bridge supports.

[0422] Figure 6This is a structural schematic diagram of the risk assessment device for bridge supports provided in this application. The bridge supports include: multiple single supports, such as... Figure 6 As shown, the risk assessment device 400 for bridge supports provided in this embodiment includes:

[0423] The acquisition module 401 is used to acquire the first point cloud data, visible light image and positioning data corresponding to each single pillar;

[0424] The construction module 402 is used to construct a 3D model of each pillar based on the first point cloud data, visible light image and positioning data corresponding to each pillar.

[0425] The determination module 403 is used to determine the defect severity factor, deterioration trend factor and detection confidence factor for each single support based on the three-dimensional model of the support corresponding to each single support.

[0426] The determination module 403 is also used to determine the first risk level of each single pillar based on the defect severity factor, deterioration trend factor and detection confidence factor corresponding to each single pillar;

[0427] The determination module 403 is also used to determine the second risk level of the bridge support based on the first risk level of each individual support.

[0428] Optionally, the device may also include: a processing module 404;

[0429] The processing module 404 is used to perform point cloud preprocessing on the first point cloud data corresponding to each single pillar to obtain the second point cloud data corresponding to each single pillar.

[0430] The processing module 404 is also used to perform point cloud registration processing on the second point cloud data corresponding to each single pillar based on the positioning data of each single pillar, so as to obtain the third point cloud data corresponding to each single pillar.

[0431] The construction module 402 is specifically used to construct a 3D model of each pillar based on the visible light image and the third point cloud data corresponding to each pillar.

[0432] Optionally, the processing module 404 is also used to unfold the 3D model of each single pillar to obtain a 2D panoramic view of each single pillar.

[0433] The determination module 403 is specifically used to determine the defect severity factor, deterioration trend factor, and detection confidence factor for each single support based on the two-dimensional panoramic image of the support corresponding to each single support.

[0434] Optionally, the determination module 403 is also used to determine the first risk value of any single pillar based on the defect severity factor, deterioration trend factor and detection confidence factor corresponding to the single pillar.

[0435] The acquisition module 401 is also used to acquire a single-pillar risk level correspondence table, which includes: multiple first risk value ranges and the first level corresponding to each first risk value range;

[0436] The determining module 403 is also used to determine the second risk value range corresponding to the first risk value from multiple first risk value ranges;

[0437] The determination module 403 is specifically used to determine the first level corresponding to the second risk value range as the first risk level of the single pillar.

[0438] Optionally, the processing module 404 is also used to perform a level comparison process on the first risk level of each single pillar to obtain a third risk level, wherein the level of the third risk level is greater than the level of other first risk levels.

[0439] The determination module 403 is also used to determine the baseline value of the level limit and the superimposed value of the level limit based on the third risk level;

[0440] The processing module 404 is also used to perform weighted summation on the first risk value corresponding to each single pillar to obtain the first value;

[0441] The determination module 403 is also used to multiply the first value by the level limit superposition value and then add the level limit base value to obtain the second value;

[0442] The determination module 403 is specifically used to compare the second value with the preset upper limit value to obtain the second risk value, and to determine the second risk level based on the second risk value.

[0443] Optionally, the acquisition module 401 is also used to acquire a bridge support risk level correspondence table, which includes: multiple third risk value ranges and the second level corresponding to each third risk value range.

[0444] The determining module 403 is also used to determine the fourth risk value range corresponding to the second risk value from multiple third risk value ranges;

[0445] The determination module 403 is specifically used to determine the second level corresponding to the fourth risk value range as the second risk level.

[0446] Optionally, the apparatus may also include: a generation module 405;

[0447] The generation module 405 is used to generate a first alarm message when the second risk level is level one risk. The first alarm message is used to indicate that the bridge support should be dealt with and repaired immediately.

[0448] The generation module 405 is also used to generate a second alarm message when the second risk level is level two. The second alarm message is used to instruct the bridge support to be monitored and / or maintained according to the first preset cycle.

[0449] The generation module 405 is also used to generate a risk assessment report when the second risk level is level three risk;

[0450] The generation module 405 is also used to generate a health assessment report when the second risk level is level four.

[0451] The risk assessment device for bridge supports provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0452] Figure 7 A structural schematic diagram of the risk assessment device for bridge supports provided in this application. Figure 7 As shown, this application provides a risk assessment device for bridge piers. The risk assessment device 500 for bridge piers includes: a receiver 501, a transmitter 502, a processor 503, and a memory 504.

[0453] Receiver 501 is used to receive instructions and data;

[0454] Transmitter 502 is used to send commands and data;

[0455] Memory 504 is used to store instructions executed by the computer;

[0456] The processor 503 is used to execute computer execution instructions stored in the memory 504 to implement the various steps of the task processing method in the above embodiments. For details, please refer to the relevant descriptions in the foregoing task processing method embodiments.

[0457] Optionally, the memory 504 can be either standalone or integrated with the processor 503.

[0458] When the memory 504 is set up independently, the electronic device also includes a bus for connecting the memory 504 and the processor 503.

[0459] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the risk assessment method for bridge supports as described above.

[0460] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned risk assessment method for bridge supports.

[0461] It will be understood by those skilled in the art that all or some of the steps, systems, or apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software may be distributed on a computer-readable medium, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0462] The technical solutions of this application have been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it is readily understood by those skilled in the art that the scope of protection of this application is obviously not limited to these specific embodiments. The above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A risk assessment method for bridge piers, characterized in that, The bridge support comprises: multiple single support columns, and the method includes: Acquire the first point cloud data, visible light image, and positioning data corresponding to each of the single pillars; Based on the first point cloud data, visible light image and positioning data corresponding to each single pillar, construct a three-dimensional model of the pillar corresponding to each single pillar; Based on the three-dimensional model of each pillar, determine the defect severity factor, degradation trend factor, and detection confidence factor for each pillar. The first risk level of each single pillar is determined based on the defect severity factor, degradation trend factor, and detection confidence factor corresponding to each single pillar. The second risk level of the bridge support is determined based on the first risk level of each individual support; The process of determining the second risk level of the bridge support based on the first risk level of each individual support includes: The first risk level of each single pillar is compared to obtain the third risk level, and the level of the third risk level is greater than the level of the other first risk levels. Based on the third risk level, a baseline value for the level limit and a superimposed value for the level limit are determined; the first risk value corresponding to each single pillar is weighted and summed to obtain a first value; wherein, the first risk value corresponding to each single pillar is determined based on the defect severity factor, deterioration trend factor, and detection confidence factor corresponding to the single pillar; The first value is multiplied by the superimposed level limit value, and then the level limit benchmark value is added to obtain the second value; the second value is compared with the preset upper limit value to obtain the second risk value, and the second risk level is determined based on the second risk value.

2. The method according to claim 1, characterized in that, The step of constructing a 3D model of each pillar based on the first point cloud data, visible light image, and positioning data corresponding to each pillar includes: Point cloud preprocessing is performed on the first point cloud data corresponding to each single pillar to obtain the second point cloud data corresponding to each single pillar; Based on the positioning data of each single pillar, point cloud registration processing is performed on the second point cloud data corresponding to each single pillar to obtain the third point cloud data corresponding to each single pillar. Based on the visible light image corresponding to each single pillar and the third point cloud data corresponding to each single pillar, a three-dimensional model of the pillar corresponding to each single pillar is constructed.

3. The method according to claim 1, characterized in that, The step of determining the defect severity factor, degradation trend factor, and detection confidence factor for each individual support pillar based on its corresponding 3D model includes: The 3D model of each single pillar is unfolded to obtain a 2D panoramic view of each single pillar. Based on the two-dimensional panoramic view of each pillar, the defect severity factor, degradation trend factor, and detection confidence factor corresponding to each pillar are determined.

4. The method according to claim 1, characterized in that, The determination of the first risk level for each single pillar based on its defect severity factor, degradation trend factor, and detection confidence factor includes: For any single pillar, a first risk value for the single pillar is determined based on the defect severity factor, degradation trend factor, and detection confidence factor corresponding to the single pillar. Obtain a single-pillar risk level correspondence table, which includes: multiple first risk value ranges, and a first level corresponding to each first risk value range; From the plurality of first risk value ranges, determine the second risk value range corresponding to the first risk value; The first level corresponding to the second risk value range is determined as the first risk level of the single pillar.

5. The method according to claim 4, characterized in that, Determining the second risk level based on the second risk value includes: Obtain a bridge support risk level correspondence table, which includes: multiple third risk value ranges, and a second level corresponding to each third risk value range; From the plurality of third risk value ranges, determine the fourth risk value range corresponding to the second risk value; The second level corresponding to the fourth risk value range is determined as the second risk level.

6. The method according to claim 1, characterized in that, The method further includes: When the second risk level is level one, a first alarm message is generated, which is used to instruct the immediate handling and repair of the bridge support. When the second risk level is level two, a second alarm message is generated. The second alarm message is used to instruct the bridge support to be monitored and / or maintained according to the first preset cycle. If the second risk level is classified as Level 3 risk, a risk assessment report will be generated. If the second risk level is level four, a health assessment report is generated.

7. A risk assessment device for bridge supports, characterized in that, The bridge supports include: multiple single supports, including: The acquisition module is used to acquire the first point cloud data, visible light image and positioning data corresponding to each of the single pillars; The construction module is used to construct a 3D model of each single pillar based on the first point cloud data, visible light image and positioning data corresponding to each single pillar. The determination module is used to determine the defect severity factor, degradation trend factor, and detection confidence factor for each single support based on the three-dimensional model of the support corresponding to each single support. The determining module is further configured to determine the first risk level of each single pillar based on the defect severity factor, deterioration trend factor, and detection confidence factor corresponding to each single pillar; The determining module is further configured to determine the second risk level of the bridge support based on the first risk level of each of the single supports; The processing module is used to perform a level comparison process on the first risk level of each single pillar to obtain a third risk level, wherein the level of the third risk level is greater than the level of the other first risk levels. The determining module is further configured to determine the baseline value of the level limit and the superimposed value of the level limit based on the third risk level; The processing module is further configured to perform a weighted summation of the first risk value corresponding to each single pillar to obtain a first value; wherein the first risk value corresponding to each single pillar is determined based on the defect severity factor, deterioration trend factor, and detection confidence factor corresponding to the single pillar; The determining module is further configured to multiply the first value by the level limit superposition value and then add the level limit benchmark value to obtain the second value. The determining module is specifically used to compare the second value with the preset upper limit value to obtain a second risk value, and to determine the second risk level based on the second risk value.

8. A risk assessment device for bridge piers, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-6.

Images