Pier underwater structure inspection system based on unmanned underwater robot

Abstract

The invention relates to the technical field of structure inspection, in particular to a bridge pier underwater structure inspection system based on an unmanned underwater robot, which comprises an inspection period adjustment module, an anomaly recognition module, an automatic supplementary shooting module, a damage evaluation module and a bridge pier anomaly evaluation module. According to the invention, bridge pier underwater structure inspection is executed through the unmanned underwater robot, the inspection period is dynamically adjusted according to the traffic flow, fine management of inspection work is realized, the inspection frequency can be adjusted according to actual use conditions and external influence factors, potential structure problems are effectively prevented and found, and the inspection efficiency is improved. The damage area can be quickly and accurately calibrated by image acquisition and automatic anomaly recognition performed by the unmanned robot, supplementary shooting is performed when necessary, the image quality is improved, the data reliability is ensured, and the influence of damage on the structural safety can be accurately evaluated by comparing the positions and overlap ratios of the damage area and the key bearing area of the bridge pier, so that the safety of the bridge pier is improved. And a scientific basis is provided for maintenance decision.

Description

Technical Field

[0001] The present invention relates to the technical field of structural inspection, and particularly to an underwater structural inspection system for bridge piers based on an unmanned underwater vehicle. Background Art

[0002] The technical field of structural inspection involves regular inspection and assessment of the integrity, safety, and durability of various structures such as building structures, transportation infrastructure, water conservancy projects, and large industrial facilities. This technical field includes multiple technical means such as crack detection, corrosion detection, settlement monitoring, and fatigue damage detection, covering methods such as ultrasonic testing, magnetic particle testing, impact echo method, eddy current testing, three-dimensional laser scanning, acoustic emission testing, underwater sonar imaging, and visual imaging testing. The technical field of structural inspection focuses on defect identification, damage degree assessment, and life prediction of structures, aiming to achieve quantitative assessment of structural states and determination of safety risks through the combination of various detection means and analysis models, providing technical support for subsequent reinforcement, repair, or replacement.

[0003] Among them, the underwater structural inspection system for bridge piers is a system used for regular inspection of the structural components of bridge piers located underwater, specifically for detecting cracks, corrosion, scouring, peeling, structural deformation, etc. of the underwater structures of bridge piers. This system can complete visual modeling of the underwater area of bridge piers, automatic defect identification, and quantitative assessment of damage degree through the combination of multiple underwater detection technologies and image processing algorithms. Its purpose is to provide accurate technical basis for maintenance decision-making, risk control, and service life prediction of the underwater part of bridges, ensuring the long-term safe and stable operation of the overall bridge structure.

[0004] Traditional inspection systems rely on manual or semi-automatic detection means, which are limited by the technical level and experience of operators and are difficult to achieve high-efficiency and high-precision detection. For example, the inspection work of bridge piers by traditional inspection systems requires divers or limited automatic equipment to be carried out at fixed intervals, which is not only time-consuming and laborious but also difficult to execute in bad weather or complex hydrological environments. In addition, the fixed-interval inspection cannot be adjusted according to actual use and environmental changes, resulting in missed inspections during high-risk periods or waste of resources during low-risk periods. This rigid inspection mode restricts the rational allocation and optimization of resources, has defects in the early damage identification and timely maintenance of key parts, increases structural safety hazards, and thus shortens the service life of bridges. Summary of the Invention

[0005] The purpose of the present invention is to solve the drawbacks existing in the prior art, and to propose an underwater structural inspection system for bridge piers based on an unmanned underwater vehicle.

[0006] To achieve the above purpose, the present invention adopts the following technical solution: An underwater structural inspection system for bridge piers based on an unmanned underwater vehicle, the system includes:

[0007] The inspection cycle adjustment module obtains the bridge traffic flow data, calculates the difference between the cumulative traffic volume in the peak period and the cumulative traffic volume in the low period, invokes the water flow shear force parameter, calculates the difference between the shear force increment value and the safety critical shear force increment threshold, adjusts the inspection cycle of the bridge pier, and generates the dynamic adjustment result of the inspection cycle;

[0008] The anomaly recognition module, based on the dynamic adjustment result of the inspection cycle, uses an unmanned underwater vehicle to inspect the underwater structure of the bridge pier, obtains the underwater inspection image, identifies the anomaly area by comparing with the known bridge pier anomaly images, and marks the anomaly area to generate the anomaly area calibration result;

[0009] The automatic reshooting module invokes the anomaly area calibration result, extracts the pixel noise index and clarity index, evaluates the shooting quality of the anomaly area. If the shooting quality is lower than the set quality threshold, it adjusts the light source brightness and exposure time, and reshoots in place to generate the local anomaly reshot image;

[0010] The damage assessment module invokes the local anomaly reshot image, extracts the damage center position, compares it with the key area of the bridge pier, calculates the coincidence degree of the damage area, and combines the damage type to calculate the damage severity, and generates the damage assessment information.

[0011] The improvement of the present invention is that the dynamic adjustment result of the inspection cycle includes the cycle adjustment coefficient, the flow difference threshold trigger state and the water flow shear force increment state. The anomaly area calibration result is specifically the anomaly boundary position parameter, the anomaly shape determination identifier and the anomaly area center point coordinates. The local anomaly reshot image includes the reshot light source brightness level, the reshot exposure time setting value and the reshot area imaging data set. The damage assessment information is specifically the damage type weight value, the sensitive area coincidence coefficient and the comprehensive influence path coefficient.

[0012] The improvement of the present invention is that the inspection cycle adjustment module includes:

[0013] The traffic flow calculation sub-module obtains the bridge traffic flow data, divides the time interval according to the peak period and the low period, calculates the difference between the cumulative traffic volume in the peak period and the cumulative traffic volume in the low period, and obtains the peak and low valley traffic difference;

[0014] The shear force increment calculation sub-module invokes the peak and low valley traffic difference, obtains the water flow shear force parameter, calculates the difference between the current shear force increment value and the safety critical shear force increment threshold according to the safety critical shear force increment threshold, and combines the preset bridge pier inspection cycle and the peak and low valley traffic difference, using the formula:

[0015]

[0016] Performs operations to obtain the cycle adjustment coefficient;

[0017] Among them, T adj represents the cycle adjustment coefficient, Q d represents the normalized value of the cumulative traffic volume in the peak period, Q p represents the normalized value of the cumulative traffic volume in the trough period, where S i represents the normalized value of the change in water flow velocity in the i-th time period, C s represents the threshold of the safety critical shear force increment, F d represents the normalized value of the current shear force increment value, F s represents the normalized value of the shear force reference benchmark value, and n represents the total number of time periods in which the change in water flow velocity participates in the calculation;

[0018] The cycle adjustment output sub-module determines the cycle adjustment direction according to the cycle adjustment coefficient, combines the preset pier inspection cycle, calculates the adjusted cycle value, adjusts the pier inspection cycle, and generates a dynamic adjustment result of the inspection cycle.

[0019] The improvement of the present invention is that the anomaly recognition module includes:

[0020] The underwater image acquisition sub-module, based on the dynamic adjustment result of the inspection cycle, schedules the unmanned underwater vehicle to perform underwater inspection of the pier structure within the specified time period, collects a continuous sequence of underwater image frames in the bottom structure area of the pier, establishes an image sequence library, and obtains the underwater image sequence of the pier;

[0021] The feature matching degree calculation sub-module calls the underwater image sequence of the pier, combines the known abnormal images of the pier, obtains the gray mean value, texture direction mean value and boundary curvature sequence of the detection area, compares them with the gray mean value, texture direction mean value and boundary curvature sequence of the reference area, and uses the formula:

[0022]

[0023] Performs operations to obtain the matching degree value of the abnormal area;

[0024] Among them, M d represents the matching degree of the abnormal area, G d represents the normalized value of the gray mean value of the detection area, G r represents the normalized value of the gray mean value of the reference area, T d represents the normalized value of the texture direction mean value of the detection area, T r represents the normalized value of the texture direction mean value of the reference area, C d,j represents the j-th boundary curvature value of the detection area, C r,j represents the j-th boundary curvature value of the reference area, and N represents the number of boundary curvature sampling points;

[0025] The area marking output sub-module compares the abnormal area matching degree value with a preset matching degree threshold, identifies the areas where the matching degree value is greater than the threshold, combines the central coordinates of the detection area, marks the positions and boundary ranges of the abnormal areas, and generates the calibration results of the abnormal areas.

[0026] The improvements of the present invention are that the automatic reshooting module includes:

[0027] The image block extraction sub-module calls the calibration results of the abnormal areas, divides pixel grids for the abnormal areas, generates a set of image blocks with a fixed size according to the pixel grid coordinates, extracts the gray matrix of the image blocks, calculates the gray variance value and the mean gradient value of the image blocks, and obtains a set of basic image parameters;

[0028] The quality evaluation sub-module compares the gray variance value with the image clarity reference value according to the set of basic image parameters, makes a difference judgment on the mean gradient value and the pixel noise reference value, combines the two judgment results to grade the quality of the image blocks, labels each image block with a quality grade label, and generates a set of image quality grade labels;

[0029] The video reshooting sub-module calls the set of image quality grade labels, screens out the image blocks labeled with low quality grades, extracts the spatial position information of the screened image blocks, adjusts the exposure time and light source brightness parameters of the underwater unmanned vehicle shooting device according to the position coordinates, performs in-situ reshooting, and generates local abnormal reshooting images.

[0030] The improvements of the present invention are that the damage assessment module includes:

[0031] The structure parameter extraction sub-module based on the local abnormal reshooting images, segments each damage area in the image, extracts the boundary lines of the damage areas, calculates the boundary curvature values, counts the pixel proportion of the damage areas in the whole image, and identifies the central position coordinates of the damage areas, and obtains a set of damage structure parameters;

[0032] The overlap degree calculation sub-module according to the set of damage structure parameters, calls the spatial boundary data of the shear force concentration area, axial pressure change area and bending resistance bearing area of the bridge pier, calculates the spatial overlap range between the damage area and the key area, calculates the ratio of the overlap area to the total area of the corresponding key area, and obtains the damage area overlap degree value;

[0033] The severity assessment sub-module calls the damage area overlap degree value and the set of damage structure parameters, extracts the boundary curvature, area proportion, spatial overlap degree of the damage area and the corresponding damage type, and uses the formula:

[0034]

[0035] Performs operations to obtain the damage severity value of the damage area and generates damage assessment information;

[0036] Among them, S zr represents the injury severity value, C zr represents the normalized value of the boundary curvature of the injury area, A zr represents the area ratio of the injury area, R zr represents the coincidence degree value between the injury area and the key area, K zr represents the coincidence degree sensitive reference value corresponding to the injury area type, D zr represents the injury type influence coefficient.

[0037] The improvement of the present invention is that the system further includes:

[0038] The pier abnormality assessment module calls the injury assessment information, screens the areas where the injury severity exceeds the set threshold, counts the injury types and distribution ranges, combines the number and central tendency of the injury areas, calculates the pier abnormality influence degree, and generates a pier abnormality assessment result;

[0039] The pier abnormality assessment result specifically refers to the injury distribution density, the total number of injuries, and the injury central tendency coefficient.

[0040] The improvement of the present invention is that the pier abnormality assessment module includes:

[0041] The high-risk area screening sub-module extracts the severity value of each injury area based on the injury assessment information, compares each area with the set severity threshold, screens the areas exceeding the threshold, extracts the injury type label and spatial range of the corresponding areas, and generates a high-severity area set;

[0042] The distribution parameter extraction sub-module calls the high-severity area set, counts the number of areas of each injury type, extracts the centroid coordinate set of each area, and calculates the density change trend value of the centroid distribution to obtain the injury distribution characteristic quantity;

[0043] The influence degree calculation sub-module calls the injury distribution parameter set, extracts the number of injuries, distribution density, and central tendency coefficient of each type, combines the screening area set, and uses the formula:

[0044]

[0045] Performs operations to obtain the pier abnormality influence degree and obtains the pier abnormality assessment result;

[0046] Among them, I b represents the pier abnormality influence degree, N as,k represents the number of the k-th type of injury area, D as,k represents the distribution density of the k-th type of injury area, T as,k represents the central tendency coefficient of the k-th type of injury area, K represents the total number of injury types, S as,kis the average severity of the k-th type of damage area.

[0047] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0048] In the present invention, an unmanned underwater vehicle is used to perform inspections on the underwater structure of the bridge pier, and the inspection cycle is dynamically adjusted according to the traffic flow, realizing the refined management of the inspection work. The inspection frequency can be adjusted according to the actual usage situation and external influencing factors, effectively preventing and detecting potential structural problems. The image acquisition and automatic anomaly recognition carried out by the unmanned robot can quickly and accurately mark the damaged area, and perform supplementary shooting when necessary to improve the image quality and ensure the reliability of the data. By comparing the position and overlap degree of the damaged area with the key load-bearing area of the bridge pier, the impact of the damage on the structural safety can be accurately evaluated, providing a scientific basis for the maintenance decision-making. The efficiency and accuracy of the inspection are improved through automated and intelligent means, reducing the uncertainty and risk of manual operations, and significantly enhancing the timeliness and safety guarantee ability of bridge maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is the system flow chart of the present invention;

[0050] Figure 2 is the flow chart of the inspection cycle adjustment module of the present invention;

[0051] Figure 3 is the flow chart of the anomaly recognition module of the present invention;

[0052] Figure 4 is the flow chart of the automatic supplementary shooting module of the present invention;

[0053] Figure 5 is the flow chart of the damage assessment module of the present invention;

[0054] Figure 6 is the flow chart of the bridge pier anomaly assessment module of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0056] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by terms such as "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0057] Please refer to Figure 1 , the present invention provides a technical solution: a pier underwater structure inspection system based on an unmanned underwater vehicle, the system includes:

[0058] The inspection cycle adjustment module obtains the bridge traffic flow data, accumulates it segment by segment according to the peak period and the trough period, calculates the difference between the cumulative traffic volume in the peak period and the cumulative traffic volume in the trough period, invokes the water flow shear force parameter, calculates the difference value between the shear force increment value and the safety critical shear force increment threshold, combines the preset pier inspection cycle, calculates the pier inspection cycle adjustment value, adjusts the pier inspection cycle, and generates the inspection cycle dynamic adjustment result;

[0059] The anomaly recognition module, based on the inspection cycle dynamic adjustment result, uses an unmanned underwater vehicle to inspect the underwater structure of the pier, obtains the underwater inspection image, calculates the matching degree of the anomaly area by comparing with the known pier anomaly images according to the characteristics of the area, identifies the anomaly area, and marks the anomaly area to generate the anomaly area calibration result;

[0060] The automatic reshooting module invokes the anomaly area calibration result, extracts the local area image block, extracts the pixel noise index and the clarity index, evaluates the shooting quality of the anomaly area. If the shooting quality is lower than the set quality threshold, it adjusts the light source brightness and exposure time and reshoots in situ to generate the local anomaly reshooting image;

[0061] The damage assessment module invokes the local anomaly reshooting image, extracts the boundary curvature and area ratio of the damage area, and extracts the damage center position, compares it with the key areas of the pier. The key areas of the pier include the shear force concentration area, the axial pressure change area and the bending resistance bearing area, calculates the damage area coincidence degree, combines the damage type, calculates the damage severity, and generates the damage assessment information;

[0062] The pier anomaly assessment module invokes the damage assessment information, screens the areas where the damage severity exceeds the set threshold, counts the damage type and distribution range, combines the number and central tendency of the damage areas, calculates the pier anomaly influence degree, and generates the pier anomaly assessment result.

[0063] The dynamic adjustment results of the inspection cycle include the cycle adjustment coefficient, the triggering status of the flow difference threshold, and the increment status of the water flow shear force. The abnormal area calibration results are specifically the abnormal boundary position parameters, the abnormal shape determination identifier, and the central point coordinates of the abnormal area. The local abnormal supplementary images include the brightness level of the supplementary light source, the set value of the supplementary exposure time, and the imaging data set of the supplementary area. The damage assessment information is specifically the damage type weight value, the coincidence coefficient of the sensitive area, and the comprehensive influence path coefficient. The abnormal assessment results of the bridge pier specifically refer to the damage distribution density, the total number of damages, and the damage concentration trend coefficient.

[0064] Please refer to Figure 2 , the inspection cycle adjustment module includes:

[0065] The traffic flow calculation sub-module obtains the bridge traffic flow data, divides the time interval according to the peak period and the trough period, calculates the difference between the cumulative traffic volume in the peak period and the cumulative traffic volume in the trough period, and obtains the peak-trough traffic difference;

[0066] The traffic flow calculation sub-module obtains the bridge traffic flow data and needs to call the vehicle counting equipment deployed at the bridge deck entrances and exits in the bridge traffic monitoring system. This equipment generates a set of vehicle passing data every 5 minutes. The acquisition object is the motor vehicles passing through the bridge deck. The recorded content includes the timestamp, vehicle type, passing direction, and vehicle speed parameters. The daily traffic volume data is summarized through the data upload module. During execution, the 24 hours of each day need to be divided into several time intervals first. The peak period is set as 7:00–9:00 and 17:00–19:00, and the trough period is set as 1:00–5:00. Calculate the cumulative vehicle traffic volume in the corresponding time periods respectively. For each peak period or trough period, its cumulative traffic volume is obtained through the following operations: sum up the number of vehicles passing through every 5 minutes in this time period. Taking the time period 7:00–9:00 as an example, if the average traffic volume per 5 minutes is 42 vehicles, then the cumulative traffic volume in this peak period is (2 hours × 60 / 5) × 42 = 1008 vehicles. The trough period is set as 1:00–5:00, that is, 4 hours, a total of 48 5-minute units. If the average traffic volume is 12 vehicles, then the cumulative traffic volume in the trough period is 48 × 12 = 576 vehicles. Subsequently, normalize the above cumulative traffic volumes in the peak period and the trough period. The normalization reference value is set as the maximum designed traffic capacity of the bridge, specifically set as the upper limit of the traffic capacity of this bridge per hour is 800 vehicles, determined based on the number of one-way traffic lanes, lane width, and speed limit standard stated in the traffic design drawings, and obtained by simulating the actual traffic flow density. This value is cyclically adjusted and updated according to the traffic volume statistical value during holidays and the highest traffic peak value over the years. The currently set daily maximum single-period traffic capacity is 800 × 2 = 1600 vehicles. Based on this, calculate the normalized value in the peak period as Q d = 1008 / 1600 = 0.63, and the normalized value in the trough period is Q p= 576 / 1600 = 0.36, and then the normalized value difference between the two is calculated as the peak-to-trough traffic difference, i.e., Q d -Q p = 0.63 - 0.36 = 0.27. The above difference reflects the fluctuation range of traffic flow at different times, and can show the difference in traffic pressure on the bridge at different times through numerical values, and finally serves as one of the inputs for the subsequent calculation of shear force increment.

[0067] The shear force increment calculation sub-module calls the peak-to-trough traffic difference, obtains the water flow shear force parameters, calculates the difference between the current shear force increment value and the safety critical shear force increment threshold according to the safety critical shear force increment threshold, combines the preset pier inspection period and the peak-to-trough traffic difference, and uses the formula:

[0068]

[0069] Performs operations to obtain the period adjustment coefficient;

[0070] Among them, T adj represents the period adjustment coefficient, Q d represents the normalized value of the cumulative traffic volume in the peak period, Q p represents the normalized value of the cumulative traffic volume in the trough period, where S i represents the normalized value of the change in water flow velocity in the i-th time period, C s represents the safety critical shear force increment threshold, F d represents the normalized value of the current shear force increment value, F s represents the normalized value of the shear force reference benchmark value, and n represents the total number of time periods in which the change in water flow velocity participates in the calculation;

[0071] The shear force increment calculation sub-module calls the previously obtained peak-to-trough traffic difference of 0.27, and at the same time obtains the water flow velocity data from the flow velocity monitoring equipment deployed in the waters under the bridge. The equipment records the water flow velocity with a sampling period of 15 minutes. Suppose n = 4 in a certain time period, that is, the data selected for 1 hour are 1.2 m / s, 1.5 m / s, 1.7 m / s, and 2.1 m / s respectively. Based on these data, the change in water flow velocity in adjacent time periods is calculated as 0.3, 0.2, and 0.4 m / s respectively. These differences are normalized, and the normalization reference value is set as the historical maximum flow velocity change range of 0.5 m / s in the monitoring area. This value is obtained from the upper limit of the 95% confidence interval of the daily maximum change statistics in the cross-section of the water area where the pier foundation of this bridge is located in the past 5 years, and is used as the limit representative value of the change in hydrological disturbance intensity. This value has periodic fluctuations in the flood season and the dry season. Currently, 0.5 m / s is selected as the static parameter for participating in the normalization process, and the corresponding normalization results are S1 = 0.3 / 0.5 = 0.6, S2 = 0.2 / 0.5 = 0.4, S3 = 0.4 / 0.5 = 0.8, and the summation term is calculated Subsequently, obtain the safety critical shear force increment threshold C s , which is the maximum allowable shear force growth limit set by the bridge structure design unit. It is derived from the critical shear stress limit provided in the structural fatigue calculation and obtained through the combination of material test data and finite element stress simulation analysis. The currently selected value is 0.55, dimensionless after unit normalization. The value is set based on the equivalent safety stress increment limit of the pier concrete-wrapped steel structure at a temperature of 35°C and a water flow velocity of 2.5 m / s. The adjustment period is updated annually through evaluation. The current shear force increment F d can be obtained by converting the flow velocity data through an empirical relationship. Let the currently converted value be 0.47, and the reference benchmark value F s = 0.30. This benchmark value represents the average shear force increment acting on the pier surface under normal water flow conditions, measured under the basic working condition where the flow velocity is stable at 1.2 m / s, which is the standard average value during the low water period in spring over the years and is not adjusted dynamically with the seasons, but is corrected every three years according to the water flow evolution trend. After normalization, it is also substituted. Calculate the difference between the current shear force increment value and the reference shear force as |F d - F s | = |0.47 - 0.30| = 0.17. Substitute the above data into the formula:

[0072]

[0073] Obtain the period adjustment coefficient of 0.675. This value reflects the combined influence degree of the bridge traffic pressure and water flow impact on the pier stability, providing a quantitative basis for the subsequent period adjustment sub-module. The advantage of the formula is that by calculating the product of the sum of the peak-to-valley traffic difference and the water flow velocity change amount, and combining the safety critical shear force increment and the current shear force change difference value as the denominator, the adjustment coefficient can simultaneously reflect the coupling degree of traffic load change and fluid shear influence.

[0074] The period adjustment output sub-module determines the period adjustment direction according to the period adjustment coefficient and in combination with the preset pier inspection period, calculates the adjusted period value, adjusts the pier inspection period, and generates the dynamic adjustment result of the inspection period;

[0075] The period adjustment output sub-module receives the period adjustment coefficient T adj= 0.675, combined with the preset inspection cycle of the bridge. Assuming the original cycle is 30 days, by comparing the cycle adjustment coefficient with the reference interval, the reference interval is set to [0.3, 0.5], which is determined by the average standard deviation range of the historical five-year cycle adjustment data of this bridge. A value lower than 0.3 indicates a stable structural state, and the upper limit of 0.5 indicates the boundary where shear risk appears. The current value of 0.675 exceeds the upper limit of 0.5, meaning the inspection demand has increased. If the cycle adjustment coefficient is greater than 0.5, it is determined that the cycle is shortening. Calculate the adjusted cycle according to the adjustment ratio. Let the adjustment ratio function be (1 - T adj ), then the adjustment coefficient is 1 - 0.675 = 0.325. Therefore, the adjusted cycle is 30×0.325 = 9.75 days, which is rounded up to 10 days, and the result is output to the inspection schedule system to generate the dynamic adjustment result of the inspection cycle. The adjustment ratio function is the direct expression of the cycle shortening rate, and it reflects the correction intensity to the original cycle by subtracting the comprehensive shear risk index from 1. The value range of this coefficient is (0, 1). The current value of 0.325 indicates the degree of shortening the original cycle by 32.5%, ensuring that the cycle adjustment operation has sufficient response strength. This result shows that the current state of the bridge pier requires an increased detection frequency. The adjusted inspection cycle value of 10 days replaces the original cycle of 30 days for dynamically updating the maintenance plan.

[0076] Please refer to Figure 3 , the anomaly recognition module includes:

[0077] Based on the dynamic adjustment result of the inspection cycle, the underwater image acquisition sub-module schedules an unmanned underwater vehicle to conduct an inspection of the underwater structure of the bridge pier within a specified time period, collects a continuous sequence of underwater image frames in the bottom structure area of the bridge pier, establishes an image sequence library, and obtains the underwater image sequence of the bridge pier;

[0078] Based on the inspection cycle adjustment result output by the aforementioned cycle adjustment sub-module, after receiving the adjusted inspection cycle value of 10 days, the underwater image acquisition sub-module generates a corresponding task scheduling instruction through the bridge maintenance scheduling system to control the unmanned underwater vehicle to enter the working state within the 10th day. The specific operation time window is set during 9:00 to 11:00 in the morning when the water flow is relatively stable. The scheduling control instruction consists of a task serial number, an inspection area number, a collection time period, and preset path points. After receiving the instruction, the robot moves to the bottom area of the specified pier according to the path points, and cruises at a constant speed of 0.5 m / s within a radius of 5 meters to take pictures. The frequency of collecting underwater image frames is set to 25 frames per second, and the continuous collection duration is 20 minutes, obtaining a total of 20×60×25 = 30000 image frames. The image frames are numbered and organized into an image sequence library according to the time series. Each frame of the image is attached with the current water depth, attitude angle, and shooting timestamp. The image format is uniformly transcoded into grayscale images for subsequent processing. The image resolution is set to 1920×1080, and the spatial resolution is set to 1 pixel corresponding to an actual size of 2 millimeters. After the images are denoised, sharpened, and equalized, they enter the subsequent feature calculation link, and finally the acquisition of the underwater image sequence of the pier is completed.

[0079] The feature matching degree calculation sub-module calls the underwater image sequence of the pier, combines it with the known abnormal images of the pier, obtains the gray mean value, texture direction mean value, and boundary curvature sequence of the detection area, and compares them with the gray mean value, texture direction mean value, and boundary curvature sequence of the reference area. The formula is used:

[0080]

[0081] Calculate the matching degree value of the abnormal area through operation;

[0082] Among them, M d represents the matching degree of the abnormal area, G d represents the normalized value of the gray mean value of the detection area, G r represents the normalized value of the gray mean value of the reference area, T d represents the normalized value of the texture direction mean value of the detection area, T r represents the normalized value of the texture direction mean value of the reference area, C d,j represents the jth boundary curvature value of the detection area, C r,j represents the jth boundary curvature value of the reference area, and N represents the number of boundary curvature sampling points;

[0083] The feature matching degree calculation sub-module calls the underwater image data in the above image sequence library, combines it with the known abnormal area image of the pier as the reference image, selects the boundary range of the area to be detected and the reference area to be the same and the shooting angle of the image frames to be similar, and uniformly crops the image block size to 256×256 pixels. First, calculate the gray mean value G of the image block in the detection aread Perform an operation by summing the grayscale values of all pixels within the image block and then dividing by the total number of pixels. Assume the range of pixel grayscale values is from 0 to 255. If the total sum of grayscale values in a certain image block is 3276800, then the grayscale mean is G d = 3276800 / (256×256) = 50.0, corresponding to the grayscale mean G of the reference region r is 45.0. For the grayscale mean normalization process, the maximum grayscale value of 255 is used as the benchmark, that is, after normalization, G d = 0.196, G r = 0.176, the texture direction mean T d 、T r respectively represent the statistical values of the gradient directions in the image. The calculation method is to perform Sobel gradient calculation on each pixel neighborhood to obtain the direction angle, and then calculate the mean within the region. Assume T d = 62.5°, T r = 60.0°. After conversion to radians, they are 1.091 and 1.047 respectively. The normalization benchmark is set to π, and the normalized values are T d = 1.091 / π ≈ 0.347, T r = 1.047 / π ≈ 0.333, the boundary curvature sequence C d,j 、C r,j respectively represent the curvature sampling points on the boundaries of the detection region and the reference region. The curvature calculation is based on the tangent change rate of the boundary contour points. Assume the number of sampling points N = 5, and the curvature values of the five detection points are [0.18, 0.20, 0.22, 0.24, 0.26], and the reference region is [0.15, 0.19, 0.21, 0.23, 0.25]. The absolute value of the difference for each item is [0.03, 0.01, 0.01, 0.01, 0.01]. After summation, it is 0.07, and the average difference is 0.07 / 5 = 0.014. Substitute into the formula

[0084]

[0085] to obtain a matching degree value of 0.9745. This value is the final operation output result. The closer the value is to 1, the higher the similarity degree between the detection region and the reference abnormal region. The current value indicates that the detection image and the reference abnormal image have a relatively high similarity. The normalization benchmark value of the grayscale mean in the participating items is set to 255, which is defined based on the maximum value of the 8-bit encoding of the image grayscale. This value is a fixed value and does not change with the scene; the texture direction normalization is capped at π radians, which comes from the physical definition range of the direction angle; the boundary curvature is not normalized, and the direct sampling values are used for participation in the matching. N is the curvature sampling quantity, and this value is set to 5, which is jointly determined by the boundary smoothness and the contour point density. If the complexity of the image boundary increases, the N value should be increased to 10 to enhance the sensitivity to contour differences.

[0086] The region marking output sub-module compares the abnormal region matching degree value with a preset matching degree threshold, identifies the regions where the matching degree value is greater than the threshold, combines the central coordinates of the detection region, marks the position and boundary range of the abnormal region, and generates the abnormal region calibration result;

[0087] The region marking output sub-module calls the aforementioned abnormal region matching degree value of 0.9745 and compares it with the preset matching degree threshold. This threshold is set to 0.90 and is derived from the fitting analysis process of manual sample annotation and image comparison results. After analyzing the distribution of the matching degree values in the comparison results of 100 groups of abnormal images and normal images, the matching degree judgment threshold is set with the 95% quantile as the lower bound. If the matching degree value is higher than this value, it indicates that the detection region has structural feature similarity with the reference abnormal region. The current matching degree value of 0.9745 > 0.90, so it is determined that this region is an abnormal region. Subsequently, the central point coordinates of the detection region in the image frame are read, and a region with a radius of 32 pixels is set as the boundary in pixels in the image, corresponding to an actual distance of 32×2mm = 64mm. Using this coordinate and the bounding box parameters, it is marked in the image sequence, and the position information and boundary range of the abnormal region are marked for this frame of image, and the abnormal region calibration result is generated. This result will be used as the original input data in subsequent image recognition statistics and maintenance suggestion generation.

[0088] Please refer to Figure 4 , the automatic reshooting module includes:

[0089] The image block extraction sub-module calls the abnormal region calibration result, divides the abnormal region into pixel grids, generates a set of image blocks with a fixed size according to the pixel grid coordinates, extracts the gray matrix of the image blocks, calculates the gray variance value and mean gradient value of the image blocks, and obtains the basic parameter group of the image;

[0090] The image block extraction sub-module calls the aforementioned abnormal region calibration result. First, it obtains the central point coordinates and boundary pixel coordinate values of the abnormal region in the image frame marked as abnormal, divides the abnormal region into equally spaced pixel grids according to the pixel coordinates, with each grid side length set to 64 pixels, generates multiple grid index regions, and then extracts the image pixel data within each grid region to obtain fixed-size image blocks, with the size uniformly 64×64 pixels. The image blocks are read in grayscale, and the pixel value range is 0–255. The corresponding gray matrix data is extracted for each image block, and the gray matrix is defined as a two-dimensional array G i,j , where i and j are the row and column indices in the image block respectively. For example, the random 5-row data in the gray matrix of a certain image block is as follows:

[0091] [[52, 53, 55, 58, 60]], [[51, 54, 57, 59, 61]], [[50, 52, 54, 56, 58]], [[48, 50, 52, 53, 55]], [[47, 48, 50, 51, 53]]. The method for calculating the grayscale variance value is to first calculate the grayscale mean value G of the image block avg , assuming that the sum of all pixel values is 208896 and the number of pixels is 64×64 = 4096, then there is G avg = 208896 / 4096 = 51.0. The grayscale variance value is the average of the squares of the differences between all pixel values and the mean value, that is Assuming that the obtained value is 36.0, the mean gradient value is calculated by taking the absolute value of the grayscale difference between adjacent pixels in the image block. Both horizontal and vertical directions are involved in the summation, and then divided by the number of participating items. Assuming that the final mean gradient value is 8.4, the image basic parameter group consists of the grayscale variance value 36.0 and the mean gradient value 8.4.

[0092] The quality assessment sub-module compares the grayscale variance value with the image clarity reference value according to the image basic parameter group, judges the difference between the mean gradient value and the pixel noise reference value, combines the two judgment results to grade the quality of the image block, labels the quality grade label for each image block, and generates an image quality grade label group;

[0093] The quality assessment sub-module receives the image basic parameter group and quantitatively grades the quality of the image block. First, it compares the grayscale variance value with the image clarity reference value. The image clarity reference value is set to 25.0, which is set according to the median of the grayscale variance statistics of the images with distinguishable contours by the human eye collected by the robot platform under the conditions of a water depth of 5 meters and a water flow velocity of 1.5 m / s. This value is stable under the conditions of a fixed resolution (1920×1080) of the image acquisition device and a fixed lens focal length (6 mm). If different devices are used, this value needs to be reset. The current grayscale variance value of the image block is 36.0, which is higher than the reference value and meets the clarity requirements. Then, it judges the difference between the mean gradient value and the pixel noise reference value. The pixel noise reference value is set to 5.0, which is the mean value of the random offset of the pixels measured from the standard static image obtained when the light source of the shooting device is turned off, reflecting the degree of local brightness fluctuation in the image. Under the parameters of the current device with an exposure time of 1 / 50 second and an ISO sensitivity of 800, the measured value is stable at 5.0. The current mean gradient value is 8.4, which is higher than the reference value, indicating that the density of the image edge information exceeds the noise influence level. Therefore, the quality level of this image block can be judged as "high", and the quality grade label is set to "A". If any parameter does not reach the reference value, it is marked as "B" or "C" level. According to the rule: those with a grayscale variance value lower than 25.0 and a mean gradient value lower than 5.0 are "C", and the rest are "B". Finally, an image quality grade label group is generated, which consists of each image block and its corresponding quality grade label.

[0094] The image supplementary shooting module calls the image quality grading tag group, screens and marks the image blocks of low quality level, extracts the spatial position information of the screened image blocks, adjusts the exposure time and light source brightness parameters of the underwater unmanned vehicle shooting device according to the position coordinates, performs in-situ supplementary shooting, and generates local abnormal supplementary shooting images;

[0095] The image supplementary shooting module calls the aforementioned image quality grading tag group, screens and marks the image blocks marked as "C" level from it. The screening action is to traverse all the data of the image block tag group and take out the entries with the corresponding grade field value of "C". Extract the upper left corner coordinate position and the image block size parameter of the image block in each record, calculate the center point coordinate of the image block. For example, if the upper left corner is (640, 1280) and the size is 64×64 pixels, then the center point coordinate is (672, 1312). Through the mapping relationship between the coordinate system of the underwater unmanned vehicle body and the image coordinates, the image center point coordinate is converted into the physical space coordinate. Suppose the transformation coefficient is 1 pixel corresponding to 2 millimeters, then the above coordinates correspond to the actual position of (1.344m, 2.624m). Adjust the parameters of the robot shooting device according to this physical position. Specifically, the exposure time is extended from 1 / 50 second to 1 / 30 second, the light source brightness is increased from the original setting of 800 lumens to 1200 lumens, and the light source color temperature remains unchanged at 5000K. After the adjustment is completed, the robot fixes the current shooting position and re-performs image acquisition. The size of the supplementary shooting image block is the same as the original one Figure 1 consistent, and the naming maintains a one-to-one correspondence with the original image block. Finally, the content of the supplementary shooting image is generated, which is used as the local abnormal supplementary shooting image for subsequent replacement and reconstruction of the image sequence.

[0096] Please refer to Figure 5 , the damage assessment module includes:

[0097] The structure parameter extraction sub-module is based on the local abnormal supplementary shooting image, segments each damage area in the image, extracts the boundary line of the damage area, calculates the boundary curvature value, counts the pixel proportion of the damage area in the whole image, and identifies the center position coordinate of the damage area to obtain the damage structure parameter set;

[0098] The structural parameter extraction sub-module calls the local abnormal supplementary shooting images. After obtaining the calibrated boundaries of the abnormal areas in each frame of the image, it performs image segmentation operations on each abnormal area, extracts the boundary contour line segments. The contour extraction is based on the calculation of the image gray gradient and the determination of the connectivity of edge pixels, obtaining a set of contour points of a closed or semi-closed boundary. Suppose the number of boundary points extracted from a certain abnormal area is 120. A continuous point sequence is constructed according to the sequential distribution of the contour points in the image coordinate system. The curvature is calculated for any three consecutive points in this point sequence. The curvature calculation method used is the reciprocal of the radius of the circumscribed circle formed by three points. Suppose the curvature values of five representative points are [0.15, 0.20, 0.23, 0.18, 0.21]. They are normalized, and the normalization reference value is set to 0.25. This value is the maximum historical value of the curvature of the conventional crack boundary on the surface of the pier concrete, which is the upper limit of the 95% interval obtained from the statistical curvature manually marked in the underwater image data in the past three years. The normalized curvature values are [0.6, 0.8, 0.92, 0.72, 0.84] respectively. The average value is taken to get C zr = 0.776. Subsequently, the area ratio of the abnormal area is calculated. The total number of pixels extracted from this area is set to 3200 pixels. The resolution of the entire image is 1920×1080, and the total number of pixels is 2,073,600. Then the area ratio is A zr = 3200 /

[0099] 2,073,600 ≈ 0.00154. Then, the geometric center of the boundary point set of this area is extracted as the center point coordinates. For example, the horizontal pixel range is (640, 704), and the vertical range is (1280, 1344). Then the center coordinates are (672, 1312). Finally, a damage structure parameter set is formed, including the normalized value C zr = 0.776, the area ratio A zr = 0.00154, and the center position coordinates (672, 1312).

[0100] The coincidence degree calculation sub-module, according to the damage structure parameter set, calls the spatial boundary data of the shear force concentration area, axial pressure change area, and flexural bearing area of the pier, calculates the spatial overlap range between the damaged area and the key area, and calculates the ratio of the overlapping area to the total area of the corresponding key area to obtain the coincidence degree value of the damaged area;

[0101] The coincidence degree calculation sub-module calls the central coordinates and boundary range information in the above-mentioned damaged structure parameter set, reads the boundary data of the defined key areas in the pier structure model, including the two-dimensional image mapping bounding boxes of the shear concentration area, the axial pressure change area, and the flexural bearing area. Assume that the boundary of the shear concentration area in the current image coordinate system is a rectangular area [(600, 1200) – (800, 1400)], with an area of 400×200 = 80,000 pixels, and the current damaged area boundary range is [(640, 1280) –

[0102] (704, 1344)], with an area of 64×64 = 4,096 pixels. The overlapping area coordinate range of the two is [(640, 1280) –

[0103] (704, 1344)], with an area of 64×64 = 4,096 pixels. Calculate the ratio of the overlapping area to the key area area as R zr = 4,096 / 80,000 = 0.0512, that is, the coincidence degree is 5.12%. If the current damaged area spans multiple key areas, calculate the coincidence ratio with each area separately and extract the maximum value as the final coincidence degree value. This value represents the spatial influence degree of the current abnormal area on the key structure area.

[0104] The severity assessment sub-module calls the coincidence degree value of the damaged area and the damaged structure parameter set, extracts the boundary curvature, area ratio, spatial coincidence degree of the damaged area, and the corresponding damage type, and uses the formula:

[0105]

[0106] Perform operations to obtain the damage severity value of the damaged area and generate damage assessment information;

[0107] Among them, S zr represents the damage severity value, C zr represents the normalized value of the boundary curvature of the damaged area, A zr represents the area ratio of the damaged area, R zr represents the coincidence degree value of the damaged area and the key area, K zr represents the coincidence degree sensitive reference value corresponding to the damaged area type, D zr represents the damage type influence coefficient;

[0108] The severity assessment sub-module calls the aforementioned damaged structure parameter set and the coincidence degree value of the damaged area, extracts the parameter C zr = 0.776, A zr = 0.00154, R zr = 0.0512, and calls the damage type influence coefficient D zr, assume that the currently recognized damage type is "longitudinal crack", and the influence coefficient of this type is set to 1.8. The basis for setting the influence coefficient is that in structural experiments, longitudinal cracks have a more serious impact on the bearing capacity of axially compressed members than transverse cracks. It is obtained by fitting the damage grade distribution and the critical value of structural instability. 1.8 is the standard value of the influence strength of longitudinal cracks. Then call the coincidence sensitivity reference value K corresponding to the type zr , which is set to 0.035. The basis for setting this reference value is that when the overlapping area of longitudinal cracks and the shear force concentration area exceeds 3.5% of the total area, it has an obvious impact on the structural safety. This value comes from the critical coincidence ratio of the ultimate bearing capacity model of shear-compression members and the local force simulation of the structure. Substitute it into the formula

[0109]

[0110] to obtain the damage severity value S zr ≈0.002115. This value is the quantitative expression result of the damage severity of this area. The small value indicates that although there are cracks in the current damaged area and it falls into the key structural area, the affected area is small. The advantage of the formula is that by introducing the product of the normalized boundary curvature and the area ratio to construct the damage scale evaluation basis, and correcting the risk range through the coincidence difference, and at the same time introducing the damage type coefficient as the strength weighting term, the final result can take into account the comprehensive influence degree of damage form, location and type on the safety of the pier structure, and improve the resolution and structural relevance of quantitative identification.

[0111] Please refer to Figure 6 , the pier anomaly assessment module includes:

[0112] The high-risk area screening sub-module extracts the severity value of each damaged area based on the damage assessment information, compares each area with the set severity threshold, screens the areas that exceed the threshold, extracts the damage type label and spatial range of the corresponding areas, and generates a high-severity area set;

[0113] The high-risk area screening sub-module calls the aforementioned damage assessment information, extracts the severity value of each damaged area item by item. Suppose there are 5 damaged areas in the current image, and their severity values are 0.0021, 0.0054, 0.0098, 0.0009, and 0.0072 respectively. Set the severity threshold to 0.005. This threshold is set according to the standard crack warning value of the pier concrete structure and the damage tolerance of the underwater pressure-bearing area, and is set by selecting the 75th percentile value corresponding to the safety boundary in the previous 600 sets of structural failure samples to ensure convergence and stability of the risk boundary. Compare the severity values of each area item by item according to this, and determine that the areas with severity values greater than 0.005 are high-risk areas. The current 2nd, 3rd, and 5th items are 0.0054, 0.0098, and 0.0072 respectively, all of which are greater than the threshold and are screened as high-risk areas. Mark the damage types corresponding to these three items as "longitudinal crack", "erosion", and "hollow drum" respectively, and their spatial boundary coordinates are [(640,1280)–(704,1344)], [(880,960)–(944,1024)], and [(1200,1500)–(1264,1564)] respectively, and store them in the high-severity area set in turn.

[0114] The distribution parameter extraction sub-module calls the high-severity area set, counts the number of areas of each damage type, extracts the set of centroid coordinates of each area, and calculates the density change trend value of the centroid distribution to obtain the damage distribution characteristic quantity;

[0115] Based on the high-severity area set, the distribution parameter extraction sub-module counts the number of areas of each damage type. Suppose there are 3 types in the current high-risk area set, including 2 "longitudinal cracks", 1 "erosion", and 1 "hollow drum". Calculate the total number and mark it. Then extract the centroid coordinates of each area. The centroid coordinates are obtained by taking the median of the boundary coordinates. For example, the center points of the "longitudinal crack" areas are (672,1312) and (1340,980), and store their centroid coordinate set as an array [(672,1312),(1340,980)]. Calculate the density change trend value for each type of centroid coordinate set. The density change trend value is defined as the ratio of the standard deviation to the mean of the shortest distance between centroids, that is, D var =std(d ij ) / mean(d ij ). Taking the two points of "longitudinal crack" as an example, calculate the Euclidean distance as Since it is a single distance, the standard deviation is 0 and the density change trend is 0, indicating a high concentration of this type of damage. If there are more centroid points, calculate the average and standard deviation after calculating the distances between each pair in turn. Finally, integrate the quantity, centroid coordinates, and density trend values of each type into the damage distribution characteristic quantity.

[0116] The influence degree calculation sub-module calls the damage distribution parameter set, extracts the quantity, distribution density, and central tendency coefficient of each type of damage, combines with the set of screening areas, and uses the formula:

[0117]

[0118] Performs operations to obtain the abnormal influence degree of the bridge pier and gets the abnormal evaluation result of the bridge pier;

[0119] Among them, I b represents the abnormal influence degree of the bridge pier, N as,k represents the quantity of the k-th type of damage area, D as,k represents the distribution density of the k-th type of damage area, T as,k represents the central tendency coefficient of the k-th type of damage area, K represents the total number of damage types, and S as,k is the average value of the damage severity of the k-th type of damage area;

[0120] The influence degree calculation sub-module calls the aforementioned distribution parameter set, extracts the area quantity N as,k , distribution density D as,k , central tendency coefficient T as,k and the average severity value S as,k , assuming that there are currently K = 3 types of damage in total, and the specific values are as follows: longitudinal crack type: quantity N as,1 = 2, density D as,1 = 0.0025, trend coefficient T as,1 = 0.1, severity mean S as,1 = 0.00575; erosion type: quantity N as,2 = 1, density D as,2 = 0.0018, trend coefficient T as,2 = 0.25, severity mean S as,2 = 0.0098; hollowing type: quantity N as,3 = 1, density D as,3 = 0.0016, trend coefficient T as,3 = 0.32, severity mean S as,3 = 0.0072, substitute into the formula:

[0121]

[0122] Expand the calculation as follows:

[0123]

[0124] Finally, obtain the abnormal influence degree value I b≈0.00000815. This result indicates that the overall influence range of the current high-risk area of the pier is relatively small, mainly caused by individual high-severity damages and local aggregation characteristics. The advantage of the formula lies in that by introducing the square root of the trend coefficient to adjust the weight of the damage aggregation effect and coupling the calculation of density, quantity, and average severity value, the evaluation result takes into account the spatial distribution, intensity, and category differences of damages simultaneously.

[0125] The above are only the preferred embodiments of the present invention and do not limit the present invention in other forms. Any person skilled in the art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. An underwater structure inspection system for bridge piers based on an unmanned underwater vehicle, characterized in that The system includes: The inspection cycle adjustment module obtains the bridge traffic flow data, calculates the difference between the cumulative traffic volume in the peak period and the cumulative traffic volume in the trough period, invokes the water flow shear force parameter, calculates the difference value between the shear force increment value and the safety critical shear force increment threshold, adjusts the inspection cycle of the bridge pier, and generates the dynamic adjustment result of the inspection cycle; The anomaly recognition module, based on the dynamic adjustment result of the inspection cycle, uses an unmanned underwater vehicle to inspect the underwater structure of the bridge pier, obtains the underwater inspection images, identifies the abnormal area by comparing with the known abnormal images of the bridge pier, and marks the abnormal area to generate the calibration result of the abnormal area; The automatic reshooting module invokes the calibration result of the abnormal area, extracts the pixel noise index and the clarity index, evaluates the shooting quality of the abnormal area. If the shooting quality is lower than the set quality threshold, it adjusts the light source brightness and exposure time and performs reshooting in-situ to generate the local abnormal reshot images; The damage assessment module invokes the local abnormal reshot images, extracts the damage center position, compares it with the key area of the bridge pier, calculates the coincidence degree of the damage area, and combines the damage type to calculate the damage severity to generate the damage assessment information.

2. The pier underwater structure inspection system based on an unmanned underwater vehicle according to claim 1, wherein The dynamic adjustment result of the inspection cycle includes the cycle adjustment coefficient, the triggering state of the flow difference threshold, and the water flow shear force increment state. The calibration result of the abnormal area is specifically the abnormal boundary position parameter, the abnormal shape determination identifier, and the central point coordinates of the abnormal area. The local abnormal reshot images include the reshot light source brightness level, the reshot exposure time setting value, and the reshot area imaging data set. The damage assessment information is specifically the damage type weight value, the sensitive area coincidence coefficient, and the comprehensive influence path coefficient.

3. The underwater structure inspection system for bridge piers based on an unmanned underwater vehicle according to claim 1, wherein The inspection cycle adjustment module includes: The traffic flow calculation sub-module obtains the bridge traffic flow data, divides the time interval according to the peak period and the trough period, calculates the difference between the cumulative traffic volume in the peak period and the cumulative traffic volume in the trough period, and obtains the peak-trough traffic difference; The shear force increment calculation sub-module invokes the peak-trough traffic difference, obtains the water flow shear force parameter, calculates the difference between the current shear force increment value and the safety critical shear force increment threshold according to the safety critical shear force increment threshold, and combines the preset inspection cycle of the bridge pier and the peak-trough traffic difference, using the formula: Performs operations to obtain the cycle adjustment coefficient; Among them, T adj represents the cycle adjustment coefficient, Q d represents the normalized value of the cumulative traffic volume during the peak period, Q p represents the normalized value of the cumulative traffic volume during the off-peak period, where S i represents the normalized value of the change in water flow velocity within the i-th time period, C s represents the threshold value of the safety critical shear force increment, F d represents the normalized value of the current shear force increment value, F s represents the normalized value of the shear force reference benchmark value, and n represents the total number of time periods in which the change in water flow velocity participates in the calculation; The cycle adjustment output sub-module, according to the cycle adjustment coefficient, combines the preset inspection cycle of the bridge pier, determines the cycle adjustment direction, calculates the adjusted cycle value, adjusts the inspection cycle of the bridge pier, and generates the dynamic adjustment result of the inspection cycle.

4. The underwater structure inspection system for bridge piers based on an unmanned underwater vehicle according to claim 1, characterized in that, The anomaly recognition module includes: The underwater image acquisition sub-module, based on the dynamic adjustment result of the inspection cycle, schedules an unmanned underwater vehicle to inspect the underwater structure of the bridge pier within a specified time period, collects a continuous underwater image frame sequence of the bottom structure area of the bridge pier, establishes an image sequence library, and obtains the underwater image sequence of the bridge pier; The feature matching degree calculation sub-module invokes the underwater image sequence of the bridge pier, combines the known abnormal images of the bridge pier, obtains the gray mean value, texture direction mean value, and boundary curvature sequence of the detection area, and compares them with the gray mean value, texture direction mean value, and boundary curvature sequence of the reference area, using the formula: The operation obtains the matching degree value of the abnormal area; Among them, M d represents the matching degree of the abnormal area, G d represents the normalized value of the average gray value of the detection area, G r represents the normalized value of the average gray value of the reference area, T d represents the normalized value of the average texture direction of the detection area, T r represents the normalized value of the average texture direction of the reference area, C d,j represents the j-th boundary curvature value of the detection area, C r,j represents the j-th boundary curvature value of the reference area, and N represents the number of boundary curvature sampling points; The area marking output sub-module compares the matching degree value of the abnormal area with a preset matching degree threshold, identifies the area where the matching degree value is greater than the threshold, combines the center coordinates of the detection area, marks the position and boundary range of the abnormal area, and generates the calibration result of the abnormal area.

5. The underwater structure inspection system for bridge piers based on an unmanned underwater vehicle according to claim 1, wherein The automatic reshooting module includes: The image block extraction sub-module calls the calibration result of the abnormal area, divides the pixel grid of the abnormal area, generates a set of image blocks with a fixed size according to the pixel grid coordinates, extracts the gray matrix of the image blocks, calculates the gray variance value and the mean gradient value of the image blocks, and obtains the basic parameter set of the images; The quality evaluation sub-module compares the gray variance value with the image clarity reference value according to the basic parameter set of the images, makes a difference judgment on the mean gradient value and the pixel noise reference value, combines the two judgment results to grade the quality of the image blocks, labels each image block with a quality grading label, and generates a group of image quality grading labels; The image reshooting sub-module calls the group of image quality grading labels, screens out the image blocks labeled with low quality levels, extracts the spatial position information of the screened image blocks, adjusts the exposure time and light source brightness parameters of the underwater unmanned vehicle shooting device according to the position coordinates, performs in-situ reshooting, and generates a local abnormal reshooting image.

6. The underwater structure inspection system for bridge piers based on an unmanned underwater vehicle according to claim 1, characterized in that, The damage assessment module includes: The structural parameter extraction sub-module segments each damage area in the image based on the local abnormal reshooting image, extracts the boundary line of the damage area, calculates the boundary curvature value, counts the pixel proportion of the damage area in the whole image, and identifies the center position coordinates of the damage area, and obtains the damage structure parameter set; The overlap degree calculation sub-module calls the spatial boundary data of the shear force concentration area, axial pressure change area and bending resistance bearing area of the pier according to the damage structure parameter set, calculates the spatial overlap range between the damage area and the key area, and calculates the ratio of the overlap area to the total area of the corresponding key area to obtain the overlap degree value of the damage area; The severity assessment sub-module calls the overlap degree value of the damage area and the damage structure parameter set, extracts the boundary curvature, area proportion, spatial overlap degree and corresponding damage type of the damage area, and uses the formula: The operation obtains the damage severity value of the damage area and generates damage assessment information; Among them, S zr represents the severity value of the damage, C zr represents the normalized value of the boundary curvature of the damaged area, A zr represents the area ratio of the damaged area, R zr represents the coincidence degree value between the damaged area and the key area, K zr represents the coincidence degree sensitive reference value corresponding to the damaged area type, D zr represents the damage type influence coefficient.

7. The underwater structure inspection system for bridge piers based on an unmanned underwater vehicle according to claim 1, wherein, The system further includes: The pier abnormality assessment module calls the damage assessment information, screens out the areas where the damage severity exceeds the set threshold, counts the damage type and distribution range, combines the number and central tendency of the damage areas, calculates the pier abnormality influence degree, and generates the pier abnormality assessment result; The pier abnormality assessment result specifically refers to the damage distribution density, the total number of damages and the damage central tendency coefficient.

8. The pier underwater structure inspection system based on an unmanned underwater vehicle according to claim 7, characterized in that The pier abnormality assessment module includes: The high-risk area screening sub-module extracts the severity value of each damage area based on the damage assessment information, compares each area with the set severity threshold, screens out the areas that exceed the threshold, extracts the damage type label and spatial range of the corresponding areas, and generates a set of high-severity areas; The distribution parameter extraction sub-module calls the high-severity region set, counts the number of regions of each type of damage, extracts the set of centroid coordinates of each region, calculates the density change trend value of the centroid distribution, and obtains the damage distribution characteristic quantity; The influence degree calculation sub-module calls the damage distribution parameter set, extracts the number of damages of each type, the distribution density and the central tendency coefficient, combines the screened region set, and uses the formula: Calculate to obtain the abnormal influence degree of the pier and get the abnormal evaluation result of the pier; Among them, I b represents the abnormal influence degree of the pier, N as,k represents the number of the k-th type of damage area, D as,k represents the distribution density of the k-th type of damage area, T as,k represents the central tendency coefficient of the k-th type of damage area, K represents the total number of damage types, S as,k is the average value of the damage severity of the k-th type of damage area.

Images