Edge computing-based decision support system for structure maintenance using drones

Abstract

The present invention relates to an edge computing-based structural maintenance decision support system utilizing drones, and more specifically, to an edge computing-based structural maintenance decision support system utilizing drones that analyzes image data acquired by an inspection drone in real time at an edge computing device and integrates this data at a central management server to support advanced decision-making, such as maintenance priorities, work packages, and budget linkage analysis. The edge computing-based structural maintenance decision support system utilizing a drone according to the present invention can not only detect damaged objects of a structure in real time and calculate damage dimensions and risk levels using an inspection drone and an edge computing device, but also provide a decision support system that goes beyond a simple damage detection system by having a central management server reconstruct the damage analysis results into a form that can be directly utilized by a manager for maintenance decision-making.

Description

Technology Field

[0001] The present invention relates to an edge computing-based structural maintenance decision support system utilizing drones, and more specifically, to an edge computing-based structural maintenance decision support system utilizing drones that analyzes image data acquired by an inspection drone in real time at an edge computing device and integrates this data at a central management server to support advanced decision-making, such as maintenance priorities, work packages, and budget linkage analysis. Background Technology

[0002] Large-scale infrastructure such as bridges, dams, and transmission towers may suffer various types of damage, including cracks, spalling, exposed rebar, exposed steel, and cross-sectional loss, due to long-term use and exposure to the external environment.

[0003] Since such damage can affect the durability and safety of the structure, it is important to accurately identify the location, size, progression, and structural risk of the damage, and to perform appropriate maintenance measures in a timely manner based on this information.

[0004] Traditionally, the general method involved an inspector visiting the site of a structure in person to perform a visual inspection or inspection using photographic equipment, and then determining the extent of damage and establishing a maintenance plan based on the collected data.

[0005] However, these conventional technologies suffered from reduced inspection efficiency when the inspection scope of the structure was wide or the inspection target was located at high altitudes, underwater, or in difficult-to-access locations, and had the problem that the interpretation of inspection results and the determination of maintenance priorities relied on the inspector's experience and the manager's subjective judgment.

[0006] Recently, technologies have been proposed that utilize drones to acquire surface images of structures and employ edge computing or AI-based analysis techniques to detect damaged objects and calculate damage dimensions and risk levels. However, since these technologies primarily focus on damage detection and condition assessment itself, they have limitations in that they do not adequately provide decision support functions—such as generating maintenance flags, assigning priorities, analyzing progress trends based on historical comparisons, making execution decisions that account for budget and resource constraints, and map-based visualization—which are necessary for managers to incorporate damage analysis results into actual maintenance plans.

[0007] In other words, while conventional technology can provide information on the existence and risk level of damaged objects, it has the problem of failing to provide a systematic system that supports managers in intuitively and quickly determining which parts among multiple damages throughout the structure should be prioritized for action in what order and manner. Prior art literature

[0008] Republic of Korea Registered Patent 10-1866781 Republic of Korea Registered Patent 10-2095643 The problem to be solved

[0009] The present invention aims to solve the aforementioned conventional problems by providing a system that supports managers in making maintenance decisions more quickly and rationally regarding the order and method of priority response to damaged parts of a structure. This is achieved by detecting damaged objects in a structure in real time through inspection drones and edge computing devices and calculating damage dimensions and risk levels, while a central management server receives the damage analysis results to generate maintenance flags, maintenance priorities, and damage progression information, and intuitively visualizing this information on a manager terminal.

[0010] Furthermore, the present invention aims to enable proactive maintenance decision-making beyond a simple assessment of the current state by analyzing changes in damage or risk levels by comparing past and current inspection data, and by providing maintenance priorities that even consider the rate of damage progression.

[0011] In addition, the present invention provides a decision support environment that allows a manager to grasp the maintenance status and priority areas at the overall structural level at a glance by visually displaying damage locations, damage grades, risk scores, maintenance flags, and priorities on a structural map or schematic diagram.

[0012] Furthermore, the present invention provides an actionable decision support function that can be directly utilized in actual maintenance tasks by determining whether repairs are possible, whether inspections are prioritized, or whether budget adjustments are necessary by reflecting information on available maintenance budgets or resources, and further generating implementation recommendations by clustering multiple damaged objects into maintenance work packages. means of solving the problem

[0013] An edge computing-based structural maintenance decision support system utilizing a drone according to the present invention for achieving the above-mentioned purpose comprises: an inspection drone including a drone body, a camera installed on the drone body for acquiring a surface image of a structure to be inspected, a position sensor and an attitude sensor for measuring the position and attitude of the inspection drone, a distance measuring sensor for measuring the distance to the surface of the structure to be inspected, and a data collection interface for collecting image and measurement data acquired from the camera and sensors and transmitting them to an external device; and an edge computing device including a damage detection unit for detecting at least one damaged object among cracks, spalling, rebar exposure, steel exposure, and cross-sectional loss of a structure based on image data received from the inspection drone, a dimension calculation unit for converting the dimension information of pixels within the image corresponding to the detected damaged object into actual physical defect dimensions by reflecting the attitude information of the inspection drone and the distance information to the structure to be inspected, a risk calculation unit for calculating the risk of damage using the physical defect dimensions and weights for each type of damage, and a judgment unit for classifying the damage grade into multiple stages of normal, caution, warning, and danger according to the calculated risk and determining whether maintenance is required. A central management server linked with the above edge computing device is provided.

[0014] The central management server of the edge computing-based structural maintenance decision support system utilizing a drone according to the present invention comprises: a result receiving unit that receives damage detection results, damage dimension information, risk level, and judgment results received from the edge computing device; a maintenance flag generating unit that generates at least one maintenance flag among repair necessity, inspection request, regular inspection, and continued observation for the structure subject to inspection based on the judgment result; a maintenance priority generating unit that generates maintenance priorities for a plurality of damaged objects or a plurality of management unit sections by reflecting at least one of the risk level, damage grade, location, and importance of the structural member for the damaged object; a damage progression trend analysis unit that calculates a change in damage or a change in risk level by comparing past inspection data and current inspection data for the same location or the same management unit section; an notification display unit that displays the damage detection results, damage dimension information, risk level, and judgment results in the form of a notification to an administrator terminal; and a support unit that displays the location of the damaged object on a structural map or schematic diagram, and visually displays the damage grade, risk score, maintenance flag, maintenance priority, and damage progression trend information corresponding to the location of each damaged object, thereby enabling the administrator to make maintenance decisions regarding the order and method of priority response to the damaged area. It may include a decision support unit and a data storage unit that stores the damage detection result, damage dimension information, risk level, judgment result, maintenance flag, maintenance priority, and inspection history data.

[0015] The central management server of the edge computing-based structural maintenance decision support system utilizing a drone according to the present invention may further include a location mapping unit that converts the damage location into location information of at least one management unit section among a span, girder, pier, main tower, deck plate, panel, or survey unit of the structure to be inspected, using at least one of GPS coordinates, vision-based local coordinates, structural drawing data, and inspection path metadata.

[0016] The damage progression trend analysis unit of the edge computing-based structural maintenance decision support system utilizing a drone according to the present invention calculates the amount of damage change or risk change by comparing past inspection data and current inspection data for the same location or the same management unit section, and if the amount of damage change or risk change exceeds a preset threshold, it can adjust the maintenance priority upward or generate a warning notification.

[0017] The decision support unit of the edge computing-based structural maintenance decision support system utilizing a drone according to the present invention may provide a map-type interface that visually displays the location of damage on a structural map or schematic diagram, and a maintenance priority list-type interface that includes a ranking, location, damage grade, risk score, and maintenance flag for a plurality of damaged objects or a plurality of management unit sections.

[0018] The central management server of the edge computing-based structural maintenance decision support system utilizing a drone according to the present invention may further include a budget determination unit that receives information on available maintenance budget or resources and determines whether maintenance is possible, whether inspection is a priority, or whether budget adjustment is necessary by comparing it with the necessity of action according to the maintenance priority.

[0019] The central management server of the edge computing-based structural maintenance decision support system utilizing a drone according to the present invention clusters a plurality of damaged objects into a single maintenance work package based on at least one of spatial proximity of the damage location, identity or similarity of the damage type, and commonality of applicable repair methods and equipment, and can generate maintenance priorities and implementation recommendations at the package level by reflecting the cumulative risk, damage progression trend, importance of structural members, and estimated repair resource requirements for each maintenance work package. Effects of the invention

[0020] The edge computing-based structural maintenance decision support system utilizing a drone according to the present invention can not only detect damaged objects of a structure in real time and calculate damage dimensions and risk levels using an inspection drone and an edge computing device, but also provide a decision support system that goes beyond a simple damage detection system by having a central management server reconstruct the damage analysis results into a form that can be directly utilized by a manager for maintenance decision-making.

[0021] In addition, the edge computing-based structural maintenance decision support system utilizing a drone according to the present invention can generate action-direction information, such as the need for repair, inspection required, regular inspection, and continued observation, based on the damage analysis results of the maintenance flag generation unit, so that the manager can immediately identify what level of response is required for each type of damage.

[0022] In addition, the edge computing-based structural maintenance decision support system utilizing a drone according to the present invention enables a maintenance priority generation unit to generate maintenance priorities for multiple damaged objects or management units by reflecting risk, damage grade, damage location, and the importance of structural members, thereby making it possible to make a rational judgment regarding where to first allocate limited maintenance resources at the overall structural level.

[0023] In addition, the edge computing-based structural maintenance decision support system utilizing a drone according to the present invention enables proactive maintenance that reflects not only the current state but also the speed of damage progression and deterioration trends, as the damage progression trend analysis unit compares past inspection data with current inspection data to calculate the amount of change in damage or risk level and reflects the results in priority adjustment or warning notification generation.

[0024] In addition, the edge computing-based structural maintenance decision support system utilizing a drone according to the present invention allows the decision support unit to mark damage locations on a structural map or schematic diagram and visually display damage grade, risk score, maintenance flag, priority, and damage progression trend information corresponding to each damage location, so that the manager can intuitively grasp the overall damage status and action priority of the structure.

[0025] In addition, the edge computing-based structural maintenance decision support system utilizing a drone according to the present invention can provide a decision support function that can be directly utilized in the actual budget execution and maintenance planning process, as the budget judgment unit can determine whether maintenance is possible, whether inspection is a priority, or whether budget adjustment is necessary by reflecting information on the budget or resources available for maintenance.

[0026] In addition, the edge computing-based structural maintenance decision support system utilizing a drone according to the present invention can cluster multiple damaged objects into maintenance work packages and generate maintenance priorities and execution recommendations at the package level, thereby enabling the establishment of an efficient maintenance work plan that considers site accessibility, method commonality, and equipment utilization.

[0027] Consequently, the present invention not only provides the results of structural damage detection as mere warning information, but also has the effect of supporting managers in determining priorities, methods of action, budget execution, and implementation plans regarding structural maintenance in a more intuitive, rapid, and rational manner. Brief explanation of the drawing

[0028] FIG. 1 is a schematic diagram showing an edge computing-based structural maintenance decision support system utilizing a drone according to the present invention. FIG. 2 is a drawing showing an inspection drone of an edge computing-based structural maintenance decision support system utilizing a drone according to the present invention. FIG. 3 is a block diagram showing the overall configuration of an edge computing-based structural maintenance decision support system utilizing a drone according to the present invention. FIG. 4 is a diagram showing the processing process at an edge computing device of an edge computing-based structural maintenance decision support system utilizing a drone according to the present invention. FIG. 5 is a diagram showing a specific processing structure in an edge computing device of an edge computing-based structural maintenance decision support system utilizing a drone according to the present invention. FIG. 6 is a diagram illustrating the process in which a central management server of an edge computing-based structural maintenance decision support system utilizing a drone according to the present invention proposes maintenance flags, maintenance priorities, and recommendations based on results received from an edge computing device. Specific details for implementing the invention

[0029] Hereinafter, an edge computing-based structural maintenance decision support system utilizing a drone according to a preferred embodiment of the present invention will be described in detail with reference to the attached drawings.

[0030] FIGS. 1 to 6 illustrate an edge computing-based structural maintenance decision support system utilizing a drone according to the present invention. Referring to FIGS. 1 to 6, the edge computing-based structural maintenance decision support system utilizing a drone according to the present invention may include an inspection drone (100) that flies around a structure to be inspected and acquires images of the structure's surface, an edge computing device (200) that detects damaged objects and calculates damage dimensions and risk levels based on images and sensor data received from the inspection drone (100), and a central management server (300) that receives analysis results transmitted from the edge computing device (200) and supports maintenance decision-making.

[0031] The inspection drone is equipped with a communication module for transmitting inspection information, including dimensions and risk levels calculated by an edge computing device (200), and various sensing and control information to a central management server and communicating with the central management server. In particular, the central management server (300) can be linked with a manager terminal (400) that provides visual or notification-type information to a manager.

[0032] The structures subject to inspection may be bridges, main towers, piers, girders, decks, slabs, panels, or similar structures. For example, locations such as the lower decks of bridges, the sides of girders, the surfaces of piers, the outer surfaces of main towers, abutments, and connections may be representative areas subject to inspection where cracks, spalling, exposed rebar, or exposed steel may occur repeatedly.

[0033] The present invention is characterized by the ability to detect damaged objects, calculate dimensions, and calculate risk levels through a drone (100) and an edge computing device (200), but in particular, the central management server (300) does not stop at providing the results as simple warning information, but rather reconstructs them into maintenance flags, maintenance priorities, damage progression trends, and execution recommendations to support manager decision-making. That is, the present invention is a decision support system that goes beyond merely showing what damage exists and enables a manager to easily decide which location, in what order, and in what manner to prioritize response.

[0034] The inspection drone (100) may include a drone body (110), a camera (120) installed on the drone body (110), a position sensor (130) for measuring the position of the inspection drone (100), a posture sensor (140) for measuring the attitude of the inspection drone (100), a distance measuring sensor (150) for measuring the distance to the surface of the structure to be inspected, and a data collection interface (160) that collects data obtained from the camera (120) and sensors and transmits it to an external device.

[0035] The camera (120) can acquire surface images of the structure in real time or sequentially, and, for example, can acquire continuous images at a level of 24 frames to 60 frames per second.

[0036] The position sensor (130) may be a GPS sensor, a satellite navigation sensor, or a position measuring device equivalent thereto, and the attitude sensor (140) may be an IMU, a gyroscope sensor, an accelerometer sensor, or a position measuring device equivalent thereto.

[0037] The distance measuring sensor (150) may be a laser rangefinder, LiDAR, an ultrasonic sensor, or a distance measuring device equivalent thereto.

[0038] The data collection interface (160) can transmit image, position, attitude, and distance data acquired from the camera (120) and sensors to the edge computing device (200) via wired or wireless means.

[0039] The data collection interface (160) can combine the timestamp corresponding to the video frame and the timestamp of the sensor data to transmit the video and sensor data at the same point in time so that they are aligned with each other. Accordingly, the dimension calculation unit (220) described later can perform a more accurate dimension calculation using the posture information and distance information at the exact time when a specific damaged object was captured.

[0040] For example, when an inspection drone (100) moves along the lower part of a bridge span and photographs the deck plate and girder, the camera (120) can continuously acquire images of the surface of the structure, the position sensor (130) can measure the current position of the drone, the attitude sensor (140) can measure the tilt and direction of the drone, and the distance measuring sensor (150) can measure the distance to the surface of the structure in real time. This data can be collected as a single inspection unit dataset and transmitted to an edge computing device (200).

[0041] The edge computing device (200) may include a damage detection unit (210), a dimension calculation unit (220), a risk calculation unit (230), and a judgment unit (240).

[0042] The edge computing device (200) can detect damaged objects, calculate dimensions, and calculate risk, and in the present invention, the result is used as input data for decision support processing of the central management server (300).

[0043] The damage detection unit (210) can detect damaged objects of the structure based on image data received from the inspection drone (100).

[0044] The damaged object may include, for example, at least one of cracks, spalling, rebar exposure, steel exposure, and cross-sectional loss.

[0045] The damage detection unit (210) can generate information regarding the location, shape, and size of a damaged object by applying a deep learning-based object detection model or a segmentation model. For example, for crack damage, a linear bounding box or a segmentation mask can be generated, and for delamination and cross-sectional loss, an area-type contour or a pixel-unit segmentation area can be generated. In the case of rebar exposure or steel exposure, a metallic exposure area extended in the longitudinal direction or an irregular exposure area can be simultaneously extracted in the form of a bounding box and a mask.

[0046] For example, in a video frame captured by a camera (120) while an inspection drone (100) is flying up along the bridge surface, the damage detection unit (210) can detect a thin and elongated object as a crack and detect an area where the surface has detached and appears rough as peeling. Additionally, if linear steel or rebar exposure is confirmed within the peeling area, it can be further classified as rebar exposure. In this way, multiple damaged objects can be detected simultaneously within a single video frame.

[0047] The dimension calculation unit (220) can convert pixel dimension information on an image corresponding to a damaged object detected by the damage detection unit (210) into actual physical defect dimensions. To this end, the dimension calculation unit (220) can reflect the posture information obtained from the posture sensor (140) and the distance information obtained from the distance measurement sensor (150).

[0048] For example, for crack damage, the width or length in pixel units can be converted to actual mm units, and for delamination or cross-sectional loss, pixel-based area information can be converted to actual area units. For example, the dimension calculation unit (220) can calculate the actual width of the crack as 0.36 mm by applying a scaling factor that reflects distance and attitude information when the distance between the drone and the structure is 4.5 m and the maximum width of a specific crack is measured as 6 pixels in the image. In addition, for the delamination area, the mask area corresponding to 12,000 pixels can be converted to an actual 0.045 m².

[0049] The physical defect dimensions calculated in this way are not merely reference values, but can be utilized as direct input values ​​for the risk calculation and maintenance flag determination described later.

[0050] The dimension calculation unit (220) can perform dynamic scaling operations that correct image scale changes and perspective distortions caused by the shooting distance and shooting attitude of the inspection drone (100), rather than simply calculating the physical defect dimensions by multiplying the number of pixels by a certain constant. To this end, the dimension calculation unit (220) can calculate an image scale factor at the time when a specific damaged object is photographed by using Roll, Pitch, and Yaw information obtained from the attitude sensor (140) and distance information to the surface of the structure obtained from the distance measuring sensor (150). For example, even for a crack with the same width of 0.3 mm, the number of pixels in the image may appear different when the drone photographs from a position 2 m away from the structure compared to when it photographs from a position 5 m away.

[0051] Additionally, if the drone photographs the surface of the structure in an inclined direction rather than from the front, the actual defect dimensions may be observed reduced or distorted in the image. Accordingly, the dimension calculation unit (220) can correct the pixel-based dimension information of the damaged object to the actual physical defect dimensions by combining a basic scale factor based on distance information and an angle correction factor based on attitude information.

[0052] The dimension calculation unit (220) can convert at least one of the maximum width, average width, or extended length of a crack object into a physical dimension for crack damage, and can convert the total number of pixels of a damage mask into an actual area unit for area-type damage such as delamination or cross-sectional loss. For example, in the case of a crack, a maximum width of 6 pixels can be converted to 0.36 mm after distance and orientation correction, and in the case of delamination, a damage area of ​​12,000 pixels can be calculated as 0.045 m². In this way, the dimension calculation unit (220) can calculate appropriate physical quantities such as width, length, or area according to the type of damage and provide them to the subsequent risk calculation unit (230).

[0053] Additionally, the dimension calculation unit (220) can determine the final physical defect dimensions by integrating the dimension calculation results for each frame into an average value, a maximum value, or a weighted average value when the same damaged object is repeatedly detected in consecutive video frames. Accordingly, measurement errors caused by instantaneous vibration of the drone, changes in the shooting angle, or video noise can be reduced.

[0054] The risk calculation unit (230) can calculate the risk of damage using the physical defect dimensions calculated by the dimension calculation unit (220) and the weights for each type of damage. For example, even if the physical defect dimensions are the same, cracks, spalling, rebar exposure, and cross-sectional loss may have different effects on structural risk, so the risk calculation unit (230) can apply different weights for each type of damage.

[0055] For example, a crack of 0.35 mm and metal exposure damage appearing with a width of 0.35 mm may not be treated structurally the same. The risk calculation unit (230) may calculate a different risk score by applying a relatively medium weight to the crack and a higher weight to the rebar exposure or cross-section loss. For example, even if the dimensions are the same, the risk can be differentiated such as a crack with a risk of 0.58 and a rebar exposure with a risk of 0.81.

[0056] The risk calculation unit (230) may not uniformly determine the risk using only the calculated physical defect dimensions and damage types, but may generate a more detailed normalized risk score by referring to a stored condition evaluation standard table or a risk mapping standard for each damage. For example, for crack damage, the lower and upper limits of the risk score may be set according to which standard range the actual crack width corresponds to, and a continuous risk score may be calculated by performing linear interpolation within the range.

[0057] For example, if the crack width is in the range of 0.3mm or more and less than 0.5mm, the risk calculation unit (230) sets the reference score corresponding to 0.3mm to 0.5 and the reference score corresponding to 0.5mm to 0.8, and if the currently calculated crack width is 0.4mm, it can generate a risk of 0.65 by reflecting the relative position within the range. Accordingly, even within the same "boundary" grade, a 0.31mm crack and a 0.49mm crack can be quantified as having different risk levels.

[0058] In addition, for damages with a high potential for reduced structural load-bearing capacity, such as exposed rebar or exposed steel, non-linear weights may be applied instead of simple linear interpolation. For example, if the damage detection unit (210) extracts the corrosion pixel distribution ratio or the metal exposure ratio together, the risk calculation unit (230) may be configured to rapidly increase the risk level from the point when the ratio exceeds a preset threshold. Accordingly, even for damages of the same size, severe corrosion, exposed steel, or cross-sectional loss damage may be reflected with a higher risk level.

[0059] The risk calculation unit (230) may assign additional density weights rather than treating the risk of individual damages simply independently when multiple damaged objects exist in a dense cluster within the same investigation unit. For example, if multiple thin cracks exist within the same area, even if each individually is at a "caution" level, the risk calculation unit (230) may calculate a higher collective risk for the entire clustered set of damages.

[0060] As such, the present invention can perform multidimensional risk calculation that reflects not only the absolute dimensions of a single damaged object but also the type of damage, degree of corrosion, and cluster characteristics.

[0061] The judgment unit (240) can classify the damage grade into multiple stages of normal, caution, warning, and danger according to the risk calculated by the risk calculation unit (230) and determine whether maintenance is required.

[0062] The judgment results may include, for example, the damage level, whether maintenance is required, whether an early inspection is required, or whether observation should be continued.

[0063] If the risk score is less than 0.2, it can be classified as "Normal"; if it is 0.2 or more but less than 0.5, it can be classified as "Caution"; if it is 0.5 or more but less than 0.75, it can be classified as "Warning"; and if it is 0.75 or more, it can be classified as "Dangerous". Of course, these ranges may vary depending on the type of structure or management policy. For example, if the risk score of a specific crack is 0.61, the judgment unit (240) can classify it as "Warning" and determine that "prompt inspection is needed," and if the risk score of a specific rebar exposure is 0.86, it can classify it as "Dangerous" and determine that "repair is needed."

[0064] The damage detection results, damage dimension information, risk level, and judgment results generated in this way can be transmitted to the central management server (300).

[0065] The central management server (300) may include a result receiving unit (310), a maintenance flag generating unit (320), a maintenance priority generating unit (330), a damage progression trend analysis unit (340), a notification display unit (350), a decision support unit (360), and a data storage unit (370). If necessary, the central management server (300) may further include a location mapping unit (380), a budget judgment unit (390), and a maintenance work package generating unit (395).

[0066] In the present invention, the central management server (300) is not a simple storage server, but a core component that converts the analysis results generated by the edge computing device (200) into maintenance decision-making information from an administrator's perspective. That is, the central management server (300) can support the establishment of a practical maintenance plan by reorganizing the damage analysis results into maintenance flags, priorities, trends, budget judgments, and execution recommendations.

[0067] The result receiving unit (310) can receive damage detection results, damage dimension information, risk level, and judgment results from the edge computing device (200). The result receiving unit (310) can structure the data by damage object, by structural member, or by management unit.

[0068] For example, a single damaged object can be managed as a single record along with a structure identifier, damage location, damage type, damage dimensions, risk level, damage grade, time of capture, and related image information. The record can be stored in a format such as "Bridge A P3-P4 span, G2 girder, crack, width 0.42mm, risk level 0.61, boundary, time of capture 2026-03-12 10:25:32".

[0069] The maintenance flag generation unit (320) can generate a maintenance flag for the structure to be inspected based on the judgment result transmitted from the judgment unit (240).

[0070] Maintenance flags may be at least one of, for example, Repair Needed, Inspection Required, Scheduled Inspection, and Continued Observation. For instance, a "Repair Needed" flag may be generated if the damage grade corresponds to "Dangerous" or if the risk score exceeds a preset threshold. Conversely, a "Inspection Required" flag may be generated if the absolute risk is moderate but there is a high probability of future deterioration. If the risk is relatively low and the trend of change is gradual, a "Scheduled Inspection" or "Continued Observation" flag may be assigned.

[0071] More specifically, the maintenance flag generation unit (320) can consider the current absolute risk level and the results of the damage progression trend analysis together. For example, if the current risk level is 0.46, which is a "caution" level, but the increase rate over the past month is high and there is a high possibility of entering a "warning" level in the future, a "request for inspection" flag can be generated instead of a "regular inspection" flag. In this way, the maintenance flag can be an action-direction type of information that suggests the actual direction of action for the manager, rather than a simple translation of the status level.

[0072] The maintenance priority generation unit (330) can generate maintenance priorities for multiple damaged objects or multiple management units by reflecting at least one of the risk level, damage grade, location of damage, and importance of structural members. For example, even if damage has the same risk score, damage occurring to structurally important members such as piers, girders, main towers, or deck slabs may have a higher maintenance priority. Additionally, if the damage is located in an area where traffic loads are concentrated, a connection, a vulnerable section, or a location where deterioration occurs rapidly during repeated inspections, the priority may be adjusted upward.

[0073] The maintenance priority generation unit (330) can generate a final priority score by combining a risk score, a structural member importance weight, and a recent change rate weight. For example, crack A can be calculated with a final priority score of 54 points by applying a risk score of 0.62, a member importance score of 0.8, and a change rate weight of 1.1, and rebar exposure B can be calculated with a final priority score of 82 points by applying a risk score of 0.79, a member importance score of 1.0, and a change rate weight of 1.2. In this case, the central management server (300) can set rebar exposure B as a priority response target over crack A.

[0074] Priorities can be generated not only at the level of damaged objects but also at the level of management units. For example, if multiple damages are concentrated within the same span, a management unit-level priority can be generated for the entire span.

[0075] The maintenance priority generation unit (330) can calculate the final maintenance priority by comprehensively reflecting the current risk level, damage grade, damage progression trend, importance of structural members, damage density within the same area, and traffic or operational impact if necessary, rather than simply listing priorities for each individual damaged object. For example, the priority score can be calculated by multiplying the risk score by the member importance weight and progression trend weight, or by adding a damage cluster correction value to it.

[0076] In one embodiment, the maintenance priority generation unit (330) can perform a calculation for individual damage in the form of [Priority Score = Risk × Importance Weight × Change Rate Weight]. Accordingly, damage that rapidly deteriorates in structurally important members, even if the current risk is somewhat low, may be assigned a high priority. Conversely, damage that is currently high but is limited to non-structural parts and progresses slowly may be assigned a relatively low priority.

[0077] For example, damage to exposed rebar located at the bottom of the main girder of a bridge can be calculated with a high priority score by applying a risk level of 0.79, a member importance weight of 1.0, and a change rate weight of 1.2, while at the same time, minor cracks on the side of the abutment can have a relatively low priority score by applying a risk level of 0.52, a member importance weight of 0.7, and a change rate weight of 1.0. Accordingly, the central management server (300) may propose the former as a priority repair or detailed inspection target rather than taking action on the latter first.

[0078] In addition, the maintenance priority generation unit (330) can generate priorities at the level of management units as well as at the level of damaged objects. For example, if multiple cracks, spalling, and rebar exposures are concentrated within the same span or the same girder section, the cumulative risk of the entire management unit, as well as the individual score of each damaged object, can be calculated to generate a priority at the management unit level, such as "P3-P4 span / G2 girder section priority inspection." This allows the manager to establish a maintenance plan by considering both individual damages and the risk of the entire section together.

[0079] The damage progression trend analysis unit (340) can calculate the amount of damage change or the amount of risk change by comparing past inspection data and current inspection data for the same location or the same management unit. For example, if the width of a specific crack is 0.25 mm at the time of past inspection and 0.42 mm at the time of current inspection, the amount of increase and the rate of increase can be calculated. In addition, if the risk score increases from 0.45 in the past to 0.78 in the present, the amount of change can also be calculated.

[0080] The damage progression trend analysis unit (340) may raise the maintenance priority or generate a separate warning notification if the amount of damage change or the amount of risk change exceeds a preset threshold. For example, if the current absolute risk level is at the "caution" level but the risk score has rapidly increased from 0.21 to 0.34 to 0.49 in the last three inspections, the damage progression trend analysis unit (340) may raise the maintenance priority of the damage and generate a "request inspection" or "review early repair" notification.

[0081] In this way, the central management server (300) can make a judgment that reflects not only the current absolute risk but also the speed of progression of damage, so it is possible to preemptively manage damage that has not yet reached a threshold.

[0082] The notification display unit (350) can display the damage detection result, damage dimension information, risk level, and judgment result in the form of a notification on the administrator terminal (400). For example, if a specific damage is classified as "repair needed," the notification display unit (350) can display the location, type, risk level, grade, and related image information of the damage in the form of a popup or a list.

[0083] For example, a summary notification such as "G2 girder, risk of rebar exposure, risk level 0.86, repair needed" can be immediately provided to the manager terminal (400). In addition, the notification display unit (350) can generate a collective notification in addition to individual notifications to draw the manager's attention when multiple damages are detected consecutively in the same management unit within a certain period of time.

[0084] The decision support unit (360) can support a manager in making maintenance decisions regarding how to respond to damaged areas in order and manner by displaying damage locations on a map or schematic diagram of the structure and visually displaying damage grade, risk score, maintenance flag, maintenance priority and damage progression information corresponding to each damage location.

[0085] The decision support unit (360) may provide a map interface that displays the location of damage in the form of a marker on a map or schematic of a structure. Each marker may be displayed in a different color or icon depending on the maintenance flag or damage grade. For example, "repair needed" may be visualized as a red circular marker, "inspection required" as a yellow triangular marker, and "regular inspection" as a blue square marker.

[0086] Additionally, the decision support unit (360) may provide a list-type interface that displays ranks, locations, damage grades, risk scores, and maintenance flags in a table format for multiple damaged objects or multiple management units. For example, it may be displayed in a sorted format such as: 1st priority: "P3-P4 span, G2 girder, rebar exposed, risk 0.86, repair needed", 2nd priority: "P2 pier, crack, risk 0.64, inspection required".

[0087] In addition, when a specific damaged object is selected, the decision support unit (360) can provide a detailed screen showing the past inspection history, current status, change trends, and recommended action methods for the damage, such as immediate repair, repair after detailed inspection, and shortening the inspection cycle. Through this, the manager can not only see whether damage exists but also determine what actual action to take and when.

[0088] The decision support unit (360) is not limited to a screen configuration that simply lists and displays information, but can present actionable decision scenarios to the manager by linking maintenance flags, priorities, and progress trend analysis results generated by the central management server (300). For example, regarding a specific damaged object, it can provide one or more of the following response methods in the form of recommendations: "immediate repair," "detailed inspection within one month," "shortening of regular inspection cycle," and "review for inclusion in next budget."

[0089] Additionally, the decision support unit (360) can be configured to allow the manager terminal (400) to selectively display detailed information by damage location. For example, if a specific red marker is selected on a map interface, information on the span or girder where the damage is located, the history of the last three inspections, the current risk level, the amount of change in the damage dimensions, a maintenance flag, a priority score, and recommended action methods can be displayed together in the form of a detailed panel. Accordingly, the manager can view the basis information necessary for making decisions on actual maintenance execution, beyond simple visual warnings, on a single screen.

[0090] As another example, the decision support unit (360) can provide a comparison screen that selects multiple damaged objects and compares them. For example, by displaying a comparison of the risk level, damage change rate, importance of absence, and expected action method of the first priority damage and the second priority damage, it can support the manager in making a more rational judgment on which damage to respond to first in a situation of limited resources.

[0091] Furthermore, the decision support unit (360) can reflect the results of the budget judgment unit (390) or the maintenance work package generation unit (395) described later, and present not only recommendations for a single damage criterion but also execution plans at the management unit level or work package level. For example, execution recommendations such as "Initiate Package 1 first," "Review Package 2 next budget," and "Re-evaluate Package 3 after detailed inspection" may be displayed together. Accordingly, the present invention can function not as a simple monitoring and alarm system, but as a decision system that supports the establishment of actual maintenance plans.

[0092] The data storage unit (370) can store damage detection results, damage dimension information, risk level, judgment result, maintenance flag, maintenance priority, and inspection history data. It can also store structure identifiers, member information, damaged object identifiers, inspection times, related images, and manager action history. For example, the data storage unit (370) can store "initial detection time," "recent inspection time," "recent 3 measurements," "recent 3 risk levels," "recently assigned maintenance flag," and "manager action status" for a specific damaged object. Accordingly, the damage progression trend analysis unit (340) and the decision support unit (360) can perform an analysis connecting the past and the present.

[0093] The central management server (300) may optionally further include a location mapping unit (380).

[0094] The location mapping unit (380) can convert the damaged location into at least one management unit location information among the spans, girders, piers, main towers, decks, panels, or survey units of a structure using at least one of GPS coordinates, structural drawing data, and inspection path metadata. For example, if the damaged object is converted into a management unit such as "P3-P4 span," "G2 girder," or "5th panel" rather than simple GPS coordinates, the manager can identify the damaged location based on criteria that match the actual maintenance practice system. In particular, even when GPS errors exist or a drone flies over a section where GPS reception is unstable, such as the lower part of a structure, the damaged location can be more accurately identified based on the management unit by combining vision-based local coordinates and drawing data.

[0095] The central management server (300) may further include a budget judgment unit (390).

[0096] The budget judgment unit (390) receives information on the budget or resources available for maintenance and can determine whether maintenance is possible, whether inspection is a priority, or whether budget adjustment is necessary by comparing it with the necessity of action according to the maintenance priority. For example, even if multiple damages are simultaneously classified as "needs maintenance," the budget judgment unit (390) can allocate the budget sequentially starting from the damages with the highest priority, and for damages exceeding the budget range, provide a judgment result such as reflecting in the next budget, deferring after priority inspection, or requiring an urgent budget review. For example, within the first budget range, it can determine that the first and second ranked damages are capable of immediate maintenance, and for damages ranked third or lower, it can propose "next budget review after detailed inspection."

[0097] The budget judgment unit (390) can generate a budget execution scenario by considering not only the current budget and the estimated maintenance costs, but also the priority calculated by the maintenance priority generation unit (330) and the package unit of the maintenance work package generation unit (395). For example, while maintenance costs may be dispersed and inefficient for individual damaged objects, the total budget requirement can be reduced by grouping spatially adjacent damaged objects that can be processed by the same equipment into a package and taking action simultaneously. Accordingly, the budget judgment unit (390) can provide a judgment result that considers not only the feasibility of repair based on simple individual damage criteria but also the budget efficiency at the package level.

[0098] For example, damages A, B, and C included in Package 1 would cost a total of 12 million won each if repaired individually, but could be repaired for a total of 9 million won if repaired collectively using the same equipment and method. In this case, the budget judgment unit (390) recommends Package 1 as a priority repair target and can generate a judgment result that reflects the execution efficiency of the package unit rather than the ranking of individual damage units.

[0099] The central management server (300) may optionally further include a maintenance task package creation unit (395).

[0100] The maintenance work package generation unit (395) can group multiple damaged objects into one maintenance work package based on at least one of the spatial proximity of the location of the damage, the sameness or similarity of the type of damage, and the commonality of applicable repair methods and equipment.

[0101] For example, if multiple peeling and cracking damages located close to each other within the same span can be treated with the same access equipment and repair method, they can be grouped into a single work package. Subsequently, the maintenance work package generation unit (395) can generate maintenance priorities and implementation recommendations at the package level by reflecting the cumulative risk, damage progression trend, importance of structural members, and estimated repair resource requirements for each package.

[0102] More specifically, if individual damages A, B, and C can be resolved within one day using the same work vehicle and method, they can be grouped into a single package and presented as "Package 1: Immediate Repair Recommended." Conversely, if damage D requires different access equipment or traffic control, it can be classified as "Package 2: Separate Planning Required." Accordingly, managers can establish maintenance execution plans at the package level, which is closer to actual on-site work units, rather than at the individual damage level.

[0103] The maintenance work package generation unit (395) can generate a maintenance work package by not simply grouping spatially close damaged objects, but by also considering the accessible equipment for each damaged object, the estimated construction time, the sameness of the construction method, whether traffic control is necessary, and the conditions for ensuring the safety of workers. For example, even if they exist within the same span, if one damage can be accessed only by an aerial work vehicle and the other damage requires separate bridge underside inspection equipment, the two damages can be classified into different work packages.

[0104] Additionally, the maintenance work package generation unit (395) can calculate package priorities by reflecting the cumulative risk and progress trends, as well as the estimated maintenance resource requirements for each package. For example, Package A may be classified as "Immediate Repair Recommendation" because it has a high cumulative risk and is immediately accessible, allowing for action within a short period, while Package B may be proposed as "Include in Next Repair Plan After Detailed Inspection" because it has a somewhat lower cumulative risk but has the potential for rapid deterioration in the future.

[0105] Additionally, the administrator terminal (400) can provide a two-way communication interface with the central management server (300), such as inputting budget information, feedback on maintenance results, or manual adjustment of priorities.

[0106] As such, the present invention goes beyond the determination of individual damage units and can even provide implementation recommendations at the work package level suitable for actual on-site construction and maintenance work.

[0107] The edge computing-based structural maintenance decision support system using a drone according to the present invention can be applied in the same way to various structures that are difficult to access, such as bridges, the outer walls of large buildings, the surface of dams, and wind turbine towers.

[0108] The edge computing-based structural maintenance decision support system using a drone according to the present invention described above has been explained with reference to the attached drawings, but this is merely illustrative, and those skilled in the art will understand that various modifications and equivalent alternative embodiments are possible therefrom.

[0109] Therefore, the true scope of technical protection of the present invention should be determined solely by the technical concept of the appended claims. Explanation of the symbols

[0110] 100: Inspection Drone 110: Drone body 120: Camera 130: GPS 140: Attitude sensor 150: Distance measuring sensor 160: Data Collection Interface 200: Edge computing devices 210: Damage detection unit 220: Dimension Calculation Section 230: Risk Calculation Unit 240: Judgment Department 300: Central Administration Server 310: Result receiving unit 320: Maintenance flag creation section 330: Maintenance Priority Generation Section 340: Damage Progression Trend Analysis Department 350: Notification display 360: Decision Support Department 370: Data storage unit 380: Location mapping section 390: Budget Judgment Department 395: Maintenance Task Package Creation Section 400: Administrator Terminal

Claims

Claim 1 An inspection drone comprising a drone body, a camera installed on the drone body for acquiring a surface image of a structure to be inspected, a position sensor and an attitude sensor for measuring the position and attitude of the drone body, a distance measuring sensor for measuring the distance to the surface of the structure to be inspected, and a data collection interface for collecting image and measurement data acquired from the camera and sensors and transmitting them to an external device; an edge computing device comprising a damage detection unit for detecting at least one damaged object among cracks, spalling, rebar exposure, steel exposure, and cross-sectional loss of a structure based on image data received from the inspection drone, a dimension calculation unit for converting the dimension information of pixels within the image corresponding to the detected damaged object into actual physical defect dimensions by reflecting the attitude information of the inspection drone and the distance information to the structure to be inspected, a risk calculation unit for calculating the risk of damage using the physical defect dimensions and weights by damage type, and a judgment unit for classifying the damage grade into multiple stages of normal, caution, warning, and danger according to the calculated risk and determining whether maintenance is required; and a central management server linked with the edge computing device.The central management server comprises: a result receiving unit that receives damage detection results, damage dimension information, risk level, and judgment results transmitted from the edge computing device; a maintenance flag generating unit that generates at least one maintenance flag among repair necessity, inspection request, regular inspection, and continued observation for the structure subject to inspection based on the judgment results; a maintenance priority generating unit that generates maintenance priorities for a plurality of damaged objects or a plurality of management unit sections by reflecting at least one of the risk level, damage grade, location, and importance of the structural member for the damaged object; a damage progression trend analysis unit that calculates a change in damage or a change in risk level by comparing past inspection data and current inspection data for the same location or the same management unit section; an notification display unit that displays the damage detection results, damage dimension information, risk level, and judgment results in the form of a notification to an administrator terminal; a decision support unit that displays the location of the damaged object on a structural map or schematic diagram, and visually displays the damage grade, risk score, maintenance flag, maintenance priority, and damage progression trend information corresponding to the location of each damaged object, thereby supporting the administrator in making maintenance decisions regarding the order and method of priority response to the damaged area; and the damage detection results, damage dimensions An edge computing-based structural maintenance decision support system utilizing a drone, comprising a data storage unit that stores information, risk, judgment results, maintenance flags, maintenance priorities, and inspection history data, wherein the decision support unit provides a map-type interface that visually displays damage locations on a structural map or schematic diagram, and a maintenance priority list-type interface that includes rankings, locations, damage grades, risk scores, and maintenance flags for a plurality of damaged objects or a plurality of management unit sections. Claim 2 An edge computing-based structural maintenance decision support system utilizing a drone, characterized in that, in claim 1, the central management server further includes a position mapping unit that converts the location of the damaged object into location information of at least one management unit section among a span, girder, pier, main tower, deck plate, panel, or survey unit of the structure to be inspected using at least one of GPS coordinates, vision-based local coordinates, structural drawing data, and inspection path metadata. Claim 3 An edge computing-based structural maintenance decision support system utilizing a drone, characterized in that, in claim 1, the damage progression trend analysis unit calculates a change in damage or a change in risk by comparing past inspection data and current inspection data for the same location or the same management unit section, and if the change in damage or the change in risk exceeds a preset threshold, the maintenance priority is raised or a warning notification is generated. Claim 4 delete Claim 5 ◈Claim 5 was abandoned upon payment of the registration fee.◈ An inspection drone comprising: a drone body; a camera installed on the drone body for acquiring a surface image of a structure to be inspected; a position sensor and an attitude sensor for measuring the position and attitude of the drone body; a distance measuring sensor for measuring the distance to the surface of the structure to be inspected; and a data collection interface that collects image and measurement data acquired from the camera and sensors and transmits them to an external device; an edge computing device comprising: a damage detection unit that detects at least one damaged object among cracks, spalling, rebar exposure, steel exposure, and cross-sectional loss of a structure based on image data received from the inspection drone; a dimension calculation unit that converts the dimension information of pixels within the image corresponding to the detected damaged object into actual physical defect dimensions by reflecting the attitude information of the inspection drone and the distance information to the structure to be inspected; a risk calculation unit that calculates the risk of damage using the physical defect dimensions and weights by damage type; and a judgment unit that classifies the damage grade into multiple stages of normal, caution, warning, and danger according to the calculated risk and determines whether maintenance is required; and a central management server linked with the edge computing device;The central management server comprises: a result receiving unit that receives damage detection results, damage dimension information, risk level, and judgment results transmitted from the edge computing device; a maintenance flag generating unit that generates at least one maintenance flag among repair necessity, inspection request, regular inspection, and continued observation for the structure subject to inspection based on the judgment results; a maintenance priority generating unit that generates maintenance priorities for a plurality of damaged objects or a plurality of management unit sections by reflecting at least one of the risk level, damage grade, location, and importance of the structural member for the damaged object; a damage progression trend analysis unit that calculates a change in damage or a change in risk level by comparing past inspection data and current inspection data for the same location or the same management unit section; an notification display unit that displays the damage detection results, damage dimension information, risk level, and judgment results in the form of a notification to an administrator terminal; a decision support unit that displays the location of the damaged object on a structural map or schematic diagram, and visually displays the damage grade, risk score, maintenance flag, maintenance priority, and damage progression trend information corresponding to the location of each damaged object, thereby supporting the administrator in making maintenance decisions regarding the order and method of priority response to the damaged area; and the damage detection results, damage dimensions An edge computing-based structural maintenance decision support system utilizing a drone, characterized by including a data storage unit that stores information, risk level, judgment result, maintenance flag, maintenance priority, and inspection history data, and further including a budget judgment unit in which the central management server receives information on available maintenance budget or resources and determines whether maintenance is possible, whether an inspection is a priority, or whether budget adjustment is necessary by comparing it with the necessity of action according to the maintenance priority. Claim 6 ◈Claim 6 was abandoned upon payment of the registration fee.◈ An inspection drone comprising: a drone body; a camera installed on the drone body for acquiring a surface image of a structure to be inspected; a position sensor and an attitude sensor for measuring the position and attitude of the drone body; a distance measuring sensor for measuring the distance to the surface of the structure to be inspected; and a data collection interface that collects image and measurement data acquired from the camera and sensors and transmits them to an external device; an edge computing device comprising: a damage detection unit that detects at least one damaged object among cracks, spalling, rebar exposure, steel exposure, and cross-sectional loss of a structure based on image data received from the inspection drone; a dimension calculation unit that converts the dimension information of pixels within the image corresponding to the detected damaged object into actual physical defect dimensions by reflecting the attitude information of the inspection drone and the distance information to the structure to be inspected; a risk calculation unit that calculates the risk of damage using the physical defect dimensions and weights by damage type; and a judgment unit that classifies the damage grade into multiple stages of normal, caution, warning, and danger according to the calculated risk and determines whether maintenance is required; and a central management server linked with the edge computing device; and The central management server comprises: a result receiving unit that receives damage detection results, damage dimension information, risk level, and judgment results received from the edge computing device; a maintenance flag generating unit that generates at least one maintenance flag among repair need, inspection request, regular inspection, and continued observation for the structure subject to inspection based on the judgment result; a maintenance priority generating unit that generates maintenance priorities for a plurality of damaged objects or a plurality of management unit sections by reflecting at least one of the risk level, damage grade, location, and importance of the structural member for the damaged object; and a damage progression trend analysis unit that calculates a change in damage amount or a change in risk amount by comparing past inspection data and current inspection data for the same location or the same management unit section.An edge computing-based structural maintenance decision support system utilizing a drone, comprising: an alert display unit that displays the damage detection results, damage dimension information, risk level, and judgment results in the form of an alert to an administrator terminal; a decision support unit that displays the location of a damaged object on a structural map or schematic diagram and visually displays damage grade, risk score, maintenance flag, maintenance priority, and damage progression trend information corresponding to the location of each damaged object to support the administrator in making maintenance decisions regarding the order and method of priority response to the damaged area; and a data storage unit that stores the damage detection results, damage dimension information, risk level, judgment results, maintenance flag, maintenance priority, and inspection history data; wherein the central management server clusters the plurality of damaged objects into a single maintenance work package based on at least one of the spatial proximity of the damage location, the identity or similarity of the damage type, and the commonality of applicable repair methods and equipment for the plurality of damaged objects, and generates package-unit maintenance priorities and execution recommendations by reflecting the cumulative risk level, damage progression trend, importance of structural members, and estimated repair resource requirements for each maintenance work package.

Images