Intelligent Decision-Making Method and System for Road-Air Integrated Highway Maintenance Operations

By using an intelligent decision-making method for integrated road-air highway maintenance operations, and utilizing coded optical anchor point patterns to achieve collaborative inspections at both the road and air ends, the problems of unstable target positioning and inconsistent data have been solved, improving inspection efficiency and decision reliability, and realizing continuous support from inspection to decision-making.

CN123070567APending Publication Date: 2026-11-10浙江省机电设计研究院有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611234046.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-14
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

In existing highway maintenance and inspection, the operational processes of roadside inspection and aerial drone inspection are inconsistent, the target positioning is unstable, the data format is not uniform, and it is difficult to form a structured decision that can be directly dispatched, which affects the efficiency of maintenance response and the rationality of resource allocation.

Method used

An intelligent decision-making method for highway maintenance operations integrating road and air is adopted. The roadside inspection platform generates and projects coded optical anchor point patterns, which are then identified and decoded by the air-end UAV platform to determine the relative pose and scale relationship. The decision-making platform performs intelligent evaluation and outputs the maintenance operation decision results.

Benefits of technology

It improves the reliability and efficiency of air-rail collaborative operations, ensures the standardization and reproducibility of supplementary data collection, achieves continuous support from inspection to decision-making, and enhances the timeliness of maintenance response and the rationality of resource allocation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123070567A_ABST
    Figure CN123070567A_ABST
Patent Text Reader

Abstract

This invention relates to the field of highway maintenance technology, specifically to an intelligent decision-making method and system for integrated road-air highway maintenance operations. In this invention, a roadside inspection platform acquires first inspection data such as roadside images, point clouds, and positioning attitudes to determine suspected maintenance target areas. It then projects coded optical anchor point patterns carrying acquisition modes and scale levels near the targets. An air-based UAV platform acquires and decodes image data containing the anchor points, calculates relative pose and scale relationships based on the anchor point geometric features, and performs supplementary acquisition according to the defined mode and scale constraints to form supplementary data. The decision-making platform intelligently evaluates the supplementary data to obtain defect information, outputs treatment priorities and treatment instructions, and can generate evidence packages and completion confirmations to reduce redundant supplementary acquisitions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of highway maintenance technology, specifically to an intelligent decision-making method and system for integrated road and air highway maintenance operations. Background Technology

[0002] With the continuous growth of my country's highway mileage and the increasing traffic load, highway maintenance has gradually shifted from relying on manual inspections and experience-based judgment to mechanization, informatization, and intelligence. Traditional methods often involved maintenance personnel periodically inspecting and recording road defects on foot or by vehicle, combined with paper records or simple information systems for handling, which limited efficiency and timeliness. Subsequently, vehicle-mounted cameras, laser measurement, positioning and navigation, and geographic information systems were gradually introduced, forming a semi-automatic detection and defect database management system based on inspection vehicles. In recent years, driven by the development of artificial intelligence recognition, multi-source sensing, and low-altitude unmanned technology, drones have been applied to the inspection of complex structures such as bridges, slopes, and tunnel entrances. Some regions have begun to explore collaborative operations between inspection vehicles and drones to improve coverage and operational safety, promoting the evolution of highway maintenance towards refined, predictive, and platform-based decision-making.

[0003] Current highway maintenance inspection and decision-making systems still suffer from several shortcomings. First, roadside inspections and aerial drone inspections often employ different operational procedures and data formats. Task handover relies on coordinate commands, verbal communication, or experience-based point selection, leading to unstable target positioning, especially in scenarios prone to drift such as under bridges, tunnel entrances, and obstructed areas, resulting in difficulties in finding points or misalignment. Second, the shooting angles, distances, and resolutions of drone-based supplementary data collection lack uniform constraints, causing inconsistencies in the scale of the same defects and making evidence incomparable, hindering subsequent severity assessments and prioritization. Third, while inspection data is large, effective evidence is insufficient, making it difficult to generate structured decision-making outputs that can be directly dispatched, impacting maintenance response efficiency and the rationality of resource allocation. Therefore, an integrated road-air intelligent decision-making method and system for highway maintenance operations is needed to address these issues. Summary of the Invention

[0004] (I) Technical problems to be solved: In view of the shortcomings of the existing technology, the present invention provides an intelligent decision-making method and system for integrated road and air highway maintenance operations, which solves the above-mentioned problems.

[0005] (II) Technical Solution: To achieve the above objectives, the present invention provides the following technical solution: a road-air integrated intelligent decision-making method for highway maintenance operations, applied to a road-air collaborative system including a roadside inspection platform, an airborne unmanned aerial vehicle (UAV) platform, and a decision-making platform, wherein the method includes: S1. The roadside inspection platform acquires the first inspection data of the road to determine the target area of ​​the suspected maintenance target, and generates a coded optical anchor pattern corresponding to the target area. The coded optical anchor pattern carries at least the acquisition mode and scale level. The roadside inspection platform projects the coded optical anchor pattern on the projection plane near the target area. S2. The air-to-ground UAV platform acquires image data containing the coded optical anchor point pattern, identifies and decodes the coded optical anchor point pattern to obtain the acquisition mode and the scale level; determines the relative pose and scale relationship of the air-to-ground UAV platform relative to the projection plane based on the geometric features of the coded optical anchor point pattern, and performs supplementary acquisition on the suspected maintenance target according to the acquisition mode and the relative pose and scale relationship to obtain supplementary acquisition data; S3. The decision-making platform performs intelligent evaluation of the suspected maintenance targets based on the supplementary data to obtain disease information, and outputs maintenance operation decision results based on the disease information. The maintenance operation decision results include at least the treatment priority and treatment instructions.

[0006] Furthermore, the first inspection data includes at least one or more of roadside image data, point cloud data, positioning data, and attitude data. The roadside inspection platform determines the road location description information of the target area based on the first inspection data. The road location description information includes one or more of the following: road number, driving direction, lane identification, mileage marker, or geographic coordinates.

[0007] Furthermore, in addition to carrying the acquisition mode and the scale level, the coded optical anchor pattern also carries target identification information corresponding to the target area, and carries one or more of the following: priority, timestamp, validity period, and verification information.

[0008] Furthermore, the roadside inspection platform projects the coded optical anchor pattern onto the projection plane using near-infrared projection, and the projection duration of the coded optical anchor pattern is a preset duration, which is 1 to 5 seconds.

[0009] Furthermore, the projection plane can be any one of the following: the surface of road ancillary facilities, the surface of bridge components, the surface of tunnel ancillary components, or the surface of a vehicle-mounted calibration plate. When there is no available projection plane near the target area, the roadside inspection platform unfolds the vehicle-mounted calibration plate and projects the coded optical anchor pattern onto the surface of the vehicle-mounted calibration plate.

[0010] Furthermore, the coded optical anchor point pattern has a preset geometric size and includes a set of feature points for pose calculation. The set of feature points includes at least four corner features. The air-end UAV platform determines the relative pose based on the corner features and determines the scale relationship based on the preset geometric size. It also determines the execution parameters for supplementary acquisition based on the scale relationship. The execution parameters include at least one or more of the following: flight distance, flight altitude, gimbal pitch angle, and shooting angle.

[0011] Furthermore, the acquisition mode includes one or more of orthophoto supplementary acquisition, side-view supplementary acquisition, and close-range detail supplementary acquisition. The scale level corresponds to the target pixel scale range or the target shooting distance range. The air-end UAV platform performs the supplementary acquisition when the range corresponding to the scale level is met.

[0012] Furthermore, the air-end UAV platform binds the supplementary data with the decoded acquisition mode, scale level, and relative pose or scale relationship to generate an evidence package and sends it to the decision platform. The evidence package includes at least one or more of the following: target identification information, keyframe index, pose or scale summary, and verification information.

[0013] Furthermore, after completing the supplementary data collection, the air-end UAV platform sends a completion confirmation message, which is either an optical confirmation code or a wireless confirmation message. After receiving the completion confirmation message, the roadside inspection platform marks the target area as having been supplemented to avoid repeatedly triggering supplementary data collection for the same target area.

[0014] This invention also provides an intelligent decision-making system for integrated road-air highway maintenance operations, used to execute the intelligent decision-making method for integrated road-air highway maintenance operations, including a roadside inspection platform, an airside unmanned aerial vehicle (UAV) platform, and a decision-making platform, wherein: The roadside inspection platform includes a first inspection data acquisition module, a target area determination module, and an anchor point pattern generation and projection module. It is used to acquire the first inspection data of the road and determine the target area of ​​the suspected maintenance target based on the first inspection data, generate a coded optical anchor point pattern corresponding to the target area, and project the coded optical anchor point pattern on a projection plane near the target area. The coded optical anchor point pattern carries at least the acquisition mode and scale level. The air-to-ground unmanned aerial vehicle (UAV) platform includes an image data acquisition module, an anchor point recognition and decoding module, a pose and scale determination module, and a supplementary acquisition module. It is used to acquire image data containing the coded optical anchor point pattern, identify and decode the coded optical anchor point pattern to obtain the acquisition mode and the scale level, determine the relative pose and scale relationship of the air-to-ground UAV platform relative to the projection plane based on the geometric features of the coded optical anchor point pattern, and perform supplementary acquisition on the suspected maintenance target according to the acquisition mode and the relative pose and scale relationship to obtain supplementary acquisition data. The decision-making platform includes an intelligent assessment module and a decision output module, which are used to intelligently assess the suspected maintenance targets based on the supplementary data to obtain disease information, and output maintenance operation decision results based on the disease information. The maintenance operation decision results include at least the treatment priority and treatment instructions.

[0015] (III) Beneficial Effects: Compared with the prior art, the present invention provides an intelligent decision-making method and system for integrated road and air highway maintenance operations, which has the following beneficial effects: 1. This road-air integrated intelligent decision-making method and system for highway maintenance operations projects coded optical anchor point patterns with acquisition modes and scale settings near suspected maintenance targets by a roadside inspection platform. This allows the air-based UAV platform to directly identify and decode the anchor point patterns in the acquired image data to obtain a clear intention for supplementary acquisition. This transforms the traditional task handover, which relies on coordinate commands or manual communication, into on-site perceptible and executable instructions. This significantly reduces the risks of unclear target orientation, difficulty in finding points, and misalignment during the collaborative process. Especially in environments where positioning is prone to drift, such as under bridges, tunnel entrances, and obstructed areas, it can still achieve stable triggering and consistent execution, improving the reliability and operational efficiency of road-air collaboration.

[0016] 2. The intelligent decision-making method and system for integrated road and air highway maintenance operations determines the relative pose and scale relationship of the airborne UAV platform relative to the projection plane by utilizing the geometric features of the coded optical anchor point pattern. Based on the acquisition mode and the relative pose and scale relationship, supplementary acquisition is performed on suspected maintenance targets. This gives the UAV supplementary acquisition process quantifiable attitude and scale constraints, avoiding the problem of incomparable evidence caused by inconsistent shooting angles, distances and resolutions under different times or operating conditions. This improves the standardization and reproducibility of the supplementary acquisition data, facilitating the accurate implementation of subsequent severity assessment, risk classification and priority ranking.

[0017] 3. The intelligent decision-making method and system for integrated road-air highway maintenance operations inputs supplementary data obtained by the air-end UAV platform under the above constraints into the decision-making platform for intelligent evaluation and outputs maintenance operation decision results containing disposal priorities and disposal instructions. This transforms the inspection results from scattered raw data or image materials into structured decision information that can be directly dispatched, reducing the burden of manual screening and secondary verification in massive inspection data, improving the timeliness of maintenance response and the rationality of resource allocation, and thus realizing continuous support capabilities from road-air collaborative inspection to maintenance operation decision-making. Attached Figure Description

[0018] Figure 1 A flowchart illustrating the intelligent decision-making method for integrated road and air highway maintenance operations provided by this invention; Figure 2 This is a schematic diagram of the module structure of the intelligent decision-making system for integrated road and air highway maintenance operations provided by the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] To enable those skilled in the art to better understand the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Please see Figure 1 , Figure 1 A flowchart illustrating the intelligent decision-making method for integrated road and air highway maintenance operations provided by this invention; Figure 2 This is a schematic diagram of the module structure of the road-air integrated intelligent decision-making system for highway maintenance operations provided by the present invention. The road-air integrated intelligent decision-making method for highway maintenance operations is applied to a road-air collaborative system including a roadside inspection platform, an airborne unmanned aerial vehicle (UAV) platform, and a decision-making platform. The method includes: S1. The roadside inspection platform acquires the first inspection data of the road to determine the target area of ​​the suspected maintenance target, and generates a coded optical anchor pattern corresponding to the target area. The coded optical anchor pattern carries at least the acquisition mode and scale level. The roadside inspection platform projects the coded optical anchor pattern on the projection plane near the target area. S2. The air-to-ground UAV platform acquires image data containing the coded optical anchor point pattern, identifies and decodes the coded optical anchor point pattern to obtain the acquisition mode and the scale level; determines the relative pose and scale relationship of the air-to-ground UAV platform relative to the projection plane based on the geometric features of the coded optical anchor point pattern, and performs supplementary acquisition on the suspected maintenance target according to the acquisition mode and the relative pose and scale relationship to obtain supplementary acquisition data; S3. The decision-making platform performs intelligent evaluation of the suspected maintenance targets based on the supplementary data to obtain disease information, and outputs maintenance operation decision results based on the disease information. The maintenance operation decision results include at least the treatment priority and treatment instructions.

[0022] In this embodiment, the method operates in a road-air collaborative system consisting of a roadside inspection platform, an airborne UAV platform, and a decision-making platform. Its basic principle is to transform the supplementary collection task of suspected maintenance targets from traditional coordinate command-based collaboration to on-site imageable coded optical anchor point interactive collaboration. The anchor point pattern simultaneously carries the collection intention and geometric calibration benchmark, enabling the UAV to have consistent perspective and scale constraints during supplementary collection, thereby providing comparable and reproducible evidence data for subsequent intelligent assessment and operational decision-making.

[0023] On the roadside inspection platform, the first step is to detect suspected maintenance targets and determine the target area. The roadside inspection platform can be an inspection vehicle or road inspection device equipped with multiple sensors, acquiring initial inspection data through one or more of the following: vehicle-mounted cameras, LiDAR, and positioning and attitude sensors. This initial inspection data is used for rapid inspection and coarse-grained identification of the road surface and its ancillary structures, such as detecting suspected potholes, cracks, subsidence, abnormal expansion joints, missing guardrails, and traces of rockfall on slopes. Based on the initial inspection data, the roadside inspection platform determines the target area of ​​the suspected maintenance target. The target area can be described using road number, driving direction, lane identification, mileage marker, or geographic coordinates, and a corresponding identifier is assigned to the target area within the system for subsequent association. After the target area is determined, the roadside inspection platform generates a coded optical anchor point pattern corresponding to the target area and projects this pattern onto a projection plane near the target area. The coded optical anchor point pattern carries at least the acquisition mode and scale information, where the acquisition mode indicates the type of image captured by the drone, and the scale level constrains the clarity or distance level of the captured image. The acquisition mode can be set according to road structure and defect type, such as orthophoto acquisition, side-view acquisition, or close-range detail acquisition. The scale level can be set to high precision, medium precision, quick overview, etc., or mapped to the target pixel scale range and target shooting distance range. The projection plane is preferably selected at a location that is easy for the UAV to image and will not affect traffic safety, such as the surface of road ancillary facilities, bridge components, tunnel ancillary components, or the surface of a vehicle-mounted calibration board carried by a roadside inspection platform. When there is no projection plane that meets the imaging conditions near the target area, the vehicle-mounted calibration board can be deployed as a temporary projection plane. To reduce ambient light interference and minimize the impact on the driver's line of sight, near-infrared projection is preferred, and the projection duration can be set to 1 to 5 seconds to ensure that the UAV can stably capture and decode in the flight video stream.

[0024] On the airborne UAV platform side, anchor point recognition and decoding, pose and scale determination, and supplementary data acquisition are performed. The airborne UAV platform acquires image data containing the coded optical anchor point pattern through its onboard camera. The image data can be a video stream or a continuous frame sequence. The UAV identifies and decodes the anchor point pattern in the image data to obtain the acquisition mode and scale level, thereby clarifying the supplementary acquisition intention and quality requirements. Furthermore, in addition to carrying coded information, the anchor point pattern also has geometric features for pose calculation, such as corner features and preset geometric dimension boundaries. The airborne UAV platform determines its relative pose and scale relationship with respect to the projection plane based on these geometric features. The relative pose characterizes the relative distance, yaw, and pitch relationship of the UAV camera relative to the projection plane, while the scale relationship characterizes the mapping between image pixels and actual length. Based on this relative pose and scale relationship, the UAV can calculate the execution parameters for supplementary acquisition and make adaptive adjustments to ensure that the actual acquisition meets the scale level requirements. Execution parameters may include flight distance, flight altitude, gimbal pitch angle, shooting angle, and number of shots, etc. Subsequently, the drone, based on the acquisition mode and relative pose and scale relationships, conducts supplementary data acquisition on suspected maintenance targets, obtaining supplementary data. The supplementary data can be orthophoto sequences, side-view image sequences, or sets of close-range keyframes, and can also be expanded to include point clouds or multispectral images. Taking bridge pier cracks as an example, after the roadside inspection platform discovers a suspected crack in the area under the bridge, it projects anchor points and specifies side-view supplementary acquisition and high-precision scale settings. After identifying the anchor points, the drone automatically adjusts to the shooting distance and gimbal angle that meet the scale settings, acquiring continuous side-view sequences along the crack's direction, thus providing quantifiable data on the crack's length, direction, and width. As another example, in the case of suspected potholes on highway surfaces, the roadside inspection platform specifies orthophoto supplementary acquisition and medium-precision scale settings. The drone performs overhead acquisition at a height that meets the scale constraints, providing a stable scale reference for pothole boundary and area estimation, reducing incomparability issues between different batches of data.

[0025] On the decision-making platform side, intelligent assessment and operational decision output are performed. The decision-making platform can be deployed on a roadside server, cloud platform, or maintenance management center server. After receiving supplementary data, it performs intelligent assessment on suspected maintenance targets to obtain defect information. Defect information includes at least the defect type and severity or risk level, and may further include defect location description, geometric quantitative indicators, and credibility. Intelligent assessment can be achieved using image recognition, segmentation, 3D reconstruction, or rule-based assessment. The assessment model is not limited to a specific network structure. This embodiment emphasizes that the supplementary data has a higher degree of standardization under anchor point interaction and scale constraints, thereby improving assessment stability. The decision-making platform outputs maintenance operation decision results based on defect information. The maintenance operation decision results include at least the treatment priority and treatment instructions. The treatment priority can be generated by sorting according to severity or risk level. The treatment instructions may include recommended treatment methods, required resources, operation time windows, or safety control levels, which are used to directly drive maintenance dispatch or generate work orders. Through the above process, this embodiment realizes a continuous support link of road-end discovery and triggering, air-end supplementary sampling according to constraints, and decision-end evaluation and dispatching. It solves the problems of unstable target alignment, inconsistent supplementary sampling scale, and difficulty in outputting directly executable decisions in road-air coordination, enabling those skilled in the art to complete system construction and method implementation accordingly.

[0026] Furthermore, the first inspection data includes at least one or more of roadside image data, point cloud data, positioning data, and attitude data. The roadside inspection platform determines the road location description information of the target area based on the first inspection data. The road location description information includes one or more of the following: road number, driving direction, lane identification, mileage marker, or geographic coordinates.

[0027] In this embodiment, the first inspection data is used to support the rapid discovery of suspected maintenance targets, the location description of target areas, and the triggering of subsequent road-air collaborative tasks. The first inspection data includes at least one or more of the following: roadside image data, point cloud data, positioning data, and attitude data. The roadside image data can be acquired by the inspection vehicle's forward-facing camera, top-view camera, or multi-view camera, and is used to characterize two-dimensional features such as road surface texture, crack morphology, pothole boundaries, and abnormal appearance of road ancillary facilities. The point cloud data can be acquired by vehicle-mounted LiDAR or structured light depth camera, and is used to characterize the three-dimensional geometric morphology of road surface elevation undulations, subsidence steps, pothole depth trends, and structures such as guardrails. The positioning data can be acquired by a satellite positioning receiver, inertial navigation unit, or high-precision map matching module, and is used to output the geographical coordinates or mileage marker positions of the inspection vehicle. The attitude data can be acquired by an inertial navigation unit combining a gyroscope and an accelerometer, and is used to output the inspection vehicle's heading angle, pitch angle, and roll angle to assist in image stabilization, point cloud registration, or position correction.

[0028] During the target area determination process, the roadside inspection platform performs time alignment and coordinate association on the aforementioned first inspection data, mapping the detection results of suspected maintenance targets in the image or point cloud to road location description information. Specifically, when using mileage markers, the suspected maintenance target can be mapped to a combination of road number and mileage marker based on the positioning data and the road centerline mileage accumulation rules, and the driving direction and lane identification can be further recorded to clarify whether the target is located in the uphill or downhill direction and the inner or outer lane. When using geographic coordinates, latitude and longitude or planar coordinates can be directly output, while adding the road number and driving direction as semantic constraints to facilitate subsequent retrieval and dispatching. Lane identification can be determined based on the lane information of the inspection vehicle, lane line recognition results, or lane-level positioning results, for example, labeling the target area as the first lane, second lane, or emergency lane. Attitude data can be used to correct the position of the image detection box on curves, slopes, or bumpy road sections, or for extrinsic parameter compensation between the point cloud and the image, thereby improving the stability of the target area location description.

[0029] By fusing roadside image data, point cloud data, and positioning attitude data to form road location description information, the target area of ​​a suspected maintenance target can be transformed from a single visual coordinate into an engineering-executable road semantic location. This provides clear spatial constraints and retrieval basis for subsequent selection of projection planes near the target area, projection coding optical anchor point patterns, and supplementary data collection by airborne UAV platforms. Those skilled in the art can use the data sources and location description generation methods described in this embodiment to achieve stable determination and description of target areas under different road grades and different inspection equipment configurations.

[0030] Furthermore, in addition to carrying the acquisition mode and the scale level, the coded optical anchor pattern also carries target identification information corresponding to the target area, and carries one or more of the following: priority, timestamp, validity period, and verification information.

[0031] In this embodiment, the coded optical anchor pattern is used to establish a visual command carrier for the on-site target between the roadside inspection platform and the airborne UAV platform. The principle is to encode the task intent and target binding information required for supplementary data collection into an imageable pattern, and then anchor the pattern on a projection plane near the target area through projection. This allows the airborne UAV platform to directly identify and decode the image data, thereby completing the task relay and data alignment without relying on verbal commands or single coordinate commands.

[0032] Specifically, the coded optical anchor point pattern, in addition to carrying the acquisition mode and scale level, also carries target identification information corresponding to the target area. The target identification information is used to uniquely or quasi-uniquely identify the target area, avoiding false associations when multiple suspected points exist on the same road segment. The target identification information can be generated by the roadside inspection platform when determining the target area, for example, using a combination of session number and sequence number, road number and mileage marker, or a short code composed of a random number and a check digit. Preferably, the target identification information corresponds one-to-one with the target area entries recorded internally by the roadside inspection platform, enabling the supplementary data collected by the airborne UAV platform to be directly associated with the corresponding target after transmission.

[0033] The encoded optical anchor point pattern can also carry one or more of the following: priority, timestamp, validity period, and verification information, to improve the availability and robustness of air-to-ground interaction. Priority indicates the urgency of the re-sampling task, allowing the airborne UAV platform to queue or preemptively execute tasks in multi-target concurrent scenarios, such as prioritizing re-sampling of high-risk potholes or bridge cracks. The timestamp characterizes the anchor point generation time, facilitating time-series tracing and event replay after data transmission. The validity period limits the effective window of the anchor point command. After decoding, the airborne UAV platform checks the validity period; if it expires, execution is abandoned, thus avoiding mis-identification of old anchor point remnants or historical patterns leading to erroneous re-sampling. Verification information detects decoding errors. Verification information can take the form of cyclic redundancy check (CRC) codes or hash digests. After decoding, the airborne UAV platform verifies data consistency based on the verification information; if verification fails, the anchor point is discarded and the search continues, reducing the risk of misinterpretation due to ambient light interference, motion blur, etc.

[0034] In terms of pattern structure, the coded optical anchor pattern can adopt a two-dimensional coding structure with positioning borders and corner features to simultaneously meet the needs of information decoding and geometric feature extraction. Corner features can be used for subsequent pose scale determination, while positioning borders and error correction coding can improve recognition rates under tilted viewpoints or partial occlusion conditions. The coding capacity can be flexibly configured according to field combinations. When only the acquisition mode, scale level, and target identification information are needed, short codes can be used to improve recognition speed; when validity period and verification information are required, coding redundancy can be increased to enhance robustness. Through these methods, the coded optical anchor pattern achieves target binding, task intent transmission, and enhanced interaction reliability while ensuring feasibility.

[0035] Furthermore, the roadside inspection platform projects the coded optical anchor pattern onto the projection plane using near-infrared projection, and the projection duration of the coded optical anchor pattern is a preset duration, which is 1 to 5 seconds.

[0036] In this embodiment, the roadside inspection platform uses near-infrared projection to project coded optical anchor point patterns onto the projection plane. The principle behind this is to utilize the imaging characteristics of near-infrared light outside the visible light range, ensuring that the anchor point patterns can be stably captured by the imaging device of the airborne UAV platform while minimizing interference with the driver's line of sight and surrounding traffic participants. Near-infrared projection preferably operates in the 850 nm or 940 nm band. The airborne camera of the airborne UAV platform can use an image sensor with near-infrared response, or enhance the imaging of the anchor point patterns by removing the infrared cutoff filter or adding a narrowband filter with a matching band. In strong outdoor light environments, the narrowband filter can suppress visible light background, improve the contrast of the anchor point patterns, and enable the UAV to perform rapid detection and decoding even while in motion.

[0037] The projection duration of the encoded optical anchor point pattern is set to a preset duration of 1 to 5 seconds. This preset duration provides a sufficient observation window in dynamic scenarios involving the movement of the inspection vehicle and the flight of the drone, ensuring that the drone captures at least several consecutive frames from the video stream containing the complete anchor point pattern for recognition, decoding, and corner point extraction. Simultaneously, it avoids increased energy consumption or unnecessary light exposure to the environment due to prolonged projection. The preset duration can be determined by considering vehicle speed, drone camera frame rate, and processing latency of the recognition algorithm. For example, when the inspection vehicle travels at a speed of 20 to 60 kilometers per hour and the drone camera frame rate is 25 to 60 frames per second, a 1-second projection can generate 25 to 60 effective observation frames, which is usually sufficient for robust recognition even with slight motion blur. When the road environment lighting is complex or there is reflection interference on the projection plane, the projection duration can be adjusted to 3 to 5 seconds to increase the number of redundant frames and improve the decoding success rate.

[0038] In terms of implementation, the roadside inspection platform can trigger the projector to start after determining the target area and simultaneously record the start and end times of the projection, so as to perform task scheduling and log tracing when multiple targets are triggered continuously. The projection location is preferably selected near the target area where imaging is possible and does not affect traffic safety, such as the side of the guardrail post, the side of the bridge pier, or the surface of the vehicle-mounted calibration plate, so that the near-infrared pattern has a stable imaging area from the side or top angle of the UAV. By configuring near-infrared projection with a duration of 1 to 5 seconds, the visibility and recognition stability of the anchor point pattern under conditions of strong light, alternating shadows, and vehicle vibration can be improved, thereby increasing the success rate of road-air cooperative triggering.

[0039] Furthermore, the projection plane can be any one of the following: the surface of road ancillary facilities, the surface of bridge components, the surface of tunnel ancillary components, or the surface of a vehicle-mounted calibration plate. When there is no available projection plane near the target area, the roadside inspection platform unfolds the vehicle-mounted calibration plate and projects the coded optical anchor pattern onto the surface of the vehicle-mounted calibration plate.

[0040] In this embodiment, the projection plane is used to ensure that the coded optical anchor point pattern has a stable imaging foundation under different road scenarios. The principle is to provide an interface for the anchor point pattern that can be clearly captured by an airborne UAV platform, allowing the anchor points to be identified and decoded as mission information carriers, and also to serve as geometric references providing corner features and preset size references, thereby supporting subsequent calculations of relative pose and scale relationships. To adapt to various working environments in highway maintenance, the projection plane can be selected from any of the following: the surface of road ancillary facilities, the surface of bridge components, the surface of tunnel ancillary components, or the surface of a vehicle-mounted calibration plate.

[0041] In highway or ordinary road scenarios, the surfaces of road ancillary facilities are preferably the sides of guardrail posts, milestones or 100-meter markers, and the sides of anti-glare panel support structures. These surfaces typically have relatively stable flatness and maintain a safe distance from the driving lane, facilitating rapid projection by inspection vehicles during movement or short stops. Simultaneously, the drone can acquire image data containing anchor point patterns above or to the side of the road. In bridge scenarios, bridge component surfaces may include the sides of piers, abutments, the outer web of beams, or the surfaces of maintenance access components. Especially when positioning under bridges is easily affected by occlusion, projecting anchor points onto the sides of piers allows the drone to directly lock onto associated targets visually, reducing misalignment caused by relying on positioning coordinates. In tunnel and its entrance / exit scenarios, tunnel ancillary component surfaces may include tunnel entrance sidewalls, portal components, entrance sign support structures, and the outer shell surfaces of maintenance facilities. These surfaces have more stable contrast in shadow environments, which is beneficial for anchor point identification and facilitates corner point extraction and pose scale estimation by the drone in confined spaces.

[0042] When a suitable projection plane is lacking near the target area, the roadside inspection platform unfolds its vehicle-mounted calibration plate and projects coded optical anchor point patterns onto its surface, thus providing a controllable standard projection carrier. The vehicle-mounted calibration plate can be configured as a foldable or retractable structure, unfolding to form a plane of a preset size. The surface material can be a low-reflection matte material to reduce glare and highlight overflow, and a high-contrast border can be added to enhance corner point positioning stability. The vehicle-mounted calibration plate can be fixed to the side, rear, or extendable bracket of the inspection vehicle, allowing it to approach the target area without entering the driving lane and maintaining a safe distance from the ground or structures. By providing a vehicle-mounted calibration plate as a projection plane when a natural plane is lacking, it ensures that the anchor point patterns can be stably generated and recognized under various road boundary conditions, while providing a consistent geometric reference for subsequent supplementary data acquisition, avoiding decoding failures or scale estimation errors caused by an unstable projection interface.

[0043] Furthermore, the coded optical anchor point pattern has a preset geometric size and includes a set of feature points for pose calculation. The set of feature points includes at least four corner features. The air-end UAV platform determines the relative pose based on the corner features and determines the scale relationship based on the preset geometric size. It also determines the execution parameters for supplementary acquisition based on the scale relationship. The execution parameters include at least one or more of the following: flight distance, flight altitude, gimbal pitch angle, and shooting angle.

[0044] Furthermore, in addition to carrying information such as acquisition mode and scale level, the coded optical anchor point pattern is also designed as a visual reference that can be used for geometric measurement. Its core principle is to set a preset geometric size and a stable and extractable set of feature points in the anchor point pattern, so that the air-end UAV platform can simultaneously complete two types of calculations in the acquired image data. One type is information decoding to obtain the supplementary acquisition intention, and the other type is geometric calculation to obtain the relative pose and scale relationship. This transforms the supplementary acquisition from a method that relies on pilot experience or coarse coordinate navigation to a controlled parametric acquisition method.

[0045] In terms of pattern structure design, the coded optical anchor point pattern has a preset geometric size, such as a square or rectangular pattern with a side length of 80 mm to 200 mm, or a two-dimensional coded pattern with a fixed outer border size. The feature point set includes at least four corner features, which can be the four corners of the pattern's outer border, the four corners of the positioning mark, or corners composed of high-contrast geometric blocks. To improve the stability of corner point extraction in moving scenes, the corner point area can be set as a high-contrast black and white block with a certain edge width, and the outer border can be a continuous closed border to keep the corner points detectable under tilted viewing angles. The preset geometric size serves as a real-world scale benchmark and can be fixed by the system during deployment or configured with a uniform size along with the vehicle-mounted calibration board.

[0046] In terms of relative pose determination, after detecting four corner features in the image data, the air-based UAV platform can obtain the pixel position of the corner points in the image coordinate system. Combined with camera intrinsic parameters and the true geometric dimensions of the anchor point pattern, it calculates the relative pose of the UAV camera relative to the projection plane. The relative pose characterizes the distance relationship from the camera to the projection plane, the tilt angle of the camera's line of sight relative to the plane, and the rotation relationship of the camera around the normal direction. Since the projection plane and anchor point pattern remain fixed for a short period, and the corner features have stable geometric constraints, the relative pose can be updated in real-time or near real-time during UAV flight to guide the UAV in adjusting its flight position and gimbal attitude.

[0047] Regarding the determination of scale relationships, the air-based UAV platform establishes a mapping relationship from image pixels to actual lengths based on preset geometric dimensions. Specifically, the pixel scale can be obtained by the ratio of the number of pixels on the side of the anchor point pattern in the image to the actual side length, or by deriving the pixel scale from the distance parameters obtained through pose calculation. This scale relationship is used to constrain supplementary acquisitions to meet specified scale requirements, such as ensuring that the resolution of the lesion boundary in the image meets the assessment accuracy, and avoiding detail loss due to shooting from too far away or insufficient field of view due to shooting from too close.

[0048] The airborne UAV platform determines the execution parameters for supplementary data acquisition based on the aforementioned scale relationship. These execution parameters include at least one or more of the following: flight distance, flight altitude, gimbal pitch angle, and shooting angle. Flight distance and flight altitude control the relative position of the UAV to the target area. The gimbal pitch angle controls the camera's line of sight to meet the requirements of orthophoto or side-view acquisition modes. The shooting angle controls the camera's orientation and heading to obtain the desired angle for presenting the defects. For example, when the acquisition mode is orthophoto supplementary acquisition, the UAV can adjust its flight altitude according to the scale relationship so that the anchor point's side length reaches a preset pixel range in the image, and adjust the gimbal pitch angle to be nearly vertically downward, so as to form a quantifiable overhead image of the pit or subsidence area. When the acquisition mode is side-view supplementary acquisition, the UAV can maintain a suitable angle with the normal direction of the projection plane based on its relative pose, and adjust the shooting angle to move the acquisition sequence along the crack direction, so as to obtain a stable representation of the crack length and direction. By using the anchor point corner features and preset geometric dimensions to calculate the relative pose and scale relationship, and determining the re-capture execution parameters accordingly, the problem of inconsistent shooting scale and perspective between different task batches can be significantly reduced.

[0049] Furthermore, the acquisition mode includes one or more of orthophoto supplementary acquisition, side-view supplementary acquisition, and close-range detail supplementary acquisition. The scale level corresponds to the target pixel scale range or the target shooting distance range. The air-end UAV platform performs the supplementary acquisition when the range corresponding to the scale level is met.

[0050] In this embodiment, the acquisition mode and scale setting are used to standardize the supplementary acquisition process of the airborne UAV platform, ensuring that the supplementary acquisition data obtained in different scenarios maintains consistency in observation angle and imaging scale, thereby improving the reliability of subsequent quantitative assessment of defects and determination of treatment priorities. The acquisition mode includes at least one or more of orthophoto supplementary acquisition, side-view supplementary acquisition, and close-range detail supplementary acquisition. After decoding the acquisition mode in the image data, the airborne UAV platform calls the flight and shooting strategy corresponding to the acquisition mode and combines it with the scale setting for execution constraints.

[0051] In orthorectified acquisition mode, the airborne UAV platform aims to obtain top-down images that are approximately perpendicular to the road surface. This mode is suitable for defects such as potholes, subsidence, bulges, and ruts, where the primary focus is on planar contours and area quantification. The airborne UAV platform can adjust its flight altitude and gimbal attitude to ensure the target area's scale in the image meets the required level, and acquire several overlapping frames or keyframes for subsequent boundary extraction, area estimation, or orthorectified stitching. For example, when a suspected pothole is found on a highway and the priority for treatment needs to be quickly determined, orthorectified acquisition can stably obtain the pothole boundary and reduce deformation errors introduced by oblique views.

[0052] In the side-view supplementary acquisition mode, the aerial drone platform aims to obtain lateral images of the structure, suitable for observing the direction and length characteristics of defects such as bridge pier cracks, slope fissures, guardrail deformation, and tunnel portal damage. The aerial drone platform can maintain an appropriate distance and tilt angle from the target surface based on its relative pose and acquire continuous sequences along the direction of the defect, clearly presenting the continuity of cracks or deformations in the images. This mode is particularly suitable for unstable positioning areas such as under bridges or at tunnel entrances, reducing the risk of misjudgment caused by mistakenly photographing adjacent components.

[0053] In close-range detail acquisition mode, the airborne UAV platform aims to achieve high-resolution detail imaging, suitable for confirming fine-grained features such as crack width levels, pit edge fragmentation morphology, spalling details, and water seepage traces. The airborne UAV platform enters the target shooting distance range while maintaining a safe distance, and uses gimbal fine-tuning to ensure key details occupy sufficient pixel area to support severity classification and handling method selection. If necessary, autofocus or exposure parameter adjustments can be combined to improve detail clarity.

[0054] The scale settings correspond to the target pixel scale range or target shooting distance range, used to constrain the clarity and quantifiability of the supplementary data. The target pixel scale range can be represented by the range of pixels corresponding to a unit actual length, or by the range of pixel lengths in the image corresponding to the preset side length of the anchor point. The target shooting distance range can be represented by the distance interval from the UAV camera to the projection plane or target area. Before performing supplementary data acquisition, the air-based UAV platform calculates the current pixel scale or shooting distance and compares it with the corresponding scale setting range. Supplementary data acquisition is only performed if the range is met. If not, adaptive correction is performed by adjusting the flight distance, flight altitude, or gimbal pitch angle until the scale setting is met before acquisition. Through the combined constraints of acquisition mode and scale setting, the supplementary data can be made more consistent in angle and scale.

[0055] Furthermore, the air-end UAV platform binds the supplementary data with the decoded acquisition mode, scale level, and relative pose or scale relationship to generate an evidence package and sends it to the decision platform. The evidence package includes at least one or more of the following: target identification information, keyframe index, pose or scale summary, and verification information.

[0056] In this embodiment, to enable the supplementary data collection results to be quickly accessed by the decision-making platform and stably associated with the corresponding suspected maintenance targets, the airborne UAV platform binds the supplementary data with the task context information after completing the supplementary data collection, forming a structured evidence package, which is then sent to the decision-making platform. The principle behind this design is to solidify the originally scattered image or video files along with their generation conditions. This allows the decision-making platform to obtain not only visual content during intelligent assessment but also constraint information such as the acquisition mode, scale level, and pose or scale relationship. This improves the traceability and reproducibility of the quantitative assessment of disease conditions and reduces the burden of manual frame-by-frame screening and secondary verification.

[0057] Specifically, the air-based UAV platform binds the acquired data with the decoded acquisition mode, scale level, and relative pose or scale relationship. The acquired data can include image sequences, video keyframe sets, or point cloud data. The acquisition mode characterizes whether the acquisition strategy is orthophoto, side-view, or close-range detail. The scale level characterizes the pixel scale range or shooting distance range satisfied by the acquisition. The relative pose characterizes the relative distance and attitude relationship between the UAV camera and the projection plane. The scale relationship characterizes the mapping between pixels and actual length or a proportional parameter related to the shooting distance. Through this binding, the decision platform can select the appropriate recognition and quantization process based on the acquisition mode during evaluation and convert the image measurement results into actual dimensions based on the scale level or scale relationship.

[0058] The evidence package includes at least one or more of the following: target identification information, keyframe index, pose or scale summary, and verification information. The target identification information is used to correspond one-to-one with the target area determined by the roadside inspection platform, avoiding mismatches when multiple targets are accessed concurrently. The keyframe index is used to quickly locate representative frames for evaluation from the supplementary data, such as recording keyframe timestamps, frame numbers, or file offset information in the video stream, thereby reducing the decoding computation load of the decision platform and accelerating the evaluation response. The pose or scale summary is used to describe the key geometric parameters of this supplementary data in a compact manner, such as distance intervals, gimbal pitch angle intervals, pixel scale intervals, or anchor point side length pixel ranges, enabling the decision platform to quickly determine whether the data meets the evaluation accuracy requirements or whether supplementary data retry is necessary. The verification information is used to ensure the consistency of the evidence package transmission and storage. The verification information can be a cyclic redundancy check code or a hash digest. After receiving the evidence package, the decision platform can verify the verification information. If the verification fails, it can request retransmission or mark the data as unusable to avoid mis-evaluation due to link jitter or file corruption.

[0059] In terms of application, when an inspection vehicle discovers multiple suspected potholes on a highway, the airborne UAV platform can generate and upload evidence packages for each target. The decision-making platform quickly aggregates roadside preliminary inspection information and airborne supplementary evidence for the same target based on target identification information, and prioritizes high-priority targets based on keyframe indexing, thereby achieving efficient evaluation and task assignment in batch tasks. Through the structured binding and transmission of evidence packages, this embodiment establishes a clear data interface between road-air collaborative data collection and intelligent evaluation.

[0060] Furthermore, after completing the supplementary data collection, the air-end UAV platform sends a completion confirmation message, which is either an optical confirmation code or a wireless confirmation message. After receiving the completion confirmation message, the roadside inspection platform marks the target area as having been supplemented to avoid repeatedly triggering supplementary data collection for the same target area.

[0061] In this embodiment, to avoid repeated triggering of supplementary data collection for the same suspected maintenance target during air-ground coordination, which would lead to wasted UAV resources, data redundancy, and increased evaluation burden on the decision-making platform, the air-based UAV platform sends a completion confirmation message to the roadside inspection platform after completing supplementary data collection. Upon receiving the confirmation message, the roadside inspection platform marks the corresponding target area as having been supplemented. The principle behind this mechanism is to establish closed-loop management of supplementary data collection tasks, transitioning the target area from a triggered state to a completed state. This ensures that even if the same suspected target is detected again during subsequent inspections, the marking status can suppress repeated supplementary data collection triggering. The completion confirmation information can be either an optical confirmation code or a wireless confirmation message. When using an optical confirmation code, the airborne UAV platform can emit a short-lived, identifiable optical signal at a preset flashing frequency or pattern via an onboard light source. The roadside inspection platform identifies and decodes this signal using a vehicle-mounted camera or a photosensitive sensor to obtain the confirmation content. Optical confirmation codes are suitable for scenarios where the roadside and airborne platforms collaborate within a short visual range, providing a network-independent completion confirmation channel when communication links are unstable or limited. When using a wireless confirmation message, the airborne UAV platform can send a message containing target identification information and completion status to the roadside inspection platform via an onboard wireless module. The roadside inspection platform verifies the message and updates the task status upon receipt. Wireless confirmation messages are suitable for scenarios where there is a certain distance between the roadside and airborne platforms or where the line of sight is obstructed. They can be implemented via an onboard local area network, cellular network, or dedicated communication link, and can further carry supplementary data summaries or evidence packet identifiers for subsequent traceability.

[0062] Regarding the implementation of the "supplemented sampling" marker, the roadside inspection platform can record target identification information and supplemented sampling status fields in its target area management table. Upon receiving a completion confirmation message, it updates the supplemented sampling status of the corresponding target to "supplemented," and can record summary information such as completion time and supplemented sampling mode. To avoid status errors caused by false confirmations, the roadside inspection platform can perform consistency verification based on the target identification information and timestamp or verification information. For example, it can update the status only when the completion confirmation message matches a target that has been triggered in the current session and is within a valid time window. The "supplemented sampling" marker can be set with an expiration period or review conditions. For example, for high-risk defects, supplemented sampling can be retried after a preset time or a specific event, in order to balance avoiding redundancy and retaining necessary re-inspection capabilities.

[0063] For example, during continuous highway inspections, inspection vehicles may detect the same pothole area multiple times within a short distance due to road undulations and changing viewing angles. Without a completion confirmation and re-marking mechanism, this could trigger multiple drone takeoffs or multiple images of the same target, resulting in wasted power and time. By sending completion confirmation information from the air and having the roadside mark the area as re-marked, the inspection vehicle can skip the re-marking trigger when it subsequently detects the same target area, only retaining the evaluation and dispatch process for that target. This significantly reduces the probability of duplicate operations, improves the efficiency of road-air collaborative operations, and enhances the system's engineering availability.

[0064] This invention also provides an intelligent decision-making system for integrated road-air highway maintenance operations, used to execute the intelligent decision-making method for integrated road-air highway maintenance operations, including a roadside inspection platform, an airside unmanned aerial vehicle (UAV) platform, and a decision-making platform, wherein: The roadside inspection platform includes a first inspection data acquisition module, a target area determination module, and an anchor point pattern generation and projection module. It is used to acquire the first inspection data of the road and determine the target area of ​​the suspected maintenance target based on the first inspection data, generate a coded optical anchor point pattern corresponding to the target area, and project the coded optical anchor point pattern on a projection plane near the target area. The coded optical anchor point pattern carries at least the acquisition mode and scale level. The air-to-ground unmanned aerial vehicle (UAV) platform includes an image data acquisition module, an anchor point recognition and decoding module, a pose and scale determination module, and a supplementary acquisition module. It is used to acquire image data containing the coded optical anchor point pattern, identify and decode the coded optical anchor point pattern to obtain the acquisition mode and the scale level, determine the relative pose and scale relationship of the air-to-ground UAV platform relative to the projection plane based on the geometric features of the coded optical anchor point pattern, and perform supplementary acquisition on the suspected maintenance target according to the acquisition mode and the relative pose and scale relationship to obtain supplementary acquisition data. The decision-making platform includes an intelligent assessment module and a decision output module, which are used to intelligently assess the suspected maintenance targets based on the supplementary data to obtain disease information, and output maintenance operation decision results based on the disease information. The maintenance operation decision results include at least the treatment priority and treatment instructions.

[0065] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0066] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A road-air integrated intelligent decision-making method for highway maintenance operations, applied to a road-air collaborative system including a roadside inspection platform, an airborne unmanned aerial vehicle (UAV) platform, and a decision-making platform, characterized in that... The method includes: S1. The roadside inspection platform acquires the first inspection data of the road to determine the target area of ​​the suspected maintenance target, and generates a coded optical anchor pattern corresponding to the target area. The coded optical anchor pattern carries at least the acquisition mode and scale level. The roadside inspection platform projects the coded optical anchor pattern on the projection plane near the target area. S2. The air-to-ground UAV platform acquires image data containing the coded optical anchor point pattern, identifies and decodes the coded optical anchor point pattern to obtain the acquisition mode and the scale level; determines the relative pose and scale relationship of the air-to-ground UAV platform relative to the projection plane based on the geometric features of the coded optical anchor point pattern, and performs supplementary acquisition on the suspected maintenance target according to the acquisition mode and the relative pose and scale relationship to obtain supplementary acquisition data; S3. The decision-making platform performs intelligent evaluation of the suspected maintenance targets based on the supplementary data to obtain disease information, and outputs maintenance operation decision results based on the disease information. The maintenance operation decision results include at least the treatment priority and treatment instructions.

2. The intelligent decision-making method for integrated road and air highway maintenance operations according to claim 1, characterized in that: The first inspection data includes at least one or more of roadside image data, point cloud data, positioning data, and attitude data. The roadside inspection platform determines the road location description information of the target area based on the first inspection data. The road location description information includes one or more of the following: road number, driving direction, lane identification, mileage marker, or geographic coordinates.

3. The intelligent decision-making method for integrated road-air highway maintenance operations according to claim 1, characterized in that: In addition to carrying the acquisition mode and the scale level, the coded optical anchor pattern also carries target identification information corresponding to the target area, and carries one or more of the following: priority, timestamp, validity period, and verification information.

4. The intelligent decision-making method for integrated road and air highway maintenance operations according to claim 1, characterized in that: The roadside inspection platform projects the coded optical anchor pattern onto the projection plane using near-infrared projection, and the projection duration of the coded optical anchor pattern is a preset duration, which is 1 to 5 seconds.

5. The intelligent decision-making method for integrated road and air highway maintenance operations according to claim 1, characterized in that: The projection plane can be any one of the surfaces of road ancillary facilities, bridge components, tunnel ancillary components, or vehicle calibration plates. When there is no available projection plane near the target area, the roadside inspection platform unfolds the vehicle calibration plate and projects the coded optical anchor pattern onto the surface of the vehicle calibration plate.

6. The intelligent decision-making method for integrated road-air highway maintenance operations according to claim 1, characterized in that: The coded optical anchor pattern has a preset geometric size and includes a set of feature points for pose calculation. The set of feature points includes at least four corner features. The air-end UAV platform determines the relative pose based on the corner features and determines the scale relationship based on the preset geometric size. It also determines the execution parameters for supplementary acquisition based on the scale relationship. The execution parameters include at least one or more of the following: flight distance, flight altitude, gimbal pitch angle, and shooting angle.

7. The intelligent decision-making method for integrated road-air highway maintenance operations according to claim 1, characterized in that: The acquisition mode includes one or more of orthophoto supplementary acquisition, side-view supplementary acquisition, and close-range detail supplementary acquisition. The scale level corresponds to the target pixel scale range or the target shooting distance range. The air-end UAV platform performs the supplementary acquisition when the range corresponding to the scale level is met.

8. The intelligent decision-making method for integrated road-air highway maintenance operations according to claim 1, characterized in that: The air-to-ground UAV platform binds the supplementary data with the decoded acquisition mode, scale level, and relative pose or scale relationship to generate an evidence package and sends it to the decision platform. The evidence package includes at least one or more of the following: target identification information, keyframe index, pose or scale summary, and verification information.

9. The intelligent decision-making method for integrated road and air highway maintenance operations according to claim 1, characterized in that: After completing the supplementary data collection, the airborne UAV platform sends a completion confirmation message, which can be either an optical confirmation code or a wireless confirmation message. After receiving the completion confirmation message, the roadside inspection platform marks the target area as having been supplemented to avoid repeatedly triggering supplementary data collection for the same target area.

10. A road-air integrated intelligent decision-making system for highway maintenance operations, used to execute the road-air integrated intelligent decision-making method for highway maintenance operations as described in claim 1, characterized in that, This includes roadside inspection platforms, airborne drone platforms, and decision-making platforms, among which: The roadside inspection platform includes a first inspection data acquisition module, a target area determination module, and an anchor point pattern generation and projection module. It is used to acquire the first inspection data of the road and determine the target area of ​​the suspected maintenance target based on the first inspection data, generate a coded optical anchor point pattern corresponding to the target area, and project the coded optical anchor point pattern on a projection plane near the target area. The coded optical anchor point pattern carries at least the acquisition mode and scale level. The air-to-ground unmanned aerial vehicle (UAV) platform includes an image data acquisition module, an anchor point recognition and decoding module, a pose and scale determination module, and a supplementary acquisition module. It is used to acquire image data containing the coded optical anchor point pattern, identify and decode the coded optical anchor point pattern to obtain the acquisition mode and the scale level, determine the relative pose and scale relationship of the air-to-ground UAV platform relative to the projection plane based on the geometric features of the coded optical anchor point pattern, and perform supplementary acquisition on the suspected maintenance target according to the acquisition mode and the relative pose and scale relationship to obtain supplementary acquisition data. The decision-making platform includes an intelligent assessment module and a decision output module, which are used to intelligently assess the suspected maintenance targets based on the supplementary data to obtain disease information, and output maintenance operation decision results based on the disease information. The maintenance operation decision results include at least the treatment priority and treatment instructions.