Low-altitude airspace operation supervision method and system for highway maintenance inspection

Abstract

The application discloses a low-altitude airspace operation supervision method and system for highway maintenance inspection, and relates to the field of traffic road maintenance inspection control.The application iteratively clusters and merges new and old crack influencing factor data, continuously optimizes mapping model parameters, and makes crack probability prediction more in line with actual road conditions.A crack probability model and an accident level model form a "prevention-emergency" double link, the former predicts crack risks, and the latter controls vehicle speed in real time, covering the whole maintenance cycle.According to crack probability mapping results, the height of unmanned aerial vehicle (UAV) inspection is automatically adjusted, high-risk road sections are preferentially investigated, and inspection efficiency is improved.The accident level model fuses vehicle dynamic speed and crack static characteristics to generate graded alarm instructions, avoiding single factor misjudgment.According to the contribution degree of influencing factors, the crack source is repaired in a targeted manner, and invalid construction is reduced.Through accurate inspection and repair priority sorting, the flight time of the UAV and the frequency of manual inspection are reduced, and maintenance costs are reduced.

Description

Technical Field

[0001] This invention belongs to the field of traffic road maintenance and inspection control, and specifically relates to a method and system for monitoring low-altitude airspace operations for highway maintenance and inspection. Background Technology

[0002] Traditional highway maintenance and inspection rely on manual inspections, which are time-consuming and make it difficult to detect early cracks in a timely manner, leading to risk accumulation. Furthermore, relying on experience to determine the priority of crack repair can easily overlook hidden influencing factors (such as underground water seepage). In addition, traditional methods only control vehicle speed through fixed speed limit signs and do not combine them with differentiated early warning based on the dynamic characteristics of cracks. Drone inspections are conducted at fixed altitudes, resulting in repeated inspections of low-risk road sections and insufficient coverage of high-risk areas. Summary of the Invention

[0003] In response to the problems in related technologies, this invention proposes a method and system for monitoring and supervising low-altitude airspace operations for highway maintenance and inspection, in order to overcome the aforementioned technical problems existing in the existing related technologies.

[0004] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0005] This invention relates to a method for monitoring and supervising low-altitude airspace operations for highway maintenance and inspection, comprising the following steps:

[0006] S1. Obtain data on the factors influencing cracks in multiple experimental highway sections;

[0007] S2. Cluster the crack influencing factor data obtained in S1 to calculate the crack occurrence probability corresponding to each data category, and then construct a mapping model between the crack occurrence probability and crack influencing factors; then set several crack feature types.

[0008] S3. Based on the several feature types set in S2, construct a mapping model between the vehicle's speed approaching the crack, the crack feature data, and the accident severity level data.

[0009] S4. Conduct maintenance inspections on the highway section to be maintained; if cracks are found, proceed to S6; otherwise, input the corresponding crack influencing factor data into the mapping model constructed in S2.

[0010] S5. Adjust the initial altitude of the low-altitude UAV maintenance inspection based on the mapping result of S4. If the altitude is reduced to the current lowest inspection altitude and no cracks are found, merge the crack influencing factor data in S4 and S1 and repeat S2; otherwise, execute S6.

[0011] S6. Input the speed of the nearest vehicle to the crack on the highway section to be maintained and the characteristic data of the crack into the mapping model constructed in S3 for mapping; issue an alarm notification to the nearest approaching vehicle to reduce its speed based on the mapping results, and then repair and maintain the crack in combination with the mapping results in S4.

[0012] Preferably, step S1 includes the following steps:

[0013] S11. Define several types of factors that affect the formation of cracks in highways to obtain a set of highway crack influencing factor types; then select several existing highway sections that have been opened to traffic and have not undergone construction and repair after opening to obtain a set of experimental highway sections.

[0014] S12. Based on the aforementioned set of highway crack influencing factor types, obtain the crack influencing factor data for each highway segment in the experimental highway segment set, thus obtaining the experimental crack influencing factor dataset;

[0015] By integrating multi-source data such as meteorology, transportation, and geology, a correlation model can be established between the time of opening to traffic, material properties, environmental load, structural response, and the probability of crack occurrence. This model can systematically reveal the crack formation mechanism and provide a basis for decision-making to take relevant preventive measures for cracks that appear.

[0016] Preferably, step S2 includes the following steps:

[0017] S21. Set the initial number of clusters; cluster the experimental crack influencing factor dataset according to the initial number of clusters to obtain the experimental crack influencing factor classification dataset; calculate the ratio between the number of experimental highway sections with cracks corresponding to each type of experimental crack influencing factor data and the total number of experimental highway sections corresponding to each type of experimental crack influencing factor data to obtain the experimental crack occurrence probability classification dataset.

[0018] S22. Based on the experimental crack occurrence probability classification dataset and the experimental crack influencing factor classification dataset, construct a mapping model between crack influencing factor data and crack occurrence probability to obtain the final crack occurrence probability mapping model; define several feature types of cracks on highways to obtain a highway crack feature type set.

[0019] By grouping complex influencing factor data using clustering algorithms, common causes of different crack types can be systematically identified, avoiding the one-sidedness of traditional experience-based judgments. The crack occurrence probability classification dataset transforms qualitative descriptions into quantitative indicators, directly supporting maintenance priority ranking, which is more valuable for decision-making than simply counting crack lengths. The final established mapping model can reveal nonlinear relationships, providing a basis for decision-making regarding subsequent low-altitude airspace operation supervision tools (such as drones) for highway maintenance inspections. Secondly, the model output can be linked to the maintenance cost calculation module to form a two-dimensional decision matrix of "crack probability - maintenance cost". Through the probability classification dataset, the comparability of pavement crack risks under different geological conditions and climate zones is achieved for the first time, solving the decision-making bias problem caused by regional differences in traditional assessments.

[0020] Preferably, the probability mapping model for the occurrence of the final crack in S22 adopts a gradient boosting decision tree model;

[0021] The data on factors affecting cracks are of mixed type. Gradient boosting trees naturally support mixed feature inputs and can automatically handle nonlinear relationships. Furthermore, gradient boosting trees are highly interpretable and can output feature importance rankings, which makes it easier for engineers to adjust maintenance strategies.

[0022] Preferably, step S3 includes the following steps:

[0023] S31. In conjunction with the highway crack feature type set, collect the speed data of vehicles approaching the crack, the various feature data of the crack, and the severity level data of the accident corresponding to several traffic accidents caused by cracks on highways in history, to obtain the historical vehicle approaching crack speed dataset, the historical crack feature dataset, and the historical traffic accident severity level dataset.

[0024] S32. Based on the historical vehicle approach speed dataset, historical crack feature dataset, and historical accident severity level dataset, construct a mapping model between the vehicle approach speed, crack feature data, and accident severity level data to obtain the final accident severity level mapping model.

[0025] The mapping model can directly calculate the probability of accident severity at a specific speed for different crack characteristics, transforming qualitative descriptions into quantitative indicators. This allows for advance warnings to vehicles traveling in highway lanes with detected cracks, enabling them to reduce speed and ensure the safety of passengers. Combined with real-time monitoring data, the model can be embedded in the highway intelligent management system to achieve dynamic early warnings, making it more adaptable to complex road conditions than traditional fixed threshold alarms. By using a unified standard for quantifying accident severity, the model addresses the problem of subjective assessment bias in traditional accident reports, improving the comparability of data across regions and road sections.

[0026] Preferably, the final traffic accident severity level mapping model in S32 adopts a multilayer perceptron model;

[0027] Crack features, vehicle speed, and accident level are all structured data. MLP is good at processing nonlinear mappings of such numerical features. The output layer can be designed as multi-class or regression to flexibly adapt to evaluation needs.

[0028] Preferably, step S4 includes the following steps:

[0029] S41. Select the highway section to be maintained and denot it as the highway section to be maintained; set the initial altitude for low-altitude UAV maintenance inspection; dispatch the UAV to perform maintenance inspection on the highway section to be maintained according to the initial altitude for low-altitude UAV maintenance inspection; after the inspection is completed, obtain the inspection results.

[0030] When the inspection results show that there are cracks in the section of highway to be maintained, the characteristic data of the cracks are obtained based on the inspection results, in conjunction with the highway crack characteristic type set, to obtain the current crack characteristic dataset and execute S61; otherwise, execute S42 and S5.

[0031] S42. Based on the set of highway crack influencing factors, collect the crack influencing factor data corresponding to the highway section to be maintained to obtain the current crack influencing factor dataset; then set the crack occurrence probability threshold; then input the current crack influencing factor dataset into the final crack occurrence probability mapping model for mapping to obtain the current crack occurrence probability data and the current crack influencing factor contribution dataset.

[0032] The crack occurrence probability mapping model can identify high-risk road sections in advance, triggering priority inspections or manual verification, transforming passive maintenance into proactive prevention and control; by using the contribution dataset of influencing factors, the main causes of cracks can be identified and targeted improvements can be made; the current crack feature dataset can be directly connected to the historical traffic accident severity level mapping model, providing real-time basis for speed limit adjustments or traffic control; by diverting inspection tasks through probability thresholds, the scheduling pressure caused by concentrated drone operations can be avoided; early detection of hidden cracks can delay road surface damage and reduce the total life cycle maintenance cost; the accumulated crack feature and influencing factor data can be used for road design optimization, reducing crack generation from the source.

[0033] Preferably, step S5 includes the following steps:

[0034] S51, Set the current minimum inspection height;

[0035] S52. If the current probability data of crack occurrence is greater than or equal to the probability threshold of crack occurrence, the initial altitude of the low-altitude UAV maintenance inspection is reduced, and S41, S51 and S52 are repeated; otherwise, it is not necessary to reduce the initial altitude of the low-altitude UAV maintenance inspection.

[0036] If the initial altitude of the low-altitude drone maintenance inspection drops to the current lowest inspection altitude and the inspection results still show that there are no cracks in the highway section to be maintained, add the current crack influencing factor dataset to the experimental crack influencing factor dataset and repeat S21 and S22; otherwise, execute S61 in conjunction with the current crack influencing factor contribution dataset.

[0037] By dynamically adjusting the drone's altitude using probability thresholds, low-altitude, refined inspections are conducted on high-probability crack sections to reduce ineffective flight time and increase the coverage area of ​​a single inspection. When the altitude drops to the minimum threshold and no cracks are found, redundant inspections are automatically stopped to avoid energy waste. Inspection results are updated to the crack feature type set in real time, and the discovery of new cracks can be fed back into model training to improve the accuracy of the mapping model. Even if the probability model predicts no cracks, inspections are still conducted at the minimum altitude to ensure no omissions, avoiding the chain risk of "model misjudgment → crack expansion".

[0038] Preferably, step S6 includes the following steps:

[0039] S61. Based on the monitoring, obtain the speed of the nearest vehicle to the crack on the highway section to be maintained, and obtain the current nearest vehicle speed data; input the current nearest vehicle speed data and the current crack feature dataset into the final traffic accident severity level mapping model for mapping, and obtain the current traffic accident severity level data.

[0040] S62. Set the current accident severity level threshold; when the current accident severity level data is greater than or equal to the current accident severity level threshold, issue an alarm to notify the nearest approaching vehicle to reduce its speed, and repeat S61 and S62 until the current accident severity level data is less than the current accident severity level threshold.

[0041] S63. After the peak traffic period on the highway section to be maintained has passed, the section corresponding to the crack on the highway section to be maintained shall be closed. After the closure, in conjunction with the current crack influencing factor contribution dataset, the causes of the cracks on the highway section to be maintained shall be investigated and repaired in order of contribution data from high to low.

[0042] The system calculates the severity of traffic accidents in real time using a mapping model, triggers alarms based on thresholds, and forces vehicles to slow down to a safe range to prevent tire blowouts or loss of control due to high-speed impacts and cracks. It continuously monitors vehicle speed and crack status until the risk level drops below the threshold, forming a real-time closed loop of "detection-early warning-mitigation." Cracks are repaired at their root causes based on the contribution of influencing factors, avoiding a "one-size-fits-all" approach to maintenance. Construction is carried out during off-peak traffic periods to minimize traffic disruptions. Precise closures and priority screening reduce ineffective construction time and lower labor and equipment idle costs. When vehicles receive a speed-reduction command, the system simultaneously displays the risk level, enhancing public acceptance of intelligent maintenance.

[0043] The low-altitude airspace operation and monitoring system for highway maintenance and inspection includes a module for acquiring data on factors affecting cracks in experimental highway sections, a module for constructing a crack occurrence probability mapping model, a module for setting crack feature types, a module for constructing a traffic accident severity level mapping model, a module for determining cracks in highway sections to be maintained, a module for adjusting the altitude of low-altitude inspection drones for highways, and a module for vehicle alarms and road maintenance.

[0044] The present invention has the following beneficial effects:

[0045] 1. This invention iteratively clusters and merges data on influencing factors of new and old cracks, continuously optimizing the mapping model parameters to make crack probability prediction more closely match actual road conditions. The crack probability model and the accident level model form a "prevention-emergency" dual-link system, with the former predicting crack risks and the latter controlling vehicle speed in real time, covering the entire maintenance cycle. The drone inspection altitude is automatically adjusted based on the crack probability mapping results, prioritizing the inspection of high-risk road sections and improving inspection efficiency. The accident level model integrates vehicle dynamic speed and crack static characteristics to generate graded alarm commands, avoiding misjudgments based on single factors. Cracks are repaired in a targeted manner based on the contribution of influencing factors, reducing ineffective construction. Through precise inspection and repair priority ranking, the drone flight time and the frequency of manual inspections are reduced, lowering maintenance costs.

[0046] 2. In this invention, complex influencing factor data is grouped by clustering algorithm, which can systematically identify the common causes of different crack types and avoid the one-sidedness of traditional experience judgment; the crack occurrence probability classification dataset transforms qualitative description into quantitative indicators, which directly supports the ranking of maintenance priorities and is more valuable for decision-making than simply counting crack lengths.

[0047] 3. In this invention, the probability of the severity level of a traffic accident under a specific speed can be directly calculated through a mapping model, transforming qualitative descriptions into quantitative indicators. This provides a basis for issuing advance warnings to vehicles traveling on highway lanes where cracks have been detected, thereby reducing their speed and ensuring the safety of the people in the vehicles.

[0048] 4. In this invention, the severity level of a traffic accident is calculated in real time through a mapping model, and an alarm is triggered by a threshold. The vehicle is forced to slow down to a safe range to avoid tire blowout or loss of control due to high-speed impact cracks. The vehicle speed and crack status are continuously monitored until the risk level drops below the threshold, forming a real-time closed loop of "detection-early warning-mitigation".

[0049] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the embodiments of the invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating the low-altitude airspace operation and supervision method for highway maintenance and inspection according to the present invention.

[0052] Figure 2 This is a schematic diagram of the modules of the low-altitude airspace operation and monitoring system for highway maintenance and inspection according to the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the invention, and not all embodiments. Based on the embodiments of the invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the invention.

[0054] Example 1: Please refer to Figure 1 This embodiment describes a low-altitude airspace operation and supervision method for highway maintenance and inspection, comprising the following steps:

[0055] S1. Obtain data on the factors influencing cracks in multiple experimental highway sections;

[0056] S1 includes the following steps:

[0057] S11. Define several types of factors that affect the formation of cracks in highways to obtain a set of highway crack influencing factor types; then select several existing highway sections that have been opened to traffic and have not undergone construction and repair after opening to obtain a set of experimental highway sections.

[0058] The factors influencing highway cracks include: opening time (obtainable by reviewing highway construction completion reports or publicly available archives from transportation departments); asphalt mixture ratio (obtainable from asphalt suppliers' raw material ratio records; improper asphalt mixture ratios can lead to material aging, reduced viscosity, and consequently brittle cracking); compaction degree (obtainable by scanning the pavement structure layer density using road radar (GPR); substandard compaction degree may be reflected in the pavement through reflective cracks); and the number of extreme rainfall events (24-hour rainfall exceeding 250 mm) experienced since construction (obtainable from historical rainfall records from the local meteorological bureau; extreme rainfall accelerates water infiltration into cracks, causing roadbed softening and widening of cracks). The following parameters are considered after the road is completed: the number of times it has experienced extreme high temperatures (maximum temperature ≥ 40℃ within 24 hours) (this can be obtained from historical temperature records from the local meteorological bureau; thermal expansion and contraction caused by extreme high temperatures is the main cause of transverse cracks, especially in newly paved roads); the average traffic volume (this can be obtained from the daily average traffic volume statistics of the toll station ETC gantry system; long-term overloaded traffic or concentrated heavy vehicle traffic will aggravate road fatigue cracking); and the uniformity of roadbed settlement (this can be obtained by periodically leveling or assessing the regional settlement trend using InSAR (synthetic aperture radar interferometry) technology; soft soil sections (such as lake area geology) are prone to cracks due to uneven settlement, such as the soft soil section problem of the Anci Expressway).

[0059] S12. Based on the aforementioned set of highway crack influencing factor types, obtain the crack influencing factor data for each highway segment in the experimental highway segment set, thus obtaining the experimental crack influencing factor dataset;

[0060] By integrating multi-source data from meteorology, traffic, and geology, a correlation model can be established between opening time, material properties, environmental load, structural response, and the probability of crack occurrence. This model can systematically reveal the crack formation mechanism (such as the accelerated aging effect of extreme high temperature and traffic flow on asphalt), thus providing a basis for decision-making regarding subsequent preventive measures for cracks. Spatial matching of compaction and settlement data can locate weak road sections (such as the soft soil area of ​​the Anci Expressway), prioritizing radar scanning or grouting repair. Temporal analysis of traffic flow combined with crack morphology can distinguish between fatigue cracks (mainly longitudinal) and reflection cracks. Based on the network distribution of cracks, differentiated maintenance strategies can be developed. A regression model based on the historical frequency of extreme weather events and crack density can predict the pavement lifespan decline trend under future climate warming scenarios (such as an increase in the frequency of 40℃ high temperatures), providing a basis for reconstruction and expansion planning. Correlation analysis between original asphalt mix data and subsequent crack distribution can trace back construction defects (such as early brittle fracture caused by asphalt-aggregate ratio deviations), optimizing subsequent engineering material standards. By statistically analyzing the correlation between settlement uniformity and crack occurrence, early warning indicators for lake area road sections can be established (such as requiring reinforcement if monthly settlement exceeds 5mm), realizing a shift from passive maintenance to proactive prevention.

[0061] S2. Cluster the crack influencing factor data obtained in S1 to calculate the crack occurrence probability corresponding to each data category, and then construct a mapping model between the crack occurrence probability and crack influencing factors; then set several crack feature types.

[0062] S2 includes the following steps:

[0063] S21. Set the initial number of clusters; cluster the experimental crack influencing factor dataset according to the initial number of clusters to obtain the experimental crack influencing factor classification dataset; the clustering method can include K-Means algorithm, etc.; according to the experimental crack influencing factor classification dataset, calculate the ratio between the number of experimental highway sections with cracks corresponding to each type of experimental crack influencing factor data and the total number of experimental highway sections corresponding to each type of experimental crack influencing factor data to obtain the experimental crack occurrence probability classification dataset.

[0064] S22. Based on the experimental crack occurrence probability classification dataset and the experimental crack influencing factor classification dataset, construct a mapping model between crack influencing factor data and crack occurrence probability to obtain the final crack occurrence probability mapping model; define several feature types of cracks on highways to obtain a highway crack feature type set; the highway crack feature type set includes crack type, angle with the road centerline, average crack width, and average crack depth, etc.; the crack types are shown in Table 1 below:

[0065] Table 1

[0066]

[0067] The probability mapping model for the occurrence of the final crack described in S22 adopts a gradient boosting decision tree model.

[0068] S22 includes the following steps:

[0069] S221. Construct an initial crack occurrence probability mapping model and set a first training data ratio (such as 7:3 or 8:2, which can be adaptively set according to the specific actual situation); use the first training data ratio to divide the experimental crack occurrence probability classification dataset and the experimental crack influencing factor classification dataset to obtain the experimental crack occurrence probability classification data training set, the experimental crack influencing factor classification data training set, the experimental crack occurrence probability classification data test set, and the experimental crack influencing factor classification data test set.

[0070] S222. Set a first training error threshold (which can be adaptively set according to the specific actual situation); input the experimental crack influencing factor classification data training set as training data and the experimental crack occurrence probability classification data training set as training labels into the initial crack occurrence probability mapping model for training; during the training process, when the training error is less than the first training error threshold, stop training and obtain the trained crack occurrence probability mapping model; otherwise, continue training until the training error is less than the first training error threshold.

[0071] S223. Set a first test accuracy threshold (e.g., 85%, which can be adaptively set according to specific circumstances); input the experimental crack influencing factor classification data test set as test data and the experimental crack occurrence probability classification data test set as test labels into the trained crack occurrence probability mapping model for testing; after the test is completed, obtain the first test accuracy data; when the first test accuracy data is greater than or equal to the first test accuracy threshold, use the trained crack occurrence probability mapping model as the final crack occurrence probability mapping model; otherwise, return to S222 to continue training the trained crack occurrence probability mapping model and repeat S223 until the first test accuracy data is greater than or equal to the first test accuracy threshold.

[0072] The structure of the initial crack occurrence probability mapping model is shown in Table 2 below:

[0073] Table 2

[0074]

[0075] The data on factors affecting cracks are of mixed types (numerical and categorical, such as time of opening to traffic, asphalt mix ratio, etc.). Gradient boosting trees naturally support mixed feature inputs and can automatically handle nonlinear relationships. Furthermore, gradient boosting trees are highly interpretable and can output feature importance rankings (such as the contribution of "number of extreme rainfalls" to crack probability), which makes it easier for engineers to adjust maintenance strategies.

[0076] By grouping complex influencing factor data using clustering algorithms, common causes of different crack types can be systematically identified (e.g., transverse cracks corresponding to specific traffic flow-settlement combinations), avoiding the one-sidedness of traditional experience-based judgments. The crack occurrence probability classification dataset transforms qualitative descriptions into quantitative indicators (e.g., "the probability of cracks in soft soil sections reaches 72% under extreme rainfall"), directly supporting maintenance priority ranking and providing more decision-making value than simply statistically analyzing crack length. The final mapping model reveals nonlinear relationships (e.g., the threshold effect between compaction degree and crack probability), providing a basis for subsequent low-altitude airspace operation supervision of highway maintenance inspections. The system provides decision-making support for regulatory tools (such as drones); secondly, the model output can be linked to the maintenance cost calculation module to form a two-dimensional decision matrix of "crack probability - maintenance cost," for example, prioritizing the treatment of crack types with a probability > 60% and low maintenance cost; through a probability classification dataset, it achieves for the first time the comparability of pavement crack risks under different geological conditions and climate zones, solving the decision bias problem caused by regional differences in traditional assessments; in particular, by setting a set of highway crack feature types, it provides a basis for subsequently establishing a mapping model between vehicle speed, crack characteristics, and the severity of accidents on highways;

[0077] S3. Based on the several feature types set in S2, construct a mapping model between the vehicle's speed approaching the crack, the crack feature data, and the accident severity level data.

[0078] S3 includes the following steps:

[0079] S31. In conjunction with the highway crack feature type set, collect the speed data of vehicles approaching cracks (which can be obtained through corresponding monitoring), various feature data of the corresponding cracks (which can be obtained through post-event measurement), and traffic accident severity level data (which can be determined by relevant professionals through assessment of the traffic accident, such as by the number of casualties, property damage, etc.) corresponding to several traffic accidents caused by cracks on highways in history, to obtain historical vehicle approach speed dataset, historical crack feature dataset, and historical traffic accident severity level dataset.

[0080] S32. Based on the historical vehicle approach speed dataset, historical crack feature dataset, and historical accident severity level dataset, construct a mapping model between the vehicle approach speed, crack feature data, and accident severity level data to obtain the final accident severity level mapping model.

[0081] The final accident severity level mapping model in S32 adopts a multilayer perceptron model;

[0082] S32 includes the following steps:

[0083] S321. Construct an initial traffic accident severity level mapping model and set a second training data ratio (such as 7:3 or 8:2, which can be adaptively set according to the specific actual situation); use the second training data ratio to divide the historical vehicle approach speed data dataset, historical crack feature data dataset, and historical traffic accident severity level dataset into training sets for historical vehicle approach speed data, training sets for historical crack feature data, training sets for historical traffic accident severity level data, test sets for historical vehicle approach speed data, test sets for historical crack feature data, and test sets for historical traffic accident severity level data.

[0084] S322. Set a second training error threshold (which can be adaptively set according to the specific actual situation); input the historical vehicle approach speed data training set, the historical crack feature data training set as training data, and the historical traffic accident severity level data training set as training labels into the initial traffic accident severity level mapping model for training; during the training process, when the training error is less than the second training error threshold, stop training and obtain the trained traffic accident severity level mapping model; otherwise, continue training until the training error is less than the second training error threshold.

[0085] S323. Set a second test accuracy threshold (e.g., 90%, which can be adaptively set according to specific actual conditions); input the historical vehicle approach speed data test set, the historical crack feature data test set as test data, and the historical traffic accident severity level data test set as test labels into the trained traffic accident severity level mapping model for testing; after the test is completed, obtain the second test accuracy data; when the second test accuracy data is greater than or equal to the second test accuracy threshold, use the trained traffic accident severity level mapping model as the final traffic accident severity level mapping model; otherwise, return to S322 to continue training the trained traffic accident severity level mapping model and repeat S323 until the second test accuracy data is greater than or equal to the second test accuracy threshold.

[0086] The structure of the initial traffic accident severity level mapping model is shown in Table 3 below:

[0087] Table 3

[0088]

[0089] Crack features (such as width and type), vehicle speed, and accident level are all structured data. MLP is good at processing nonlinear mappings of such numerical features. The output layer can be designed as multi-classification (such as accident level 1-5) or regression (continuous risk value) to flexibly adapt to assessment needs.

[0090] The mapping model can directly calculate the probability of accident severity at a specific speed (e.g., 120 km / h) for different crack characteristics (e.g., transverse cracks with a width > 5 mm). This transforms qualitative descriptions ("prone to accidents") into quantitative indicators (e.g., "rollover probability reaches 68%)", providing advance warnings for vehicles traveling in highway lanes with detected cracks to reduce speed and ensure the safety of occupants. Combined with real-time monitoring data, the model can be embedded in the highway intelligent management system to achieve dynamic early warnings (e.g., "current speed 90 km / h + network cracks → triggering variable speed limit sign"), which is more adaptable to complex road conditions than traditional fixed threshold alarms. By using unified quantitative standards for accident severity (e.g., categorized by number of casualties and property damage), the model addresses the problem of subjective assessment bias in traditional accident reports and improves the comparability of data across regions and road sections.

[0091] S4. Conduct maintenance inspections on the highway section to be maintained; if cracks are found, proceed to S6; otherwise, input the corresponding crack influencing factor data into the mapping model constructed in S2.

[0092] S4 includes the following steps:

[0093] S41. Select the highway section to be maintained and denot it as the highway section to be maintained; set the initial altitude for low-altitude UAV maintenance inspection; dispatch the UAV to perform maintenance inspection on the highway section to be maintained according to the initial altitude for low-altitude UAV maintenance inspection; after the inspection is completed, obtain the inspection results.

[0094] When the inspection results show that there are cracks in the section of highway to be maintained, the characteristic data of the cracks are obtained based on the inspection results, in conjunction with the highway crack characteristic type set, to obtain the current crack characteristic dataset and execute S61; otherwise, execute S42 and S5.

[0095] S42. Based on the set of highway crack influencing factors, collect the crack influencing factor data corresponding to the highway section to be maintained to obtain the current crack influencing factor dataset; then set the crack occurrence probability threshold; then input the current crack influencing factor dataset into the final crack occurrence probability mapping model for mapping to obtain the current crack occurrence probability data and the current crack influencing factor contribution dataset.

[0096] Specifically, when setting the threshold for the probability of crack occurrence, one can refer to the correlation curve between crack probability and accident rate in historical data and select the probability value corresponding to the inflection point of returns. Alternatively, one can cross-validate the model's prediction accuracy. If the model's accuracy on the test set is 85%, the threshold can be set as the lower limit of the model's confidence interval (e.g., 55%) to avoid misjudgments. Furthermore, depending on the actual situation, differentiated thresholds can be set for different crack types (e.g., transverse / longitudinal cracks), with lower thresholds (e.g., 40%) used for high-risk types.

[0097] The crack occurrence probability mapping model can identify high-risk road sections in advance (e.g., "current probability ≥ 70%)", triggering priority inspections or manual verification, transforming passive maintenance into proactive prevention and control; by using the contribution dataset of influencing factors (e.g., "rainfall contribution weight accounts for 35%)", the main causes of cracks can be identified and targeted improvements can be made (e.g., strengthening drainage facilities); the current crack feature dataset can be directly connected to the historical traffic accident severity level mapping model, providing real-time basis for speed limit adjustments or traffic control; by diverting inspection tasks through probability thresholds (using conventional inspection vehicles for low-probability road sections), the scheduling pressure caused by concentrated drone operations can be avoided; early detection of hidden cracks (e.g., probabilistic model warnings but not visible to the naked eye) can delay road surface damage and reduce the total life cycle maintenance cost; the accumulated crack feature and influencing factor data can be used for road design optimization (e.g., adjusting the roadbed material ratio), reducing crack generation from the source;

[0098] S5. Adjust the initial altitude of the low-altitude UAV maintenance inspection based on the mapping result of S4. If the altitude is reduced to the current lowest inspection altitude and no cracks are found, merge the crack influencing factor data in S4 and S1 and repeat S2; otherwise, execute S6.

[0099] S5 includes the following steps:

[0100] S51. Set the current minimum inspection height; specifically, the setting of the current minimum inspection height is based on the following:

[0101] 1. The theoretical minimum height can be calculated based on the resolution of the drone camera and the minimum identifiable width of the crack (e.g., 0.2mm). For example, if the camera has 20 megapixels, and the crack needs to cover ≥10 pixels in the image, then the minimum height is ≤15 meters. 2. Comply with the Civil Aviation Administration's flight altitude restrictions for low-altitude drones (usually ≥10 meters), while also considering the effective detection range of obstacle avoidance radar (e.g., blind spots within 5 meters). 3. Data acquisition integrity requirements: Ensure that the measurement error of crack feature data (e.g., depth, length) is ≤5%, and that the minimum height ensures sufficient image coverage of the area surrounding the crack.

[0102] S52. If the current probability data of crack occurrence is greater than or equal to the probability threshold of crack occurrence, the initial altitude of the low-altitude UAV maintenance inspection is reduced, and S41, S51 and S52 are repeated; otherwise, it is not necessary to reduce the initial altitude of the low-altitude UAV maintenance inspection.

[0103] If the initial altitude of the low-altitude drone maintenance inspection drops to the current lowest inspection altitude and the inspection results still show that there are no cracks in the highway section to be maintained, add the current crack influencing factor dataset to the experimental crack influencing factor dataset and repeat S21 and S22; otherwise, execute S61 in conjunction with the current crack influencing factor contribution dataset.

[0104] By dynamically adjusting the drone's altitude using probability thresholds (e.g., from an initial 50m to 20m), low-altitude, refined inspections are conducted in high-probability crack sections (such as soft soil sections) to reduce ineffective flight time and increase the coverage area of ​​a single inspection. When the altitude drops to the minimum threshold and no cracks are found, redundant inspections are automatically stopped to avoid energy waste. Inspection results are updated to the crack feature type set in real time, and the discovery of new types of cracks (such as composite cracks) can be fed back into model training to improve the accuracy of the mapping model. Even if the probability model predicts no cracks, inspections are still conducted at the minimum altitude to ensure no omissions and avoid the chain reaction risk of "model misjudgment → crack expansion".

[0105] S6. Input the speed of the nearest vehicle to the crack on the highway section to be maintained and the characteristic data of the crack into the mapping model constructed in S3 for mapping; based on the mapping results, issue an alarm to notify the nearest approaching vehicle to reduce its speed, and then combine the mapping results in S4 to repair and maintain the crack.

[0106] S6 includes the following steps:

[0107] S61. Based on the monitoring, obtain the speed of the nearest vehicle to the crack on the highway section to be maintained, and obtain the current nearest vehicle speed data; input the current nearest vehicle speed data and the current crack feature dataset into the final traffic accident severity level mapping model for mapping, and obtain the current traffic accident severity level data.

[0108] S62. Set the current accident severity level threshold; when the current accident severity level data is greater than or equal to the current accident severity level threshold, issue an alarm to notify the nearest approaching vehicle to reduce its speed, and repeat S61 and S62 until the current accident severity level data is less than the current accident severity level threshold.

[0109] Specifically, the current threshold for the severity level of traffic accidents is set based on the following:

[0110] 1. By analyzing historical accident databases, the severity levels of traffic accidents caused by cracks at different vehicle speeds can be statistically distributed (e.g., at speeds ≥80km / h, the probability of severity level ≥4 reaches 70%), thus determining the risk threshold. In practical applications, differentiated thresholds can be set for different types of cracks (e.g., longitudinal / network cracks). For example, transverse cracks are prone to causing vehicle deviation, so the threshold can be set to level 3, while longitudinal cracks can be set to level 4. 2. Use a test set to verify the prediction accuracy of the mapping model (e.g., accuracy ≥90%), and set the threshold as the lower limit of the model's confidence interval (e.g., triggering an alarm when level ≥3) to avoid false alarms or missed alarms. 3. Refer to the speed limits for highways in the "Road Traffic Safety Law of the People's Republic of China" (e.g., 120km / h) and associate the threshold with the legal speed limit (e.g., triggering level 3 when vehicle speed ≥100km / h and crack depth ≥2cm).

[0111] S63. After the peak traffic period on the highway section to be maintained, the section corresponding to the cracks on the highway section to be maintained is closed off. After closure, based on the current dataset of contribution of factors affecting the cracks, the causes of the cracks on the highway section to be maintained are investigated and repaired in descending order of contribution. The severity level of the traffic accident is calculated in real time through a mapping model (e.g., level 1-5), and an alarm is triggered by a threshold (e.g., when level ≥ 3), forcing vehicles to reduce their speed to a safe range (e.g., speed limit of 60km / h) to avoid tire blowouts or loss of control due to high-speed impacts with the cracks. Vehicle speed and crack status are continuously monitored until the risk level drops below the threshold, forming a real-time closed loop of "detection-early warning-mitigation". The cracks are repaired in a targeted manner according to the contribution of the influencing factors (e.g., "roadbed settlement weight 40%" is given priority). Targeting the root causes of cracks avoids a "one-size-fits-all" approach to maintenance; choosing off-peak traffic periods (such as nighttime) for construction minimizes traffic disruptions, while selecting the most critical road sections to close based on real-time data; the model integrates vehicle dynamic data (speed) and crack static data (width, type) for more accurate accident risk assessment (e.g., the combination of transverse cracks and speeding carries a higher risk); reusing pre-trained accident severity mapping models eliminates the need for repeated data labeling, improving response efficiency; speed warning thresholds and accident level thresholds provide double protection against misjudgments by a single model (e.g., false crack detection leading to invalid speed limits); precise closures and priority screening reduce ineffective construction time and lower labor and equipment idle costs; when vehicles receive speed reduction commands, the risk level information is displayed simultaneously (e.g., "crack width 3cm, risk level 4"), enhancing public acceptance of intelligent maintenance.

[0112] Example 2: Please refer to Figure 2This embodiment discloses a low-altitude airspace operation and supervision system for highway maintenance and inspection. The system can implement the methods of the above embodiment, including a data acquisition module for the influencing factors of cracks in experimental highway sections, a crack occurrence probability mapping model construction module, a crack feature type setting module, a traffic accident severity level mapping model construction module, a crack judgment module for highway sections to be maintained, a highway low-altitude inspection drone altitude adjustment module, and a vehicle alarm and road maintenance module.

[0113] The experimental highway section crack influencing factor data acquisition module acquires crack influencing factor data for multiple experimental highway sections;

[0114] The crack occurrence probability mapping model construction module clusters the crack influencing factor data obtained in S1 to calculate the crack occurrence probability data corresponding to each data category, and then constructs a mapping model between the crack occurrence probability data and the crack influencing factor data.

[0115] The crack feature type setting module sets several feature types for cracks on highways;

[0116] The traffic accident severity level mapping model construction module constructs a mapping model between the vehicle's approach speed to the crack, the crack feature data, and the traffic accident severity level data based on several crack feature types set in S2.

[0117] The crack detection module for the highway section to be maintained sets the initial altitude for low-altitude UAV maintenance inspection and performs maintenance inspection on the highway section to be maintained. If cracks exist, S6 is executed; otherwise, the data on the influencing factors of cracks corresponding to the highway section to be maintained are input into the mapping model constructed in S2 for mapping.

[0118] The highway low-altitude inspection drone altitude adjustment module adjusts the initial altitude of the low-altitude drone maintenance inspection according to the mapping result in S4. If it is lowered to the current lowest inspection altitude and no cracks are displayed, the crack influencing factor data in S4 is merged with the crack influencing factor data in S1, and S2 is repeated; otherwise, S6 is executed.

[0119] The vehicle alarm and road maintenance module inputs the speed of the nearest vehicle to the crack on the highway section to be maintained and the characteristic data of the crack into the mapping model constructed in S3 for mapping; based on the mapping result, it issues an alarm to notify the nearest approaching vehicle to reduce its speed, and then combines the contribution data of each influencing factor in the mapping result in S4 to repair and maintain the crack.

[0120] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0121] The preferred embodiments of the invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A method for monitoring and supervising low-altitude airspace operations for highway maintenance and inspection, characterized in that, Includes the following steps: S1. Obtain data on the factors influencing cracks in multiple experimental highway sections; S2. Cluster the crack influencing factor data obtained in S1 to calculate the crack occurrence probability corresponding to each data category, and then construct a mapping model between the crack occurrence probability and crack influencing factors; then set several crack feature types. S3. Based on the several feature types set in S2, construct a mapping model between the vehicle's speed approaching the crack, the crack feature data, and the accident severity level data. S4. Conduct maintenance inspections on the highway section to be maintained; if cracks are found, proceed to S6; otherwise, input the corresponding crack influencing factor data into the mapping model constructed in S2. S5. Adjust the initial altitude of the low-altitude UAV maintenance inspection based on the mapping result of S4. If the altitude is reduced to the current lowest inspection altitude and no cracks are found, merge the crack influencing factor data in S4 and S1 and repeat S2; otherwise, execute S6. S6. Input the speed of the nearest vehicle to the crack on the highway section to be maintained and the characteristic data of the crack into the mapping model constructed in S3 for mapping; issue an alarm notification to the nearest approaching vehicle to reduce its speed based on the mapping results, and then repair and maintain the crack in combination with the mapping results in S4.

2. The method for low-altitude airspace operation supervision for highway maintenance and inspection according to claim 1, characterized in that, S1 includes the following steps: S11. Define several types of factors that affect the formation of cracks in highways to obtain a set of highway crack influencing factor types; then select several existing highway sections that have been opened to traffic and have not undergone construction and repair after opening to obtain a set of experimental highway sections. S12. Based on the set of highway crack influencing factors, obtain the crack influencing factor data for each highway segment in the experimental highway segment set to obtain the experimental crack influencing factor dataset.

3. The method for low-altitude airspace operation supervision for highway maintenance and inspection according to claim 2, characterized in that, S2 includes the following steps: S21. Set the initial number of clusters; cluster the experimental crack influencing factor dataset according to the initial number of clusters to obtain the experimental crack influencing factor classification dataset; calculate the ratio between the number of experimental highway sections with cracks corresponding to each type of experimental crack influencing factor data and the total number of experimental highway sections corresponding to each type of experimental crack influencing factor data to obtain the experimental crack occurrence probability classification dataset. S22. Based on the experimental crack occurrence probability classification dataset and the experimental crack influencing factor classification dataset, construct a mapping model between crack influencing factor data and crack occurrence probability to obtain the final crack occurrence probability mapping model; define several feature types of cracks on highways to obtain a highway crack feature type set.

4. The method for low-altitude airspace operation supervision for highway maintenance and inspection according to claim 3, characterized in that: The probability mapping model for the occurrence of the final crack described in S22 adopts a gradient boosting decision tree model.

5. The low-altitude airspace operation and supervision method for highway maintenance and inspection according to claim 4, characterized in that, S3 includes the following steps: S31. In conjunction with the highway crack feature type set, collect the speed data of vehicles approaching the crack, the various feature data of the crack, and the severity level data of the accident corresponding to several traffic accidents caused by cracks on highways in history, to obtain the historical vehicle approaching crack speed dataset, the historical crack feature dataset, and the historical traffic accident severity level dataset. S32. Based on the historical vehicle approach speed dataset, historical crack feature dataset, and historical accident severity level dataset, construct a mapping model between vehicle approach speed, crack feature data, and accident severity level data to obtain the final accident severity level mapping model.

6. The method for monitoring and supervising low-altitude airspace operations for highway maintenance and inspection according to claim 5, characterized in that: The final accident severity level mapping model in S32 adopts a multilayer perceptron model.

7. The method for low-altitude airspace operation supervision for highway maintenance and inspection according to claim 6, characterized in that, S4 includes the following steps: S41. Select the highway section to be maintained and denot it as the highway section to be maintained; set the initial altitude for low-altitude UAV maintenance inspection; dispatch the UAV to perform maintenance inspection on the highway section to be maintained according to the initial altitude for low-altitude UAV maintenance inspection; after the inspection is completed, obtain the inspection results. When the inspection results show that there are cracks in the section of highway to be maintained, the characteristic data of the cracks are obtained based on the inspection results, in conjunction with the highway crack feature type set, to obtain the current crack feature dataset and execute S61. Otherwise, execute S42 and S5; S42. Based on the set of highway crack influencing factors, collect the crack influencing factor data corresponding to the highway section to be maintained to obtain the current crack influencing factor dataset; then set the crack occurrence probability threshold; then input the current crack influencing factor dataset into the final crack occurrence probability mapping model for mapping to obtain the current crack occurrence probability data and the current crack influencing factor contribution dataset. S6 includes the following steps: S61. Based on the monitoring, obtain the speed of the nearest vehicle to the crack on the highway section to be maintained, and obtain the current nearest vehicle speed data; input the current nearest vehicle speed data and the current crack feature dataset into the final traffic accident severity level mapping model for mapping, and obtain the current traffic accident severity level data. S62. Set the current accident severity level threshold; when the current accident severity level data is greater than or equal to the current accident severity level threshold, issue an alarm to notify the nearest approaching vehicle to reduce its speed, and repeat S61 and S62 until the current accident severity level data is less than the current accident severity level threshold. S63. After the peak traffic period on the highway section to be maintained has passed, the section corresponding to the crack on the highway section to be maintained shall be closed. After closure, in conjunction with the current contribution data of the crack influencing factors, the causes of the cracks on the highway section to be maintained shall be investigated and repaired in order of the contribution data from high to low.

8. The method for low-altitude airspace operation supervision for highway maintenance and inspection according to claim 7, characterized in that, S5 includes the following steps: S51, Set the current minimum inspection height; S52. If the current probability data of crack occurrence is greater than or equal to the probability threshold of crack occurrence, the initial altitude of the low-altitude UAV maintenance inspection is reduced, and S41, S51 and S52 are repeated; otherwise, it is not necessary to reduce the initial altitude of the low-altitude UAV maintenance inspection. If the initial altitude of the low-altitude drone maintenance inspection drops to the current lowest inspection altitude and the inspection results still show that there are no cracks in the highway section to be maintained, add the current crack influencing factor dataset to the experimental crack influencing factor dataset and repeat S21 and S22; otherwise, execute S61 in conjunction with the current crack influencing factor contribution dataset.

9. A system for implementing the low-altitude airspace operation and supervision method for highway maintenance and inspection as described in any one of claims 1-8, characterized in that: It includes modules for acquiring data on factors affecting cracks in experimental highway sections, constructing a crack occurrence probability mapping model, setting crack feature types, constructing a traffic accident severity level mapping model, determining cracks in highway sections under maintenance, adjusting the altitude of low-altitude inspection drones for highways, and vehicle alarms and road maintenance.

Images