Disaster damage assessment method and system based on unmanned aerial vehicle cooperation

Abstract

The invention discloses a disaster damage assessment method and system based on unmanned aerial vehicle cooperation, and relates to the technical field of road network disaster damage intelligent management, and the method comprises the steps: determining the damage potential and importance factors of each road section through constructing a road network disaster influence priori model based on road basic data, geological environment information and disaster information, and determining the damage potential and importance factors of each road section; task priority division and dynamic planning of a multi-unmanned aerial vehicle inspection path are realized; the unmanned aerial vehicle collects road visual data in cooperative flight, generates a structural damage score based on semantic feature extraction, and calculates a traffic adjustment damage index in combination with a road segment importance factor, thereby quantitatively characterizing the influence of structural damage on traffic operation; and by summarizing the traffic adjustment damage indexes of each road section, a toughness index reflecting the overall traffic capacity reduction degree of the road network is obtained. Therefore, full-process intelligent evaluation from disaster perception to road network toughness quantification is realized, and the precision and timeliness of post-disaster road monitoring are remarkably improved.

Description

Technical Field

[0001] This application relates to the field of intelligent road network disaster loss management technology, and in particular to a disaster loss assessment method and system based on UAV collaboration. Background Technology

[0002] Road transportation infrastructure is the foundation of the national economy and social operation. After extreme natural disasters, problems such as road collapses, bridge damage, and tunnel leaks often lead to traffic disruptions and even secondary disasters. Current disaster damage assessments rely primarily on manual surveys, supplemented by satellite remote sensing and fixed monitoring equipment. While these methods provide some feedback, the vast road network, potential road blockages in disaster areas, low efficiency of manual patrols, and safety hazards all contribute to the problem. Satellite remote sensing suffers from limited spatial resolution, cloud cover, and insufficient timeliness, making it impossible to identify detailed localized damage in a short period, thus affecting the accuracy of emergency decision-making.

[0003] With the development of drone technology, using drones to quickly acquire images of disaster areas and conduct disaster damage assessments is gradually becoming a trend. Drones can fly to areas inaccessible to humans, taking real-time photos of disaster sites and providing intelligence for rescue teams. Current applications mostly focus on aerial photography and image analysis using single drones or a small number of drones, achieving some results in scenarios such as building collapses and flood boundaries. However, for targets like road networks, which have spatial extension and connectivity characteristics, single-drone operations struggle to cover a large area of ​​the road network and analyze results in a timely manner within a limited time. Furthermore, the communication bandwidth between the drone and the ground station is limited, and transmitting large amounts of high-definition images can cause delays. Summary of the Invention

[0004] This application provides a disaster damage assessment method, system, storage medium, computer program product, and electronic device based on UAV collaboration, which aims to at least solve the problems of slow response, limited coverage, and incomplete information acquisition in the current related technologies for post-disaster road traffic infrastructure assessment.

[0005] In a first aspect, embodiments of this application provide a disaster damage assessment method based on UAV collaboration. The method includes: constructing a priori model of road network disaster impact based on basic road network data, geological environment information, and historical maintenance records of a target road network, combined with real-time acquired disaster information, to assess the damage potential index and road segment importance factors of roads; calculating the priority index of each road segment in the target road network according to the priori model, and dynamically planning and allocating the inspection task areas and flight paths of multiple UAVs based on the status information of available UAVs, so that high-priority road segments are covered first; and controlling the multiple UAVs according to the allocated inspection task areas and flight paths. The system conducts coordinated flights to collect visual sensing data of roads and the surrounding environment using onboard sensors. The drones also extract visual feature information related to road structural damage from the collected visual sensing data to generate corresponding compressed data packets. The system receives and fuses these compressed data packets from multiple drones, calculates a structural damage score reflecting the structural safety status of each road segment based on the visual feature information corresponding to each segment, and calculates a traffic adjustment damage index representing the degree of impact of road damage on traffic operation for each segment based on the structural damage scores and corresponding road segment importance factors. Finally, it summarizes and calculates a road network resilience index reflecting the overall decline in the capacity of the target road network.

[0006] Secondly, embodiments of this application provide a disaster damage assessment system based on UAV collaboration. The system includes: a disaster information modeling unit, used to construct a priori model of road network disaster impact based on basic road network data, geological environment information, and historical maintenance records of the target road network, combined with real-time acquired disaster information, for assessing the damage potential index and road segment importance factors of roads; a dynamic task allocation unit, used to calculate the priority index of each road segment in the target road network according to the priori model of road network disaster impact, and dynamically plan and allocate the inspection task areas and flight paths of multiple UAVs based on the status information of available UAVs, so that high-priority road segments are covered first; and a collaborative data acquisition unit, used to control the multiple UAVs according to the allocated inspection task areas. The system coordinates flight paths and domains to collect visual sensing data of roads and surrounding environments using onboard sensors. The drone also extracts visual feature information related to road structural damage from the collected visual sensing data to generate corresponding compressed data packets. A structural damage scoring unit receives and fuses the compressed data packets from multiple drones, calculating a structural damage score reflecting the structural safety status of each road segment based on the visual feature information corresponding to each segment. A resilience index assessment unit calculates a traffic adjustment damage index representing the degree of impact of road damage on traffic operation for each road segment based on the structural damage score and corresponding road segment importance factors, and summarizes and calculates a road network resilience index reflecting the overall decline in the capacity of the target road network.

[0007] Thirdly, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the unmanned aerial vehicle-based disaster assessment method of any embodiment of this application.

[0008] Fourthly, embodiments of this application provide a storage medium storing a computer program thereon, characterized in that, when the program is executed by a processor, it implements the steps of the disaster damage assessment method based on UAV collaboration of any embodiment of this application.

[0009] Fifthly, embodiments of this application provide a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the disaster damage assessment method based on UAV collaboration of any embodiment of this application.

[0010] The disaster damage assessment method and system based on UAV collaboration provided in this application can achieve at least the following technical effects: (1) By constructing a priori model of road network disaster impact that combines basic road network data, geological environment information, historical maintenance records, and real-time disaster information, the potential for road damage and the importance factors of road segments were quantitatively characterized. Based on this, priority calculation guided by this model can specifically identify key road segments that need to be inspected first, and dynamically plan inspection tasks and flight paths in combination with the real-time status information of UAVs. Thus, a data-driven task scheduling mechanism is formed, which makes the inspection sequence and flight path distribution of UAV swarms self-consistently matched with the road damage risk and traffic importance, significantly improving the overall resource utilization rate and time response efficiency of inspections, and ensuring that high-value road segments are assessed at an early stage.

[0011] (2) In the UAV collaborative operation and data processing stage, the road network is covered in sections by multi-UAV collaborative flight. Visual feature extraction and data compression are performed at the UAV end, and high-resolution images are converted into compact structured feature data packets before being transmitted to the ground system. This effectively reduces communication bandwidth usage, alleviates link congestion when multiple UAVs transmit back simultaneously, and retains the core information for structural damage identification, thus improving the efficiency of disaster damage analysis. In addition, after the multi-source feature data is fused on the ground, the multi-angle observation results from different UAVs can be combined to make a robust damage score for the road structure safety status, thereby improving the accuracy and consistency of the assessment results.

[0012] (3) By combining the structural damage scores of each road segment with the corresponding importance factors, a traffic adjustment damage index was established. Based on this, the overall road network resilience index was calculated to quantify the comprehensive impact of disasters on the road network's traffic capacity, thus realizing the transformation from local damage identification to global traffic performance assessment. As a result, the system can provide intuitive and quantifiable evidence for emergency decision-making, enabling measures such as road closures, detours, and resource allocation to have clear priorities and quantitative support.

[0013] This technical solution achieves high timeliness, high accuracy, and decision-making output for post-disaster road network damage assessment. It forms a closed-loop mechanism in the post-disaster emergency monitoring system, from data perception and drone swarm task allocation to comprehensive assessment. This enables the perception results of the drone swarm to directly serve traffic emergency management, significantly improving the efficiency and scientific nature of road traffic infrastructure damage assessment. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 A flowchart illustrating an example of a disaster damage assessment method based on UAV collaboration according to an embodiment of this application is shown; Figure 2 A flowchart illustrating an example of constructing a priori model of road network disaster impacts according to an embodiment of this application is shown. Figure 3 A flowchart illustrating an example of dynamic task allocation for multiple unmanned aerial vehicles according to an embodiment of this application is shown. Figure 4 A flowchart illustrating an example of a drone locally generating compressed data packets according to an embodiment of this application is shown. Figure 5 A flowchart illustrating an example of calculating a structural damage score according to an embodiment of this application is shown. Figure 6 A flowchart illustrating an example of calculating a road network resilience index according to an embodiment of this application is shown. Figure 7 A comparative diagram illustrating an example of how inspection coverage time varies with the number of drones is shown; Figure 8 A schematic diagram illustrating the spatial distribution of a traffic adjustment damage index is shown; Figure 9A structural block diagram of an example of a disaster damage assessment system based on drone collaboration according to an embodiment of this application is shown. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] Currently, some experts and scholars have proposed intelligent risk assessment technologies for road network disasters, but these technologies suffer from slow big data processing speeds, difficulty in integrating risk characteristics from different sources, and inflexible adjustment of early warning rules.

[0018] Some researchers have proposed an embedded semantic feature extraction framework to reduce the communication load in UAV disaster monitoring tasks. This framework identifies object categories in images through semantic segmentation, and then a binary mask predictor selects regions useful for specific downstream tasks, transmitting only filtered semantic information, thus reducing data volume. The framework was validated on a flood image dataset, reducing the amount of transmitted data by approximately 85% while maintaining classification accuracy. However, this method still relies on a general semantic segmentation model and is not specifically optimized for the structural features of road infrastructure. Furthermore, its downstream tasks mainly focus on building collapses or water feature identification, and do not address fine-grained damage such as road cracks and subsidence. Additionally, the algorithm is only applicable to data collected by a single UAV and does not consider multi-UAV collaborative operations or the overall assessment of the road network.

[0019] Following Hurricane Helen in 2024, the U.S. Geological Survey used small drones to explore landslide-damaged roads along the Blue Ridge Highway. Operators carried small drones with integrated video cameras to sections of road blocked by trees, acquiring on-site images to inform subsequent flight planning for larger drones and advanced sensors. This demonstration work shows that drones offer advantages in accessibility and safety at landslide sites; however, they rely on manual operation and visual analysis, lacking automated damage identification and mission planning processes, and failing to quantify road conditions.

[0020] Some literature also describes the practice of using drones for damage assessment after hurricanes. Specifically, after a hurricane, roads are disrupted and power lines collapse. Traditional manual and manned aircraft assessments are both slow and dangerous, while drones equipped with RGB cameras, thermal imaging, multispectral, and LiDAR sensors can quickly and safely complete large-scale post-disaster assessments. The blog lists various applications of drones, such as residential and commercial damage assessment, flood and coastal erosion detection, and infrastructure inspection, emphasizing that drones can obtain high-resolution images and real-time data transmission. However, this literature mainly describes hardware configuration and general workflows (planning, flight, data acquisition, processing, and distribution), without providing specific details for road infrastructure damage assessment, nor addressing issues such as multi-drone collaboration, network road coverage, and bandwidth limitations.

[0021] The UAV-PDD2023 dataset is an open-source drone aerial photography dataset for detecting road pavement distress. Built by institutions such as Hebei University of Technology, it includes numerous examples of road damage, encompassing six categories: longitudinal cracks, transverse cracks, network cracks, diagonal cracks, repairs, and potholes. This dataset provides rich training samples for algorithm research and is significant for road maintenance and low-cost monitoring. However, it focuses on local crack detection, primarily evaluating target detection accuracy, lacking assessment of road structural safety, and failing to consider damage scenarios under disaster conditions or the relative importance of roads within the traffic network.

[0022] Some literature has proposed a "pre-set site selection model for UAVs in traffic accident assessment considering road dispersion." This study analyzes the non-periodic traffic congestion caused by traffic accidents and suggests that reducing accident assessment time by 1 minute can reduce vehicle delays by approximately 5 minutes. Specifically, an optimization model targeting accident handling costs and congestion reduction is proposed, employing an improved simulated annealing algorithm (including multi-neighborhood strategy, adaptive neighborhood size, and taboo list) to determine UAV site locations. Research shows that UAVs can quickly reach the accident scene, collect images using high-definition cameras and sensors, and transmit them back in real time, effectively shortening assessment time and reducing the risk of secondary accidents. The literature also discusses the construction of UAV sites in locations such as public infrastructure and commercial rooftops, and proposes large-scale simulations in complex scenarios. While this research demonstrates the advantages of rapid UAV response in traffic accident emergency response, it primarily focuses on site deployment and scheduling algorithms, without addressing the assessment of large-scale road damage caused by disasters; the simulated annealing algorithm used is a general optimization method and does not incorporate road structural safety characteristics or disaster intensity factors.

[0023] In summary, current technologies for road damage assessment using drones have the following key shortcomings: 1) lack of collaborative planning and rapid coverage mechanisms for road networks; 2) lack of damage assessment methods that combine road structure, safety functions, and traffic importance; 3) neglect of task-driven information compression under communication bandwidth limitations; and 4) insufficient evaluation of overall road network resilience.

[0024] It should be understood that the above description of the relevant technologies is intended only to help the public better understand the inventive spirit and motivation of this application, and is not intended to limit this application. Furthermore, the technical solutions described in the above-mentioned relevant technologies are not prior art, and may also be undisclosed technical solutions, such as those under research or in the laboratory stage.

[0025] The technical solutions in this application, including the collection, storage, use, processing, transmission, provision, and disclosure of users' personal information, comply with relevant laws and regulations and do not violate public order and good morals.

[0026] Figure 1 A flowchart illustrating an example of a drone-based disaster damage assessment method according to an embodiment of this application is shown.

[0027] Regarding the execution subject of the method in the embodiments of this application, it can be any controller or processor with computing or processing capabilities, such as the controller of a road network disaster damage early warning platform, which can complete the coverage survey, damage identification and network resilience evaluation of a large-scale road network in a short period of time after a disaster occurs.

[0028] In some examples, it may be integrated into an electronic device or terminal through software, hardware, or a combination of both, and the type of terminal or electronic device may be diverse.

[0029] like Figure 1 As shown, in step S110, a priori model of road network disaster impact is constructed based on the basic road network data, geological environment information, and historical maintenance records of the target road network, combined with real-time disaster information, to assess the damage potential index and road segment importance factor of the road.

[0030] In some implementations, spatial topology information of the target road network (e.g., road segment length, pavement structure, bridge and tunnel distribution, traffic level, etc.) is extracted from the road infrastructure database, and geological and topographic environmental parameters of the corresponding area, such as soil type, slope distribution, river network density, and historical records of landslides and collapses, are loaded. Simultaneously, historical maintenance archives, including pavement damage records, structural reinforcement time, and maintenance frequency, are accessed.

[0031] Specifically, after a real-time disaster occurs, disaster information from meteorological and earthquake monitoring centers, emergency platforms, and other channels is acquired, including data on rainfall intensity, seismic intensity, flood levels, and surface deformation. Using multi-factor weighting or machine learning methods (e.g., risk estimation models based on gradient boosting or Bayesian inference), a damage potential index for each road segment is comprehensively calculated, reflecting its likelihood of structural damage under the current disaster situation. Simultaneously, the importance factor of each road segment is calculated based on its traffic level, regional substitutability, and key node attributes (e.g., bridges, tunnels, and connecting road sections of transportation hubs). Finally, these two factors are combined to form a priori model of road network disaster impact, establishing a mapping relationship between pre-disaster static attributes and in-disaster dynamic risks. This allows the system to predict the potential risk level and strategic importance of each road segment before a comprehensive disaster assessment. Guided by this priori model, random or average coverage is avoided, improving the operational focus and information acquisition efficiency of the UAV swarm.

[0032] In step S120, priority indicators of each road segment in the target road network are calculated based on the prior model of road network disaster impact, and the inspection task areas and flight paths of multiple drones are dynamically planned and allocated in combination with the status information of available drones, so that high-priority road segments are covered first.

[0033] In some implementations, a road segment priority index is generated by weighted calculation based on the damage potential index and importance factor of each road segment. For example, a weighted linear combination or fuzzy comprehensive evaluation can be used to ensure that the weights of importance and risk can be dynamically adjusted at different stages of the disaster.

[0034] Subsequently, the status information of the drone fleet is acquired, such as the current battery level, payload capacity, communication signal strength, and remaining flight time of each drone. Using this information, a multi-drone mission planning algorithm based on constraint optimization or reinforcement learning is employed to spatially divide the inspection task area and generate paths. The spatial connectivity of the road network is considered during the planning process to avoid repeated flights on the same road segment, and a dynamic update mechanism is used to adjust the paths in real time during drone execution to cope with sudden weather or communication anomalies.

[0035] Therefore, by combining road risk perception with real-time UAV status perception, the risk and importance results in the prior model are transformed into executable inspection strategies for UAVs, so that high-priority areas are covered first. The dynamic allocation mechanism can significantly improve the early coverage rate and information update frequency of high-value road sections, while reducing invalid flights in low-value areas, thereby improving the overall assessment efficiency and operational timeliness.

[0036] In step S130, multiple drones are controlled to fly collaboratively according to the assigned inspection task area and flight path, so as to use the onboard sensors to collect visual sensing data of the road and the surrounding environment.

[0037] Here, drones are also used to extract visual feature information related to road structural damage from the collected visual sensing data to generate corresponding compressed data packets. Multiple drones are used to conduct collaborative flight inspections of the target area, collecting high-resolution data on road damage through airborne visual sensors, and performing feature extraction and data compression on the drone itself, thus reducing communication load at the source.

[0038] In some implementations, the UAV enters the designated inspection area according to an assigned path and uses its onboard high-definition visible light or infrared camera to acquire surface images. Some models can combine structured light or LiDAR sensors to acquire point cloud information, improving the accuracy of 3D structure reconstruction. During the acquisition process, the onboard processing unit performs real-time image preprocessing (including distortion correction, registration, and illumination equalization). Subsequently, visual features related to road damage, such as crack lines, pothole outlines, subsidence edges, and slope slippage features, are extracted based on convolutional neural networks or lightweight edge computing algorithms. The extraction results are encapsulated into compressed data packets in the form of structured feature vectors or codes, retaining only the core feature information related to structural damage assessment, rather than the complete original image data. The compressed data packets are transmitted to the ground control center in real-time or semi-real-time via a wireless link.

[0039] By performing feature extraction and compression on the UAV end, the amount of data transmitted is significantly reduced, avoiding latency or packet loss issues caused by bandwidth limitations, and enabling simultaneous backhaul from multiple UAVs. Meanwhile, since feature extraction is completed near the source, key indicators for damage assessment can be obtained in a very short time, achieving higher real-time performance and scalability.

[0040] In step S140, compressed data packets from multiple drones are received and fused, and a structural damage score reflecting the structural safety status of each road segment is calculated based on the visual feature information corresponding to each road segment.

[0041] In some implementations, the ground control center receives compressed data packets from various UAVs, performs time synchronization and georegistration to ensure consistency of multi-source data for the same road segment. Subsequently, multi-view feature fusion algorithms (e.g., multi-layer feature overlay or confidence-weighted fusion mechanisms) are used to synthesize the observation results from different UAVs. After fusion, the structural damage score for each road segment is calculated based on the extracted fusion features. This score, for example, is calculated using pattern recognition or regression models, reflecting parameters such as crack density, deformation amplitude, and abnormal characteristics of supporting structures, thus forming a unified comprehensive assessment of the road's structural safety status.

[0042] Thus, through a multi-source fusion mechanism, single-view misjudgments and perception blind spots are eliminated, and enhanced spatial consistency across drones is achieved. Structural damage scores not only improve recognition accuracy but also possess spatiotemporal traceability, making road safety status assessment results more stable and reliable.

[0043] In step S150, based on the structural damage score of each road segment and the corresponding road segment importance factor, the traffic adjustment damage index, which characterizes the degree of impact of road damage on traffic operation, is calculated for each road segment, and the road network resilience index, which reflects the degree of decline in the overall traffic capacity of the target road network, is calculated.

[0044] In some implementations, a traffic adjustment damage index is calculated based on the structural damage score and corresponding importance factor of each road segment. This index comprehensively considers factors such as the road segment's topological location in the road network, the feasibility of alternative routes, and the proportion of traffic capacity it carries, to characterize the impact of damage to a particular road segment on overall traffic operation. Furthermore, through network flow analysis methods or simulation calculations, the passability ratio and average travel cost of the entire road network under different damage states are assessed, thus forming a road network resilience index to quantify the degree of decline in the overall road network's capacity and its recovery potential in the post-disaster phase. The road network resilience index provides emergency command departments with a scientific and quantitative basis, guiding the formulation of road closure, detour, and repair sequences, thereby improving the rapid recovery capability and resource allocation efficiency of the post-disaster transportation system.

[0045] Through the embodiments of this application, high timeliness, high accuracy and decision-making output of post-disaster road network damage assessment are achieved. High-risk road sections in disaster areas can be prioritized for identification and coverage, data transmission and processing latency are significantly reduced, structural damage identification results are more accurate, and the overall resilience of the road network is quantitatively characterized, providing real-time and operable decision-making basis for emergency traffic management.

[0046] Figure 2 A flowchart illustrating an example of constructing a priori model of road network disaster impacts according to an embodiment of this application is shown.

[0047] like Figure 2 As shown, in step S210, disaster observation data and basic geological environment data at the time of the corresponding disaster event that match the target road network coverage area are obtained, and a corresponding geological vulnerability field is constructed.

[0048] Specifically, the acquired data is aligned with a unified coordinate system and a unified timestamp, and then resampled to a unified resolution raster set; each original raster layer is normalized within a range, and multiple vulnerability factors are synthesized into a geological vulnerability field based on weight vectors.

[0049] Here, spatial coordinate reprojection is performed based on a unified geographic coordinate system to eliminate coordinate deviations between different data sources; disaster event time series data and geological observation time series data are aligned based on a unified timestamp to ensure physical consistency at the same time. In some implementations, multi-source data are mapped to a raster set with a unified resolution using a bilinear interpolation resampling algorithm to achieve spatial scale uniformity and computational comparability.

[0050] Furthermore, the severity of road damage depends not only on the intensity of the disaster but also on the geological environment's response to the impact. Therefore, a geological vulnerability field is constructed by weighted fusion of multiple vulnerability factors to reflect the comprehensive spatial distribution of the geological environment's sensitivity. Vulnerability factors are parameters used to measure and describe the vulnerability of roads or regions to disaster impacts. They reflect the sensitivity and resilience of a specific area to disasters under different geographical, structural, or environmental conditions, such as soil type, slope, and road type. This is helpful in assessing and predicting the extent of damage and the recovery capacity of a region during a disaster.

[0051] In some implementations, the original geological vulnerability factor grid layer (such as slope, lithology, soil moisture, groundwater depth, etc.) is normalized; an interval linear normalization operator is used to linearly map each factor value to the interval [0,1]; and the geological vulnerability field is calculated by weighted superposition according to the importance of geological features. Equation (1) In the formula, This represents the two-dimensional spatial coordinates of the disaster analysis area in a unified geographic coordinate system. A time index representing the moment of the disaster event, used to indicate the temporal updates of the model; Indicates time Next position The geological vulnerability field function value at a given location is used to describe the sensitivity of the geological environment of the area to disaster impacts; This represents the total number of vulnerability factors used to calculate geological vulnerability. Indicates the first Geological vulnerability factors at location The original raster value at that location, Indicates the first The weighting coefficients of each geological vulnerability factor are used to characterize the relative influence of that factor on the geological vulnerability field. This represents the interval normalization operator, used to linearly map the raster values ​​of each factor to [0,1].

[0052] In equation (1), It can be determined by principal component analysis or expert weighting methods. Furthermore, the construction process of the aforementioned geological vulnerability field can be carried out at each disaster event. This process is repeated to achieve temporal dynamic updates of the geological vulnerability field during disaster evolution. Thus, discrete geological attribute factors are uniformly projected into a continuous vulnerability distribution field, and through adaptive weight allocation, a reasonable expression of the influence degree of different geological factors is achieved.

[0053] In step S220, the intensity corresponding to the disaster type is normalized to the corresponding disaster intensity field. .

[0054] Specifically, Indicates time Next position The disaster intensity distribution field describes the spatial distribution characteristics of disaster energy. The spatial distribution of disaster intensity directly determines the energy density of roads affected by disasters. To achieve a unified scale of calculation, it is necessary to map the physical quantities of disaster types (such as rainfall, seismic intensity, landslide displacement, etc.) into a disaster intensity field.

[0055] In some implementations, an intensity normalization index can be selected based on the type of disaster (e.g., daily rainfall distribution for rainfall intensity, peak ground acceleration (PGA) for earthquakes), mapping the original intensity values ​​to [0,1] using an interval linear normalization formula; for each grid point Calculate the disaster intensity field function It reflects the relative spatial distribution characteristics of disaster energy, realizes a unified characterization method for the energy distribution of different types of disasters, and provides an external driving force input for the geological vulnerability field.

[0056] In step S230, for each road segment in the target road network, a fixed-width buffer polygon of the centerline of the road segment is taken, and the damage potential index is calculated.

[0057] Here, the disaster potential of a road segment depends not only on the intensity of the disaster, but also on the vulnerability of the geological conditions. These factors are spatially coupled, and the disaster risk level of each road segment can be comprehensively calculated through integral calculations.

[0058] Specifically, using the road centerline as a reference, buffer polygon regions of fixed width are generated on both sides of it, and then coupled integration is performed on the disaster intensity field and the geological vulnerability field within the buffer region: Equation (2) In the formula, For indexing road segments in the road network, Indicates the first The buffer polygon area generated by the road segment centerline with a fixed width is used to spatially map the area data to the linear road segment; Represents a buffer zone area, and Positions The disaster intensity field and geological vulnerability field values ​​at the location, Represents the differential area element in a plane coordinate system; Indicates the first Road section at time The damage potential index reflects the degree of coupling between disaster intensity and geological vulnerability in the spatial neighborhood of the road section. The higher the value, the higher the disaster potential.

[0059] In Equation (2), the integral value is normalized by dividing the buffer area to obtain the damage potential index. The higher the value, the stronger the superposition of disaster energy and vulnerability in the spatial neighborhood of the road segment, and the greater the disaster potential. Thus, the mapping of area data to linear road objects is realized by spatial integration, which quantitatively characterizes the disaster-geological coupling intensity of the spatial neighborhood of each road segment.

[0060] In step S240, for each road segment, the baseline traffic flow of the road segment is obtained based on historical traffic volume statistics, the network centrality of the road segment is calculated based on the road network map, and the level index of the road segment is obtained from the road function level. These are then normalized and linearly synthesized into a road segment importance factor.

[0061] In some implementations, the average traffic flow of each road segment is extracted from a historical traffic database. Network centrality indices (such as betweenness centrality and proximity centrality) for each road segment are calculated in the road network map. Level indices are defined based on road function levels (such as arterial roads, secondary arterial roads, and local roads). Finally, the above three factors are normalized and linearly weighted to synthesize the results.

[0062] Equation (3) In the formula, These represent the traffic flow weight, network structure weight, and functional level weight of the road segment, respectively. ; Indicates the first Baseline traffic flow for the road segment Indicates the first The centrality index of a road segment in a road network map is used to measure the degree of impact of the road segment on the overall traffic efficiency of the network. Indicates the first Functional level indicators of road sections Indicates the first Importance factors of road sections; , and These represent the maximum values ​​of traffic flow, network centrality, and grade index across the entire network segment, respectively, and are used for normalization processing.

[0063] In equation (3), the weighting coefficients The importance factor can be determined based on traffic planning experience or through regression learning. The calculation process of the importance factor depends not only on the severity of the disaster, but also on the functional importance of road segments in the network structure and traffic system, making the model closer to the actual post-disaster recovery priority logic.

[0064] In step S250, the combination of the damage potential index and the importance factor of each road segment is used. As prior parameter pairs for road segments at the time of a disaster event, they constitute a prior model of the disaster impact on the road network.

[0065] By combining two core indicators—disaster potential and road importance—a priori model of the disaster impact on the target road network is constructed. Specifically, the disaster impact of each road segment is... As input pairs, attribute mapping relationships are established on the road network topology, with each road segment serving as a node feature vector. This vector is used as a priori influence feature of the road network. Thus, a comprehensive coupled model considering disaster intensity, geological conditions, and the transportation network is constructed, enabling the output prior model to achieve a spatial fusion expression of multidimensional disaster factors.

[0066] Figure 3 An operational flowchart illustrating an example of dynamic task allocation for multiple unmanned aerial vehicles according to an embodiment of this application is shown.

[0067] like Figure 3 As shown, in step S310, the priority index of each road segment is calculated based on the damage potential index and road segment importance factor of each road segment in the prior model of road network disaster impact.

[0068] Road inspections should prioritize covering road sections with higher potential for damage and greater criticality within the network. To this end, a comprehensive priority is calculated based on the damage potential index and road section importance factor given by the prior model, and the comparability of indicators with different dimensions and scales is ensured through normalization and weighting.

[0069] Specifically, for each road segment Read If any indicator is missing, neighborhood interpolation or historical sliding window mean is used to fill the gap, to avoid ranking distortion caused by individual outliers, and the denominator is normalized by a given maximum value to maintain a scale consistent with the prior model.

[0070] Equation (4) In the formula, and These are the damage potential weight and the importance weight, respectively, and satisfy the following conditions: ; Used to characterize the The relative priority of road sections in post-disaster inspections; and These represent the maximum values ​​of the damage potential index and importance factor across all network segments, respectively, and are used for normalization.

[0071] here, and It can be set or adjusted according to actual business needs. For example, a higher setting can be used in the initial rapid investigation phase. (Preferential to disaster relief), while setting higher limits during the traffic recovery phase. (Preferential to network functionality); weights can be preset by an expert database or iteratively updated based on historical assessment and verification results. This unifies disaster risk and network criticality onto the same priority scale, ensuring objective and consistent ranking.

[0072] In addition, robustness measures can be performed to... Perform a light smoothing (such as a first-order adjacency average or a 3-point moving average of adjacent road segments) to reduce the impact of scattered noise on the ranking.

[0073] In step S320, a task scheduling set is constructed based on the priority indicators of each road segment, and a feasible task allocation constraint model for UAVs is established by combining the available flight time, real-time remaining power, payload capacity and communication signal strength of each UAV.

[0074] Here, to avoid incorporating physically unexecutable combinations into the optimization, we first use the hard constraints of drones to determine the feasibility of the "road segment-drone" binary relationship, forming a task scheduling set and an executable mapping. Under the feasibility constraints, we optimize task allocation with the objective of "maximizing the overall benefit of priority coverage".

[0075] Equation (5) The constraints are: Equation (6) In the formula, Indicates the first Does the drone execute the first The inspection task for a road section takes a value of 1 when it is executed, and a value of 0 otherwise. This means that each road segment's task can be assigned to at most one drone at any given time; Indicates the first The drone carried out the first The total voyage of the inspection mission on the road segment is used to constrain its executable mission scope; For the first The average cruising speed of the drone; Indicates the first The available flight time of the drone during the current scheduling cycle. Indicates the first The drone's current remaining battery power. This indicates the minimum safe power threshold required for the drone to perform its mission.

[0076] In some implementations, this is achieved by collecting real-time data on each drone's available flight time, average cruising speed, current remaining battery power, and communication link quality (signal strength, reachable base stations). For each candidate route segment... With drones The following rules apply:

[0077] Range: If the expected flight distance Exceed If the reachable mileage is [not specified], then set [the appropriate mileage]. Not selectable (equivalent to removing the pair); Battery level: If the battery level is below the safe threshold after the task is completed. It was deemed infeasible. Communication: If the task area is not within the communication coverage and the strategy requires real-time feedback, it is deemed infeasible (this factor can be relaxed or ignored in offline data collection scenarios).

[0078] By using the above method, the difficulty of solving the problem is reduced by using prior filtering of infeasible road segment tasks-UAV pairs, and all choices entering the optimizer are combinations that are physically reachable, energy feasible, and link available.

[0079] In step S330, the feasible task allocation constraint model for UAVs is solved to determine the cooperative flight path of multiple UAVs, so that high-priority road segments are covered first.

[0080] In terms of task allocation optimization computation, MILP (Mixed Integer Linear Programming) can be used to find exact solutions when the scale is small; when the scale is large, a two-stage heuristic algorithm is adopted: Phase A: Based on an auction / bid allocation or a modified version of the Hungarian algorithm, an approximate feasible solution is obtained; Phase B: Tabu search / simulated annealing / local 2-opt are used to refine the allocation space and improve the target value. This two-phase heuristic effectively improves the processing efficiency of large-scale road network data.

[0081] Regarding safety and constraint consistency, if the battery level or link status of any drone changes during the solution process, an incremental recalculation is triggered (only the affected drones are recalculated). (a set of tasks), ensuring real-time performance. For tasks that are not assignable due to constraints but... For high-altitude sections of road, a "manual intervention prompt" will be displayed, making them priority targets for emergency response teams or ground vehicles to inspect.

[0082] Through the embodiments of this application, the flight time, power, communication and payload of the scheduled drone are filtered for feasibility before entering the optimization process to avoid unreachable solutions and achieve a cost-benefit balance. In addition, the objective function and priority construction are used to ensure "value priority" and ensure maximum coverage benefits.

[0083] Figure 4 A flowchart illustrating an example of a drone locally generating compressed data packets according to an embodiment of this application is shown.

[0084] like Figure 4 As shown, in step S410, fast filtering and structured optical flow detection are performed on the image data and point cloud data collected by the UAV to identify the road surface area, and the collected current image is compared and analyzed with the pre-disaster baseline image to calculate the damage correlation factor corresponding to each pixel.

[0085] In some implementations, image data and point cloud data simultaneously acquired by a camera and LiDAR on a UAV are spatiotemporally registered to obtain unified projection coordinates. Gaussian fast filtering is applied to the image data to remove noise and enhance edge continuity; radius neighborhood statistical filtering is applied to the point cloud data to eliminate sparse outliers. Furthermore, based on the changes in optical flow field and spatial gradient distribution between adjacent frames, continuous planar regions are extracted; then, combined with the point cloud normal vector distribution, flat surface regions are screened to identify the road surface mask. Finally, the point cloud planar detection results are fused with the image semantic segmentation mask, and a Boolean intersection operation is used to obtain the final binary map of the road region. This ensures that subsequent damage feature extraction focuses only on the road structure region, avoiding the introduction of interfering information and reducing data volume.

[0086] Furthermore, comparing pre-disaster and post-disaster road images can reflect changes in structural surfaces, such as cracks, erosion, and potholes. To achieve this, the relative brightness difference of each pixel between the pre-disaster baseline image and the currently acquired image is calculated, and combined with local gradient weights, to obtain damage-related factors, thereby quantitatively characterizing the potential damage level at each location.

[0087] Specifically, the current image is first spatially registered with the corresponding pre-disaster baseline image, ensuring a one-to-one correspondence between pixels at the same geographical location in the two images. Then, for each pixel location, the following formula is calculated:

[0088] Equation (7) In the formula, Indicates the location The current pixel value at that location, Indicates the location Pre-disaster baseline image pixel values, Indicates the location Pixel gradient magnitude at that location To prevent extremely small positive values ​​with a denominator of zero; Indicates the location Damage-related factors at the site.

[0089] In Equation (7), by combining the difference ratio and gradient weight, it is possible to capture both brightness changes and enhance structural damage features (such as crack edges).

[0090] Preferably, to eliminate the influence of changes in illumination, local mean normalization can be introduced, that is, before calculating the difference, the mean is normalized. and Perform local brightness balancing processing separately. Use a 3×3 mean to verify the initial values. Spatial smoothing is performed to reduce interference from isolated noise points. Illumination normalization and smoothing operations effectively suppress misjudgments of non-structural changes (such as shadows or dust).

[0091] In step S420, based on the statistical distribution of damage-related factors, the sensitivity threshold for damage-sensitive regions is determined using an information entropy adaptive threshold strategy.

[0092] Because image characteristics vary significantly under different disaster scenarios, fixed thresholds are insufficient to accurately segment damage-sensitive areas. Therefore, adaptive damage detection is achieved by statistically analyzing the distribution characteristics in the current images and dynamically determining the judgment threshold based on the principle of information entropy.

[0093] Specifically, the mean and standard deviation of the damage-related factors for all pixels in the current image are calculated, and the threshold function is defined accordingly. Equation (8) In the formula, and These represent the mean and standard deviation of the damage-related factors for all pixels in the current image, respectively. The threshold for sensitive judgment, It is an adjustable coefficient that can be set based on the complexity of the disaster and the sensitivity of the detection (e.g., 0.8 to 1.0 for mild disasters and 1.2 to 1.5 for severe disasters).

[0094] In some implementations, a sliding window is used in the image sequence. Smooth updates are performed to prevent sudden changes between different frames from causing flickering in the detection results, thus achieving smooth threshold updates.

[0095] In step S430, based on the damage-related factors and sensitivity thresholds corresponding to each pixel, the corresponding region is determined as a damage-sensitive region, and a compressed data packet is generated using the visual feature information of the damage-sensitive region.

[0096] Specifically, when At that time, the corresponding area is identified as a damage-sensitive area, and a compressed data packet is generated based on the visual feature information of the damage-sensitive area. To reduce the burden of transmission and storage, only the visual feature information of these high-value areas can be retained, and a compressed data packet can be generated to achieve lightweight information uploading and backend recognition.

[0097] In some implementations, when Pixels are marked as damage-sensitive regions, and adjacent regions are merged using connected component analysis. Morphological closing operations are performed on each connected component to smooth the boundaries and remove isolated noise points to optimize the boundaries.

[0098] Regarding the description of visual feature information, it can be, on the one hand, conventional image pixel statistical features, such as grayscale statistical features (mean, variance, skewness, etc.), texture features (LBP, GLCM contrast, homogeneity), and geometric features (region area, principal direction, shape moments). Then, using region-level feature encoding (such as feature descriptor compression based on binary quantization or vector quantization), the extracted region features are encapsulated into compressed data packets. On the other hand, visual feature information can also include the distribution of damage-related factors extracted from UAV images and their corresponding sensitivity thresholds, which, through combination and encapsulation into compressed data, can support the intuitive display of damage distribution information.

[0099] Furthermore, metadata (including UAV ID, image timestamp, geographic coordinates, and regional confidence level) is appended to each data packet to facilitate fusion and matching by the ground receiver.

[0100] Through the embodiments of this application, targeted compression and efficient transmission of key damaged areas are achieved, significantly reducing data redundancy; in addition, structural, textural, and geometric features are preserved, providing complete input for backend automatic identification and severity assessment.

[0101] Figure 5 A flowchart illustrating an example of calculating structural damage scores according to an embodiment of this application is shown.

[0102] like Figure 5 As shown, in step S510, compressed visual feature data packets from multiple drones are received, and the data packets are geographically mapped and decoded to obtain the distribution of damage-related factors for each road segment and their corresponding sensitivity judgment thresholds.

[0103] In some implementations, the ground receiving platform receives compressed visual feature data packets from multiple UAVs. Each data packet contains a visual feature matrix, image timestamp, UAV pose information, and shooting parameters. During decoding, the original feature matrix is ​​recovered based on the metadata. Based on UAV attitude calculation and GPS / IMU data, the image coordinates are mapped to a unified geographic coordinate system (e.g., WGS84 or UTM). Weighted fusion is performed on the overlapping sampling areas of different UAVs, using a linear weighting method based on viewpoint weights for pixel-level merging, thereby obtaining a globally consistent distribution map of damage-related factors. Furthermore, spatial interpolation is performed on the sensitivity thresholds within each image frame to ensure a continuous transition in threshold distribution between different flight paths.

[0104] This enables seamless stitching of different drone data at the geographic coordinate level, forming a visual damage distribution map covering the entire network, and also avoids local deviations caused by differences in shooting angles, improving the consistency and comparability of damage factors.

[0105] In step S520, for each road segment in the target road network, a damage-sensitive subdomain is determined within the buffer zone of that road segment.

[0106] It should be noted that structural damage to road segments typically exhibits spatial concentration. By extracting high-response damage areas within the buffer zone of each road segment, the spatial distribution range of structural anomalies can be effectively identified.

[0107] Specifically, a fixed-width buffer polygon region is generated for each road centerline. This is used to indicate the detectable spatial range of the road segment.

[0108] Select within the buffer region that satisfies The set of pixels constitutes the damage-sensitive subdomain: Equation (9) In the formula, Indicates the first Damage-sensitive areas within the buffer zone of the road segment.

[0109] In some implementations, if a road segment is covered by multiple overlapping drones, the mean field of their damage-related factors is used for determination to eliminate duplicate detections in overlapping areas. Furthermore, Gaussian filtering or polygon fitting can be used to smooth the boundaries, making the damage area boundaries continuous. This allows for accurate separation of the actual damaged area from background noise on the road surface, preserving the spatial continuity of the damage morphology and improving the geometric accuracy of damage identification.

[0110] In step S530, the proportion of damage area and the average value of the over-threshold intensity are calculated based on the damage-sensitive subdomain.

[0111] The percentage of damaged area represents the proportion of the damaged area of ​​the road relative to the overall buffer zone area, and is an important indicator reflecting the spatial extent of the damage.

[0112] Equation (10) In the formula, Indicates the first The area of ​​the damage-sensitive zone within the buffer zone of the road segment. Indicates the first The percentage of damaged area on the road section.

[0113] In some implementations, pixel-by-pixel statistics are performed on the damage-related factor raster data mapped to geographic coordinates, counting the number of pixels in the damaged area and multiplying it by the cell area, thereby quantitatively reflecting the spatial coverage of road surface damage.

[0114] Furthermore, the damage intensity is quantified by calculating the average excess value of the portion exceeding the threshold within the damage-sensitive subdomain.

[0115] Equation (11) In the formula, Indicates the first The average intensity of exceeding the threshold for road sections.

[0116] Formula (11) effectively characterizes the strength difference and severity within the damaged area, distinguishes between minor cracks (low excess) and deep damage (high excess), and achieves graded judgment.

[0117] In step S540, the structural damage score of each road segment is calculated based on the calculated proportion of damaged area and the average value of the over-threshold intensity.

[0118] It should be noted that the percentage of damaged area and mean intensity These respectively reflect the "range" and "intensity" of the damage. To achieve unified quantification of structural damage, a linearly weighted fusion structural damage scoring model is adopted, combining the two into a comprehensive index according to their weights:

[0119] Equation (12) In the formula, These represent the area proportion fusion weight and the intensity contribution fusion weight, respectively. , Indicates the first Structural damage score of the road segment.

[0120] In equation (12), for scenarios with large-area but shallow damage (such as water damage or flooding), the following is added: For scenarios with concentrated deep damage (such as craters and fractures), improve Weights can be obtained by fitting historical data or dynamically updated by expert experience or machine learning models. Furthermore, weights can also be... Normalized to the [0,1] interval, threshold grading intervals are set (e.g., 0.2, 0.5, 0.8), and classification labels of "mild damage", "moderate damage", and "severe damage" are output to achieve quantitative structural damage assessment through multi-factor fusion. Through the embodiments of this application, a continuous numerical structural safety characterization is obtained by measuring both damage area and damage intensity.

[0121] Figure 6 A flowchart illustrating an example of calculating a road network resilience index according to an embodiment of this application is shown.

[0122] like Figure 6 As shown, in step S610, the traffic adjustment damage index of each road segment is calculated based on the structural damage score of each road segment and the corresponding road segment importance factor.

[0123] The structural damage score describes the degree of damage to a road segment, while the road segment importance factor describes the weight of the impact of the damaged road segment on traffic function within the network. Multiplying these two factors maps structural layer damage to traffic layer impact. Thus, when either factor approaches zero, the traffic impact should also approach zero; and when both factors increase simultaneously, the impact should monotonically increase.

[0124] Specifically, the traffic adjustment damage index for a single road segment is expressed by the following formula: Equation (13) In the formula, Indicates the first Traffic adjustment damage index for road sections.

[0125] Furthermore, indexed by time Archive This facilitates subsequent time-series evaluation and rolling updates.

[0126] In step S620, based on the traffic adjustment damage index of each road segment, the degree of decline in the traffic capacity of the target road network is summarized and calculated to obtain the road network resilience index, which is inversely correlated with the degree of decline in the traffic capacity of the road network.

[0127] It should be noted that the overall performance decline of the road network is related to the intensity of the impact of damaged road segments on traffic and their spatial coverage. Specifically, the traffic adjustment damage index dispersed across various road segments can be aggregated into a single network index by using road segment length as the weight and taking a weighted average across the entire network; and a 1-weighted mean method can be used to make the index inversely correlated with the degree of traffic capacity decline.

[0128] Specifically, the calculation is performed using the following formula: Equation (14) In the formula, This represents a network resilience index for the target road network. For the first The length of the road segment This represents the total number of road segments in the target road network.

[0129] The denominator in equation (14) is the baseline for the entire network length, ensuring scale independence. Furthermore, for temporarily closed or repaired road sections, the traffic adjustment damage index for the corresponding road section can be set to zero, naturally preventing further reduction of the index. After calculation, Archived data serves as a network layer performance indicator. Furthermore, to suppress short-term fluctuations, [measures can be taken]. Perform sliding window smoothing (e.g., averaging over 3-5 evaluation periods).

[0130] In the embodiments of this application, The index is inversely correlated with the degree of traffic capacity reduction; the higher the value, the smaller the overall traffic adjustment damage index for each road segment, indicating that the network as a whole still possesses high connectivity and redundant travel paths. In other words, when the local structural damage or traffic obstruction caused by a disaster is limited, A high value indicates that the road network has good absorption and recovery capabilities, and can maintain overall traffic stability through diversion or detours from adjacent road segments. Conversely, when multiple key road segments are damaged simultaneously and network connectivity decreases, A significant decrease in the value indicates a weakening of the road system's resilience and a marked decline in traffic capacity. Therefore, it can not only quantitatively describe the robustness of the post-disaster road network but also dynamically reflect the impact of disaster damage on the operational performance of the transportation system, providing a basis for subsequent repair priority decisions and resource allocation.

[0131] In some examples of embodiments of this application, when the repair of the first road segment in the target road network is detected, the traffic adjustment damage index corresponding to the first road segment is set to zero to eliminate the impact of the first road segment on task allocation and road network resilience index calculation.

[0132] More specifically, the system receives structural repair status indicators from drone inspection results or road monitoring nodes. When the first [item] in the target road network... The repair status marker for each road segment is set to "Completed," triggering the resilience index update process. The traffic adjustment damage index for that road segment is then adjusted. Set it to 0. This removes the negative impact term for this road segment from the network-wide weighted average, ensuring that the subsequent network resilience index reflects the degree of traffic recovery after repair.

[0133] In addition, when a secondary disaster event is detected in the target road network, the disaster intensity distribution field is recalculated, and the damage potential index of each road segment is corrected based on the updated disaster intensity distribution field to update the road network resilience index.

[0134] Here, when the system receives new disaster observation data (such as rainstorms, floods, landslides, aftershocks, etc.), it determines whether it is a secondary disaster event by comparing timestamps and spatial matching. If it is confirmed to have occurred, it enters the disaster intensity distribution field recalculation process.

[0135] Specifically, based on the latest observational data, a new disaster intensity distribution field is generated using spatial interpolation or geographically weighted regression methods. On this basis, the damage potential index for each road segment is corrected, and the structural damage score and traffic adjustment damage index for each road segment are recalculated based on the corrected damage potential index. Subsequently, network layer indicators are automatically triggered. Updated to reflect the secondary impact of secondary disasters on overall traffic capacity.

[0136] In some implementations, each update is accompanied by a timestamp, and the system retains the corresponding timestamp. This allows for trend analysis. If multiple disasters occur consecutively, the resilience decline rate can be calculated based on version differences. Thus, the system achieves a shift from "static assessment" to "dynamic evolution perception," enabling real-time reflection of the road network's resilience recovery process.

[0137] To verify the effectiveness of the IDAS-RNR (Intelligent Disaster-Damage Assessment System for Road Network Resilience) provided in this application, a series of simulation experiments were designed. By comparing with existing methods, the coverage efficiency, bandwidth consumption, and disaster assessment accuracy were evaluated.

[0138] In terms of the experimental environment, a virtual road network containing 25 road segments (5×5 grid) was constructed, and a random disaster intensity field and vulnerability distribution were generated. Based on the calculated values ​​of each road segment... and This will be the priority. Six drones will be deployed, each with the same flight speed and sensor configuration, with takeoff points distributed around the network. We will implement the following three schemes:

[0139] 1. Baseline Plan: All drones will cover areas according to a simple principle of equal division, with each drone responsible for several consecutive road segments, without considering... and It also does not perform semantic compression.

[0140] 2. Improvement Solution 1: Adopt the NIWDA task allocation algorithm, but directly transmit the complete image back during the data transmission stage without using DRF compression.

[0141] 3. The IDAS-RNR method of this invention: integrates NIWDA (Network-Integrated UAV WorkloadDynamic Allocation) task allocation, DRF (Damage Relevance Factor) semantic compression, and SDS / TDI evaluation model.

[0142] Evaluation metrics include: Coverage time: the time required to complete the inspection of all key road sections; Transmitted data volume: the total amount of data transmitted by the UAV to the ground station; Damage index accuracy: the correlation coefficient between the calculated TDI and the actual damage level (the ground truth value generated by simulation); Network resilience index error: the mean absolute error between the estimated RNRS and the true value.

[0143] Figure 7 A comparative diagram illustrating an example of how inspection coverage time varies with the number of drones is shown.

[0144] like Figure 7 As shown, with the increase in the number of drones, the overall inspection coverage time of the road network shows a significant downward trend, indicating that multi-drone collaborative allocation can effectively improve task execution efficiency. Under the same number of drones, the collaborative algorithm based on the NIWDA task allocation and DRF semantic compression fusion mechanism proposed in this application is superior to the baseline task allocation scheme in terms of coverage efficiency. When the number of drones is small (e.g., 1-2 drones), since this scheme can dynamically allocate task areas according to road priority, avoiding duplicate inspections and task conflicts, the coverage time is significantly shortened compared to the baseline scheme. With the increase in the number of drones, under the condition of 6 drones, the coverage time of the baseline scheme is about 16-17 minutes, while the IDAS-RNR method is about 11-12 minutes, further reducing the coverage time by about 30%. Experimental results demonstrate that the dynamic task balancing and path collaboration based on NIWDA, as well as the reduction of backhaul waiting and scheduling lag by DRF semantic compression, achieve higher coverage efficiency and stability under different drone fleet sizes.

[0145] Figure 8 A schematic diagram illustrating the spatial distribution of a traffic adjustment damage index is shown.

[0146] like Figure 8As shown, the darker the red, the higher the Traffic Adjusted Damage Index (TDI) value of the corresponding road segment, indicating that the combined impact of structural damage and traffic restriction on the road in that area is more significant. The spatial distribution results show that high TDI values ​​in the road network are mainly concentrated in the central and upper regions (such as coordinates (0,0), (1,3), etc.), with TDI values ​​reaching 70-77. This indicates that these road segments not only have high structural damage intensity during the disaster but also occupy a critical position in the traffic network. In contrast, the TDI values ​​in the lower right edge regions (such as coordinates (4,3), (4,4), etc.) are significantly lower, only about 0.4-20, indicating that their impact on post-disaster traffic is relatively small.

[0147] Therefore, the heatmap clearly shows the gradient distribution characteristics of TDI in space, indicating that after fusing the structural damage score (SDS) and the segment importance factor (SIF), this method can reflect the differences in damage and traffic importance of different road units on a global scale, and achieve accurate quantification of the risk of road network capacity decline.

[0148] The simulation results demonstrate the advantages of the embodiments of this application. NIWDA rationally allocates tasks to UAVs while considering both disaster impact and road segment importance, prioritizing the inspection of high-risk and high-importance road segments. DRF semantic compression targets road damage characteristics, avoiding the transmission of irrelevant information, reducing bandwidth requirements, and improving overall efficiency. The structural damage scoring model and traffic adjustment damage index ensure that the assessment results not only focus on the structural damage itself but also consider its actual impact on traffic operation. The network resilience index reflects not only structural damage but also the impact on traffic operation, making the assessment results more business-valued and thus better meeting the decision-making needs of road management departments. Furthermore, this technical solution has broad application scenarios, applicable to various disaster scenarios such as earthquakes, floods, and landslides, and can be extended to damage monitoring of other infrastructure such as bridges and tunnels, providing a scientific basis for post-disaster emergency decision-making.

[0149] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of combined actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Secondly, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application. In the above embodiments, the descriptions of each embodiment have their own emphasis; for parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0150] Figure 9A structural block diagram of an example of a disaster damage assessment system based on drone collaboration according to an embodiment of this application is shown.

[0151] like Figure 9 As shown, the disaster damage assessment system 900 based on UAV collaboration includes a disaster information modeling unit 910, a dynamic task allocation unit 920, a collaborative data acquisition unit 930, a structural damage scoring unit 940, and a resilience index assessment unit 950.

[0152] The disaster information modeling unit 910 is used to construct a priori model of road network disaster impact based on the basic road network data, geological environment information and historical maintenance records of the target road network, and combined with real-time acquired disaster information, to assess the damage potential index and road segment importance factors of roads.

[0153] The dynamic task allocation unit 920 is used to calculate the priority index of each road segment in the target road network according to the prior model of road network disaster impact, and dynamically plan and allocate the inspection task area and flight path of multiple drones in combination with the status information of available drones, so that high priority road segments are covered first.

[0154] The collaborative data acquisition unit 930 is used to control the multiple drones to fly collaboratively according to the assigned inspection task area and flight path, so as to collect visual sensing data of the road and surrounding environment using the onboard sensors; wherein, the drones are also used to extract visual feature information related to road structure damage from the collected visual sensing data to generate corresponding compressed data packets.

[0155] The structural damage scoring unit 940 is used to receive and fuse the compressed data packets from multiple drones, and calculate the structural damage score reflecting the structural safety status of each road segment based on the visual feature information corresponding to each road segment.

[0156] The resilience index assessment unit 950 is used to calculate the traffic adjustment damage index, which characterizes the degree of impact of road damage on traffic operation, based on the structural damage score of each road segment and the corresponding road segment importance factor, and to summarize and calculate the road network resilience index, which reflects the degree of decline in the overall traffic capacity of the target road network.

[0157] In some embodiments, this application provides a non-volatile computer-readable storage medium storing one or more programs including execution instructions. These execution instructions can be read and executed by electronic devices (including but not limited to computers, servers, or network devices) to perform the steps of any of the above-described UAV-based disaster damage assessment methods of this application.

[0158] In some embodiments, this application also provides a computer program product, the computer program product including a computer program stored on a non-volatile computer-readable storage medium, the computer program including program instructions, which, when executed by a computer, cause the computer to perform any of the steps of the above-described UAV-based disaster damage assessment method.

[0159] In some embodiments, this application also provides an electronic device comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform steps of a disaster damage assessment method based on unmanned aerial vehicle (UAV) collaboration.

[0160] The above-described product can perform the methods provided in the embodiments of this application, and has the corresponding functional modules and beneficial effects for performing the methods. Technical details not described in detail in this embodiment can be found in the methods provided in the embodiments of this application.

[0161] The electronic devices in this application can exist in various forms, including but not limited to: mobile communication devices, ultra-mobile personal computer devices, portable entertainment devices, or other airborne electronic devices with data interaction functions.

[0162] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0163] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0164] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for disaster damage assessment based on unmanned aerial vehicle cooperation, characterized in that, The method comprises: Based on the road network basic data of the target road network, geological environment information and historical maintenance records, and combined with real-time obtained disaster information, a road network disaster impact prior model for evaluating road damage potential index and road segment importance factor is constructed; According to the road network disaster impact prior model, the priority index of each road segment in the target road network is calculated, and the patrol task area and flight path of multiple unmanned aerial vehicles are dynamically planned and allocated combined with the state information of the available unmanned aerial vehicles, so that high-priority road segments are preferentially covered; The multiple unmanned aerial vehicles are controlled to fly cooperatively according to the allocated patrol task area and flight path, so as to collect visual sensing data of the road and the surrounding environment by using the sensors carried thereon; wherein the unmanned aerial vehicle is also used to extract visual feature information related to road structure damage from the collected visual sensing data to generate corresponding compressed data packets; The compressed data packets from the multiple unmanned aerial vehicles are received and fused, and the structure damage score reflecting the road structure safety condition of each road segment is calculated based on the visual feature information corresponding to each road segment; Based on the structure damage score of each road segment and the corresponding road segment importance factor, the traffic adjustment damage index of each road segment representing the influence degree of road damage on traffic operation is calculated, and the road network resilience index reflecting the overall traffic capacity reduction degree of the target road network is calculated.

2. The method of claim 1, wherein, The road network disaster impact prior model for evaluating road damage potential index and road segment importance factor is constructed based on the road network basic data of the target road network, geological environment information and historical maintenance records, and combined with real-time obtained disaster information, comprising: Obtain disaster observation data and geological environment basic data corresponding to the target road network coverage area at the disaster event time, and perform uniform coordinate system and uniform timestamp alignment processing, and resample to a uniform resolution grid set; Interval normalization is performed on each original grid layer, and multiple vulnerability factors are integrated into a geological vulnerability field based on a weight vector: ， In the formula, represents a two-dimensional spatial coordinate point of the disaster analysis area in the unified geographic coordinate system, represents a time index of a disaster event moment, used to indicate the temporal update of the model; represents the moment , the geological vulnerability field function value at the location , used to describe the sensitivity of the regional geological environment to the impact of disasters; represents the total number of vulnerability factors used to calculate the geological vulnerability, represents the original grid value of the th geological vulnerability factor at the location , represents the weight coefficient of the th geological vulnerability factor, used to represent the relative influence degree of the factor on the geological vulnerability field; represents an interval normalization operator, used to linearly map each factor grid value to [0, 1]; normalize the intensity corresponding to the disaster species to the corresponding disaster intensity field ; representing the time instant the disaster intensity distribution field at the position , for describing the distribution characteristics of disaster energy in space; For each road segment in the target road network, a fixed-width buffer polygon of the road segment centerline is taken, and the damage potential index is calculated: ， In the formula, For indexing road segments in the road network, Indicates the first The buffer polygon area generated by the road segment centerline with a fixed width is used to spatially map the area data to the linear road segment; Represents a buffer zone area, and Positions The disaster intensity field and geological vulnerability field values ​​at the location, Represents the differential area element in a plane coordinate system; Indicates the first Road section at time The damage potential index reflects the degree of coupling between disaster intensity and geological vulnerability in the spatial neighborhood of the road section; the higher the value, the higher the disaster potential. For each road segment, the baseline traffic flow of the road segment is obtained based on historical statistical traffic volume, the network centrality of the road segment is calculated based on the road network graph, and the grade index of the road segment is obtained based on the road function grade, which is normalized and linearly integrated into the road segment importance factor: ， In the formula, respectively represent the traffic flow weight, network structure weight and function level weight of the road segment, and ; represent the reference traffic flow of the first road segment, represent the reference traffic flow of the first road segment in the road network diagram, to measure the influence degree of the road segment on the overall traffic efficiency of the network; represent the function level index of the first road segment, represent the importance factor of the first road segment; , and respectively represent the maximum value of the traffic flow, network centrality and level index in the entire network road segment, for normalization processing; a combination of the damage potential index of each section and a section importance factor The road network disaster influence prior model is constituted by the prior parameter pair of the section at the moment of the disaster event.

3. The method of claim 2, wherein, The priority index of each road segment in the target road network is calculated according to the road network disaster impact prior model, and the patrol task area and flight path of multiple unmanned aerial vehicles are dynamically planned and allocated combined with the state information of the available unmanned aerial vehicles, comprising: Based on the damage potential index and the road segment importance factor of each road segment in the road network disaster impact prior model, the priority index of each road segment is calculated: ， wherein, are the damage potential weight and the importance weight, respectively, and satisfy ; for characterizing the relative priority of a road segment in post-disaster inspection; and represent the maximum value of the damage potential index and the importance factor in the whole network, respectively, for normalization.​​ A task scheduling set is constructed according to the priority index of each road segment, and a feasible task allocation constraint model of the unmanned aerial vehicle is established combined with the available endurance time, real-time residual power, payload capacity and communication signal strength of each unmanned aerial vehicle: ， The constraint condition is: ， In the formula, Indicates the first Does the drone execute the first The inspection task for a road section takes a value of 1 when it is executed, and a value of 0 otherwise. This means that each road segment's task can be assigned to at most one drone at any given time; Indicates the first The drone carried out the first The total voyage of the inspection mission on the road segment is used to constrain its executable mission scope; For the first The average cruising speed of the drone; Indicates the first The available flight time of the drone during the current scheduling cycle. Indicates the first The drone's current remaining battery power, This indicates the minimum safe power threshold required for a drone to perform a mission; Solving the UAV feasible task allocation constraint model to determine the cooperative flight path of multiple UAVs, so that high-priority road sections are preferentially covered.

4. The method of claim 3, wherein, The visual feature information related to road structure damage is extracted from the collected visual sensing data to generate corresponding compressed data packets, including: For the image data and point cloud data collected by the UAV, fast filtering and structure light flow detection are performed to identify the road surface area, and the collected current image is compared with the reference image before the disaster to calculate the damage-related factor of each pixel: ， wherein represents a current captured pixel value at position , represents a pre-disaster baseline image pixel value at position , represents a pixel gradient magnitude at position , is a small positive value to prevent the denominator from being zero; represents a damage-related factor at position , According to the statistical distribution of the damage-related factor, an information entropy adaptive threshold strategy is used to determine the sensitive judgment threshold of the damage-sensitive area: ， wherein and respectively denote the mean and the standard deviation of the impairment-related factors of all pixels in the current image, is an adjustable coefficient, is a sensitive decision threshold; When the corresponding region is determined as a damage sensitive region, and a compression data packet is generated with visual feature information of the damage sensitive region.

5. The method of claim 4, wherein, The structure damage score reflecting the road structure safety condition of each road section is calculated based on the visual feature information corresponding to each road section, including: Receiving compressed visual feature data packets from multiple UAVs, geographically mapping and decoding the data packets to obtain the damage-related factor distribution and the corresponding sensitive judgment threshold of each road section; For each road section in the target road network, the damage-sensitive sub-domain is determined within the buffer domain of the road section: ， In the formula, indicates the damage-sensitive areas in the buffer zone of the road segment; According to the damage-sensitive sub-domain, the damage area ratio and the average intensity of the threshold value are calculated: ， ， In the formula, represents the first area of the damage-sensitive region in the buffer zone of the road segment, represents the first damage area ratio of the road segment, represents the first average of the super-threshold intensity of the road segment; Based on the calculated damage area ratio and the average intensity of the threshold value, the structure damage score of each road section is calculated: ， In the formula, respectively represent the area ratio fusion weight and the intensity contribution fusion weight, and , represents the structure damage score of the road section.

6. The method of claim 5, wherein, Based on the structure damage score of each road section and the corresponding road section importance factor, the traffic adjustment damage index of each road section is calculated, which represents the influence degree of road damage on traffic operation, and the road network resilience index reflecting the overall traffic capacity reduction degree of the target road network is calculated, including: According to the structure damage score of each road section and the corresponding road section importance factor, the traffic adjustment damage index of each road section is calculated: ， In the formula, represents the first traffic adjustment damage index of the road section; Based on the traffic adjustment damage index of each road section, the traffic capacity reduction degree of the target road network is calculated to obtain the road network resilience index, which is inversely related to the traffic capacity reduction degree of the road network: ， In the formula, a network resilience indicator of the target road network, is the first is the length of the road segment, is the total number of road segments in the target road network.

7. The method according to any one of claims 2-6, characterized in that, After calculating the road network resilience index reflecting the overall traffic capacity reduction degree of the target road network, the method further includes: In the case where it is detected that the first road section in the target road network is repaired, the traffic adjustment damage index corresponding to the first road section is set to zero to eliminate the influence of the first road section on task allocation and road network resilience index calculation; In the case where a secondary disaster event occurs in the target road network, the disaster intensity distribution field is recalculated, and the damage potential index of each road section is corrected based on the updated disaster intensity distribution field to update the road network resilience index.

8. A disaster damage assessment system based on UAV cooperation, characterized in that, The system includes: A disaster information modeling unit is configured to construct a road network disaster impact prior model for evaluating the damage potential index of the road and the road section importance factor based on the road network basic data, geological environment information and historical maintenance records of the target road network, and in combination with real-time acquired disaster information. a dynamic task allocation unit configured to calculate priority indexes of road segments in the target road network according to the road network disaster impact prior model, and dynamically plan and allocate inspection task areas and flight paths of multiple UAVs in combination with state information of the UAVs, so that high-priority road segments are preferentially covered; a cooperative data collection unit configured to control the multiple UAVs to cooperatively fly according to the allocated inspection task areas and flight paths, so as to collect visual sensing data of roads and surrounding environments by using sensors carried by the UAVs; wherein the UAVs are further configured to extract visual feature information related to road structure damage from the collected visual sensing data, so as to generate corresponding compressed data packets; a structure damage scoring unit configured to receive and fuse the compressed data packets from the multiple UAVs, and calculate structure damage scores reflecting safety conditions of road structures of road segments based on the visual feature information corresponding to the road segments; a resilience index evaluation unit configured to calculate traffic adjustment damage indexes of road segments, which represent degrees of influence of road damage on traffic operation, based on structure damage scores of the road segments and corresponding road segment importance factors, and to calculate a road network resilience index reflecting a degree of decline of overall traffic capacity of the target road network by summarizing the traffic adjustment damage indexes.

Images