A method and system for intelligent inspection of municipal roads based on unmanned aerial vehicle (UAV) arrays

Abstract

This invention discloses a method and system for intelligent inspection of municipal roads based on unmanned aerial vehicle (UAV) arrays. The method includes: constructing a municipal digital twin base, fusing multi-source data, and using convolutional neural networks for initial defect screening; employing a hybrid mode of quantum key distribution and elliptic curve encryption to enhance data security during transmission; relying on edge computing nodes to obtain the locations of UAV relay nodes and collaboratively constructing microwave links using multiple modules; measuring link quality and dynamically planning the UAV inspection array; constructing a three-level intelligent diagnostic system based on multi-algorithm coupling and performing defect diagnosis; evaluating inspection results and updating the inspection plan. This application achieves rapid inspection through UAV arrays, ensures data transmission security through encryption technology, improves transmission reliability through dynamic link aggregation, dynamically adjusts the inspection plan, and automatically optimizes the inspection path and UAV array, forming a virtuous cycle and promoting the intelligent development of municipal road inspection.

Description

Technical Field

[0001] This invention relates to the field of intelligent inspection technology using unmanned aerial vehicles (UAVs), and more particularly to an intelligent inspection method and system for municipal roads based on an array of UAVs. Background Technology

[0002] In the field of drone inspection, traditional technologies have many limitations. First, manual inspection is currently the main method, but it is extremely inefficient, making it difficult to cover large areas in a short time and almost impossible to achieve comprehensive inspection. Manual inspection relies heavily on the experience and subjective judgment of the inspectors, which can easily lead to missed inspections, false inspections, and other problems, thus delaying maintenance.

[0003] Traditional data collection relied primarily on simple measuring tools and photographic recording, resulting in limited data volume and a lack of in-depth information, making it difficult to comprehensively reflect the true state of road surface defects. Traditional technologies also lack effective data analysis tools and algorithms, hindering accurate classification, quantitative assessment, and trend prediction of defects. Measures were only taken when defects became severe. Furthermore, traditional inspection plans were typically fixed, following predetermined times and routes, unable to dynamically adjust based on real-time road conditions, defect status, and environmental changes. Traditional inspection plans also lacked consideration for different task priorities, failing to rationally allocate inspection resources based on the severity and urgency of defects.

[0004] Traditional technologies also reveal numerous problems in data transmission. Data transmission is easily limited by the network environment, resulting in slow speeds and high data loss rates. In remote areas or regions with poor network signals, the stability of data transmission is difficult to guarantee. Furthermore, the lack of effective data encryption methods means that initial disease screening data is at risk of being stolen or tampered with.

[0005] This solution addresses many shortcomings of traditional technologies by constructing a municipal digital twin platform, collecting multi-source data, and integrating various information sources, thus solving the problem of insufficient data collection in traditional methods. It optimizes data preprocessing and transmission security by utilizing edge computing and encryption technologies. It improves the flexibility and targeting of inspections through dynamic planning and optimization of inspection plans. By leveraging an AI diagnostic system, it enhances the accuracy and depth of disease diagnosis, overcoming many limitations of traditional inspections. Summary of the Invention

[0006] This invention provides a method for intelligent inspection of municipal roads based on an unmanned aerial vehicle (UAV) array, comprising:

[0007] Integrate geographic information system map data with building information model, collect and associate multi-source data, construct a municipal digital twin base, complete point cloud downsampling at the edge, extract regions of interest from image data, and use lightweight convolutional neural network algorithm to complete the initial screening of diseases;

[0008] A quantum channel is established through a quantum key distribution module, a quantum key is generated and distributed using qubits, and an elliptic curve cryptography algorithm is called to combine the quantum key with the elliptic curve public key to generate a hybrid encryption key, thereby enhancing the security of the transmission of the initial disease screening data.

[0009] Based on the task load and network topology information, the optimal microwave link parameters are calculated, the position of the UAV relay node is calculated, and the position is adjusted to build a microwave link through the collaboration of the flight control system, communication module and airborne sensors, combined with the path planning algorithm. The encrypted data is sent via microwave, and the UAV relay node completes demodulation and error correction, re-encoding and modulation and forwards to the target node to form a complete transmission link.

[0010] A microwave link topology graph is constructed based on graph theory. The link quality is quantified by the edge weight calculation formula. The genetic algorithm and Dijkstra's algorithm are integrated. A weighted graph is constructed by combining link quality and task priority, and an initial inspection path set is generated. The initial population is generated by encoding through genetic algorithm. The global path is optimized iteratively through selection, crossover, and mutation. The local shortest path between nodes is calculated using Dijkstra's algorithm. The global optimization result is combined with the local optimal path to determine the globally optimal inspection route and array.

[0011] Acquire array point cloud data of the target area, compare and match abnormal areas and historical features with the historical disease database, implement target detection algorithm to quickly detect cracks, use point cloud neural network algorithm to assess structural safety, and combine long short-term memory network and time series analysis to predict disease trends.

[0012] Based on the disease diagnosis, assess the inspection results, update the inspection plan, and repeat the first step in a cycle.

[0013] The above-described intelligent municipal road inspection method based on UAV arrays integrates geographic information system map data and building information model, collects and correlates multi-source data, constructs a municipal digital twin base, performs point cloud downsampling at the edge, extracts regions of interest from image data, and uses a lightweight convolutional neural network algorithm to complete the initial screening of road defects, including:

[0014] Construct a digital twin foundation for municipal infrastructure, integrating geographic information system maps, building information models, and inspection data, and annotating information on key infrastructure.

[0015] Point cloud downsampling and region of interest extraction are performed at the edge to retain key feature information, identify diseased areas, and generate preliminary disease screening results.

[0016] The above-described intelligent municipal road inspection method based on UAV arrays includes: constructing a microwave link topology graph based on graph theory; quantifying link quality using edge weight calculation formulas; integrating genetic algorithms and Dijkstra's algorithm; constructing a weighted graph by combining link quality and task priority to generate an initial inspection path set; generating an initial population through genetic algorithm encoding; iteratively optimizing the global path through selection, crossover, and mutation; calculating the local shortest path between nodes using Dijkstra's algorithm; and combining the global optimization result with the local optimal path to determine the globally optimal inspection route and array, including:

[0017] Microwave link topology graphs are constructed based on graph theory to measure link quality and adjust microwave link transmission.

[0018] By combining genetic algorithms and Dijkstra's algorithm, the initial inspection path combination is generated and the fitness is evaluated to generate the optimal inspection route and array.

[0019] The above-described intelligent inspection method for municipal roads based on unmanned aerial vehicle (UAV) arrays includes: acquiring point cloud data of the target area array; comparing and matching abnormal areas with historical defects in a historical defect database; implementing a target detection algorithm for rapid crack detection; using a point cloud neural network algorithm to assess structural safety; and combining long short-term memory networks and time series analysis to predict defect trends.

[0020] Acquire and compare array point cloud data with historical disease database, and synchronize multi-sensor data timestamps;

[0021] Based on the comparison results, a three-level intelligent diagnostic system is used to diagnose diseases.

[0022] The above-described intelligent inspection method for municipal roads based on unmanned aerial vehicle (UAV) arrays includes, based on comparison results, a three-level intelligent diagnostic system for diagnosing road defects, comprising:

[0023] Crack features are identified using target detection algorithms, coordinates and boundary ranges are calculated, and potential disease areas are screened.

[0024] At the regional center, multi-scale feature extraction is performed using point cloud neural network algorithms, and modeling is combined with finite element analysis and structural mechanics principles to assess the overall safety status of the structure.

[0025] By combining long short-term memory networks with time series analysis, we can capture the dynamic patterns of disease development and predict the future spread rate and severity.

[0026] A municipal road intelligent inspection system based on an array of unmanned aerial vehicles (UAVs), comprising:

[0027] Data acquisition module: used to integrate geographic information system map data and building information model, collect and associate multi-source data, build a digital twin base for municipalities, and complete point cloud downsampling at the edge, extract regions of interest from image data, and complete the initial screening of diseases using a lightweight convolutional neural network algorithm;

[0028] Key transmission module: Used to establish a quantum channel through the quantum key distribution module, generate and distribute quantum keys using qubits, call the elliptic curve cryptography algorithm, combine the quantum key with the elliptic curve public key to generate a hybrid encryption key, and enhance the security of the transmission of the initial disease screening data.

[0029] Link construction module: It is used to calculate the optimal microwave link parameters based on the task load and network topology information, calculate the position of the UAV relay node, and adjust the position to build the microwave link through the flight control system, communication module and airborne sensors in collaboration with the path planning algorithm. The encrypted data is transmitted via microwave, and the UAV relay node completes demodulation and error correction, re-encoding and modulation and forwards to the target node to form a complete transmission link.

[0030] Inspection planning module: It is used to construct microwave link topology graph based on graph theory, quantify link quality through edge weight calculation formula, integrate genetic algorithm and Dijkstra algorithm, construct weighted graph and generate initial inspection path set by combining link quality and task priority, generate initial population through genetic algorithm encoding, optimize global path through selection, crossover and mutation iteration, calculate local shortest path between nodes using Dijkstra algorithm, and combine global optimization result with local optimal path to determine global optimal inspection route and array;

[0031] Intelligent diagnostic module: used to acquire array point cloud data of target area, compare and match abnormal areas and historical features with historical disease database, implement target detection algorithm to quickly detect cracks, use point cloud neural network algorithm to assess structural safety, combine long short-term memory network and time series analysis to predict disease trend, evaluate inspection results, and update inspection plan.

[0032] The aforementioned intelligent municipal road inspection system based on a drone array includes a data acquisition module that specifically comprises:

[0033] Twin Model Construction Submodule: Used to build a digital twin foundation for municipal infrastructure, integrating geographic information system maps, building information models and inspection data, and labeling key infrastructure information;

[0034] The initial disease screening module is used to perform point cloud downsampling at the edge, extract regions of interest from the image, retain key feature information, identify disease areas, and generate initial disease screening results.

[0035] The aforementioned intelligent municipal road inspection system based on a drone array includes, in particular, an inspection planning module comprising:

[0036] Link quality calculation submodule: used to construct microwave link topology graphs based on graph theory, measure link quality, and adjust microwave link transmission;

[0037] The array planning submodule is used to integrate the genetic algorithm and the Dijkstra algorithm, encode and generate the initial inspection path combination, evaluate the fitness, and generate the optimal inspection route and array.

[0038] The aforementioned intelligent municipal road inspection system based on a drone array includes, in particular, an intelligent diagnostic module comprising:

[0039] Array point cloud data synchronization submodule: used to acquire and compare array point cloud data with historical disease database, and synchronize multi-sensor data timestamps;

[0040] Diagnosis and prediction submodule: Based on the comparison results, it is used to quickly detect cracks, assess structural safety, and predict disease trends using a three-level intelligent diagnostic system.

[0041] The beneficial effects achieved by this invention are as follows:

[0042] This solution achieves rapid, large-area inspection using a drone array, significantly improving the speed and accuracy of defect detection by combining multi-source data acquisition and AI diagnostic technology. A hybrid mode of quantum key distribution and elliptic curve cryptography enhances the security of data transmission during initial defect screening, while a dynamic link aggregation strategy intelligently selects transmission links to improve data transmission reliability and reduce the risk of data loss. The inspection plan is dynamically adjusted using robot learning algorithms and multi-objective optimization algorithms to determine the optimal inspection route and array. A three-tiered AI diagnostic system combines multiple advanced algorithms to quickly detect defects, assess structural safety, and predict defect trends. Based on inspection results and defect diagnoses, this solution automatically optimizes inspection paths and drone arrays, creating a virtuous cycle that promotes the intelligent development of municipal road inspection systems and provides more accurate data support for urban planning, construction, and management. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0044] Figure 1 This is a flowchart of an intelligent inspection method for municipal roads based on an unmanned aerial vehicle (UAV) array, provided in Embodiment 1 of this application.

[0045] Figure 2 This is a schematic diagram of a municipal road intelligent inspection system based on an unmanned aerial vehicle (UAV) array, provided in Embodiment 2 of this application. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] Example 1

[0048] like Figure 1 As shown, Embodiment 1 of this application provides a method for intelligent inspection of municipal roads based on an unmanned aerial vehicle (UAV) array, including:

[0049] S110: Integrates geographic information system map data and building information model, collects and associates multi-source data, constructs a municipal digital twin base, completes point cloud downsampling at the edge, extracts regions of interest from image data, and uses a lightweight convolutional neural network algorithm to complete the initial screening of diseases;

[0050] The multi-source data for inspection includes at least drone inspection data and urban planning and deployment data.

[0051] The system constructs a digital twin foundation for the municipality, integrates GIS maps and BIM models, and imports urban planning data. It controls drones to collect inspection data, performs preprocessing such as point cloud downsampling and image ROI extraction, and uses a lightweight CNN algorithm to complete the initial screening of defects. It combines urban planning data to eliminate false positives, stores the initial screening results, and feeds them back to the user interface.

[0052] The process of building a municipal digital twin platform, collecting multi-source data, and conducting initial disease screening includes the following sub-steps:

[0053] S111: Construct a digital twin foundation for municipal infrastructure, integrating geographic information system maps, building information models, and inspection data, and marking information on key infrastructure.

[0054] High-precision Geographic Information System (GIS) map data is loaded as the spatial framework foundation. Building Information Modeling (BIM) is imported to spatially align and integrate the three-dimensional structural information of urban infrastructure with the GIS map, achieving seamless connection under a unified coordinate system.

[0055] Next, the system imports the inspection data into the base and links it with GIS and BIM data. Critical infrastructure information, such as bridges, tunnels, and important intersections, is labeled, and their location and attributes are highlighted through a visual interface.

[0056] S112: Perform point cloud downsampling and image region of interest extraction at the edge, retain key feature information, identify diseased areas, and generate preliminary disease screening results;

[0057] After receiving the raw data collected by the drone at the edge, the point cloud downsampling module is activated. The algorithm removes redundant array point cloud data, reduces the amount of data, and retains key feature information.

[0058] Regions of interest (ROI) are extracted from image data to identify and crop out areas that may contain diseases, reducing the data size for subsequent processing.

[0059] The lightweight convolutional neural network (CNN) algorithm is invoked, and a pre-trained model deployed at the edge is used to identify disease features in the processed data, quickly determine whether there are diseases such as cracks and pits, and obtain the initial screening results of diseases.

[0060] The initial disease screening results and related data are compressed and packaged, and then transmitted to the central server via the network.

[0061] S120: Establishes a quantum channel through the quantum key distribution module, generates and distributes quantum keys using qubits, calls the elliptic curve cryptography algorithm, combines the quantum key with the elliptic curve public key to generate a hybrid encryption key, and enhances the security of the transmission of the initial disease screening data.

[0062] The system establishes a quantum channel through a quantum key distribution (QKD) module, leveraging the non-cloning and non-measurability of qubits to generate and distribute a set of quantum keys in real time. After quantum key distribution is complete, the system divides the quantum key into multiple subkey fragments and stores them in a secure key pool.

[0063] The system invokes the Elliptic Curve Cryptography (ECC) module to generate a public-private key pair based on pre-negotiated elliptic curve parameters. The quantum key fragment is then combined with the elliptic curve public key to generate a hybrid encryption key. During data encryption, the system first uses the hybrid encryption key to perform symmetric encryption on the initial disease screening data, generating encrypted data blocks.

[0064] Simultaneously, the system performs segmented encryption on the elliptic curve private key, using quantum key fragments to encrypt each segment. The encrypted data and private key segments are transmitted to the receiving end via a secure channel. The receiving system first uses the quantum key fragments to decrypt the elliptic curve private key, recovering the complete private key information.

[0065] The hybrid encryption key is regenerated using elliptic curve private key and quantum key fragments. The encrypted data block is then decrypted, and after verifying the data integrity and security, the data is transmitted to the link for decryption to obtain the initial disease screening.

[0066] S130: Calculates the optimal microwave link parameters based on the task load and network topology information, calculates the location of the UAV relay node, and adjusts the location to build a microwave link through the collaboration of the flight control system, communication module and airborne sensors, combined with the path planning algorithm. The encrypted data is transmitted via microwave. The UAV relay node completes demodulation and error correction, re-encodes and modulates, and forwards to the target node to form a complete transmission link.

[0067] The system first uses the resource management module of the edge computing node to calculate the optimal microwave link parameter configuration based on the current task load and network topology information, and then sends these parameter commands to the UAV relay node through control signals.

[0068] By working in concert with the flight control system and communication module of the UAV relay node, the UAV relay node uses onboard sensors to perceive the surrounding environment and combines a preset flight path planning algorithm to precisely adjust the position and antenna pointing angle of the UAV relay node, forming an unobstructed, high-gain microwave transmission path with the edge computing node.

[0069] Meanwhile, the system needs to acquire node data and perform data preprocessing such as normalization based on the transmission requirements of the link and environmental factors; and dynamically calculate the position of the UAV relay node among each node. The objective function for the relay node position is: in, The balance coefficient represents the effect of the sum of squared distances between dominant nodes; m represents the total number of existing nodes. The elements of the dynamic weight matrix represent the inspection priority association weights between the i-th and j-th existing nodes; The elements of the asymmetric attenuation coefficient matrix represent the anisotropic attenuation of signal transmission from node i to node j by the environment; This represents the distance between the i-th node and the j-th existing node; The impact of dominant energy consumption items; This represents the energy consumption index, which couples the UAV's flight energy consumption with its signal transmission power. This represents the distance from the i-th existing node to the relay node, where i represents the existing node and r represents the relay node. ; The terrain complexity factor is used as an adjustment parameter for the distance power factor; k: penalty coefficient, which controls the degree of delay constraint and forces relay nodes to move closer to meet the delay requirements; This represents the truncation function, which is applied when the total transmission distance is... Exceed At that time, the penalty item is activated to forcibly constrain the total signal transmission delay and prevent communication timeouts caused by relay nodes being deployed too far away; This represents the time delay tolerance threshold. This represents the signal propagation constant.

[0070] Differentiate the coordinates of the relay node with respect to the partial derivative, set the partial derivative to 0, and solve for the coordinate expression. Let... The coordinates of the relay node can be obtained: Simultaneously, the communication module of the edge computing node is activated, encoding and modulating the data to be transmitted according to the preset microwave link parameters, and sending it to the UAV relay node via microwave signal. Its communication processing module demodulates and corrects the signal, then re-encodes and modulates the signal according to the address information of the target node, and forwards the data to the target node via microwave link.

[0071] Throughout the entire link operation, the system monitors the link status in real time through the monitoring module, including key indicators such as signal strength and bit error rate. Once an anomaly is detected, such as signal attenuation or an increase in bit error rate, the system will trigger a link optimization mechanism, in which the edge computing nodes and UAV relay nodes work together to adjust the link parameters.

[0072] S140: Construct a microwave link topology graph based on graph theory, quantify link quality through edge weight calculation formula, integrate genetic algorithm and Dijkstra algorithm, construct a weighted graph by combining link quality and task priority and generate an initial inspection path set, generate an initial population by encoding through genetic algorithm, optimize the global path through selection, crossover and mutation iteration, calculate the local shortest path between nodes using Dijkstra algorithm, combine the global optimization result with the local optimal path to determine the global optimal inspection route and array;

[0073] The system constructs a microwave link topology map in real time and measures link quality using an edge weight calculation formula. It utilizes robot learning algorithms to analyze and predict link states, and dynamically plans UAV inspection schemes and arrays.

[0074] A microwave link topology graph is constructed based on graph theory. Link quality is quantified using edge weight calculation formulas. A weighted graph is constructed by integrating genetic algorithms and Dijkstra's algorithm, combining link quality and task priority to generate an initial inspection path set. An initial population is generated using genetic algorithm encoding. Global paths are iteratively optimized through selection, crossover, and mutation. Local shortest paths between nodes are calculated using Dijkstra's algorithm. The global optimization results are combined with the local optimal paths to determine the globally optimal inspection route and array. The specific steps include the following:

[0075] S141: Construct microwave link topology graphs based on graph theory, measure link quality, and adjust microwave link transmission;

[0076] A topology graph for microwave links is constructed based on graph theory, with nodes in the link (such as edge computing nodes, UAV relay nodes, etc.) as vertices and microwave connections between nodes represented as edges. Link quality is then measured using an edge weight calculation formula, as follows: Where n represents the number of link quality assessment metrics; This represents the value of the quality assessment metric for the i-th link. Indicates the first The fuzzy membership function values ​​of each indicator; This represents the Riemann curvature tensor, which reflects the degree of curvature in the link propagation space; The determinant of the Riemann curvature tensor is used to quantify the overall impact of spatial curvature on link quality; m represents the number of bit error rate sampling points. This represents the weight of the j-th bit error rate sampling point; This represents the bit error rate value at the j-th bit error rate sampling point; The position error penalty coefficient is used to adjust the degree of impact of the position error of the UAV relay node on the link quality; K represents the number of UAV relay nodes participating in signal transmission in the microwave link; This represents the actual position vector of the k-th UAV relay node; This represents the ideal position vector of the kth UAV relay node; , where S represents the ideal position vector of the kth UAV relay node; This represents a function that describes the distribution of signal strength in a two-dimensional projected region.

[0077] In this system, edges with higher weights indicate better link quality and are suitable for data transmission; edges with lower weights indicate poorer link quality and require optimization or adjustment. The system dynamically adjusts the microwave link topology based on edge weights, prioritizing links with higher weights for data transmission, thereby optimizing the overall performance and reliability of the microwave link.

[0078] S142: Combines genetic algorithm and Dijkstra algorithm to encode and generate initial inspection path combinations and perform fitness evaluation to generate the optimal inspection route and array;

[0079] The system constructs a weighted graph based on the link quality index, uses task priority as the weight adjustment coefficient, and generates an initial set of inspection paths. A genetic algorithm is introduced to generate an initial population through encoding, with each chromosome representing a possible combination of inspection paths.

[0080] The population undergoes fitness evaluation, with the fitness function considering both total path weight and task priority, prioritizing paths with high link quality and high task priority. Through selection, crossover, and mutation operations, the genetic algorithm iteratively optimizes the population, gradually approaching the global optimum.

[0081] Meanwhile, the Dijkstra algorithm is used to calculate the shortest path between each node, and the optimal local path is determined in the path combination optimized by the genetic algorithm.

[0082] By combining the global optimization results obtained from the genetic algorithm with the local optimal paths from the Dijkstra algorithm, the path weights are dynamically adjusted according to the task priority, and finally the globally optimal inspection route and array are determined.

[0083] S150: Acquire array point cloud data of the target area, compare and match abnormal areas and historical features with the historical disease database, implement target detection algorithm to quickly detect cracks, use point cloud neural network algorithm to assess structural safety, and combine long short-term memory network and time series analysis to predict disease trends.

[0084] The system acquires array point cloud data of the target area and compares it with data in a historical disease database. Through a three-tiered intelligent diagnostic system—preliminary screening, feature analysis, and deep diagnosis—diseases are gradually identified and classified. Potential disease areas are screened using preset thresholds, features are extracted and matched with historical data, and a deep learning model is used for precise diagnosis, outputting the disease type and severity.

[0085] Acquire array point cloud data of the target area, compare and match abnormal areas and historical features with a historical disease database, implement target detection algorithms for rapid crack detection, use point cloud neural network algorithms to assess structural safety, and combine long short-term memory networks and time series analysis to predict disease trends. The specific steps include the following:

[0086] S151: Acquire and compare array point cloud data with historical disease database, and synchronize multi-sensor data timestamps;

[0087] The system collects array point cloud data of the target area and acquires multi-sensor data in real time. It compares the array point cloud data with a historical disease database to obtain the comparison results; simultaneously, it performs timestamp synchronization processing on the multi-sensor data. It accurately matches abnormal areas in the array point cloud data with historical disease characteristics and performs comprehensive analysis based on multi-sensor information.

[0088] S152: Based on the comparison results, disease diagnosis is carried out with the help of a three-level intelligent diagnostic system;

[0089] Based on the comparison results between the array point cloud data and the historical disease database, disease diagnosis is carried out according to a three-level intelligent diagnosis process. The target detection algorithm (YOLOv8) is used to quickly detect cracks at the edge and initially locate the disease area; a point cloud neural network algorithm (PointNet++) is used to conduct a detailed assessment of structural safety at the center of the region and analyze the impact of the disease on the structure; finally, the development trend of the disease is predicted by combining Long Short-Term Memory (LSTM) network and time series analysis.

[0090] Based on the comparison results, a three-level intelligent diagnostic system is used to diagnose diseases, which includes the following sub-steps:

[0091] S1521: Identify crack features using a target detection algorithm, calculate coordinates and boundary range, and screen potential disease areas;

[0092] At the edge, the system deploys target detection algorithms to rapidly process the acquired images or array point cloud data. A pre-trained deep learning model identifies crack features, and a convolutional neural network extracts the texture and shape features of the cracks. Once a crack is detected, its center coordinates and boundary extent are accurately calculated, and its location information is labeled in the data. Detection results are provided in real time, quickly identifying potential disease areas.

[0093] S1522: Multi-scale feature extraction is performed at the center of the region using a point cloud neural network algorithm. Combined with finite element analysis and structural mechanics principles, the overall safety status of the structure is assessed.

[0094] At the center of the region, the system performs an in-depth assessment of structural safety using a point cloud neural network algorithm. First, multi-scale feature extraction is performed on the arrayed point cloud data to identify potential defects in the structure, such as micro-cracks, deformation areas, and potential stress concentration points.

[0095] The extracted features are mechanically modeled using finite element analysis combined with structural mechanics principles to calculate the stress distribution, deformation degree, and stability indices of the structure under the current defect state. Based on the analysis results, the overall safety status of the structure is comprehensively assessed, the existence of potential safety hazards is determined, and a detailed assessment report is generated.

[0096] S1523: Utilizing long short-term memory networks combined with time series analysis, we can capture the dynamic patterns of disease development and predict the future spread rate and severity.

[0097] Time-series information on disease characteristics, such as crack width, disease area, and structural deformation, is extracted from inspection data. The time-series data is preprocessed, including normalization and missing value imputation. The processed data is then input into an LSTM network, leveraging its powerful long short-term memory to capture the dynamic patterns of disease development.

[0098] LSTM learns the dependencies in time series data to identify key turning points and trend changes in disease development. Combined with time series analysis methods, such as the Autoregressive Moving Average (ARMA) model, the accuracy of disease trend prediction is enhanced. Finally, based on the results of LSTM and time series analysis, the development trend of the disease over a future period is predicted, including the rate of disease spread and potential severity.

[0099] S160: Based on the disease diagnosis, evaluate the inspection results, update the inspection plan, and repeat S110.

[0100] Based on the three-level intelligent disease diagnosis results and inspection data, the inspection effect is comprehensively evaluated; the digital twin model of municipal facilities and disease feature library are automatically updated according to the evaluation results, the inspection path and drone array are optimized, and S110 is continued.

[0101] Example 2

[0102] like Figure 2 As shown, Embodiment 2 of this application provides a municipal road intelligent inspection system based on an unmanned aerial vehicle (UAV) array, comprising:

[0103] Data acquisition module 21: It is used to integrate geographic information system map data and building information model, collect and associate multi-source data, build a municipal digital twin base, and complete point cloud downsampling at the edge, extract regions of interest from image data, and complete the initial screening of diseases using a lightweight convolutional neural network algorithm;

[0104] Twin Model Construction Submodule 211: Used to build a municipal digital twin foundation, integrating geographic information system maps, building information models and inspection data, and marking key infrastructure information;

[0105] Disease screening module 212: Used for point cloud downsampling at the edge, extraction of regions of interest in the image, preservation of key feature information, identification of disease areas, and generation of disease screening results.

[0106] Key transmission module 22: Used to establish a quantum channel through the quantum key distribution module, generate and distribute quantum keys using qubits, call the elliptic curve cryptography algorithm, combine the quantum key with the elliptic curve public key to generate a hybrid encryption key, and enhance the security of the transmission of the initial disease screening data.

[0107] Link construction module 23: It is used to calculate the optimal microwave link parameters based on the task load and network topology information, calculate the position of the UAV relay node, and build a microwave link by coordinating the flight control system, communication module and airborne sensors, combined with the path planning algorithm to adjust the position. The encrypted data is sent via microwave, and the UAV relay node completes demodulation and error correction, re-encoding and modulation and forwards to the target node to form a complete transmission link.

[0108] Inspection planning module 24: It is used to construct a microwave link topology graph based on graph theory, quantify the link quality through the edge weight calculation formula, integrate the genetic algorithm and Dijkstra algorithm, construct a weighted graph by combining link quality and task priority, and generate an initial inspection path set. The initial population is generated by encoding through the genetic algorithm, and the global path is iteratively optimized through selection, crossover, and mutation. The local shortest path between nodes is calculated using the Dijkstra algorithm. The global optimization result is combined with the local optimal path to determine the globally optimal inspection route and array.

[0109] Link quality calculation submodule 241: used to construct microwave link topology graphs based on graph theory, measure link quality, and adjust microwave link transmission;

[0110] Array planning submodule 242: used to integrate genetic algorithm and Dijkstra algorithm, encode and generate initial inspection path combinations and perform fitness evaluation, and generate optimal inspection route and array.

[0111] Intelligent Diagnosis Module 25: Used to acquire array point cloud data of target area, compare and match abnormal areas and historical features with historical disease database, implement target detection algorithm for rapid crack detection, use point cloud neural network algorithm to assess structural safety, combine long short-term memory network and time series analysis to predict disease trend, evaluate inspection results, update inspection plan, and execute S110 in a loop.

[0112] Array point cloud data synchronization submodule 251: used to acquire and compare array point cloud data with historical disease database, and synchronize multi-sensor data timestamps;

[0113] Diagnostic prediction submodule 252: Based on the comparison results, it is used to perform rapid crack detection, structural safety assessment and disease trend prediction with the help of a three-level intelligent diagnostic system.

[0114] Corresponding to the above embodiments, the present invention provides a computer storage medium, including: at least one memory and at least one processor;

[0115] The memory is used to store one or more program instructions;

[0116] A processor for running one or more program instructions to execute a method and system for intelligent inspection of municipal roads based on an array of unmanned aerial vehicles (UAVs).

[0117] Corresponding to the above embodiments, this embodiment of the invention provides a computer-readable storage medium containing one or more program instructions, which are executed by a processor to provide a method and system for intelligent inspection of municipal roads based on an unmanned aerial vehicle (UAV) array.

[0118] The embodiments disclosed in this invention provide a computer-readable storage medium storing computer program instructions. When the computer program instructions are executed on a computer, the computer executes the above-described intelligent inspection method and system for municipal roads based on an unmanned aerial vehicle (UAV) array.

[0119] In this embodiment of the invention, the processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0120] The various methods, steps, and logic diagrams disclosed in the embodiments of this invention can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The processor reads information from the storage medium and, in conjunction with its hardware, completes the steps of the above methods.

[0121] The storage medium can be memory, such as volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.

[0122] Among them, non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory.

[0123] Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (Synchlink DRAM, SLDRAM), and direct memory bus RAM (DRRAM).

[0124] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0125] Those skilled in the art will recognize that, in one or more of the examples above, the functions described in this invention can be implemented using a combination of hardware and software. When applied as software, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of computer programs from one place to another. Storage media can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0126] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent inspection of municipal roads based on unmanned aerial vehicle (UAV) arrays, characterized in that, include: Integrate geographic information system map data with building information model, collect and associate multi-source data, construct a municipal digital twin base, complete point cloud downsampling at the edge, extract regions of interest from image data, and use lightweight convolutional neural network algorithm to complete the initial screening of diseases; A quantum channel is established through a quantum key distribution module, a quantum key is generated and distributed using qubits, and an elliptic curve cryptography algorithm is called to combine the quantum key with the elliptic curve public key to generate a hybrid encryption key, thereby enhancing the security of the transmission of the initial disease screening data. Based on the task load and network topology information, the optimal microwave link parameters are calculated, the position of the UAV relay node is calculated, and the position is adjusted to build a microwave link through the collaboration of the flight control system, communication module and airborne sensors, combined with the path planning algorithm. The encrypted data is sent via microwave, and the UAV relay node completes demodulation and error correction, re-encoding and modulation and forwards to the target node to form a complete transmission link. A microwave link topology graph is constructed based on graph theory. The link quality is quantified by the edge weight calculation formula. The genetic algorithm and Dijkstra's algorithm are integrated. A weighted graph is constructed by combining link quality and task priority, and an initial inspection path set is generated. The initial population is generated by encoding through genetic algorithm. The global path is optimized iteratively through selection, crossover, and mutation. The local shortest path between nodes is calculated using Dijkstra's algorithm. The global optimization result is combined with the local optimal path to determine the globally optimal inspection route and array. Acquire array point cloud data of the target area, compare and match abnormal areas and historical features with the historical disease database, implement target detection algorithm to quickly detect cracks, use point cloud neural network algorithm to assess structural safety, and combine long short-term memory network and time series analysis to predict disease trends. Based on the disease diagnosis, assess the inspection results, update the inspection plan, and repeat the first step in a cycle. Specifically, a topology graph of the microwave link is constructed based on graph theory, with nodes in the link as vertices and microwave connections between nodes represented as edges. Link quality is then measured using an edge weight calculation formula, as follows: Where n represents the number of link quality assessment metrics; This represents the value of the quality assessment metric for the i-th link. Indicates the first The fuzzy membership function values ​​of each indicator; This represents the Riemann curvature tensor, which reflects the degree of curvature in the link propagation space; The determinant of the Riemann curvature tensor is used to quantify the overall impact of spatial curvature on link quality. m represents the number of bit error rate sampling points; This represents the weight of the j-th bit error rate sampling point; This represents the bit error rate value at the j-th bit error rate sampling point; The position error penalty coefficient is used to adjust the degree of impact of the position error of the UAV relay node on the link quality; K represents the number of UAV relay nodes participating in signal transmission in the microwave link; This represents the actual position vector of the k-th UAV relay node; This represents the ideal position vector of the kth UAV relay node; , where S represents the ideal position vector of the kth UAV relay node; This represents a function describing the distribution of signal intensity in a two-dimensional projected region; Edges with higher weights indicate better link quality and are suitable for data transmission; edges with lower weights indicate poorer link quality and require optimization or adjustment. The system dynamically adjusts the microwave link topology based on edge weights, prioritizing links with higher weights for data transmission, thereby optimizing the performance and reliability of the entire microwave link.

2. The intelligent inspection method for municipal roads based on unmanned aerial vehicle (UAV) arrays according to claim 1, characterized in that, Integrating geographic information system map data with building information models, collecting and associating multi-source data, constructing a municipal digital twin foundation, and performing point cloud downsampling at the edge, extracting regions of interest from image data, and using a lightweight convolutional neural network algorithm to complete the initial screening of defects, including: Construct a digital twin foundation for municipal infrastructure, integrating geographic information system maps, building information models, and inspection data, and annotating information on key infrastructure. Point cloud downsampling and region of interest extraction are performed at the edge to retain key feature information, identify diseased areas, and generate preliminary disease screening results.

3. The intelligent inspection method for municipal roads based on unmanned aerial vehicle (UAV) arrays according to claim 1, characterized in that, A microwave link topology graph is constructed based on graph theory. Link quality is quantified using edge weight calculation formulas. A weighted graph is constructed by integrating genetic algorithms and Dijkstra's algorithm, combining link quality and task priority to generate an initial inspection path set. An initial population is generated through genetic algorithm encoding. Global paths are iteratively optimized through selection, crossover, and mutation. Local shortest paths between nodes are calculated using Dijkstra's algorithm. The global optimization results are combined with the local optimal paths to determine the globally optimal inspection route and array, including: Microwave link topology graphs are constructed based on graph theory to measure link quality and adjust microwave link transmission. By combining genetic algorithms and Dijkstra's algorithm, the initial inspection path combination is generated and the fitness is evaluated to generate the optimal inspection route and array.

4. The intelligent inspection method for municipal roads based on unmanned aerial vehicle (UAV) arrays according to claim 1, characterized in that, Acquire array point cloud data of the target area, compare and match abnormal areas and historical features with a historical disease database, implement target detection algorithms for rapid crack detection, utilize point cloud neural network algorithms to assess structural safety, and combine long short-term memory networks and time series analysis to predict disease trends, including: Acquire and compare array point cloud data with historical disease database, and synchronize multi-sensor data timestamps; Based on the comparison results, a three-level intelligent diagnostic system is used to diagnose diseases.

5. A method for intelligent inspection of municipal roads based on an unmanned aerial vehicle (UAV) array according to claim 4, characterized in that, Based on the comparison results, a three-level intelligent diagnostic system is used to diagnose diseases, including: Crack features are identified using target detection algorithms, coordinates and boundary ranges are calculated, and potential disease areas are screened. At the regional center, multi-scale feature extraction is performed using point cloud neural network algorithms, and modeling is combined with finite element analysis and structural mechanics principles to assess the overall safety status of the structure. By combining long short-term memory networks with time series analysis, we can capture the dynamic patterns of disease development and predict the future spread rate and severity.

6. A municipal road intelligent inspection system based on unmanned aerial vehicle (UAV) arrays, characterized in that, include: Data acquisition module: used to integrate geographic information system map data and building information model, collect and associate multi-source data, build a digital twin base for municipalities, and complete point cloud downsampling at the edge, extract regions of interest from image data, and complete the initial screening of diseases using a lightweight convolutional neural network algorithm; Key transmission module: Used to establish a quantum channel through the quantum key distribution module, generate and distribute quantum keys using qubits, call the elliptic curve cryptography algorithm, combine the quantum key with the elliptic curve public key to generate a hybrid encryption key, and enhance the security of the transmission of the initial disease screening data. Link construction module: It is used to calculate the optimal microwave link parameters based on the task load and network topology information, calculate the position of the UAV relay node, and adjust the position to build the microwave link through the flight control system, communication module and airborne sensors in collaboration with the path planning algorithm. The encrypted data is transmitted via microwave, and the UAV relay node completes demodulation and error correction, re-encoding and modulation and forwards to the target node to form a complete transmission link. Inspection planning module: It is used to construct microwave link topology graph based on graph theory, quantify link quality through edge weight calculation formula, integrate genetic algorithm and Dijkstra algorithm, construct weighted graph and generate initial inspection path set by combining link quality and task priority, generate initial population through genetic algorithm encoding, optimize global path through selection, crossover and mutation iteration, calculate local shortest path between nodes using Dijkstra algorithm, and combine global optimization result with local optimal path to determine global optimal inspection route and array; Intelligent diagnostic module: used to acquire array point cloud data of target area, compare and match abnormal areas and historical features with historical disease database, implement target detection algorithm to quickly detect cracks, use point cloud neural network algorithm to assess structural safety, combine long short-term memory network and time series analysis to predict disease trend, evaluate inspection results, and update inspection plan; Specifically, a topology graph of the microwave link is constructed based on graph theory, with nodes in the link as vertices and microwave connections between nodes represented as edges. Link quality is then measured using an edge weight calculation formula, as follows: Where n represents the number of link quality assessment metrics; This represents the value of the quality assessment metric for the i-th link. Indicates the first The fuzzy membership function values ​​of each indicator; This represents the Riemann curvature tensor, which reflects the degree of curvature in the link propagation space; The determinant of the Riemann curvature tensor is used to quantify the overall impact of spatial curvature on link quality. m represents the number of bit error rate sampling points; This represents the weight of the j-th bit error rate sampling point; This represents the bit error rate value at the j-th bit error rate sampling point; The position error penalty coefficient is used to adjust the degree of impact of the position error of the UAV relay node on the link quality; K represents the number of UAV relay nodes participating in signal transmission in the microwave link; This represents the actual position vector of the k-th UAV relay node; This represents the ideal position vector of the kth UAV relay node; , where S represents the ideal position vector of the kth UAV relay node; This represents a function describing the distribution of signal intensity in a two-dimensional projected region; Edges with higher weights indicate better link quality and are suitable for data transmission; edges with lower weights indicate poorer link quality and require optimization or adjustment. The system dynamically adjusts the microwave link topology based on edge weights, prioritizing links with higher weights for data transmission, thereby optimizing the performance and reliability of the entire microwave link.

7. A municipal road intelligent inspection system based on an unmanned aerial vehicle (UAV) array according to claim 6, characterized in that, The data acquisition module specifically includes: Twin Model Construction Submodule: Used to build a digital twin foundation for municipal infrastructure, integrating geographic information system maps, building information models and inspection data, and labeling key infrastructure information; The initial disease screening module is used to perform point cloud downsampling at the edge, extract regions of interest from the image, retain key feature information, identify disease areas, and generate initial disease screening results.

8. A municipal road intelligent inspection system based on an unmanned aerial vehicle (UAV) array according to claim 6, characterized in that, The inspection planning module specifically includes: Link quality calculation submodule: used to construct microwave link topology graphs based on graph theory, measure link quality, and adjust microwave link transmission; The array planning submodule is used to integrate the genetic algorithm and the Dijkstra algorithm, encode and generate the initial inspection path combination, evaluate the fitness, and generate the optimal inspection route and array.

9. A municipal road intelligent inspection system based on an unmanned aerial vehicle (UAV) array according to claim 6, characterized in that, The intelligent diagnostic module specifically includes: Array point cloud data synchronization submodule: used to acquire and compare array point cloud data with historical disease database, and synchronize multi-sensor data timestamps; Diagnosis and prediction submodule: Based on the comparison results, it is used to quickly detect cracks, assess structural safety, and predict disease trends using a three-level intelligent diagnostic system.

Images