A road crack detection system and method

Abstract

The present application relates to the technical field of road detection, and relates to a road crack detection system and method, the road crack detection system mainly comprises a UAV module, an unmanned vehicle module, a multi-sensor module, a control center module and a data processing and optimization module, a UAV is used to carry out wide-range inspection on a target area, collect road image data for processing and detection and classify cracks, estimate the length and width of the cracks, transmit crack position data to the control center module, the control center generates a preliminary crack distribution map according to the data, sends crack coordinates to the unmanned vehicle, instructs the unmanned vehicle to accurately reach the crack position and mark, the unmanned vehicle obtains three-dimensional geometric data and expansion information of the crack, generates a high-precision three-dimensional crack model, predicts the expansion trend of the crack, and generates a crack health assessment report, which can effectively improve the accuracy and reliability of crack identification, reduce false detection and missed detection, and ensure the timeliness and effectiveness of road maintenance.

Description

Technical Field

[0001] This invention relates to the technical field of road inspection, and more specifically, to a road crack detection system and method. Background Technology

[0002] Cracks are a common defect in infrastructure such as roads, bridges, and tunnels. They not only affect the strength and service life of structures but can also pose a serious threat to traffic safety. Therefore, timely and accurate crack detection, and assessment of their severity and development trend, are important research directions in the field of structural health monitoring.

[0003] For example, CN114298972B discloses a method for calculating the degree of road surface damage. This method involves installing image acquisition equipment on a road surface inspection vehicle to obtain road surface images, thereby determining the road surface condition. The Road Condition Index (PCI) is then calculated in the background, eliminating the need for manual processing.

[0004] This reduces the hassle of manual on-site measurement and calculation, lowers labor costs, and improves calculation efficiency.

[0005] This method has good application value, but in practical use, image quality is greatly affected by the environment, leading to a decrease in the accuracy of crack detection. Relying solely on two-dimensional images cannot accurately measure the depth and internal propagation of cracks, especially posing a challenge for identifying deep cracks. Although this method improves the efficiency of crack detection, the real-time performance of image processing remains limited in applications covering large areas or complex regions.

[0006] There is also a public notice (CN110046584B) that discloses a road crack detection device and method based on drone inspection. When using a drone for inspection, it collects images through an image acquisition device, and then transmits them sequentially to a ground wireless image receiving platform and a back-end central station for image processing. This allows for the automatic detection and extraction of crack contours, which can be used to analyze the road surface health condition. The device is easy to operate, has high detection efficiency, saves a lot of manpower and material resources, and provides accurate analysis, making it highly practical and valuable for promotion.

[0007] While drones can cover large areas, the precise marking and depth measurement of cracks depend on image quality and processing algorithms. In complex environments, the accuracy of crack localization is affected.

[0008] It is difficult to fully assess the type and development trend of cracks, especially for cracks with complex morphology and deep cracks; accurate classification and assessment remain challenging.

[0009] Given the numerous existing problems, there is an urgent need for new road crack detection systems and methods in the market. Summary of the Invention

[0010] Therefore, the purpose of this invention is to provide a road crack detection system and method that integrates high-precision detection, advanced data fusion, optimized path planning, and accurate trend prediction to improve the efficiency and accuracy of road crack detection.

[0011] To achieve the above objectives, the present invention provides the following technical solution:

[0012] A road crack detection system, comprising:

[0013] The drone module is used to conduct large-scale road inspections, collect road image data, and perform preliminary crack detection and feature extraction using a frequency domain analysis algorithm based on Fourier transform.

[0014] The unmanned vehicle module is used to move to a designated location based on the crack coordinate information sent by the drone, and to mark and detect the crack;

[0015] The multi-sensor module, including lidar, infrared thermal imager, and piezoelectric sensor, is used to acquire three-dimensional information, depth data, and dynamic response data of crack propagation.

[0016] The control center module is used to receive data from drones and unmanned vehicles, generate 3D models of cracks, and provide maintenance suggestions.

[0017] The data processing and optimization module is used to process crack detection data in real time, and performs path planning and task scheduling through an adaptive hybrid quantum genetic algorithm.

[0018] The present invention is further configured such that the control center module includes:

[0019] The deep learning algorithm module uses convolutional neural networks and graph neural networks to process images and depth data collected by the UAV, and extracts the location, shape and depth features of the cracks;

[0020] The dynamic expansion prediction module combines the diffusion equation model with the variational autoencoder to predict the future development trend of cracks.

[0021] The present invention is further configured such that: the data processing and optimization module adopts an adaptive hybrid quantum genetic algorithm and combines it with deep reinforcement learning, adjusts task allocation based on real-time feedback, optimizes the path planning of UAVs and unmanned vehicles, and avoids path overlap.

[0022] The present invention is further configured such that: the unmanned vehicle module has a paint spraying unit for spraying fluorescent paint onto the crack location.

[0023] A method for detecting road cracks includes the following steps:

[0024] Step S1: Use drones to conduct a large-scale inspection of the target area and collect road image data;

[0025] Step S2: The acquired image data is processed using a frequency domain analysis algorithm based on Fourier transform to detect and classify cracks and estimate the length and width of the cracks.

[0026] Step S3: The UAV transmits the crack location data to the control center module, and the control center generates a preliminary crack distribution map based on the data;

[0027] Step S4: The control center sends the crack coordinates to the unmanned vehicle, instructing the unmanned vehicle to accurately reach the crack location and mark it;

[0028] Step S5: The unmanned vehicle uses lidar, infrared thermal imager, and piezoelectric sensor to measure the depth of the crack and obtain the three-dimensional geometric data and expansion information of the crack;

[0029] Step S6: Use a Bayesian inference and graph neural network fusion algorithm to fuse multi-sensor data and generate a high-precision three-dimensional crack model;

[0030] Step S7: Use an adaptive hybrid quantum genetic algorithm combined with deep reinforcement learning to dynamically optimize the paths and tasks of the drones and unmanned vehicles;

[0031] Step S8: Based on the prediction model combining the diffusion equation model and the variational autoencoder, predict the crack propagation trend;

[0032] Step S9: Generate a crack health assessment report, including crack risk score, propagation prediction and repair recommendations.

[0033] The present invention is further configured such that: in step S1, the UAV acquires images through a high-definition camera, identifies the location of the crack using a frequency domain analysis algorithm based on Fourier transform, and records the geographical coordinates of the crack using an RTK positioning system.

[0034] The present invention is further configured such that: in step S6, the multi-sensor data fusion includes Bayesian inference and graph neural network fusion, Bayesian inference is used to integrate the uncertainty information of multi-source data, and graph neural network is used to capture the complex correlation between data, so as to improve the accuracy and robustness of data fusion.

[0035] The present invention is further configured such that, in step S7, the adaptive hybrid quantum genetic algorithm combined with deep reinforcement learning specifically includes the implementation of the following mathematical model:

[0036] Quantum bits represent path planning, specifically representing the path nodes of drones and unmanned vehicles as qubits:

[0037]

[0038] in, and For complex amplitudes, the normalization condition must be met: ;

[0039] Quantum gate operations are used for genetic operations, applying quantum gates to perform crossover and mutation on path nodes. Specific operations include:

[0040] The function of Hadamard gate H:

[0041] The role of Pauli-XT door X: ;

[0042] fitness function Defined as the weighted sum of total path length and energy consumption:

[0043]

[0044] This represents the distance between adjacent nodes in the path. Represents a node energy consumption

[0045] and These are weighting coefficients, reflecting the degree of influence of path length and energy consumption on fitness, respectively.

[0046] Adaptive mechanism to dynamically adjust cross rate and variability To adapt to the needs of different optimization stages, it is defined as:

[0047]

[0048] and These are the initial crossover rate and the mutation rate, respectively. To adjust the rate, This represents the current iteration number;

[0049] Deep reinforcement learning, employing a deep Q-network state-action value function And define the reward function as:

[0050]

[0051] The weighting coefficients for task completion, energy consumption, and time in the reward function.

[0052] The current status includes the location of the drone and unmanned vehicle, remaining tasks, and energy status.

[0053] The actions to be taken include the selection of the direction and speed of movement for drones and unmanned vehicles.

[0054] This represents the network parameters of a deep Q-network;

[0055] State-action value function ) is defined as:

[0056]

[0057] in, Discount factor of square, In time + The rewards received For the current time step, Current time From then on Each time step

[0058] Parameter update rule: Minimize the loss function using the Bellman equation and gradient descent.

[0059]

[0060] The objective function is defined as minimizing the task completion time and energy consumption, specifically:

[0061]

[0062] and These are the task completion times for drones and unmanned vehicles, respectively. The total energy consumption of the system.

[0063] These respectively reflect the degree of influence of task completion time and energy consumption on the objective function.

[0064] The present invention is further configured such that the crack propagation trend prediction model includes:

[0065] The diffusion equation model, its mathematical expression is:

[0066]

[0067] in, Indicates the extent of crack propagation. Where is the diffusion coefficient. The source term represents the external factors influencing crack propagation; For time The partial derivative operator, For spatial coordinates The second-order partial derivative operator;

[0068] The variational autoencoder model, used to generate a latent representation of crack propagation, has the following mathematical expression:

[0069]

[0070] in, For crack propagation data, As latent variables, For encoder distribution, For decoder distribution, The divergence is Kullback-Leibler.

[0071] The present invention is further configured such that: the crack health assessment report includes a crack risk score and predicts the crack propagation trend; the scoring model is:

[0072]

[0073] The length of the crack, The width of the crack. The depth of the crack. The volume for crack propagation. The maximum stress at the crack tip. This represents the strain value around the crack. As the benchmark weighting coefficient, These are dynamically adjusted weighting coefficients.

[0074] Compared with the shortcomings of the prior art, the beneficial effects of the present invention are as follows:

[0075] A frequency domain analysis algorithm based on Fourier transform combined with a three-dimensional convolutional neural network (3D-CNN) is employed to achieve high-precision detection and depth estimation of road cracks. This method effectively improves the accuracy and reliability of crack identification, reduces false detections and missed detections, and ensures the timeliness and effectiveness of road maintenance.

[0076] In particular, by fusing Bayesian inference with graph neural networks (GNNs), high-precision 3D crack models can be generated by integrating data from lidar, infrared thermal imagers, and piezoelectric sensors. This technology improves the accuracy and robustness of data fusion, ensuring the comprehensiveness and accuracy of crack detection results.

[0077] In particular, by combining the diffusion equation model with a variational autoencoder (VAE), the future propagation trend of cracks can be accurately predicted. This predictive capability provides a scientific basis for road maintenance decisions, extends the service life of roads, and reduces maintenance costs. Attached Figure Description

[0078] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0079] Reference Figure 1 This invention provides a road crack detection system and method integrating unmanned aerial vehicles (UAVs) and unmanned vehicles. By combining advanced algorithm models and mathematical functions, it improves the accuracy, real-time performance, and predictive ability of crack detection. The following will describe in detail the various components of this invention and their collaborative working mechanism:

[0080] The road crack detection system mainly consists of the following modules: UAV module, unmanned vehicle module, multi-sensor module, control center module, data processing and optimization module, and user interface module.

[0081] The method for detecting road cracks includes the following steps:

[0082] Step S1: Use drones to conduct a large-scale inspection of the target area and collect road image data;

[0083] Step S2: The acquired image data is processed using a frequency domain analysis algorithm based on Fourier transform to detect and classify cracks and estimate the length and width of the cracks.

[0084] Step S3: The UAV transmits the crack location data to the control center module, and the control center generates a preliminary crack distribution map based on the data;

[0085] Step S4: The control center sends the crack coordinates to the unmanned vehicle, instructing the unmanned vehicle to accurately reach the crack location and mark it;

[0086] Step S5: The unmanned vehicle uses lidar, infrared thermal imager, and piezoelectric sensor to measure the depth of the crack and obtain the three-dimensional geometric data and expansion information of the crack;

[0087] Step S6: Use a Bayesian inference and graph neural network fusion algorithm to fuse multi-sensor data and generate a high-precision three-dimensional crack model;

[0088] Step S7: Use an adaptive hybrid quantum genetic algorithm combined with deep reinforcement learning to dynamically optimize the paths and tasks of the drones and unmanned vehicles;

[0089] Step S8: Based on the prediction model combining the diffusion equation model and the variational autoencoder, predict the crack propagation trend;

[0090] Step S9: Generate a crack health assessment report, including crack risk score, propagation prediction and repair recommendations.

[0091] The specific steps to achieve the above nine steps are as follows:

[0092] The drone module is responsible for conducting large-scale, high-efficiency inspections of the target road area, collecting high-resolution road image data and depth information. Equipped with a high-definition camera and stereo vision sensor, this module can capture real-time information about cracks on the road surface.

[0093] Image acquisition: The drone is equipped with a high-definition camera and flies at a fixed altitude and speed to acquire continuous images of the road surface.

[0094] Depth information acquisition: Using stereo vision technology or structured light equipment, the depth information of the crack is acquired to provide data support for subsequent 3D reconstruction.

[0095] Real-time positioning: Equipped with an RTK (Real-time Dynamic Positioning) system, it accurately records the location coordinates of each crack, ensuring the spatial accuracy of the data.

[0096] The unmanned vehicle module moves to a designated location based on the crack coordinate information sent by the drone, marking and finely inspecting the crack. This module has a self-healing function, automatically releasing a repair agent for initial repair when a high-risk crack is detected.

[0097] Precise positioning and navigation: Utilizing GNSS and inertial navigation systems, the unmanned vehicle accurately reaches the designated crack location.

[0098] Marking and Detection: Equipped with high-contrast, long-lasting fluorescent paint to clearly mark the location of cracks, and equipped with lidar, infrared thermal imager, piezoelectric sensor and structured light equipment for three-dimensional scanning and dynamic response data acquisition.

[0099] The multi-sensor module integrates various sensors to acquire three-dimensional information, depth data, and dynamic response data of crack propagation. Through the fusion and analysis of multi-source data, a high-precision three-dimensional crack model is generated.

[0100] LiDAR: Used to accurately measure the geometry and spatial distribution of cracks.

[0101] Infrared thermal imager: detects temperature changes around cracks and analyzes the thermodynamic characteristics of crack propagation.

[0102] Piezoelectric sensors: monitor the dynamic response of cracks and capture vibration and stress changes during crack propagation.

[0103] Structured light equipment: acquires surface structure information of cracks to assist in 3D reconstruction.

[0104] The control center module is responsible for receiving data from UAVs and unmanned vehicles, generating 3D crack models and providing maintenance suggestions through advanced data fusion algorithms. This module is also responsible for optimizing path planning and task scheduling to ensure efficient system operation.

[0105] Data reception and storage: The system receives images, depth information, and sensor data from drones and unmanned vehicles in real time via wireless communication protocols and stores them in a high-performance database.

[0106] Data fusion algorithm: A fusion algorithm of Bayesian inference and graph neural network (GNN) is adopted to comprehensively analyze multi-source data and generate a high-precision three-dimensional model of cracks.

[0107] Dynamic task scheduling: Combining adaptive hybrid quantum genetic algorithm (AHQGA) and deep reinforcement learning (DRL), the working paths and task allocation of UAVs and unmanned vehicles are optimized to ensure efficient collaborative operation of the system.

[0108] The data processing and optimization module is the core of the system. It is responsible for processing crack detection data in real time and optimizing path planning and task scheduling through advanced algorithm models to minimize task completion time and energy consumption.

[0109] Adaptive Hybrid Quantum Genetic Algorithm (AHQGA):

[0110] Quantum bits represent path planning, specifically representing the path nodes of drones and unmanned vehicles as qubits:

[0111]

[0112] in, and For complex amplitudes, the normalization condition must be met: .

[0113] Quantum gate operations: By applying quantum gates (such as Hadamard gate HHH and Pauli-X gate XXX) to perform crossover and mutation operations on path nodes, the diversity of genetic operations and global search capabilities are enhanced.

[0114] The function of Hadamard gate H:

[0115] The role of Pauli-XT door X: ;

[0116] fitness function Defined as the weighted sum of total path length and energy consumption:

[0117]

[0118] This represents the distance between adjacent nodes in the path. Represents a node energy consumption

[0119] and These are weighting coefficients, reflecting the degree of influence of path length and energy consumption on fitness, respectively.

[0120] Adaptive mechanism to dynamically adjust cross rate and variability To adapt to the needs of different optimization stages, it is defined as:

[0121]

[0122] and These are the initial crossover rate and the mutation rate, respectively. To adjust the rate, This represents the current iteration number;

[0123] Deep reinforcement learning, employing a deep Q-network state-action value function ,

[0124]

[0125] This represents the current state of the system at time t, including the positions of the drone and the unmanned vehicle, remaining tasks, and energy status.

[0126] For actions taken at time t, including the selection of the direction and speed of movement for drones and unmanned vehicles,

[0127] This represents all trainable parameters of the deep Q-network (including network weights and biases).

[0128] in, Discount factor of This allows for control over the decay of future rewards; In time + Instant rewards received For the current time step, Current time From then on Each time step

[0129] The reward function is defined as a weighted combination of task completion rate, energy consumption, and time:

[0130]

[0131] These are the weighting coefficients for task completion, energy consumption, and time in the reward function.

[0132] Strategy selection and parameter updates:

[0133] Employing an ε-greedy strategy to balance exploration and exploitation:

[0134]

[0135] Parameter updates use the Bellman equation and gradient descent to minimize the loss function. :

[0136]

[0137] For the target network parameters, periodically from It has been updated to stabilize the training process.

[0138] Comprehensive optimization process:

[0139] Initialize the population: Use AHQGA to randomly generate the qubit representation path of the initial population.

[0140] Fitness assessment: Calculate the fitness f(X) for each individual.

[0141] Quantum gate operations and genetic operations: Hadamard gates and Pauli-X gates are used to perform path crossover and mutation to generate new individuals.

[0142] Selection and evolution: Select superior individuals based on their fitness values ​​to enter the next generation.

[0143] Adaptive adjustment: Dynamically adjust the crossover rate C(t) and mutation rate M(t).

[0144] Deep reinforcement learning optimization: Utilize DQN to perform local optimization on the selected path, adjusting path details to adapt to real-time environmental changes.

[0145] Iteration Termination: When the preset number of iterations or fitness convergence is reached, the optimal path solution is output.

[0146] By combining AHQGA's global search capabilities with DRL's real-time strategy optimization, efficient collaborative operations between UAVs and unmanned vehicles can be achieved, significantly minimizing task completion time and energy consumption, and improving the overall operating efficiency and response speed of the system.

[0147] The crack propagation trend prediction model mainly consists of the following two parts: diffusion equation model and variational autoencoder.

[0148] Diffusion equation model:

[0149]

[0150] This indicates the location of the crack. and time The extent of expansion on The diffusion coefficient reflects the rate of crack propagation. The source term represents the external factors influencing crack propagation, such as external loads and environmental conditions. The coordinates are spatial coordinates, describing the distribution of cracks in one-dimensional space; The time variable describes the process of crack propagation over time.

[0151] The diffusion equation model is used to describe the diffusion of cracks along spatial coordinates inside or on the surface of a material. The extended behavior, the time derivative term on the left. The diffusion term on the right represents the rate of change of crack propagation over time. Describes the spatial diffusion characteristics of crack propagation, source term This takes into account the influence of external factors on crack propagation.

[0152] Variational autoencoder:

[0153]

[0154] For crack propagation data, As latent variables, Encoder distribution: representing given crack propagation data Latent variables The distribution For decoder distribution: This represents a given split representing a given latent variable. Time crack propagation data The generation distribution, Prior distribution: Usually set as the standard normal distribution N(0,I), Kullback-Leibler divergence: used to measure the difference between two probability distributions.

[0155] VAE uses an encoder to process crack propagation data. Mapping to latent space Then, through the decoder from the latent space Reconstructing crack propagation data By maximizing the lower bound (ELBO), VAE learns the distribution of the latent space, making the reconstructed data as close as possible to the original data, while the distribution of the latent space approximates the prior distribution. .

[0156] To improve the accuracy and robustness of crack propagation trend prediction, a hybrid prediction model is formed by combining the diffusion equation model with VAE. The specific implementation steps are as follows:

[0157] Data Acquisition: High-resolution images, depth information, and multi-sensor data of road cracks are collected using drones and unmanned vehicle modules, recording crack length, width, depth, expansion volume, stress and strain data, etc.

[0158] Data preprocessing: Preprocessing steps such as denoising and enhancement are performed on image data;

[0159] Standardize multi-sensor data to unify data format and scale;

[0160] Extract the geometric features and physical properties of the crack to prepare data for subsequent model input.

[0161] Dynamic simulation of crack propagation: Based on historical crack propagation data, the propagation behavior of cracks at different time steps is simulated using a diffusion equation model.

[0162] The diffusion equation is solved using numerical methods (such as the finite difference method or the finite element method) to predict the extent of crack propagation in future time steps. .

[0163] Source Item Modeling: Considering external influencing factors such as traffic load, climate change, and material fatigue, source terms are modeled, and historical data and environmental monitoring data are used to dynamically adjust source terms to make the model more in line with the actual situation.

[0164] Encoder Design: Construct a deep neural network as an encoder to process crack propagation data. Mapping to latent space .

[0165] The encoder outputs the mean of the latent variables. and variance Used for sampling latent variables .

[0166] Construct a deep neural network as a decoder to extract latent variables Generate reconstructed crack propagation data .

[0167] Loss function optimization:

[0168] Maximizing the lower bound of evidence (ELBO) to train the VAE:

[0169]

[0170] Backpropagation and gradient descent are used to optimize network parameters, enabling the VAE to accurately reconstruct crack propagation data while the potential spatial distribution approximates a standard normal distribution.

[0171] Construction of the hybrid model: combining the prediction results of the diffusion equation model As input to VAE, VAE is used to extract latent variables of crack propagation. latent variables It captures high-dimensional features and potential patterns of crack propagation, providing rich information for trend prediction. The prediction results are fed back to the control center module, where they are combined with other detection data to generate a comprehensive health assessment report.

[0172] To more accurately reflect the impact of crack characteristics on risk, a multi-dimensional, nonlinear, and interactive risk scoring model is proposed:

[0173]

[0174] The length of the crack, The width of the crack. The depth of the crack. The volume for crack propagation. The maximum stress at the crack tip. This represents the strain value around the crack. As the benchmark weighting coefficient, These are dynamically adjusted weighting coefficients.

[0175] Model Analysis:

[0176]

[0177] Numerator: Combined with power function and exponential function The nonlinear enhancement of the impact of crack length and width on risk score.

[0178] Denominator: Passed through the exponential decay function Adjusting crack depth The negative impact on risk scores should be mitigated to prevent scores from becoming too low as depth increases.

[0179] Interaction term of sine and cosine:

[0180] Sine function: capturing the maximum stress at the crack tip The periodic changes reflect the nonlinear effect of stress on risk scores.

[0181] Cosine function: reflects the strain value around the crack. The periodic changes reflect the dynamic impact of strain on risk scores.

[0182] Logarithmic transformation and interaction terms:

[0183] Logarithmic function: handling crack propagation volume The interaction term with stress and strain controls the rate of increase in the score under large volume and high stress and strain conditions, ensuring the rationality of the score.

[0184] Interactive items: Combining and It reflects the combined impact of multiple factors interacting on risk scores.

[0185] The dynamically adjustable weight coefficients can be dynamically adjusted through machine learning algorithms to adapt to different road conditions and material properties, thereby improving the model's adaptability and accuracy.

[0186] The final data results are displayed in the user interface module, which is a screen used for data presentation.

[0187] Example 1: Crack detection on a highway:

[0188] Table 1: Crack data collected by drone:

[0189]

[0190] Table 2: 3D crack data collected by the unmanned vehicle

[0191]

[0192] Table 3: Path planning and task scheduling optimization results

[0193]

[0194] Table 4: Crack propagation trend prediction results

[0195]

[0196] Table Explanation:

[0197] Table 1. Crack Data Collected by UAV: ​​This table records the basic information of cracks detected by UAV during highway inspection, including crack number, location coordinates, length, width, depth and collection time.

[0198] Table 2. Three-dimensional crack data collected by the unmanned vehicle: This table shows the three-dimensional data collected by the unmanned vehicle when performing fine inspection at the specified crack location, including crack number, measurement point number, spatial coordinates, stress and strain values, and measurement time.

[0199] Table 3. Path planning and task scheduling optimization results: This table records the iterative optimization results of the adaptive hybrid quantum genetic algorithm and deep reinforcement learning in the path planning and task scheduling process, including the crossover rate, mutation rate, optimal path length, total energy consumption and task completion time for each iteration.

[0200] Table 4. Crack propagation trend prediction results: This table shows the crack propagation trend prediction results based on the diffusion equation model and VAE, including crack number, current length, predicted length for the next 1 day, 3 days, and 7 days, and corresponding risk scores.

[0201] Example 2: Crack detection in a city road:

[0202] Table 5: Urban road crack data collected by drones

[0203]

[0204] Table 6: 3D crack data of urban roads collected by the unmanned vehicle

[0205]

[0206] Table 7: Results of Urban Road Route Planning and Task Scheduling Optimization

[0207]

[0208] Table 8: Crack propagation trend prediction results

[0209]

[0210] Table Explanation:

[0211] Table 5. Urban road crack data collected by UAV: ​​This table records the basic information of cracks detected by UAV during urban road inspection, including crack number, location coordinates, length, width, depth and collection time.

[0212] Table 6. Three-dimensional crack data of urban roads collected by unmanned vehicles: This table shows the three-dimensional data collected by unmanned vehicles when performing fine detection at the location of cracks in designated urban roads, including crack number, measurement point number, spatial coordinates, stress and strain values ​​and measurement time.

[0213] Table 7. Optimization Results of Urban Road Path Planning and Task Scheduling: This table records the iterative optimization results of adaptive hybrid quantum genetic algorithm and deep reinforcement learning in the process of crack detection and maintenance of urban roads, including the crossover rate, mutation rate, optimal path length, total energy consumption and task completion time for each iteration.

[0214] Table 8. Crack propagation trend prediction results: This table shows the crack propagation trend prediction results based on the diffusion equation model and VAE, including crack number, current length, predicted length for the next 1 day, 3 days, and 7 days, and corresponding risk scores.

[0215] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any ordinary changes and substitutions made by those skilled in the art within the scope of the technical solution of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for detecting road cracks, characterized in that, Includes the following steps: Step S1: Use drones to conduct a large-scale inspection of the target area and collect road image data; S2: The acquired image data is processed using a frequency domain analysis algorithm based on Fourier transform to detect and classify cracks and estimate the length and width of the cracks. Step S3: The UAV transmits the crack location data to the control center module, and the control center generates a preliminary crack distribution map based on the data; Step S4: The control center sends the crack coordinates to the unmanned vehicle, instructing the unmanned vehicle to accurately reach the crack location and mark it; Step S5: The unmanned vehicle uses lidar, infrared thermal imager, and piezoelectric sensor to measure the depth of the crack and obtain the three-dimensional geometric data and expansion information of the crack; Step S6: Use a Bayesian inference and graph neural network fusion algorithm to fuse multi-sensor data and generate a high-precision three-dimensional crack model; Step S7: Use an adaptive hybrid quantum genetic algorithm combined with deep reinforcement learning to dynamically optimize the paths and tasks of the drones and unmanned vehicles; Step S8: Based on the prediction model combining the diffusion equation model and the variational autoencoder, predict the crack propagation trend; Step S9: Generate a crack health assessment report, including crack risk score, propagation prediction, and repair recommendations. In step S1, the drone acquires images using a high-definition camera, identifies the location of the crack using a frequency domain analysis algorithm based on Fourier transform, and records the geographical coordinates of the crack using an RTK positioning system. In step S6, the multi-sensor data fusion includes Bayesian inference and graph neural network fusion. Bayesian inference is used to integrate the uncertainty information of multi-source data, and graph neural networks are used to capture complex relationships between data to improve the accuracy and robustness of data fusion. In step S7, the adaptive hybrid quantum genetic algorithm combined with deep reinforcement learning specifically includes the implementation of the following mathematical model: Quantum bits represent path planning, specifically representing the path nodes of drones and unmanned vehicles as qubits: ； in, and For complex amplitudes, the normalization condition must be met: ; Quantum gate operations are used for genetic operations, applying quantum gates to perform crossover and mutation on path nodes. Specific operations include: The function of Hadamard gate H: ; The role of Pauli-XT door X: ; fitness function Defined as the weighted sum of total path length and energy consumption: ； This represents the distance between adjacent nodes in the path. Represents a node energy consumption 1 and These are weighting coefficients, reflecting the degree of influence of path length and energy consumption on fitness, respectively. Adaptive mechanism to dynamically adjust cross rate and variability To adapt to the needs of different optimization stages, it is defined as: ； and These are the initial crossover rate and the mutation rate, respectively. 1 is for adjusting the speed. This represents the current iteration number; Deep reinforcement learning, employing a deep Q-network state-action value function And define the reward function as: ； 2 represents the weighting coefficients for task completion, energy consumption, and time in the reward function. The current status includes the location of the drone and unmanned vehicle, remaining tasks, and energy status. The actions to be taken include the selection of the direction and speed of movement for drones and unmanned vehicles. This represents the network parameters of a deep Q-network; State-action value function Defined as: ； in, Discount factor of square, In time + The rewards received For the current time step, Current time From then on Each time step Parameter update rule: Minimize the loss function using the Bellman equation and gradient descent. ； The objective function is defined as minimizing the task completion time and energy consumption, specifically: ； and These are the task completion times for drones and unmanned vehicles, respectively. The total energy consumption of the system. 4 respectively reflect the degree of influence of task completion time and energy consumption on the objective function.

2. The road crack detection method according to claim 1, characterized in that, The crack propagation trend prediction model includes: The diffusion equation model, its mathematical expression is: ； in, Indicates the extent of crack propagation. Where is the diffusion coefficient. The source term represents the external factors influencing crack propagation; For time The partial derivative operator, For spatial coordinates The second-order partial derivative operator; The variational autoencoder model, used to generate a latent representation of crack propagation, has the following mathematical expression: ； in, For crack propagation data, As latent variables, For encoder distribution, For decoder distribution, The divergence is Kullback-Leibler.

3. The road crack detection method according to claim 2, characterized in that, The crack health assessment report includes a crack risk score and a prediction of crack propagation trends. The scoring model is as follows: ； L is the length of the crack. The width of the crack. The depth of the crack. The volume for crack propagation. 1 represents the maximum stress at the crack tip. This represents the strain value around the crack. As the benchmark weighting coefficient, 2 represents the dynamically adjusted weighting coefficient.

4. A road crack detection system, used to perform the road crack detection method according to any one of claims 1-3, characterized in that, include: The drone module is used to conduct large-scale road inspections, collect road image data, and perform preliminary crack detection and feature extraction using a frequency domain analysis algorithm based on Fourier transform. The unmanned vehicle module is used to move to a designated location based on the crack coordinate information sent by the drone, and to mark and detect the crack; The multi-sensor module, including lidar, infrared thermal imager, and piezoelectric sensor, is used to acquire three-dimensional information, depth data, and dynamic response data of crack propagation. The control center module is used to receive data from drones and unmanned vehicles, generate 3D models of cracks, and provide maintenance suggestions. The data processing and optimization module is used to process crack detection data in real time, and performs path planning and task scheduling through an adaptive hybrid quantum genetic algorithm.

5. A road crack detection system according to claim 4, characterized in that, The control center module includes: The deep learning algorithm module uses convolutional neural networks and graph neural networks to process images and depth data collected by the UAV, and extracts the location, shape and depth features of the cracks; The dynamic expansion prediction module combines the diffusion equation model with the variational autoencoder to predict the future development trend of cracks.

6. A road crack detection system according to claim 5, characterized in that, The data processing and optimization module adopts an adaptive hybrid quantum genetic algorithm combined with deep reinforcement learning. Based on real-time feedback, it adjusts task allocation, optimizes path planning for UAVs and unmanned vehicles, and avoids path overlap.

7. The road crack detection system according to claim 6, characterized in that, The unmanned vehicle module has a paint spraying unit for spraying fluorescent paint onto the crack locations.

Images