A road disease and traffic facility patrol method and system, device, medium
By using multimodal data processing and fusion models, combined with deep learning networks and GIS data, the problems of low efficiency and missed detection in traditional road inspections have been solved, enabling efficient and accurate detection of road defects and traffic facilities, as well as dynamic path planning.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-26
- Publication Date
- 2026-06-05
AI Technical Summary
Traditional road inspections rely on manual labor, which is inefficient, prone to missing inspections, and makes it difficult to achieve timely and accurate detection of road defects and traffic facilities.
By employing multimodal data processing and fusion models, and combining deep learning networks such as YOLOv8, CBAM, MobileNetV4, U-Net++, and ViT, a disease and facility detection model is constructed. A comprehensive risk index is generated through a fusion function, and then corrected by combining GIS data and environmental attributes to achieve dynamic path planning.
It has enabled full-process digital detection of road defects and traffic facilities, improved inspection efficiency, reduced the subjectivity and missed detection rate of manual inspection, and provided objective risk assessment and dynamic path optimization.
Smart Images
Figure CN122156654A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent transportation technology, and more specifically, to a method, system, equipment, and medium for inspecting road defects and traffic facilities. Background Technology
[0002] The core purpose of road patrols is to ensure safe and smooth traffic flow, maintain the integrity of facilities, and prevent accidents. Through regular or dynamic patrols, problems and potential hazards such as potholes, cracks, and damaged guardrails can be identified promptly. Accurate monitoring of road conditions provides a basis for maintenance decisions. Simultaneously, it regulates road traffic order, ensures public safety, and is a key means of fulfilling road maintenance responsibilities and extending the lifespan of facilities.
[0003] However, traditional road patrols mainly rely on visual observation, hand notes, and photo reporting, which are subjective, inefficient, and prone to missed detections. Semi-automatic detection can improve efficiency through vehicle-mounted lasers, image acquisition, and manual interpretation, but it still depends on manual labor, resulting in passive and delayed maintenance decisions. Summary of the Invention
[0004] The purpose of this invention is to provide a method, system, equipment, and medium for inspecting road defects and traffic facilities, in order to solve the above-mentioned problems in the prior art.
[0005] This invention is achieved through the following technical solution: In a first aspect, the present invention provides a method for inspecting road defects and traffic facilities, comprising: Acquire multimodal data of the target road, preprocess the multimodal data, construct a disease detection model and a facility detection model, and output the disease damage index and facility damage index based on the multimodal data through the disease detection model and the facility detection model respectively; A fusion function is constructed to obtain an initial comprehensive risk index based on the disease damage index, facility damage index, and corresponding multimodal data. The location of the current target road is obtained based on GIS data. The environmental attributes of the current road are obtained based on the location. The initial comprehensive risk index is corrected based on the environmental attributes to obtain the target comprehensive risk index. The risk level of the current road is obtained based on the target comprehensive risk index. Obtain the risk levels of all roads in the target area, and perform dynamic route planning based on the road risk levels.
[0006] Preferably, the preprocessing of the multimodal data includes denoising the image data, including: The image data is modeled, and the reflection branch and illumination branch are obtained based on the Retinex-Net algorithm; A loss function is constructed for training. The reflection branch is denoised using a residual denoising network to obtain the target reflection branch. The illumination branch is enhanced to obtain the target illumination branch. The denoising results are output based on the target reflection branch and the target illumination branch.
[0007] Preferably, the construction loss function includes:
[0008] The denoising results based on the target reflection branch and the target illumination branch include:
[0009] In the formula, For loss function, For the input image, For reflection branch, For lighting branches, To output the image, For the target reflection branch, For target lighting branches.
[0010] Preferably, the construction of the disease detection model includes: A disease identification model was built based on the YOLOv8 network and the CBAM attention module, and the model was trained based on a historical image dataset of diseases. The preprocessed image is input into the disease identification model to obtain several confidence levels at different locations. Based on the disease type corresponding to the preset confidence level range, the disease type obtained from the confidence level at different locations is output. Output the current disease damage index based on the specific data for each disease type.
[0011] Preferably, the facility detection model includes: Edge features of facilities in the current image data are extracted based on the MobileNetV4 model. The U-Net++ model segmentation framework is used to locate the areas where facilities are abnormal. Global facility anomaly features are extracted based on ViT transfer learning. Output a segmentation mask, which corresponds to different abnormal situations. Collect abnormal data in abnormal areas and obtain a facility damage index based on the abnormal data.
[0012] Preferably, the output of the current disease damage index includes:
[0013] The facility damage index, derived from outlier data, includes:
[0014] Constructing the fusion function includes:
[0015] In the formula, The disease damage index. and As the first calculation weight, and For the second calculation weight, and The third calculation weight, For the calculation coefficient of the ice cone, This represents the number of locations where the icicles appeared. This represents the current average length of the ice cone. This represents the average length of ice cones for the current month. For snow cover calculation coefficient, This represents the number of locations where snow accumulation occurred. This represents the current average snow depth. Average snow cover length for the current month The calculation coefficient for loosening. The number of loose devices, This represents the average area affected by the loosened region. This represents the average area affected by each loosened area. These are the coefficients for calculating deformation. The number of devices that showed deformation. For the cost of deformation facilities, The average cost of current road infrastructure, This is the initial comprehensive risk index.
[0016] Preferably, the step of correcting the initial comprehensive risk index based on the environmental attributes includes: If the current environmental attributes do not include bridges and roads with slopes, then the initial comprehensive risk index is equal to the target comprehensive risk index; If the current environmental attributes include bridges or roads with slopes, then make corrections;
[0017] In the formula, The target is a comprehensive risk index. and The fourth calculation weight is used. The length of the bridge. This represents the average bridge length in the city where the road is located. This represents the average slope of the current slope. This represents the average slope of all slopes in the city where the current road is located.
[0018] Secondly, the present invention also provides a road defect and traffic facility inspection system for performing the above-described road defect and traffic facility inspection method, comprising: The data processing module acquires multimodal data of the target road, preprocesses the multimodal data, constructs a disease detection model and a facility detection model, and outputs disease damage index and facility damage index respectively based on the multimodal data through the disease detection model and the facility detection model; it constructs a fusion function, and obtains an initial comprehensive risk index based on the disease damage index, facility damage index and the corresponding multimodal data through the fusion function. The results output module is configured to obtain the location of the current target road based on GIS data, obtain the environmental attributes of the current road based on the location, correct the initial comprehensive risk index based on the environmental attributes to obtain the target comprehensive risk index, obtain the current road risk level based on the target comprehensive risk index, obtain the risk levels of all roads in the target area, and perform dynamic path planning based on the road risk levels.
[0019] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method for inspecting road defects and traffic facilities.
[0020] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described method for inspecting road defects and traffic facilities.
[0021] The technical solution of the present invention has at least the following advantages and beneficial effects: The method provided by this invention mainly includes: outputting disease damage indices and facility damage indices based on multimodal data through disease detection models and facility detection models respectively; obtaining an initial comprehensive risk index based on the disease damage index, facility damage index, and corresponding multimodal data through a fusion function; correcting the initial comprehensive risk index; and obtaining the current road risk level based on the target comprehensive risk index. This method enables fully digital processing, yielding relatively objective comprehensive results, minimizing the subjectivity, inefficiency, and missed detections inherent in manual inspections, and improving inspection efficiency. Attached Figure Description
[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0023] Figure 1 This is a schematic diagram of the process of the present invention; Figure 2 This is a schematic diagram of the system structure of the present invention. Detailed Implementation
[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0025] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. The naming or numbering of steps in this application does not imply that the steps in the method flow must be executed in the chronological / logical order indicated by the naming or numbering. The execution order of named or numbered process steps can be changed according to the desired technical objective, as long as the same or similar technical effect is achieved.
[0026] Please refer to Figures 1-2 The present invention provides a method for inspecting road defects and traffic facilities, comprising: S1: Acquire multimodal data of the target road, preprocess the multimodal data, construct a disease detection model and a facility detection model, and output the disease damage index and facility damage index respectively through the disease detection model and the facility detection model based on the multimodal data; Multimodal data includes various types of image data of the current road, specific parameters and location data of various anomalies, etc. Preprocessing multimodal data improves the accuracy of subsequent results.
[0027] Among them, the disease detection model can use an improved YOLOv8 network, CBAM attention module to enhance small target detection capability, BiFPN to integrate multi-scale features, and set up several types of ice and snow-specific disease detection heads.
[0028] The facility detection model can use the U-Net++ semantic segmentation network and the MobileNetV4 lightweight backbone network to integrate ViT transfer learning and knowledge distillation, supporting the identification of 8 major categories and 22 minor categories of traffic facilities, and adding "loose" and "deformed" state assessment (based on point cloud 3D pose deviation judgment).
[0029] S2: Construct a fusion function to obtain an initial comprehensive risk index based on the disease damage index, facility damage index and corresponding multimodal data; S3: Obtain the current location of the target road based on GIS data, obtain the environmental attributes of the current road based on the location, correct the initial comprehensive risk index based on the environmental attributes to obtain the target comprehensive risk index, and obtain the current road risk level based on the target comprehensive risk index; S4: Obtain all road risk levels in the target area and perform dynamic path planning based on the road risk levels.
[0030] By combining real-time detection data, prediction results, and historical disease distribution, the number of repeated inspections is reduced, and priority is given to planning snow-free road sections in snowy weather; self-supervised optimization: post-maintenance images are automatically labeled through comparative learning, the detection model is updated weekly, and the prediction model is updated every two weeks; output: the MQTT protocol reports in seconds, the management platform generates maintenance work orders with predicted trends, and supports dispatching and tracking via mobile mini-program.
[0031] The method provided by this invention mainly includes: outputting disease damage indices and facility damage indices based on multimodal data through disease detection models and facility detection models respectively; obtaining an initial comprehensive risk index based on the disease damage index, facility damage index, and corresponding multimodal data through a fusion function; correcting the initial comprehensive risk index; and obtaining the current road risk level based on the target comprehensive risk index. This method enables fully digital processing, yielding relatively objective comprehensive results, minimizing the subjectivity, inefficiency, and missed detections inherent in manual inspections, and improving inspection efficiency.
[0032] In one exemplary embodiment of the present invention, preprocessing multimodal data includes denoising image data, comprising: The image data is modeled, and the reflection branch and illumination branch are obtained based on the Retinex-Net algorithm; A loss function is constructed for training. The reflection branch is denoised using a residual denoising network to obtain the target reflection branch. The illumination branch is enhanced to obtain the target illumination branch. The denoising results are output based on the target reflection branch and the target illumination branch.
[0033] Specifically, constructing the loss function includes:
[0034] The denoising results based on the target reflection branch and the target illumination branch include:
[0035] In the formula, For loss function, For the input image, For reflection branch, For lighting branches, To output the image, For the target reflection branch, For target lighting branches.
[0036] Secondly, ice thickness estimation: Based on the infrared radiation intensity and temperature difference, a mapping model is established to accurately calculate the ice thickness on the road surface (error ≤ 2mm). Snow and ice region segmentation: Improve the Mask R-CNN network to fuse visual texture and infrared temperature features to achieve pixel-level segmentation of snow / ice (accuracy ≥94%). Disease restoration by shading: By using GAN networks to complete the disease outline in snow-covered areas, the recall rate is improved by more than 35%.
[0037] In one exemplary embodiment of the present invention, constructing a disease detection model includes: A disease identification model was built based on the YOLOv8 network and the CBAM attention module, and the model was trained based on a historical image dataset of diseases. The preprocessed image is input into the disease identification model to obtain several confidence levels at different locations. Based on the disease type corresponding to the preset confidence level range, the disease type obtained from the confidence level at different locations is output. Output the current disease damage index based on the specific data for each disease type.
[0038] In one exemplary embodiment of the present invention, the facility detection model includes: Edge features of facilities in the current image data are extracted based on the MobileNetV4 model. The U-Net++ model segmentation framework is used to locate the areas where facilities are abnormal. Global facility anomaly features are extracted based on ViT transfer learning. Output a segmentation mask, which corresponds to different abnormal situations. Collect abnormal data in abnormal areas and obtain a facility damage index based on the abnormal data.
[0039] In one exemplary embodiment of the present invention, the output of the current disease damage index includes:
[0040] The facility damage index, derived from outlier data, includes:
[0041] Constructing the fusion function includes:
[0042] In the formula, The disease damage index. and As the first calculation weight, and For the second calculation weight, and The third calculation weight, For the calculation coefficient of the ice cone, This represents the number of locations where the icicles appeared. This represents the current average length of the ice cone. This represents the average length of ice cones for the current month. For snow cover calculation coefficient, This represents the number of locations where snow accumulation occurred. This represents the current average snow depth. Average snow cover length for the current month The calculation coefficient for loosening. The number of loose devices, This represents the average area affected by the loosened region. This represents the average area affected by each loosened area. These are the coefficients for calculating deformation. The number of devices that showed deformation. For the cost of deformation facilities, The average cost of current road infrastructure, This is the initial comprehensive risk index.
[0043] An exemplary embodiment of the present invention, which modifies the initial comprehensive risk index based on the environmental attributes, includes: If the current environmental attributes do not include bridges and roads with slopes, then the initial comprehensive risk index is equal to the target comprehensive risk index; If the current environmental attributes include bridges or roads with slopes, then make corrections;
[0044] In the formula, The target is a comprehensive risk index. and The fourth calculation weight is used. The length of the bridge. This represents the average bridge length in the city where the road is located. This represents the average slope of the current slope. This represents the average slope of all slopes in the city where the current road is located.
[0045] The above model also fully considers the impact of the specific environmental attributes of the current road on the comprehensive risk index. The environmental attributes in this example include normal road surfaces without bridges and slopes, and roads with bridges or slopes. Based on the specific conditions of the bridges or slopes, a comprehensive analysis is conducted to obtain the corrected target comprehensive risk index. A classification threshold is set according to historical data and specific conditions. Different road risk levels can be classified according to the classification threshold. Based on the range of the current target comprehensive risk index, the corresponding road risk level is obtained.
[0046] This also includes setting up a disease development trend prediction model: Feature inputs: time series detection data (disease changes over 1-3 years), environmental factors (temperature and humidity, precipitation, snow cover days, freeze-thaw cycles), traffic load (traffic volume, heavy load ratio), and road structure parameters; Model architecture: ConvLSTM+Transformer dual-branch network, ConvLSTM captures temporal patterns, and Transformer extracts multi-factor interaction features; Training optimization: MAE loss function, historical data transfer learning, predicting disease severity and spread rate in the next 1 / 3 / 6 months; Output application: Prediction results are incorporated into maintenance priority determination (high-expansion-rate diseases are prioritized by 1 level).
[0047] A road defect and traffic facility inspection system, used to perform the aforementioned road defect and traffic facility inspection method, includes: The data processing module acquires multimodal data of the target road, preprocesses the multimodal data, constructs a disease detection model and a facility detection model, and outputs disease damage index and facility damage index respectively based on the multimodal data through the disease detection model and the facility detection model; it constructs a fusion function, and obtains an initial comprehensive risk index based on the disease damage index, facility damage index and the corresponding multimodal data through the fusion function. The results output module is configured to obtain the location of the current target road based on GIS data, obtain the environmental attributes of the current road based on the location, correct the initial comprehensive risk index based on the environmental attributes to obtain the target comprehensive risk index, obtain the current road risk level based on the target comprehensive risk index, obtain the risk levels of all roads in the target area, and perform dynamic path planning based on the road risk levels.
[0048] An exemplary embodiment of the present invention: Inspection vehicle: Inspection vehicle equipped with multimodal acquisition module (vehicle speed 60km / h), drone swarm supplements bridge and slope areas, acquisition frequency 25Hz, infrared-visual fusion mode activated; Data acquisition: Simultaneously acquire 4K images, lidar point clouds, IMU data, BeiDou positioning data, and infrared thermal imaging data. The infrared sensor detected a road surface temperature of -5℃, and the radiation intensity deviation in the snow-covered area was 0.3. Preprocessing: The snow-covered area was segmented by an improved Mask R-CNN (coverage of 35%), and the contours of two horizontal cracks obscured by snow were completed by a GAN network. The ice thickness was estimated to be 8 mm by lidar point cloud. Joint detection: The model identified 2 instances of frost heave (moderate damage), 3 instances of snow-covered cracks (minor damage), 1 instance of loosening of the isolation fence, and 2 instances of damage to the anti-glare panels; Location association: Generate lane-level coordinates and overlay them with a GIS map to mark the snow and ice coverage area; Trend prediction: Based on the detection data of the past two years, the local average annual freeze-thaw cycles of 20, and the proportion of heavy-duty vehicles of 30%, the model predicts that the cracks will develop into moderate damage in 6 months, with a frost heave expansion rate of 5 mm / month. Results output: The management platform generates a priority level 1 maintenance work order, recommending that it be handled within 1 month. After the work order is dispatched by the mobile app, snow removal and crack repair are completed within 3 hours. The model includes samples for iteration.
[0049] Another exemplary embodiment of the present invention: Inspection vehicle: electric patrol vehicle (speed 30km / h), data collection frequency 15Hz, covering 10km of urban main roads; Data acquisition: Simultaneously acquire multimodal data, and call up the detection data of this road section in the past 3 years, the annual precipitation of 800mm provided by the meteorological department, the number of snow days of 12 days, and the daily traffic volume of 5,000 vehicles (heavy load account for 15%) provided by the transportation department. Joint inspection revealed 1 pothole (moderate damage, area 0.8㎡), 2 manhole covers subsided, and 1 set of pedestrian crossing traffic lights damaged; Trend prediction: The model uses the ConvLSTM branch to capture the size changes of the potholes in the last 6 detections (from 0.3㎡ to 0.8㎡), and the Transformer branch integrates precipitation and traffic flow factors to predict that the pothole area will reach 1.2㎡ (severe damage) in 3 months. Path optimization: Improve A The algorithm avoids already-handled road sections and plans the optimal inspection route for the remaining 5km, reducing the distance by 12%. Closed-loop optimization: After a follow-up inspection two months later, the actual area of the pit was 1.1㎡, with a prediction deviation of 8%. The model automatically adjusted the ConvLSTM branch weights, and the prediction accuracy was improved to 92% in the next iteration.
[0050] Another exemplary embodiment of the present invention: Detection vehicle: light patrol vehicle (speed 40km / h), data collection frequency 20Hz, covering 20km of rural roads; Joint detection: The model identified 3 blurred milestones, 2 blocked drainage ditches, 1 deformed section of wave-shaped guardrail, and 1 tilted speed measuring device; Status assessment: Based on the laser radar point cloud computing data showing a deformation of 5cm and the speed measuring device showing an 8° tilt angle, the guardrail was determined to be in a "deformed" or "tilted" state. Trend forecast: Due to the failure to address the blockage in the drainage ditch in a timely manner, and considering the local annual rainfall of 600mm, it is predicted that road subsidence (slight to moderate) will occur in 3 months. Results output: The maintenance work order recommends dredging the drainage ditch within 2 weeks. After the treatment, the model records the restoration of the facility status, and the accuracy of traffic facility detection is improved by 2%.
[0051] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0052] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. This computer software product, stored in a storage medium, includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0053] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for inspecting road defects and traffic facilities, characterized in that, include: Acquire multimodal data of the target road, preprocess the multimodal data, construct a disease detection model and a facility detection model, and output the disease damage index and facility damage index based on the multimodal data through the disease detection model and the facility detection model respectively; A fusion function is constructed to obtain an initial comprehensive risk index based on the disease damage index, facility damage index, and corresponding multimodal data. The location of the current target road is obtained based on GIS data. The environmental attributes of the current road are obtained based on the location. The initial comprehensive risk index is corrected based on the environmental attributes to obtain the target comprehensive risk index. The risk level of the current road is obtained based on the target comprehensive risk index. Obtain the risk levels of all roads in the target area, and perform dynamic route planning based on the road risk levels.
2. The method for inspecting road defects and traffic facilities according to claim 1, characterized in that, The preprocessing of multimodal data includes denoising the image data, including: The image data is modeled, and the reflection branch and illumination branch are obtained based on the Retinex-Net algorithm; A loss function is constructed for training. The reflection branch is denoised using a residual denoising network to obtain the target reflection branch. The illumination branch is enhanced to obtain the target illumination branch. The denoising results are output based on the target reflection branch and the target illumination branch.
3. The method for inspecting road defects and traffic facilities according to claim 2, characterized in that, The construction loss function includes: The denoising results based on the target reflection branch and the target illumination branch include: In the formula, For loss function, For the input image, For reflection branch, For lighting branches, To output the image, For the target reflection branch, For target lighting branches.
4. The method for inspecting road defects and traffic facilities according to claim 2, characterized in that, The construction of the disease detection model includes: A disease identification model was built based on the YOLOv8 network and the CBAM attention module, and the model was trained based on a historical image dataset of diseases. The preprocessed image is input into the disease identification model to obtain several confidence levels at different locations. Based on the disease type corresponding to the preset confidence level range, the disease type obtained from the confidence level at different locations is output. Output the current disease damage index based on the specific data for each disease type.
5. The method for inspecting road defects and traffic facilities according to claim 4, characterized in that, The facility detection model includes: Edge features of facilities in the current image data are extracted based on the MobileNetV4 model. The U-Net++ model segmentation framework is used to locate the areas where facilities are abnormal. Global facility anomaly features are extracted based on ViT transfer learning. Output a segmentation mask, which corresponds to different abnormal situations. Collect abnormal data in abnormal areas and obtain a facility damage index based on the abnormal data.
6. The method for inspecting road defects and traffic facilities according to claim 5, characterized in that, The output of the current disease damage index includes: The facility damage index, derived from outlier data, includes: Constructing the fusion function includes: In the formula, The disease damage index. and As the first calculation weight, and For the second calculation weight, and The third calculation weight, For the calculation coefficient of the ice cone, This represents the number of locations where the icicles appeared. This represents the current average length of the ice cone. This represents the average length of ice cones for the current month. For snow cover calculation coefficient, This represents the number of locations where snow accumulation occurred. This represents the current average snow depth. Average snow cover length for the current month The calculation coefficient for loosening. The number of loose devices, This represents the average area affected by the loosened region. This represents the average area affected by each loosened area. These are the coefficients for calculating deformation. The number of devices that showed deformation. For the cost of deformation facilities, The average cost of current road infrastructure, This is the initial comprehensive risk index.
7. The method for inspecting road defects and traffic facilities according to claim 6, characterized in that, The correction of the initial comprehensive risk index based on the environmental attributes includes: If the current environmental attributes do not include bridges and roads with slopes, then the initial comprehensive risk index is equal to the target comprehensive risk index; If the current environmental attributes include bridges or roads with slopes, then make corrections; In the formula, The target is a comprehensive risk index. and The fourth calculation weight is used. The length of the bridge. This represents the average bridge length in the city where the road is located. This represents the average slope of the current slope. This represents the average slope of all slopes in the city where the current road is located.
8. A road defect and traffic facility inspection system, characterized in that, A method for inspecting road defects and traffic facilities according to any one of claims 1-7, comprising: The data processing module acquires multimodal data of the target road, preprocesses the multimodal data, constructs a disease detection model and a facility detection model, and outputs disease damage index and facility damage index respectively based on the multimodal data through the disease detection model and the facility detection model; it constructs a fusion function, and obtains an initial comprehensive risk index based on the disease damage index, facility damage index and the corresponding multimodal data through the fusion function. The results output module is configured to obtain the location of the current target road based on GIS data, obtain the environmental attributes of the current road based on the location, correct the initial comprehensive risk index based on the environmental attributes to obtain the target comprehensive risk index, obtain the current road risk level based on the target comprehensive risk index, obtain the risk levels of all roads in the target area, and perform dynamic path planning based on the road risk levels.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for inspecting road defects and traffic facilities as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when executed by a processor, the computer program implements a method for inspecting road defects and traffic facilities as described in any one of claims 1-7.