Road crack detection method and system based on fused image

Through multi-source data fusion and stress field reconstruction technology, the accuracy problem of road crack detection under complex lighting conditions was solved, high-precision crack identification and prediction were achieved, and road maintenance efficiency and repair effects were improved.

CN120689311APending Publication Date: 2025-09-23BINHAI BAY BRANCH OF DONGGUAN CITY URBAN MANAGEMENT & COMPREHENSIVE LAW ENFORCEMENT BUREAU
View PDF 0 Cites 16 Cited by

Patent Information

Application Number
CN202510786067.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-06-12
Publication Date
2025-09-23

AI Technical Summary

Technical Problem

Existing road crack detection methods based on single-modal images have reduced recognition accuracy under complex lighting conditions, making it difficult to accurately locate reflective cracks. They also lack the ability to reconstruct stress fields and cannot effectively predict the crack propagation process.

Method used

Using multi-source data fusion technology, combined with visible light images, infrared thermal imaging data and lidar point cloud data, we can perform three-dimensional road reconstruction and crack curvature analysis, reconstruct the reflected crack stress field, and predict the crack expansion trend, and finally repair it through thermal blocking technology.

Benefits of technology

It improves the accuracy and adaptability of road crack detection, achieves high-precision identification and prediction in complex environments, reduces noise interference, and improves road maintenance efficiency and the durability of repairs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120689311A_ABST
    Figure CN120689311A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of road crack detection, in particular to a road crack detection method and system based on a fused image. The method comprises the following steps: acquiring road multi-source monitoring data including a visible light image, infrared thermal imaging data and laser radar point cloud data, and performing multi-modal image fusion and road three-dimensional point cloud reconstruction to generate a fused road image and road three-dimensional modeling data; performing crack curvature analysis based on the fused road image to generate crack curvature data; performing reflection crack contour recognition and positioning on the fused road image through the crack curvature data to generate reflection crack initial positioning data; obtaining road base material data; and performing reflection crack stress field reconstruction on the road area according to the reflection crack initial positioning data to obtain a reflection crack stress field. According to the invention, through multi-modal fusion, curvature identification, stress field modeling and crack channel analysis, the accuracy and strain of road reflection crack detection are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of road crack detection, and in particular to a road crack detection method and system based on fused images. Background Art

[0002] In the field of intelligent identification of road defects, especially the detection of reflective cracks, existing detection methods based on single-modal images (such as RGB or grayscale images) suffer from significant recognition accuracy degradation under complex lighting conditions. Reflective cracks usually occur above existing cracks or structural joints. Under the combined effects of ambient light interference, asphalt reflections, and specular highlights, their texture features are easily submerged by the surrounding background, exhibiting atypical forms such as low contrast and overlapping artifacts. This causes traditional algorithms based on edge enhancement or deep learning classification to produce a large number of false detections and missed detections in highly reflective areas. This type of problem is particularly prominent in complex scenes such as crack intersections, residual water stains, and surface contamination. The highlighted areas caused by reflections lack obvious differences in spectral characteristics from real cracks, resulting in a sharp drop in discrimination under a single image channel. The main bottlenecks of existing technologies are reflected in three aspects: First, reflective cracks are difficult to accurately locate through image enhancement or edge extraction due to their low contrast and blurred edges; second, existing methods lack the ability to reconstruct the stress field based on the mechanical properties of the materials in the crack area, and cannot provide an essential explanation for the generation and expansion process of reflective cracks; third, current crack repair is mostly based on visible damage on the crack surface, ignoring the penetration and evolution of cracks between road structure layers. Summary of the Invention

[0003] Based on this, it is necessary to provide a road crack detection method and system based on fused images to solve at least one of the above technical problems.

[0004] To achieve the above object, a road crack detection method based on fused images is provided, the method comprising the following steps:

[0005] Step S1: Acquire multi-source road monitoring data, including visible light images, infrared thermal imaging data, and lidar point cloud data, and perform multimodal image fusion and road 3D point cloud reconstruction to generate fused road images and road 3D modeling data;

[0006] Step S2: performing crack curvature analysis based on the fused road image to generate crack curvature data; performing reflected crack contour identification and positioning on the fused road image using the crack curvature data to generate reflected crack initial positioning data;

[0007] Step S3: Acquire road base material data; reconstruct the reflection crack stress field of the road area based on the initial reflection crack positioning data to obtain the reflection crack stress field; predict the crack expansion trend of the road base material data based on the reflection crack stress field to generate crack expansion trend prediction data;

[0008] Step S4: Use the crack expansion trend prediction data to perform reflection crack excitation analysis on the road 3D modeling data to generate reflection crack interlayer crossing channel data; use the reflection crack interlayer crossing channel data to thermally block the crack channel in the road area to perform road crack repair operations.

[0009] By fusing visible light images, infrared thermal imaging data, and lidar point cloud data, this method significantly improves the quality of information captured about road surfaces and internal structures, providing richer feature information for subsequent crack identification and analysis, and enhancing the accuracy of the perception of complex road defects. Crack curvature analysis combined with fused images enables precise contour identification and initial positioning of reflective cracks. Compared to traditional image recognition methods, this method can effectively reduce noise interference and improve positioning accuracy, making it particularly suitable for highly heterogeneous road surfaces. The introduction of a stress field reconstruction mechanism can simulate crack stress distribution and development trends in conjunction with road base material data, enabling intelligent prediction of future crack expansion paths and providing data support for preventive maintenance. Crack interlayer crossing channel data generated based on the prediction results can be used to simulate potential paths for cracks to penetrate downward, thereby providing precise targets for thermal blocking intervention and achieving more targeted crack closure and repair. Through automated data processing and intelligent analysis, this method enables intelligent coordination of the entire process of road crack detection, prediction, and repair, significantly improving road maintenance efficiency and decision-making timeliness, while reducing labor costs. Combining crack excitation analysis with thermal blocking technology to form a closed-loop response mechanism effectively blocks the path of continued crack expansion, fundamentally improving the durability and stability of road repairs and extending road service life. Therefore, this invention improves the accuracy and adaptability of road reflective crack detection through multimodal fusion, curvature recognition, stress field modeling, and crack channel analysis.

[0010] Preferably, step S1 includes the following steps:

[0011] Step S11: synchronously collecting visible light images and infrared thermal imaging data of the road through a vehicle-mounted multispectral imaging device;

[0012] Step S12: using a mobile laser radar scanner to obtain three-dimensional point cloud data of the road surface;

[0013] Step S13: performing radiometric correction and geometric registration on the visible light image to generate standard visible light image data;

[0014] Step S14: performing temperature field mapping and noise suppression processing on the infrared thermal imaging data to generate thermal stress distribution data;

[0015] Step S15: reconstructing the three-dimensional point cloud of the road based on the three-dimensional point cloud data of the road surface to generate three-dimensional road modeling data;

[0016] Step S16: Perform multi-scale fusion of the standard visible light image, thermal stress distribution data, and road 3D modeling data to generate a fused road image.

[0017] The present invention uses a vehicle-mounted multispectral imaging device to synchronously collect visible light and infrared thermal imaging data, ensuring that the multi-source images have temporal and spatial consistency, significantly reducing subsequent image registration errors, and improving the overall data fusion quality. Using mobile lidar scanning to obtain high-resolution point cloud data of the road surface and perform three-dimensional modeling, the road structure and surface micromorphology can be accurately restored, enhancing the modeling capabilities of complex road morphologies. Visible light images are subjected to radiation correction and geometric registration to generate standard image data, eliminating external interference factors such as illumination and viewing angle, and improving the accuracy and robustness of subsequent image analysis. Infrared thermal imaging data is converted into thermal stress distribution data, and through noise suppression, potential thermal anomaly areas in the road structure are extracted, providing a key basis for early crack identification and stress analysis. Image information (visible light), physical field information (thermal stress), and geometric information (three-dimensional modeling) are fused to generate a fused road image. This multi-scale information fusion method breaks through the limitations of a single data source and significantly enhances the identifiability and diagnostic accuracy of road diseases. The entire step S1 builds a stable, high-quality road multi-source perception data processing process, providing full-scale, high-quality input data for subsequent crack identification, stress field analysis, evolution prediction and repair strategies, forming the basis for a closed technical loop.

[0018] Preferably, step S15 includes the following steps:

[0019] Step S151: performing density equalization on the three-dimensional point cloud data of the road surface to generate point cloud equalization data;

[0020] Step S152: performing ground segmentation on the three-dimensional point cloud data of the road surface based on the point cloud equalization data, distinguishing the road surface and eliminating the non-road surface, and generating road main surface segmentation data;

[0021] Step S153: performing normal estimation on the road main surface segmentation data to generate road surface normal information data; performing curved surface reconstruction on the road according to the road surface normal information data to generate road surface curved surface reconstruction data;

[0022] Step S154: Structural meshing is performed on the road surface curved surface reconstruction data to obtain road three-dimensional modeling data.

[0023] The present invention performs density equalization processing on the original three-dimensional point cloud data, effectively eliminating the problem of uneven point density caused by speed fluctuations, environmental occlusion, etc. during the acquisition process, ensuring the uniform distribution of point cloud data, and improving data integrity and subsequent processing stability. The road surface and non-road surface (such as vegetation, vehicles, slopes, etc.) in the point cloud data are distinguished by the ground segmentation algorithm, the main surface area of ​​the road is accurately extracted, and irrelevant interference data is eliminated from the source, providing pure input data for subsequent modeling. First, the normal vector of the road surface is estimated to accurately depict the geometric change trend of the road; then, the surface is reconstructed based on the normal information, which can truly restore the undulations and detailed structure of the road surface and improve the physical authenticity and expression ability of the model. By structurally meshing the reconstructed surface, not only does the model have a standardized topological structure, which is convenient for use in simulation analysis (such as crack evolution and stress propagation), but it also improves the data compression rate and transmission efficiency of the model, which is conducive to deployment in smart terminals or edge devices. The entire S15 process, from data optimization, main surface extraction, geometry recovery to structured modeling, builds a systematic 3D road modeling process, ensuring that the final generated 3D road modeling data is highly accurate, robust, and applicable, providing a solid geometric foundation for subsequent crack excitation analysis, trend prediction, and thermal blockage repair.

[0024] Preferably, step S2 includes the following steps:

[0025] Step S21: extracting texture abnormality areas based on the fused road image, and performing regional core division on the texture abnormality areas to obtain texture abnormality core area images and texture abnormality edge area images;

[0026] Step S22: Analyze the texture morphology and direction of the texture abnormality core area image to generate the core area texture feature data; calculate the directional consistency and grayscale gradient concentration of the texture abnormality edge area image to obtain the edge area texture feature data;

[0027] Step S23: Performing 3D projection on the fused road image using the core area texture feature data and the edge area texture feature data, and extracting the elevation value and normal vector change after projection; calculating the curvature of the texture abnormal area based on the elevation value and normal vector change to generate crack curvature data;

[0028] Step S24: Obtain a historical pavement structure design drawing; perform cross-layer structural crack feature matching on the historical pavement structure design drawing based on the crack curvature data to generate reflected crack matching data; perform crack coordinate conversion and positioning on the road using the reflected crack matching data to generate reflected crack initial positioning data.

[0029] The present invention extracts texture anomaly areas by fusing road images and divides them into core areas and edge areas, effectively focusing on potential crack occurrence areas, reducing redundant calculations, and improving the local accuracy and stability of target analysis. By utilizing the texture morphology and direction of the core area, as well as the directional consistency and grayscale gradient concentration of the edge area, a hierarchical and distinguishable texture feature system is formed, which significantly improves the perception of fine cracks and heterogeneous texture morphology. The two-dimensional texture anomaly area is projected into the three-dimensional model, and the elevation change and normal vector change are combined to achieve accurate calculation of the crack curvature, which not only improves the accuracy of crack geometric change detection, but also achieves a deep fusion of texture and geometric information. By using historical pavement structure design drawings and crack curvature data for cross-layer feature matching, typical structural cracking patterns belonging to "reflection cracks" can be effectively identified; at the same time, high-precision initial positioning of reflection cracks is achieved through the coordinate conversion mechanism. The entire step S2 is not only based on surface image analysis, but also combines historical structural drawings and three-dimensional models to achieve a triple information fusion of texture, geometry, and structure. This multi-dimensional perspective enhances the system's recognition depth and positioning accuracy of structural cracks, facilitating early intervention and precise repair.

[0030] Preferably, performing cross-layer structural crack feature matching on the historical pavement structure design drawing based on the crack curvature data in step S24 includes:

[0031] Extract the layer structure from the historical pavement structure design drawings to generate pavement layer profile data;

[0032] Perform projection transformation on the crack curvature data to generate curvature structure mapping data;

[0033] Perform inter-layer trajectory overlap analysis on the curvature structure mapping data and the pavement layer profile data to generate curvature layer coupling map data;

[0034] The commonality of structural features is determined on the curvature layer coupling map data to generate reflection feature consistency data;

[0035] The fracture path consensus reconstruction is performed on the curvature layer coupling map data based on the reflection feature consistency data to generate reflection fracture matching data.

[0036] The present invention extracts the layered structure of historical pavement structure design drawings to generate pavement layer profile data containing different structural layers (such as surface layer, base layer, subbase layer, cushion layer, etc.), laying a structural foundation at the atlas level for the subsequent layered positioning and analysis of curvature characteristics. By projecting the crack curvature data, curvature structure mapping data reflecting its spatial distribution in the pavement longitudinal section dimension is generated, which realizes the natural transition from two-dimensional crack information to the three-dimensional depth domain of the structure, and enhances the spatial consistency and alignment accuracy of the data. By performing trajectory coupling analysis on the curvature structure mapping data and the pavement layer profile data, curvature layer coupling atlas data is generated, which can identify potential channels for cracks to penetrate from the surface layer downward and capture the cross-layer propagation trend and implicit trajectory of structural cracking. Based on the curvature layer coupling atlas, the commonality of structural features is judged, and characteristic patterns such as path overlap, trend consistency, and angle variation of cracks in different hierarchical structures are extracted to generate reflection feature consistency data, which can effectively identify the typical "reflection crack" type caused by differences in material layer boundaries. Using consistent reflection feature data to reconstruct a consensus on crack paths and generate reflective crack matching data, this not only improves the depth resolution of crack location but also provides high-precision spatial input for subsequent path prediction, structural intervention, and repair planning. The entire matching process forms a logical loop from surface crack curvature to the evolution mechanism between structural layers, effectively supporting chain-based traceability analysis of cracks from "characterization to penetration to root cause," significantly improving the scientific and accurate nature of intelligent road disease diagnosis.

[0037] Preferably, step S3 includes the following steps:

[0038] Step S31: Acquire road base material data;

[0039] Step S32: Deconstructing the material elastic characteristics of the road base material data to generate material elastic modulus field data;

[0040] Step S33: performing crack region meshing on the initial positioning data of the reflected crack to generate crack finite element discrete mesh data; performing stress field solving on the crack finite element discrete mesh data based on the material elastic modulus field data to generate the reflected crack stress field;

[0041] Step S34: extracting the stress gradient direction of the reflected crack stress field, and performing nonlinear coupling prediction on the road base material data to generate crack extension trend prediction data.

[0042] This method acquires road base material data to provide the necessary material property support for subsequent stress analysis and crack propagation prediction, ensuring comprehensive and multidimensional data and facilitating accurate simulation of crack behavior in real-world engineering environments. By deconstructing the material's elastic characteristics, it generates material elastic modulus field data, precisely characterizing the road base's response under different stress conditions. This enhances the fidelity and accuracy of stress field calculations and ensures that the model accurately reflects the material's mechanical properties. By meshing the crack region, it generates finite element discrete mesh data for the cracks, significantly improving the spatial scope and resolution of crack analysis. Combining this with the material elastic modulus field data for stress field calculation provides the physical context for subsequent crack propagation trend prediction, enhancing the scientific nature and accuracy of the analysis. The stress gradient direction of the reflected crack stress field is extracted and combined with nonlinear coupling predictions of the road base material to generate crack propagation trend prediction data. Through in-depth analysis of crack stress variations, crack propagation direction and velocity can be accurately predicted, providing a scientific basis for road repair decision-making. Predicting crack propagation trends not only provides information on the crack's location and morphology at future points in time but also guides engineering design, maintenance decisions, and road repair, significantly improving the predictability and efficiency of road maintenance. Step S3 comprehensively analyzes the elastic characteristics of the base material, the crack stress field, and its propagation trend, forming a comprehensive material-stress-crack relationship model. This model not only improves the durability assessment capabilities of road structures but also provides data support for the development of intelligent repair plans.

[0043] Preferably, step S34 includes the following steps:

[0044] Step S341: performing multi-scale tensor differential processing on the reflected fracture stress field data to generate fracture principal stress gradient tensor data; performing directional field stability analysis on the fracture principal stress gradient tensor data to generate stress stability directional domain data;

[0045] Step S342: performing dynamic modulus partitioning on the road base material data to generate layered modulus mapping data;

[0046] Step S343: performing nonlinear interactive fusion on the stress stability direction domain data and the layered modulus mapping data to generate crack advancement probability field data;

[0047] Step S344: performing spatial entropy evolution on the fracture propagation probability field data to generate fracture extension trend prediction data.

[0048] The present invention generates crack principal stress gradient tensor data by performing multi-scale tensor differential processing on reflected crack stress field data. This method can precisely capture stress gradient variations at different scales within the crack region, further improving the accuracy of crack stress analysis. By analyzing the directional field stability of this data, stress stability directional domain data is generated, effectively identifying potential directions of crack propagation and providing accurate crack development trend predictions. Dynamic modulus partitioning is performed on road base material data to generate layered modulus mapping data, providing information on the dynamic distribution of material properties in different regions for crack propagation analysis. This layered processing accurately reflects the response characteristics of different material layers under different stress conditions, helping to improve the accuracy of simulated road structure responses. Crack propagation probability field data is generated by performing nonlinear interactive fusion of stress stability directional domain data and layered modulus mapping data. This processing method considers the complex interaction between material properties and stress variations, providing more comprehensive and accurate predictions of crack propagation, enhancing the ability to identify and predict crack propagation trends, and helping to better understand the probability and risk of crack propagation. Spatial entropy evolution is performed on the crack propagation probability field data to generate crack propagation trend prediction data. The evolution of spatial entropy effectively assesses the uncertainty of crack growth and optimizes the prediction process. This allows the prediction of crack growth trends to not only account for physical and stress conditions but also incorporate spatial variability, thereby improving the accuracy and robustness of the model. Through the coordinated efforts of these steps, the entire process significantly improves the accuracy of crack growth trend predictions, particularly accurately reflecting the stress response characteristics of complex road substructures. This provides strong data support for road maintenance managers, helping them make more informed decisions during actual repair and maintenance work.

[0049] Preferably, step S4 includes the following steps:

[0050] Step S41: performing spatial coupling mapping processing on the road 3D modeling data using the crack extension trend prediction data to generate crack 3D path prediction data;

[0051] Step S42: analyzing the path topology of the three-dimensional crack path prediction data, and extracting interlayer penetration nodes from the three-dimensional crack path prediction data based on the results of the path topology analysis to generate interlayer path node atlas data;

[0052] Step S43: performing heat flux simulation on the layer-penetrating path node atlas data to generate path thermal conductivity penetration data; performing dynamic thermal resistance control analysis on the initial positioning data of the reflective crack based on the path thermal conductivity penetration data to generate crack channel thermal blocking data;

[0053] Step S44: performing structural response feedback on the initial positioning data of the reflected crack according to the crack channel thermal blocking data to obtain reflected crack blocking feedback data to perform the road crack repair operation.

[0054] The present invention uses crack expansion trend prediction data to perform spatial coupling mapping on road three-dimensional modeling data to generate crack three-dimensional path prediction data. This process can highly integrate crack expansion prediction information with the three-dimensional structure of the road, provide accurate crack path prediction, and provide a clear and comprehensive path reference for subsequent repair work, significantly improving the predictability and efficiency of crack repair operations. By analyzing the path topology of the crack three-dimensional path prediction data and further analyzing the crack path based on this, the through-layer path node map data is extracted. This analysis can accurately identify the key nodes and potential channels for crack expansion, provide more detailed spatial information for subsequent crack repair work, and help determine repair priorities and strategies. By performing heat flux simulation on the through-layer path node map data, path thermal conductivity penetration data is generated. This simulation can simulate the thermal conductivity characteristics of cracks in different paths, evaluate the impact of cracks on heat flow, and provide a more accurate physical model for crack repair. In actual repair, this data can help optimize the selection of repair materials and the design of thermal isolation effects. After analyzing the path thermal conductivity penetration data, dynamic thermal resistance control analysis is performed on the initial reflected crack location data based on this data to generate crack channel thermal blocking data. This process evaluates the thermal resistance characteristics of the crack channel and dynamically adjusts the thermal resistance based on the different crack states, ensuring that the repair work can effectively isolate heat flow and prevent further crack expansion. This improves the reliability of the repair work and ensures longer-term repair results. Based on the crack channel thermal blocking data, structural response feedback is performed on the initial reflected crack location data to generate reflected crack blocking feedback data. This feedback mechanism can timely adjust the repair plan based on the actual repair results, ensuring real-time feedback on the crack status at each step of the repair process, thereby more effectively executing road crack repair operations.

[0055] Preferably, step S43 includes the following steps:

[0056] Step S431: performing multi-physics thermal response modeling on the layer-penetrating path node atlas data to generate local node thermal coupling model data; performing time-series thermal disturbance simulation on the local node thermal coupling model data to generate thermal flow response sequence data;

[0057] Step S432: performing three-dimensional thermal field inversion calculation on the heat flow response sequence data to generate path thermal conductivity penetration data;

[0058] Step S433: Clustering and identifying non-uniform heat-sensitive areas on the path thermal conductivity penetration data to generate high-risk heat penetration cluster data; performing positional thermal gradient cross-analysis on the high-risk heat penetration cluster data and the initial positioning data of the reflective cracks to generate dynamic thermal resistance configuration data;

[0059] Step S434: performing a global thermal control simulation of the crack channel on the dynamic thermal resistance configuration data to generate crack channel thermal blocking data.

[0060] The present invention generates local node thermal coupling model data by performing multi-physics thermal response modeling on the node atlas data of the through-layer path, and generates heat flow response sequence data through time-series thermal disturbance simulation. This modeling process can simulate the influence of different heat sources on the crack path, accurately predict the response of thermal disturbance to crack development, and provide very detailed thermal response information for subsequent crack repair. This step greatly improves the accuracy of thermal response prediction and provides a more targeted thermal control solution for repair work. By performing three-dimensional thermal field inversion calculation on the heat flow response sequence data, path thermal conductivity penetration data is generated. This inversion calculation can more comprehensively and accurately evaluate the characteristics of heat conduction in the crack path by fully considering the changes in heat flow, and provide a scientific basis for further thermal resistance control. This method greatly improves the depth and accuracy of thermal conductivity penetration analysis and provides high-quality data support for the formulation of thermal control strategies during the repair process. By clustering and identifying non-uniform thermal sensitive areas of the path thermal conductivity penetration data, high-risk heat penetration cluster data is generated. This cluster identification can automatically identify high-temperature conduction areas in the crack channel, thereby discovering potential thermal damage risks in advance. Through further cross-analysis of thermal gradients at specific locations, dynamic thermal resistance configuration data is generated, providing specific data for risk assessment and dynamic adjustment of thermal resistance for repair operations. This analysis process can effectively warn of thermal issues related to crack expansion and provide a more precise risk control solution for crack repair strategies. Based on the dynamic thermal resistance configuration data, a global thermal control simulation of the crack channel is performed to generate thermal blocking data for the crack channel. This global thermal control simulation can adjust the thermal resistance configuration in real time throughout the crack channel, ensuring effective heat isolation in the crack channel and preventing further crack expansion due to heat flow. This step significantly improves the thermal protection effect during the repair process and ensures the long-term stability and reliability of the repair results.

[0061] In this specification, a road crack detection system based on fused images is provided, which is used to perform the above-mentioned road crack detection method based on fused images. The road crack detection system based on fused images includes:

[0062] The image fusion module is used to obtain multi-source road monitoring data, including visible light images, infrared thermal imaging data, and lidar point cloud data, and perform multimodal image fusion and road 3D point cloud reconstruction to generate fused road images and road 3D modeling data;

[0063] The crack location module is used to perform crack curvature analysis based on the fused road image and generate crack curvature data; the crack curvature data is used to identify and locate the reflected crack contours of the fused road image and generate initial reflected crack location data;

[0064] The crack stress simulation module is used to obtain road base material data; reconstruct the reflected crack stress field of the road area based on the initial location data of the reflected cracks to obtain the reflected crack stress field; predict the crack expansion trend of the road base material data based on the reflected crack stress field to generate crack expansion trend prediction data;

[0065] The layer penetration blocking module is used to use the crack expansion trend prediction data to perform reflection crack excitation analysis on the road 3D modeling data to generate reflection crack interlayer penetration channel data; the reflection crack interlayer penetration channel data is used to thermally block the crack channel in the road area to perform road crack repair operations.

[0066] The present invention's beneficial effect lies in integrating visible light imagery, infrared thermal imaging data, and LiDAR point cloud data, synthesizing multiple sensory information to provide comprehensive road monitoring data. This multimodal data fusion not only effectively improves road detection accuracy but also enables identification of road defects in different environments, such as the interplay between temperature changes and physical structure. By performing multimodal image fusion and road point cloud reconstruction on this data, the module generates fused road images and 3D road modeling data, significantly improving the accuracy of the road models. These precise 3D models serve as the basis for subsequent crack analysis and repair operations, reducing errors and improving data reliability. The fused images and 3D data provide multi-angle information for subsequent analysis. Combining infrared thermal imaging with LiDAR data effectively reveals potential cracks and other road damage. This fusion solution ensures comprehensive and accurate road status, providing strong data support for crack identification and repair solutions. By performing crack curvature analysis based on the fused road images, the module generates precise crack curvature data, which is then used to identify and locate reflective crack contours. This step not only accurately identifies the location of road cracks but also efficiently locates the crack edges, providing clear targets for subsequent repair work. Analyzing crack curvature accurately identifies crack morphology and growth trends, ensuring high-precision crack identification. This is particularly important for locating cracks in complex road environments, avoiding misidentification or missed detections often associated with traditional methods and improving crack detection reliability. By acquiring road base material data and combining it with initial reflected crack location data, the module reconstructs the crack stress field and generates a reflected crack stress field. This process accurately simulates the stress distribution of cracks on the road surface, providing high-quality input data for crack growth trend prediction. Based on the reflected crack stress field data, the module predicts crack growth trends and generates crack growth trend prediction data. This prediction can identify potential crack growth issues in advance, providing early warning before repair work begins, enabling targeted repair plans to be developed and preventing further damage from crack expansion. By using the crack growth trend prediction data for reflected crack excitation analysis, the module generates reflected crack interlayer crossover channel data, effectively identifying crack propagation paths. This step helps determine whether cracks pose a risk of cross-layer penetration and provides specific guidance for repair work. By performing thermal blockage analysis on the crack channel, the module can thermally control crack growth and facilitate road crack repair. Thermal blocking analysis helps provide effective thermal isolation during the crack repair process, preventing further crack expansion due to thermal diffusion and ensuring the durability of the repair. By accurately analyzing the interlayer cross-channels of the crack and implementing thermal blocking, the module can provide more intelligent repair strategies, reduce unnecessary intervention during the repair process, improve repair efficiency, and ensure the long-term stability of the repair results.Therefore, the present invention improves the accuracy and adaptability of road reflective crack detection through multimodal fusion, curvature recognition, stress field modeling and crack channel analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 Schematic diagram of the steps of a road crack detection method based on fusion images;

[0068] Figure 2 for Figure 1 Detailed implementation steps of step S2 in FIG.

[0069] Figure 3 for Figure 1 Detailed implementation steps of step S4 in FIG. DETAILED DESCRIPTION

[0070] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0071] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0072] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0073] To achieve this, please refer to Figures 1 to 3 , a road crack detection method based on fusion image, the method comprising the following steps:

[0074] Step S1: Acquire multi-source road monitoring data, including visible light images, infrared thermal imaging data, and lidar point cloud data, and perform multimodal image fusion and road 3D point cloud reconstruction to generate fused road images and road 3D modeling data;

[0075] Step S2: performing crack curvature analysis based on the fused road image to generate crack curvature data; performing reflected crack contour identification and positioning on the fused road image using the crack curvature data to generate reflected crack initial positioning data;

[0076] Step S3: Acquire road base material data; reconstruct the reflection crack stress field of the road area based on the initial reflection crack positioning data to obtain the reflection crack stress field; predict the crack expansion trend of the road base material data based on the reflection crack stress field to generate crack expansion trend prediction data;

[0077] Step S4: Use the crack expansion trend prediction data to perform reflection crack excitation analysis on the road 3D modeling data to generate reflection crack interlayer crossing channel data; use the reflection crack interlayer crossing channel data to thermally block the crack channel in the road area to perform road crack repair operations.

[0078] By fusing visible light images, infrared thermal imaging data, and lidar point cloud data, this method significantly improves the quality of information captured about road surfaces and internal structures, providing richer feature information for subsequent crack identification and analysis, and enhancing the accuracy of the perception of complex road defects. Crack curvature analysis combined with fused images enables precise contour identification and initial positioning of reflective cracks. Compared to traditional image recognition methods, this method can effectively reduce noise interference and improve positioning accuracy, making it particularly suitable for highly heterogeneous road surfaces. The introduction of a stress field reconstruction mechanism can simulate crack stress distribution and development trends in conjunction with road base material data, enabling intelligent prediction of future crack expansion paths and providing data support for preventive maintenance. Crack interlayer crossing channel data generated based on the prediction results can be used to simulate potential paths for cracks to penetrate downward, thereby providing precise targets for thermal blocking intervention and achieving more targeted crack closure and repair. Through automated data processing and intelligent analysis, this method enables intelligent coordination of the entire process of road crack detection, prediction, and repair, significantly improving road maintenance efficiency and decision-making timeliness, while reducing labor costs. Combining crack excitation analysis with thermal blocking technology to form a closed-loop response mechanism effectively blocks the path of continued crack expansion, fundamentally improving the durability and stability of road repairs and extending road service life. Therefore, this invention improves the accuracy and adaptability of road reflective crack detection through multimodal fusion, curvature recognition, stress field modeling, and crack channel analysis.

[0079] In the embodiment of the present invention, reference Figure 1FIG. 1 is a schematic flow chart of a method for detecting road cracks based on fused images according to the present invention. In this example, the method for detecting road cracks based on fused images includes the following steps:

[0080] Step S1: Acquire multi-source road monitoring data, including visible light images, infrared thermal imaging data, and lidar point cloud data, and perform multimodal image fusion and road 3D point cloud reconstruction to generate fused road images and road 3D modeling data;

[0081] Step S2: performing crack curvature analysis based on the fused road image to generate crack curvature data; performing reflected crack contour identification and positioning on the fused road image using the crack curvature data to generate reflected crack initial positioning data;

[0082] Step S3: Acquire road base material data; reconstruct the reflection crack stress field of the road area based on the initial reflection crack positioning data to obtain the reflection crack stress field; predict the crack expansion trend of the road base material data based on the reflection crack stress field to generate crack expansion trend prediction data;

[0083] Step S4: Use the crack expansion trend prediction data to perform reflection crack excitation analysis on the road 3D modeling data to generate reflection crack interlayer crossing channel data; use the reflection crack interlayer crossing channel data to thermally block the crack channel in the road area to perform road crack repair operations.

[0084] In one embodiment of the present invention, a multi-sensor platform is used to comprehensively monitor roads. The specific configuration includes: a visible light camera with a resolution of 4096×2160 and a field of view of 120°; an infrared thermal imager with a temperature measurement range of -20°C to 150°C and a thermal sensitivity of 0.05°C; and a lidar (LiDAR) with a rotation frequency of 10Hz and a single-frame point cloud of approximately 100,000 points. The inspection vehicle patrols a 100-meter road section at a speed of 5 km / h, collecting data every 1 meter. After data acquisition, a deep convolutional neural network (such as ResNet50) is used to extract texture and temperature features from the visible light and infrared images. The lidar point cloud is then registered and downsampled using the ICP (Iterative Closest Point) algorithm and Voxel Grid filtering. This is then combined with a deep learning-based multimodal image fusion network (such as DenseFuse) to achieve image-level information fusion. This ultimately generates a spatially consistent fused image that combines both textural and thermal characteristics, and a three-dimensional point cloud model of the road with an accuracy of ±1 cm. The crack identification process involves two core steps: crack curvature analysis and reflection crack location. First, after Gaussian filtering and edge enhancement of the fused image, curvature is extracted using the Hessian matrix, with a curvature threshold of 0.15 mm⁻¹ to identify areas of curvature anomalies. Next, a crack identification model based on Mask R-CNN is used to classify these anomaly areas. Reflection crack paths are selected based on crack length (>0.5 m) and directional consistency. In a test section, 25 initial crack paths were detected. Ultimately, through reflection symmetry analysis and comparison with historical structural crack data, eight were identified as typical reflection cracks, ranging in length from 0.8 m to 3.2 m and in width from 1.5 mm to 6 mm. Initial reflection crack location data was generated. Road structural layer information was obtained, including a 5 cm thick asphalt surface layer and a 20 cm thick cement-stabilized macadam base layer. Material parameters included an elastic modulus of 2.5 GPa for the surface layer and 4.0 GPa for the base layer, with tensile strengths of 2.0 MPa and 3.5 MPa, respectively. Using a modeling method based on the finite element software ABAQUS, a two-dimensional cross-sectional model containing the crack path was constructed. The stress distribution in the reflective crack area under a standard vehicle axle load (100kN) was simulated and analyzed. The results showed that the maximum principal tensile stress at the crack tip was 1.8MPa, which is close to the material limit. Based on this, the Paris-Erdogan crack propagation model (C = 1.5×10 -8 , m = 3.1), combined with stress intensity factor ΔK = 0.9 MPa·m 1 / 2, predicting that the crack would expand to 2.3m after 100,000 equivalent load cycles, with a crack width increase of approximately 0.8mm. This generated crack expansion trend prediction data. Using this crack expansion trend prediction data, combined with a 3D point cloud model, depth projection and voxel segmentation algorithms (such as 3D U-Net) were used to construct a three-dimensional spatial model of the cracks and identify whether they formed interlayer channels. Three cracks were identified in the example road section with penetration depths exceeding 15mm, forming effective interlayer channels. Subsequently, based on a thermal blocking repair strategy, epoxy-modified asphalt with a permeability ≥20mm, a thermal curing time of 30min, and an initial bond strength ≥1.2MPa was used as the repair material. In the thermal blocking numerical simulation, a thermal-stress coupling analysis was performed using COMSOL. The results showed that the peak stress at the crack tip was reduced by 25% and the crack growth rate decreased by approximately 40%. Finally, this thermal blocking repair method was implemented on a 30-35m test section, supplemented by high-precision thermal sensor monitoring. After three months, the crack width remained ≤1mm, with no new crack expansion observed, verifying the feasibility and effectiveness of this implementation method in crack suppression and repair.

[0085] Preferably, step S1 includes the following steps:

[0086] Step S11: synchronously collecting visible light images and infrared thermal imaging data of the road through a vehicle-mounted multispectral imaging device;

[0087] Step S12: using a mobile laser radar scanner to obtain three-dimensional point cloud data of the road surface;

[0088] Step S13: performing radiometric correction and geometric registration on the visible light image to generate standard visible light image data;

[0089] Step S14: performing temperature field mapping and noise suppression processing on the infrared thermal imaging data to generate thermal stress distribution data;

[0090] Step S15: reconstructing the three-dimensional point cloud of the road based on the three-dimensional point cloud data of the road surface to generate three-dimensional road modeling data;

[0091] Step S16: Perform multi-scale fusion of the standard visible light image, thermal stress distribution data, and road 3D modeling data to generate a fused road image.

[0092] In an embodiment of the present invention, synchronous acquisition is performed during the road inspection process by using a vehicle-mounted multispectral imaging system. The system configuration includes: a visible light industrial camera with a resolution of 4096×2160, a frame rate of 30fps, and a focal length of 12mm; an uncooled infrared thermal imager with an infrared band of 8-14μm and a thermal sensitivity better than 50mK; the synchronization control module uses the PTP (Precision Time Protocol) protocol to ensure that the acquisition accuracy is controlled within ±10ms. The inspection vehicle travels at a constant speed of 5km / h and performs synchronous image acquisition on the road area every 0.5m to ensure that the image coverage density meets the spatial resolution of 2 frames per square meter. A rotating multi-line laser radar system (such as Velodyne HDL-32E) is used, with a scanning frequency of 10Hz, a vertical resolution of 1.33°, 360° coverage in the horizontal direction, and a point cloud acquisition accuracy of ±2cm. The point cloud data and image data are paired and stored through the time synchronization module, and a frame of 3D road point cloud data is generated every time it travels 1 meter, with a point density of approximately 10 Points / frame. Perform radiation correction on the original visible light image and use the grayscale correction plate to perform linear regression fitting on the image grayscale (R 2>0.95) to remove the effects of illumination variations. Geometric registration was then performed using an image registration algorithm based on SIFT (Scale-Invariant Feature Transform) and RANSAC (Random Sample Consensus). A homography transformation between the image and the main view of the laser point cloud was achieved, with the registration error kept within 1.5 pixels. A standardized visible light image dataset was then output. Temperature field mapping was performed on the original infrared image, converting grayscale values ​​to actual temperature values ​​using a linear response model. The calibration formula is: T = a·G+b, where T is the temperature (°C), G is the infrared image grayscale value, a = 0.04, and b = -5. For noise suppression, a combination of temporal median filtering and spatial bilateral filtering was used to remove random noise caused by jitter and differences in thermal conduction. Finally, a spatially continuous thermal stress distribution map was generated and rasterized into a thermal stress data matrix with a resolution of 0.1m × 0.1m. The point cloud obtained in step S12 undergoes preprocessing, including VoxelGrid filtering (with a voxel grid size set to 0.05m) to reduce data volume and remove dangling and non-ground points. Subsequently, a 3D model of the road surface is generated using a Poisson reconstruction algorithm or a plane-fitting-based triangulated meshing method. This generates 3D road model data containing elevation, longitudinal profile, and crack and depression features, with a modeling accuracy better than ±1cm. The fusion process utilizes a Multi-Channel Residual Fusion Network (MCRFN). The network inputs are: the RGB channels of the visible light image, the grayscale channels of the thermal stress distribution map, and the elevation grayscale map of the point cloud modeling projection. The network uses three layers of convolutional layers plus residual units to extract features from each modality. In the second stage, an attention mechanism (CBAM module) guides the fusion of information between the different modalities, outputting a fused road image of 512×512 pixels. This image preserves the texture details of the visible light image while also displaying the characteristics of areas of concentrated thermal stress. It also embeds 3D geometric structure information, facilitating subsequent crack identification and structural analysis.

[0093] Preferably, step S15 includes the following steps:

[0094] Step S151: performing density equalization on the three-dimensional point cloud data of the road surface to generate point cloud equalization data;

[0095] Step S152: performing ground segmentation on the three-dimensional point cloud data of the road surface based on the point cloud equalization data, distinguishing the road surface and eliminating the non-road surface, and generating road main surface segmentation data;

[0096] Step S153: performing normal estimation on the road main surface segmentation data to generate road surface normal information data; performing curved surface reconstruction on the road according to the road surface normal information data to generate road surface curved surface reconstruction data;

[0097] Step S154: Structural meshing is performed on the road surface curved surface reconstruction data to obtain road three-dimensional modeling data.

[0098] In an embodiment of the present invention, a voxel grid filter is used to downsample the original lidar point cloud to address the uneven density problem, ensuring a uniform distribution of point clouds in each region. The specific operations are as follows: setting the voxel side length to 0.05m; segmenting the original point cloud into multiple voxel blocks; selecting the center point or centroid point in each voxel block as the representative point; generating point cloud density-balanced data, reducing the number of point clouds from the original approximately 1.2 million points per frame to approximately 400,000 points per frame, improving subsequent processing efficiency while preserving structural features. Road surface extraction is performed on the density-balanced point cloud data, using the RANSAC (Random Sample Consensus) algorithm for ground fitting and non-ground point removal: setting the minimum in-plane point ratio to 70% and the distance threshold to 0.02m; extracting points with residuals less than the threshold from the plane model as the main road surface; and removing 3D objects with a height exceeding 0.3m, such as vehicles, curbs, and trees. The main road surface segmentation data containing the main road area is generated in the .pcd or .las format. Normal vectors were estimated from the segmented road surface data to capture subtle road inclination and curvature features. The K-neighborhood normal estimation algorithm from the PCL (Point Cloud Library) was used with the following parameters: the number of neighborhood points, k, was set to 30; the normal vectors were uniformly oriented upwards in the LiDAR coordinate system; and the output normal data was formatted as a vector field. Based on this normal information, the Moving Least Squares (MLS) surface fitting algorithm was used to reconstruct the road point cloud. A Gaussian kernel was used as the kernel function, with a bandwidth, σ, set to 0.1m. The interpolation accuracy was set to 0.005m. The reconstructed surface retained road waveform features, such as rutting and bumps, and other microstructures, generating road surface reconstruction data. Structured mesh generation is performed on the surface reconstruction data. A terrain modeling method based on the Delaunay triangulation algorithm is used to construct a regular or irregular triangular mesh (TIN). The mesh side length range is set to [0.05m, 0.15m]. The mesh nodes record the spatial position (x, y, z) and the curvature change. The modeling data structure is output as a .obj or .stl three-dimensional mesh file. At the same time, geometric parameters such as road width, slope, and camber are extracted to support subsequent thermo-mechanical analysis and simulation, ultimately obtaining the road 3D modeling data points.

[0099] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0100] Step S21: extracting texture abnormality areas based on the fused road image, and performing regional core division on the texture abnormality areas to obtain texture abnormality core area images and texture abnormality edge area images;

[0101] Step S22: Analyze the texture morphology and direction of the texture abnormality core area image to generate the core area texture feature data; calculate the directional consistency and grayscale gradient concentration of the texture abnormality edge area image to obtain the edge area texture feature data;

[0102] Step S23: Performing 3D projection on the fused road image using the core area texture feature data and the edge area texture feature data, and extracting the elevation value and normal vector change after projection; calculating the curvature of the texture abnormal area based on the elevation value and normal vector change to generate crack curvature data;

[0103] Step S24: Obtain a historical pavement structure design drawing; perform cross-layer structural crack feature matching on the historical pavement structure design drawing based on the crack curvature data to generate reflected crack matching data; perform crack coordinate conversion and positioning on the road using the reflected crack matching data to generate reflected crack initial positioning data.

[0104] In the embodiment of the present invention, the fused road image generated in step S1 is used to detect texture anomaly areas using the gray-level co-occurrence matrix (GLCM) + local binary pattern (LBP) algorithm: GLCM features such as image contrast, entropy, and energy are calculated in a sliding window with an image scale of 512×512 pixels; sub-blocks with abnormal LBP distribution in the texture anomaly area are extracted; clustering is performed based on region growing and connected domain analysis methods to divide the core area and edge area: the area of ​​the core area is generally controlled to be 15 to 40 pixels2; the width of the edge area is 2 to 5 pixels outside the core boundary; finally, images of the texture anomaly core area and edge area are generated, and the region mask data format (such as .png+.json annotation file) is output. Texture feature extraction processing is performed on the core and edge areas separately: Fourier transform (FFT) and structure tensor analysis are used to jointly extract the main texture direction and morphological features; the main texture strike angle (0° to 180°) and morphological compactness are obtained; and the output is a core area texture feature data matrix. The standard deviation of texture orientation is calculated using the Orientation Coherence coefficient. The Sobel operator is used to extract grayscale gradients and calculate the Gradient Concentration Index (GCI). This output is a texture feature matrix for the edge region. The coordinates of the texture anomaly region are projected onto the 3D road model using the camera's intrinsic and extrinsic parameter matrix (obtained through multi-view geometry calibration). The corresponding elevation (Z) change and surface normal vector are extracted from the projected point cloud. The curvature (κ) of the crack region is calculated using the following formula: κ = Δθ / Δs, where Δθ is the angle change between the normal vectors (in radians) and Δs is the arc length between the curvature calculation points (in meters). For each texture anomaly patch, a mean curvature value (typically in the range of 0.01 to 0.12 rad / m) is output, and crack curvature data for the corresponding spatial location is generated. Obtain the historical structural design drawing of the road (such as CAD format or vector structure drawing) and convert it into a unified spatial coordinate system; use the template matching + ICP point cloud alignment algorithm based on the crack curvature data and the material boundary layer or cutting seam characteristics in the historical structure drawing to achieve cross-layer crack feature matching; the matching results generate reflected crack matching data in a format that includes a matching score (a match is considered successful when the matching degree is greater than 85%) and a spatial offset value; convert the matching area into GIS coordinates or WGS84 coordinates in the 3D modeling coordinate system and output it as the initial positioning data of the reflected crack.

[0105] Preferably, performing cross-layer structural crack feature matching on the historical pavement structure design drawing based on the crack curvature data in step S24 includes:

[0106] Extract the layer structure from the historical pavement structure design drawings to generate pavement layer profile data;

[0107] Perform projection transformation on the crack curvature data to generate curvature structure mapping data;

[0108] Perform inter-layer trajectory overlap analysis on the curvature structure mapping data and the pavement layer profile data to generate curvature layer coupling map data;

[0109] The commonality of structural features is determined on the curvature layer coupling map data to generate reflection feature consistency data;

[0110] The fracture path consensus reconstruction is performed on the curvature layer coupling map data based on the reflection feature consistency data to generate reflection fracture matching data.

[0111] In an embodiment of the present invention, a historical pavement structure design drawing (usually in .dxf, .dwg or .tiff format) is imported into a CAD / GIS parsing module; vector graphics primitive recognition and color / line layer separation algorithms are used to extract the boundaries representing different structural layers (such as surface layer, base layer, subbase layer, etc.) in the drawing; thickness information and material labels between structural layers are extracted; bedding boundary lines are converted into a unified longitudinal coordinate system (such as X is path distance and Y is depth) to generate pavement layer profile data (in a format such as a two-dimensional curve set or GeoJSON structure); examples of measured parameters include: the thickness of the asphalt surface layer is 8 cm; the granular base layer is 18 cm; and the thickness of the water-stable layer is 28 cm. Using the crack spatial coordinate data and curvature value data obtained in step S23, the three-dimensional curvature points are transformed into the hierarchical profile coordinate system through orthographic projection. The projection algorithm, based on multi-view geometry (MVG) and projection matrix calculation, maps the crack line in the profile. The result is curvature structure mapping data, which records the crack's trajectory and corresponding curvature value in each structural layer. For a measured example, at a longitudinal distance of 5.2 m, the crack extends from a surface layer curvature of 0.068 rad / m to a base layer curvature of 0.042 rad / m. Trajectory overlap analysis is performed on the curvature structure mapping data and the pavement hierarchical profile data: Bezier curve interpolation is used to fit the crack trajectory. The extension length and direction consistency of the crack line segments between different structural layers are analyzed. If the curvature trajectory changes continuously between two layers (Δκ≤0.03 rad / m), the trajectory is considered coupled. The output result is curvature layer coupling atlas data, indicating the coupling segments and fracture trend segments. Statistical and structural pattern recognition analysis was performed on the curvature layer coupling atlas data. The crack initiation angle, extension length, and material boundary angle of each layer were extracted. A support vector machine (SVM) coupled with a hierarchical clustering algorithm was used to identify commonalities between different crack segments across structural layers. If the coupling paths exhibited similar morphologies across multiple profiles, they were identified as homologous reflection cracks. The results were output as reflection feature consistency data, including a path similarity score (e.g., 0 to 1, with a threshold of 0.85) and a position offset. Based on this reflection feature consistency data, the fracture path was reconstructed retrospectively within the 3D modeling data. A particle filtering algorithm was used for path fitting optimization. A multi-solution path set was generated based on the matching scores between known fracture trajectory points and reflection paths. A weighted average reconstruction of the path set was performed to output a unique consensus fracture path. Finally, reflection crack matching data was generated, including the fracture start and end points, path direction, interlayer penetration location, and reconstruction accuracy (e.g., RMSE < 5 cm).

[0112] Preferably, step S3 includes the following steps:

[0113] Step S31: Acquire road base material data;

[0114] Step S32: Deconstructing the material elastic characteristics of the road base material data to generate material elastic modulus field data;

[0115] Step S33: performing crack region meshing on the initial positioning data of the reflected crack to generate crack finite element discrete mesh data; performing stress field solving on the crack finite element discrete mesh data based on the material elastic modulus field data to generate the reflected crack stress field;

[0116] Step S34: extracting the stress gradient direction of the reflected crack stress field, and performing nonlinear coupling prediction on the road base material data to generate crack extension trend prediction data.

[0117] In this embodiment of the present invention, the physical and mechanical properties of road base materials are obtained through the following methods: on-site coring experiments (drilling depth is generally 30cm to 50cm); combining geological radar tomography (GPR) with FWD drop weight deflection testing; and extracting material type (such as cement-stabilized crushed stone, graded gravel, etc.), particle composition, moisture content, and density. The output is a structured material data table, including the following example parameters: Material type: cement-stabilized crushed stone; Density: 97.4%; Moisture content: 4.1%; Density: 2.32g / cm 3Initial elastic modulus: E = 1850 MPa. Elastic modulus field modeling is performed using a material mechanics model: the model is based on the heterogeneous elasticity theory (HET). Elastic modulus fitting is performed at the sampling points, and a spatially continuous field is generated using three-dimensional interpolation (such as Kriging or IDW). The nonlinear effects of material moisture content and density changes on the modulus are considered, and the elastic modulus distribution is modified. The generated material elastic modulus field data is in the form of a three-dimensional grid (e.g., each 1m×1m×0.1m unit), which supports finite element calculation input. Example: The elastic modulus of the (3,4,1) unit is E = 1705 MPa. Based on the initial location data of the reflective crack obtained in step S24, the crack path and surrounding influence radius (e.g., 1m) are extracted. The area surrounding the crack is discretized using finite element methods using Delaunay triangulation or quadrilateral meshing algorithms. The grid size is preferably 10 cm×10 cm to improve accuracy. The output is the crack finite element discrete mesh data, including nodes, boundary conditions, and material properties. Map the elastic modulus field data to a discrete grid; set boundary constraints (such as fixing the road edge and applying a distributed load Q to the wheel load area); use the finite element method (FEM) to solve two-dimensional / three-dimensional stress; solve the control equation: [K]{u}={F}; where K is the stiffness matrix, u is the displacement vector, and F is the node force; the solution output is the reflected crack stress field, including the spatial distribution of σ_x, σ_y, and τ_xy; example results: σ_max=1.42MPa at the crack starting point and τ_xy=0.26MPa at the crack tail. Extract the principal stress direction in the crack stress field (such as based on stress tensor eigenvalue analysis); calculate the stress gradient direction near each crack node. A vector field is formed; the potential extension path is clustered and calibrated according to the gradient direction of the stress concentration area (σ≥1.25MPa). The material elastic modulus field and the crack stress field are used as input features to construct a crack extension prediction model: BP neural network + LSTM time series fitting module is used; the input features include σ, τ, E. Material moisture; output is the coordinates and direction vector of the crack extension path points; output is crack expansion trend prediction data, visualized as a probability cloud map of the crack expansion path within the next 30 cm range.

[0118] Preferably, step S34 includes the following steps:

[0119] Step S341: performing multi-scale tensor differential processing on the reflected fracture stress field data to generate fracture principal stress gradient tensor data; performing directional field stability analysis on the fracture principal stress gradient tensor data to generate stress stability directional domain data;

[0120] Step S342: performing dynamic modulus partitioning on the road base material data to generate layered modulus mapping data;

[0121] Step S343: performing nonlinear interactive fusion on the stress stability direction domain data and the layered modulus mapping data to generate crack advancement probability field data;

[0122] Step S344: performing spatial entropy evolution on the fracture propagation probability field data to generate fracture extension trend prediction data.

[0123] In an embodiment of the present invention, multi-scale analysis is performed on stress field data from reflective cracks. By calculating stress variations at different scales, the principal stress directions and patterns of variation within the cracks can be identified. Specifically, differential processing at different scales is used to extract information about the rate of change of stress and the directionality of the principal stresses. This process helps determine the stress variation trend within the crack region. Next, the stability of the crack region is assessed by analyzing the consistency of stress directions. Stability analysis is performed on the stress directions within each region, and regions with strong stress direction consistency are identified. These regions are typically the preferred paths for crack propagation, providing an effective basis for subsequent crack development trend prediction. Dynamic modulus zoning is performed based on the physical properties of the road base material, such as elastic modulus and density. Geological survey data is used to analyze the elastic properties of the materials in different regions and, based on this data, to divide them into different modulus zones. Because the elastic properties of materials change with load, this zoning process is dynamic and adjusts to changes in the stress field. The resulting dynamic modulus data is mapped to the spatial distribution of the road base material to form layered modulus maps. These maps provide a detailed representation of the modulus variations within each region and provide a material property basis for crack propagation trend prediction. In this stage, data from the previous two steps are combined to construct a nonlinear interactive model, incorporating stress direction domain and material modulus mapping data. This approach identifies potential crack propagation paths and calculates the probability of crack development in different regions. This process incorporates multiple factors, such as the initial crack location, the strength of the stress field, and the elastic response of the material. Based on the model's calculations, a crack propagation probability field is generated. This data reflects the likelihood of crack propagation in different regions and helps predict which areas are most likely to experience crack propagation. This probability field can be used to guide road maintenance decisions and prioritize high-risk areas. Information entropy analysis of the spatial distribution of the crack propagation probability field assesses the uncertainty of crack development in different regions. Regions with higher spatial entropy values ​​indicate more dispersed and uncertain crack propagation paths, while regions with lower spatial entropy indicate more stable and well-defined crack propagation paths. As cracks propagate over time, changes in entropy values ​​reflect the crack's evolution. By analyzing the evolutionary trends of entropy values, the potential and direction of future crack propagation can be predicted. This process helps determine crack evolution trends and predict crack propagation regions based on entropy values. Based on the evolution of spatial entropy, crack growth trend prediction data is generated. This data can provide guidance for road crack repair and maintenance, helping decision-makers determine crack growth priorities and maintenance strategies. The final crack growth trend prediction data can effectively support subsequent repair work and the planning of prevention and control measures.

[0124] As an example of the present invention, refer to Figure 3 As shown, in this example, step S4 includes:

[0125] Step S41: performing spatial coupling mapping processing on the road 3D modeling data using the crack extension trend prediction data to generate crack 3D path prediction data;

[0126] Step S42: analyzing the path topology of the three-dimensional crack path prediction data, and extracting interlayer penetration nodes from the three-dimensional crack path prediction data based on the results of the path topology analysis to generate interlayer path node atlas data;

[0127] Step S43: performing heat flux simulation on the layer-penetrating path node atlas data to generate path thermal conductivity penetration data; performing dynamic thermal resistance control analysis on the initial positioning data of the reflective crack based on the path thermal conductivity penetration data to generate crack channel thermal blocking data;

[0128] Step S44: performing structural response feedback on the initial positioning data of the reflected crack according to the crack channel thermal blocking data to obtain reflected crack blocking feedback data to perform the road crack repair operation.

[0129] In an embodiment of the present invention, three-dimensional road modeling data and crack expansion trend prediction data are collected. The crack expansion trend prediction data can be derived from historical crack expansion data or prediction data obtained through methods such as finite element analysis. Spatial mapping algorithms, such as nearest neighbor interpolation and layered grid interpolation, are used to couple the three-dimensional road modeling data with the crack expansion trend prediction data. The purpose of coupled mapping is to generate three-dimensional crack path prediction data based on the geometric characteristics of the road and the crack expansion trend. Through spatial coupled mapping, the crack expansion path under different conditions is predicted, and three-dimensional path data is generated, including information such as the crack direction and expansion depth. Graph theory methods (such as the Dijkstra algorithm) are used to perform path topology analysis on the three-dimensional crack path prediction data to determine the connectivity of the cracks, the branching structure of the path, and potential intersections. Based on the results of the crack path topology analysis, interlayer nodes in the crack path (i.e., the portion where the crack crosses different geological layers or road structural layers) are extracted. These nodes are key areas that affect crack expansion or repair strategies. The extracted interlayer penetration nodes are mapped to generate interlayer path node map data. This data includes the crack's interlayer location, the intersection level of the path segments, and the relative location and importance of the nodes. Heat flux simulation is performed on the interlayer path node map data. Using a heat conduction model (such as finite element heat conduction analysis), the heat flow distribution along the crack path is simulated. Heat flow simulation can help evaluate the thermal conductivity of the crack path and the thermal effects that influence crack propagation. Heat flux simulation generates path thermal conductivity penetration data, which includes important indicators such as the thermal conductivity and thermal penetration degree of the crack path. Dynamic thermal resistance control analysis is performed on the initial reflected crack location data using the path thermal conductivity penetration data to calculate the change in thermal impedance of the crack channel. Dynamic thermal resistance control helps understand whether the crack will further propagate or be blocked due to heat transfer. Based on the results of the thermal resistance control analysis, crack channel thermal blockage data is generated, indicating which crack channels have been effectively blocked due to heat flow. Based on the crack channel thermal blockage data, structural response feedback analysis is performed on the initial reflected crack location data to evaluate the structural response under different conditions. The structural response includes changes in crack location, the effects of temperature gradients, and changes in bearing capacity. Structural response feedback generates reflected crack blocking feedback data. This data helps repair personnel understand the actual crack propagation situation and provides feedback to guide the optimization of repair strategies. Based on this crack blocking feedback data, crack repair operations are initiated. These repair operations may include crack grouting, structural reinforcement, and temperature regulation.

[0130] Preferably, step S43 includes the following steps:

[0131] Step S431: performing multi-physics thermal response modeling on the layer-penetrating path node atlas data to generate local node thermal coupling model data; performing time-series thermal disturbance simulation on the local node thermal coupling model data to generate thermal flow response sequence data;

[0132] Step S432: performing three-dimensional thermal field inversion calculation on the heat flow response sequence data to generate path thermal conductivity penetration data;

[0133] Step S433: Clustering and identifying non-uniform heat-sensitive areas on the path thermal conductivity penetration data to generate high-risk heat penetration cluster data; performing positional thermal gradient cross-analysis on the high-risk heat penetration cluster data and the initial positioning data of the reflective cracks to generate dynamic thermal resistance configuration data;

[0134] Step S434: performing a global thermal control simulation of the crack channel on the dynamic thermal resistance configuration data to generate crack channel thermal blocking data.

[0135] In an embodiment of the present invention, a thermal coupling model for local nodes is constructed by performing multi-physics field analysis on the node atlas data of the fracture path, using a coupled thermodynamic model and combining the thermophysical properties of materials (such as thermal conductivity, specific heat capacity, density, etc.). This model needs to consider the thermal interaction between the fracture path and the surrounding geological layers, as well as the thermal response characteristics of different materials. Through multi-physics modeling, thermal response data of local nodes under different environmental conditions are obtained. This data describes the coupled effects of heat conduction, thermal radiation, and thermal convection at each node. Based on the data from the local node thermal coupling model, a time-series thermal perturbation simulation is performed. Different thermal perturbation sources (such as ambient temperature changes, fracture extension, etc.) are introduced into the simulation to simulate the propagation of thermal perturbations at the nodes. Through simulation calculations, thermal flow response data in a time series is obtained. This data reflects the changes and responses of heat flow at different time points, helping to predict the thermal impact of the fracture path. An inversion algorithm (such as Bayesian inversion, genetic algorithm, or least squares method) is used to perform three-dimensional thermal field inversion calculations on the thermal flow response sequence data. Through inversion techniques, the known heat flow data is used to infer the thermal field distribution around the fracture path. From the three-dimensional thermal field inversion results, path thermal conductivity and penetration data are extracted. These data include thermal conductivity and penetration heat flux density along the fracture path, providing important insights for subsequent thermal control analysis. The path thermal conductivity and penetration data are then clustered to identify non-uniform thermally sensitive regions. Density-based spatial clustering algorithms (such as DBSCAN) or K-means clustering are used to identify regions exhibiting high thermal sensitivity during heat flow. The cluster analysis results are presented as high-risk thermal penetration clusters, representing areas where heat accumulation occurs during heat transfer, accelerating crack propagation. Identifying these high-risk areas is crucial for subsequent crack repair. A positional thermal gradient intersection analysis is performed on the high-risk thermal penetration cluster data and the initial location data of the reflected cracks. By calculating the intersection of the thermal gradients between the fracture path and the high-risk thermal penetration region, it is assessed whether the crack will propagate due to heat flow. Based on the results of the thermal gradient intersection analysis, dynamic thermal resistances are configured. These data are used to adjust the thermal resistance of the fracture channel to prevent heat from propagating through the high-risk region, thereby reducing the risk of crack propagation. Based on the dynamic thermal resistance configuration data, a global thermal control simulation of the fracture channel is performed using thermal control simulation software (such as COMSOL or ANSYS). The simulation process takes into account the thermal characteristics of the fracture path, the direction of the fracture, the possibility of fracture expansion, and its thermal interaction with the surrounding geological layers. Through global thermal control simulation, the thermal blocking effect in the fracture channel is calculated, and it is predicted which fracture channels will be blocked by heat flow and which areas will cause crack expansion or aggravation due to thermal effects. Based on the simulation results, the fracture channel thermal blocking data is generated. This data will be used to guide the strategy of crack repair, especially in repair scenarios involving temperature changes or heat conduction, helping the repair team optimize the repair design and construction plan.

[0136] In this specification, a road crack detection system based on fused images is provided, which is used to perform the above-mentioned road crack detection method based on fused images. The road crack detection system based on fused images includes:

[0137] The image fusion module is used to obtain multi-source road monitoring data, including visible light images, infrared thermal imaging data, and lidar point cloud data, and perform multimodal image fusion and road 3D point cloud reconstruction to generate fused road images and road 3D modeling data;

[0138] The crack location module is used to perform crack curvature analysis based on the fused road image and generate crack curvature data; the crack curvature data is used to identify and locate the reflected crack contours of the fused road image and generate initial reflected crack location data;

[0139] The crack stress simulation module is used to obtain road base material data; reconstruct the reflected crack stress field of the road area based on the initial location data of the reflected cracks to obtain the reflected crack stress field; predict the crack expansion trend of the road base material data based on the reflected crack stress field to generate crack expansion trend prediction data;

[0140] The layer penetration blocking module is used to use the crack expansion trend prediction data to perform reflection crack excitation analysis on the road 3D modeling data to generate reflection crack interlayer penetration channel data; the reflection crack interlayer penetration channel data is used to thermally block the crack channel in the road area to perform road crack repair operations.

[0141] The present invention's beneficial effect lies in integrating visible light imagery, infrared thermal imaging data, and LiDAR point cloud data, synthesizing multiple sensory information to provide comprehensive road monitoring data. This multimodal data fusion not only effectively improves road detection accuracy but also enables identification of road defects in different environments, such as the interplay between temperature changes and physical structure. By performing multimodal image fusion and road point cloud reconstruction on this data, the module generates fused road images and 3D road modeling data, significantly improving the accuracy of the road models. These precise 3D models serve as the basis for subsequent crack analysis and repair operations, reducing errors and improving data reliability. The fused images and 3D data provide multi-angle information for subsequent analysis. Combining infrared thermal imaging with LiDAR data effectively reveals potential cracks and other road damage. This fusion solution ensures comprehensive and accurate road status, providing strong data support for crack identification and repair solutions. By performing crack curvature analysis based on the fused road images, the module generates precise crack curvature data, which is then used to identify and locate reflective crack contours. This step not only accurately identifies the location of road cracks but also efficiently locates the crack edges, providing clear targets for subsequent repair work. Analyzing crack curvature accurately identifies crack morphology and growth trends, ensuring high-precision crack identification. This is particularly important for locating cracks in complex road environments, avoiding misidentification or missed detections often associated with traditional methods and improving crack detection reliability. By acquiring road base material data and combining it with initial reflected crack location data, the module reconstructs the crack stress field and generates a reflected crack stress field. This process accurately simulates the stress distribution of cracks on the road surface, providing high-quality input data for crack growth trend prediction. Based on the reflected crack stress field data, the module predicts crack growth trends and generates crack growth trend prediction data. This prediction can identify potential crack growth issues in advance, providing early warning before repair work begins, enabling targeted repair plans to be developed and preventing further damage from crack expansion. By using the crack growth trend prediction data for reflected crack excitation analysis, the module generates reflected crack interlayer crossover channel data, effectively identifying crack propagation paths. This step helps determine whether cracks pose a risk of cross-layer penetration and provides specific guidance for repair work. By performing thermal blockage analysis on the crack channel, the module can thermally control crack growth and facilitate road crack repair. Thermal blocking analysis helps provide effective thermal isolation during the crack repair process, preventing further crack expansion due to thermal diffusion and ensuring the durability of the repair. By accurately analyzing the interlayer cross-channels of the crack and implementing thermal blocking, the module can provide more intelligent repair strategies, reduce unnecessary intervention during the repair process, improve repair efficiency, and ensure the long-term stability of the repair results.Therefore, the present invention improves the accuracy and adaptability of road reflective crack detection through multimodal fusion, curvature recognition, stress field modeling and crack channel analysis.

[0142] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0143] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A road crack detection method based on fusion image, characterized in that: The following steps are involved: Step S1: Acquire multi-source road monitoring data, including visible light images, infrared thermal imaging data, and lidar point cloud data, and perform multimodal image fusion and road 3D point cloud reconstruction to generate fused road images and road 3D modeling data; Step S2: performing crack curvature analysis based on the fused road image to generate crack curvature data; performing reflected crack contour identification and positioning on the fused road image using the crack curvature data to generate reflected crack initial positioning data; Step S3: Acquire road base material data; reconstruct the reflection crack stress field of the road area based on the initial reflection crack positioning data to obtain the reflection crack stress field; predict the crack expansion trend of the road base material data based on the reflection crack stress field to generate crack expansion trend prediction data; Step S4: using the crack extension trend prediction data to perform reflection crack excitation analysis on the road 3D modeling data to generate reflection crack interlayer crossing channel data; By reflecting the data of the channel crossing between crack layers, the crack channel in the road area is thermally blocked to perform road crack repair operations.

2. The road crack detection method based on fusion image according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: synchronously collecting visible light images and infrared thermal imaging data of the road through a vehicle-mounted multispectral imaging device; Step S12: using a mobile laser radar scanner to obtain three-dimensional point cloud data of the road surface; Step S13: performing radiometric correction and geometric registration on the visible light image to generate standard visible light image data; Step S14: performing temperature field mapping and noise suppression processing on the infrared thermal imaging data to generate thermal stress distribution data; Step S15: reconstructing the three-dimensional point cloud of the road based on the three-dimensional point cloud data of the road surface to generate three-dimensional road modeling data; Step S16: Perform multi-scale fusion of the standard visible light image, thermal stress distribution data, and road 3D modeling data to generate a fused road image.

3. The road crack detection method based on fusion image according to claim 2, characterized in that: Step S15 includes the following steps: Step S151: performing density equalization on the three-dimensional point cloud data of the road surface to generate point cloud equalization data; Step S152: performing ground segmentation on the three-dimensional point cloud data of the road surface based on the point cloud equalization data, distinguishing the road surface and eliminating the non-road surface, and generating road main surface segmentation data; Step S153: performing normal estimation on the road main surface segmentation data to generate road surface normal information data; performing curved surface reconstruction on the road according to the road surface normal information data to generate road surface curved surface reconstruction data; Step S154: Structural meshing is performed on the road surface curved surface reconstruction data to obtain road three-dimensional modeling data.

4. The road crack detection method based on fusion image according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: extracting texture abnormality areas based on the fused road image, and performing regional core division on the texture abnormality areas to obtain texture abnormality core area images and texture abnormality edge area images; Step S22: Analyze the texture morphology and direction of the texture abnormality core area image to generate the core area texture feature data; calculate the directional consistency and grayscale gradient concentration of the texture abnormality edge area image to obtain the edge area texture feature data; Step S23: Performing 3D projection on the fused road image using the core area texture feature data and the edge area texture feature data, and extracting the elevation value and normal vector change after projection; calculating the curvature of the texture abnormal area based on the elevation value and normal vector change to generate crack curvature data; Step S24: Obtain a historical pavement structure design drawing; perform cross-layer structural crack feature matching on the historical pavement structure design drawing based on the crack curvature data to generate reflected crack matching data; perform crack coordinate conversion and positioning on the road using the reflected crack matching data to generate reflected crack initial positioning data.

5. The road crack detection method based on fusion image according to claim 4, characterized in that: In step S24, matching the cross-layer structural crack characteristics of the historical pavement structure design drawing based on the crack curvature data includes: Extract the layer structure from the historical pavement structure design drawings to generate pavement layer profile data; Perform projection transformation on the crack curvature data to generate curvature structure mapping data; Perform inter-layer trajectory overlap analysis on the curvature structure mapping data and the pavement layer profile data to generate curvature layer coupling map data; The commonality of structural features is determined on the curvature layer coupling map data to generate reflection feature consistency data; The fracture path consensus reconstruction is performed on the curvature layer coupling map data based on the reflection feature consistency data to generate reflection fracture matching data.

6. The road crack detection method based on fusion image according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: Acquire road base material data; Step S32: Deconstructing the material elastic characteristics of the road base material data to generate material elastic modulus field data; Step S33: performing crack region meshing on the initial positioning data of the reflected crack to generate crack finite element discrete mesh data; performing stress field solving on the crack finite element discrete mesh data based on the material elastic modulus field data to generate the reflected crack stress field; Step S34: extracting the stress gradient direction of the reflected crack stress field, and performing nonlinear coupling prediction on the road base material data to generate crack extension trend prediction data.

7. The road crack detection method based on fusion image according to claim 6, characterized in that: Step S34 includes the following steps: Step S341: performing multi-scale tensor differential processing on the reflected fracture stress field data to generate fracture principal stress gradient tensor data; performing directional field stability analysis on the fracture principal stress gradient tensor data to generate stress stability directional domain data; Step S342: performing dynamic modulus partitioning on the road base material data to generate layered modulus mapping data; Step S343: performing nonlinear interactive fusion on the stress stability direction domain data and the layered modulus mapping data to generate crack advancement probability field data; Step S344: performing spatial entropy evolution on the fracture propagation probability field data to generate fracture extension trend prediction data.

8. The road crack detection method based on fusion image according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: performing spatial coupling mapping processing on the road 3D modeling data using the crack extension trend prediction data to generate crack 3D path prediction data; Step S42: analyzing the path topology of the three-dimensional crack path prediction data, and extracting interlayer penetration nodes from the three-dimensional crack path prediction data based on the results of the path topology analysis to generate interlayer path node atlas data; Step S43: performing heat flux simulation on the layer-penetrating path node atlas data to generate path thermal conductivity penetration data; performing dynamic thermal resistance control analysis on the initial positioning data of the reflective crack based on the path thermal conductivity penetration data to generate crack channel thermal blocking data; Step S44: performing structural response feedback on the initial positioning data of the reflected crack according to the crack channel thermal blocking data to obtain reflected crack blocking feedback data to perform the road crack repair operation.

9. The road crack detection method based on fusion image according to claim 8, characterized in that: Step S43 includes the following steps: Step S431: performing multi-physics thermal response modeling on the layer-penetrating path node atlas data to generate local node thermal coupling model data; performing time-series thermal disturbance simulation on the local node thermal coupling model data to generate thermal flow response sequence data; Step S432: performing three-dimensional thermal field inversion calculation on the heat flow response sequence data to generate path thermal conductivity penetration data; Step S433: Clustering and identifying non-uniform heat-sensitive areas on the path thermal conductivity penetration data to generate high-risk heat penetration cluster data; performing positional thermal gradient cross-analysis on the high-risk heat penetration cluster data and the initial positioning data of the reflective cracks to generate dynamic thermal resistance configuration data; Step S434: performing a global thermal control simulation of the crack channel on the dynamic thermal resistance configuration data to generate crack channel thermal blocking data.

10. A road crack detection system based on fusion images, characterized in that: For executing the road crack detection method based on fused images as claimed in claim 1, the road crack detection system based on fused images comprises: The image fusion module is used to obtain multi-source road monitoring data, including visible light images, infrared thermal imaging data, and lidar point cloud data, and perform multimodal image fusion and road 3D point cloud reconstruction to generate fused road images and road 3D modeling data; The crack location module is used to perform crack curvature analysis based on the fused road image and generate crack curvature data; the crack curvature data is used to identify and locate the reflected crack contours of the fused road image and generate initial reflected crack location data; The crack stress simulation module is used to obtain road base material data; reconstruct the reflected crack stress field of the road area based on the initial location data of the reflected cracks to obtain the reflected crack stress field; predict the crack expansion trend of the road base material data based on the reflected crack stress field to generate crack expansion trend prediction data; The layer penetration blocking module is used to use the crack expansion trend prediction data to perform reflection crack excitation analysis on the road 3D modeling data to generate reflection crack interlayer penetration channel data; the reflection crack interlayer penetration channel data is used to thermally block the crack channel in the road area to perform road crack repair operations.

Citation Information

Cited By

  • Three-dimensional identification modeling method and system for power grid channel

    CN120951700A

  • A three-dimensional identification modeling method and system for a power grid channel

    CN120951700B

  • Regional tectonic stress risk quantitative analysis method for tunnel engineering

    CN120995730A

  • Digital twinning fused material-equipment collaborative crack repair design method and equipment

    CN121031239A

  • Pavement crack detection device and method based on machine vision

    CN121095229A