Road disease detection method, device and equipment and storage medium

Abstract

The application discloses a road disease detection method, device and equipment and a storage medium. The method comprises the following steps: traversing road picture data of vehicles driving on a road in a preset period and taking pictures by a vehicle-mounted image device; inputting the road picture data into a pre-trained road basic information recognition model to recognize road basic information, wherein the road basic information comprises a road boundary and a road marking line; in response to the road boundary containing a valid pavement range, inputting the road picture data into a pre-trained disease detection model to output an initial detection result of whether a disease exists; performing spatial logic verification based on the initial detection result and the road basic information to obtain an effective detection result of whether a disease exists, and outputting a disease detection result in combination with road network data. According to the scheme, the effective pavement range and the marking line position are accurately defined first, then the disease is accurately recognized, and the spatial relationship between the disease and the marking line is introduced for secondary verification, so that the road disease detection accuracy is improved.

Description

Technical Field

[0001] This application relates to the field of intelligent detection technology, and more specifically, to a method, apparatus, equipment, and storage medium for detecting road defects. Background Technology

[0002] As road maintenance and management evolves towards intelligence and precision, automatic road defect identification technology based on computer vision has become a research hotspot. Current mainstream methods primarily rely on a single end-to-end deep learning detection model, treating defect identification as a general object detection task and directly training the model using large-scale labeled data. However, current solutions lack a structural understanding of the road scene and fail to explicitly distinguish between different semantic regions such as drivable road surfaces and medians. This leads to numerous surface defects occurring in non-driving areas such as curbs, sidewalks, and medians being misclassified as valid road defects. Defects in road marking areas are also prone to being missed or misclassified, resulting in large detection errors and insufficient positioning accuracy. Summary of the Invention

[0003] This application provides a method, apparatus, device, and storage medium for detecting road defects, in order to at least solve the technical problem of low accuracy in road defect detection in related technologies.

[0004] According to one aspect of the embodiments of this application, a method for detecting road defects is provided, comprising: Iterate through road image data captured by vehicles traveling on the road using in-vehicle imaging equipment within a preset time period; The road image data is input into a pre-trained road basic information recognition model to identify the road basic information, which includes road boundaries and road markings. In response to the fact that the road boundary includes the effective road surface area, the road image data is input into a pre-trained disease detection model, and an initial detection result of whether or not a disease exists is output. Based on the initial detection results and the road infrastructure information, spatial logic verification is performed to obtain valid detection results for the presence or absence of defects. The defect detection results are then output in conjunction with the road network data.

[0005] In one implementation, spatial logic verification is performed based on the initial detection results and the road infrastructure information to obtain a valid detection result indicating the presence or absence of defects, including: Obtain the candidate regions of diseases from the initial detection results; Determine whether the candidate area of ​​the defect is located within the effective road surface area defined by the road boundary; When the candidate area of ​​the defect is located within the effective road surface area, the intersection-combination ratio of the candidate area of ​​the defect and the road markings is calculated; When the intersection ratio of the candidate disease area and the road marking is less than a preset threshold, the candidate disease area is determined to be a valid disease.

[0006] In one implementation, it further includes: When the candidate area of ​​a defect is located within the effective road surface area, and the candidate area of ​​a defect and the road markings do not intersect, the candidate area of ​​a defect is determined to be a valid defect.

[0007] In one implementation, the output of defect detection results is combined with road network data, including: Based on the timestamp information in the road image data, the driving trajectory data of the photographed vehicle in the corresponding time period is obtained; The driving trajectory data is matched with the road network data to obtain information on road sections with effective defects; Associate the disease types corresponding to the effective diseases, generate and output disease detection results that include disease types and road segment information.

[0008] In one implementation, it further includes: Based on the geometric characteristics and disease type of the effective diseases, calculate their severity index; Based on the road grade and current traffic flow characteristics in the road segment information of the effective defects, the risk level of the effective defects is obtained; The severity score of the effective disease is obtained by weighted summation based on the severity index and risk level, and the severity score is output.

[0009] In one implementation, before inputting the road image data into a pre-trained road infrastructure information recognition model, the method further includes: The image data containing road damage is manually annotated to form an annotated dataset. The annotations include the effective boundary of the road, various road markings and their types. When the road has defects within its effective area, the labeling content also includes the defect type and the defect range boundary; A YOLO neural network model is trained based on the road boundary and road marking data in the labeled dataset to obtain a trained road basic information recognition model, which is used to identify the road boundary and road marking.

[0010] In one implementation, before inputting the road image data into the pre-trained disease detection model, the method further includes: A YOLO neural network model is trained based on the disease types and disease range boundary data of the labeled dataset to obtain a trained disease detection model. The disease detection model is used to identify disease types and disease extent boundaries.

[0011] According to another aspect of the embodiments of this application, a road defect detection device is provided, comprising: The image acquisition module is used to traverse road image data captured by vehicles traveling on the road through in-vehicle imaging equipment within a preset time period; The road information recognition module is used to input the road image data into a pre-trained road basic information recognition model to identify the road basic information, which includes road boundaries and road markings. The defect detection module is used to respond to the fact that the road boundary includes the effective road surface area, input the road image data into the pre-trained defect detection model, and output the initial detection result of whether there is a defect; The verification output module is used to perform spatial logic verification based on the initial detection results and the road basic information to obtain a valid detection result of whether there are defects, and output the defect detection result in combination with the road network data.

[0012] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the road defect detection method described above through the computer program.

[0013] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer program, which is configured to execute the above-described road defect detection method when running.

[0014] The technical solutions provided in this application embodiment may include the following beneficial effects: This application constructs a collaborative computational pipeline for a road infrastructure information identification model and a road defect detection model. First, the road infrastructure information identification model is used to analyze road boundaries and road markings. Then, combined with industry standards, defects are identified within the effective road surface area. A spatial relationship verification mechanism between defects and road markings significantly suppresses road marking interference and false alarms from outside the road. Simultaneously, by matching vehicle trajectories with a high-precision road network, defects are accurately mapped from image coordinates to specific road segments and lanes. Ultimately, this achieves beneficial effects such as improved detection accuracy, enhanced positioning precision, and compliance of defect classification, systematically overcoming the engineering implementation bottlenecks of traditional end-to-end methods, including high false detection rates, ambiguous positioning, and disconnect from standards. Attached Figure Description

[0015] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1This is a flowchart of a road defect detection method according to an embodiment of this application; Figure 2 This is a schematic diagram of a model training method according to an embodiment of this application; Figure 3 This is a schematic diagram of a disease detection process according to an embodiment of this application; Figure 4 This is a schematic diagram of a road surface distress scenario according to an embodiment of this application; Figure 5 This is a schematic diagram of a road defect detection device according to an embodiment of this application; Figure 6 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0016] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0017] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0018] The road defect detection method of this application embodiment will be described in detail below with reference to the accompanying drawings. Figure 1 As shown, the method mainly includes the following steps: S101 iterates through road image data captured by vehicles traveling on the road using in-vehicle imaging equipment within a preset time period.

[0019] In one implementation, a detection system deployed on a cloud or edge computing platform initiates a data collection task, automatically traversing vehicles on the road connected to the platform, such as logistics vehicles, taxis, and private cars, and collecting road image data through in-vehicle imaging equipment within the most recent preset time period, such as 24 hours.

[0020] S102 inputs road image data into a pre-trained road basic information recognition model to identify road basic information, including road boundaries and road markings.

[0021] During implementation, the collected road image data can be preprocessed, such as adjusting image resolution and normalizing. The preprocessed, standardized images are then input into a pre-trained road infrastructure information recognition model. This model uses a YOLO-based instance segmentation network to output the recognized road infrastructure information, including the identified road boundaries and road markings.

[0022] In an exemplary scenario, this application uses structured data output by a road infrastructure information recognition model to accurately deconstruct the semantic and geometric information of a road scene: It outputs road boundary masks identifying elements such as drivable areas, shoulders, and medians to define the effective range of defect detection. It also outputs multiple categories of road markings, including lane lines, edge lines, stop lines, pedestrian crossings, and directional arrows, along with their geometric shapes. Each instance includes its category label, contour coordinates, and confidence level. Together, these constitute the spatial constraint benchmark for defect analysis.

[0023] S103 responds to the fact that the road boundary includes the effective road surface area, inputs road image data into the pre-trained defect detection model, and outputs the initial detection result of whether defects exist.

[0024] After the system determines that there is a valid road surface area in the road boundary recognition results, it triggers the defect detection process, inputting the original road image data into the pre-trained YOLO defect detection model. This model is trained on a dataset labeled with domestic road maintenance standards and outputs multiple defect candidate regions and their initial detection results in real time. Each result includes the defect bounding box coordinates, confidence score, and predefined defect type.

[0025] In one embodiment, the types of defects include, but are not limited to, cracks, blocky fissures, potholes, ruts, subsidence, loosening, oil seepage, and poor repair.

[0026] S104 performs spatial logic verification based on the initial detection results and road infrastructure information to obtain valid detection results for the presence of defects, and outputs defect detection results in combination with road network data.

[0027] In one implementation, spatial logic verification is performed based on the initial detection results and road infrastructure information to obtain a valid detection result indicating whether a defect exists. This includes obtaining candidate defect areas from the initial detection results; determining whether the candidate defect areas are located within the effective road surface area defined by the road boundary; calculating the intersection-union ratio (IUU) between the candidate defect areas and road markings when the candidate defect areas are located within the effective road surface area; and determining the candidate defect areas as valid defects when the IUU is less than a preset threshold.

[0028] Specifically, all candidate areas of defects are extracted from the initial detection results output by the defect detection model. Then, the geometric contour of each candidate area of ​​defects is spatially verified against the road boundary output by the road basic information recognition model. By comparing the candidate area of ​​defects with the road boundary area, it is determined whether the defect is located within the effective road surface area.

[0029] For candidate areas of road defects that pass the pavement extent verification, a spatial relationship analysis is further performed between these areas and the geometric contours of various identified road markings (such as lane lines, edge lines, zebra crossings, etc.). The intersection-union ratio (IUGR) between the defect area and each type of marking area is calculated. If the IUGR of the defect area with any marking area is greater than or equal to a preset interference threshold, the candidate area is determined to be likely a false detection caused by incomplete, damaged, or reflective markings, and is therefore filtered out. Only candidate areas of defects whose IUGR with all marking areas is lower than this threshold are retained, and these are ultimately identified as valid defects.

[0030] This implementation method employs two levels of spatial logic constraints to first ensure that the defects are located within the treatable road surface, and then eliminates interfering areas that are visually easily confused with road markings. This significantly improves the accuracy and practicality of defect detection, making the system more suitable for the needs of actual road maintenance scenarios.

[0031] In one embodiment, the method further includes determining the candidate area of ​​a defect as a valid defect when the candidate area of ​​the defect is located within the effective road surface area and the candidate area of ​​the defect and the road markings do not intersect.

[0032] When a candidate pavement defect does not spatially intersect with the geometric contours of any road markings, the system directly identifies the candidate defect as a valid defect without requiring intersection-union calculation. This logic is specifically designed for rapid verification of typical pavement defects that do not overlap with road markings, such as potholes located in the center of the lane or blocky cracks far from road markings, reducing computational complexity while ensuring detection accuracy.

[0033] In one implementation, after valid defects are detected, the defect detection results are output in combination with road network data, including obtaining the driving trajectory data of the photographed vehicle in the corresponding time period based on the timestamp information in the road image data; and matching the driving trajectory data with the road network data to obtain the road segment information of valid defects.

[0034] Specifically, based on the standardized timestamps embedded in the road image files and the corresponding unique vehicle identifiers, the trajectory sequence of the vehicle within a set time window before and after the image capture time is extracted from the onboard equipment or trajectory database. This sequence includes information such as latitude and longitude coordinates and speed. Subsequently, a map matching algorithm is used to match the trajectory sequence with high-precision road network data in real time to determine the precise road segment number, driving direction, and lane number information of the vehicle at the time of capture.

[0035] This application improves the positioning accuracy of defect detection results through road network matching, positioning down to the lane level, and achieving precise binding of defects, road sections, and lanes, providing a directly locatable construction basis for maintenance operations.

[0036] Furthermore, by associating the valid defects with the corresponding defect types, defect detection results containing defect types and road segment information are generated and output. For example... Figure 4 The image shows the road surface defects identified in the embodiments of this application, which can accurately identify the defect types and areas of specific road sections.

[0037] In one optional implementation, the detected defect type and the road segment information corresponding to the effective defect are output. For example, the output results include: the defect type is "pothole" and it is located "in the second lane of G42 Shanghai-Nanjing Expressway, 1.2 meters away from the center line of the lane".

[0038] In an optional implementation, the method further includes calculating a severity index based on the geometric characteristics and type of the effective disease. For example, geometric characteristics such as disease area, length, and width are obtained based on the boundaries of the effective disease. Subsequently, the severity index is calculated using a weighted formula based on a predefined disease type weight table (e.g., pits have a higher base weight than cracks) and a geometric feature grading threshold. For example, a pit with an area of ​​0.8 square meters and a depth of 4 centimeters might receive an index of 85, while a crack that is 2 meters long and 3 millimeters wide might receive an index of 30.

[0039] Furthermore, based on the road grade and current traffic flow characteristics in the effective road segment information, the risk level of the effective defects is obtained. First, based on the road segment information obtained from road network matching, the road grade coefficient is extracted. The risk level can be queried from a preset lookup table, such as a weight of 1.0 for highways, 0.8 for national highways, and 0.6 for municipal roads. Real-time traffic flow data is also integrated. For example, during peak hours, a defect located in the overtaking lane of a highway will receive a higher risk level because it affects high-speed vehicles and has high traffic volume. Conversely, the risk level of the same defect will be significantly lower if it appears on a municipal auxiliary road with low traffic volume at night.

[0040] Furthermore, based on the severity index and risk level, a weighted sum is obtained to obtain the severity score of the effective disease, and the severity score is output. The severity index and risk level are multiplied by configurable weighting coefficients, such as a severity weight of 0.7 and a risk level weight of 0.3, and then the weighted sum is obtained to obtain the comprehensive severity score.

[0041] When outputting the detection results, the severity score of the effective diseases can be output together, and the terminal can perform orderly processing based on the severity score.

[0042] To facilitate understanding of the detection process in the embodiments of this application, the following description is provided in conjunction with the appendix. Figure 3 Further description.

[0043] like Figure 3 As shown, the system iterates through all image data acquired by vehicles equipped with front-facing cameras within a certain time frame. A road infrastructure information recognition model is used to determine the effective road area and road markings. If a road area exists, a pavement defect detection model is used to determine if defects exist. If defects exist and are located within the road area, it is determined whether they overlap with the boundaries of road markings identified in the image. Results with defects and an intersection-union ratio (IUU) less than a threshold (e.g., less than 0.5), or no intersection with road markings, are selected as detected effective pavement defects. Further, based on the image capture time, vehicle trajectories are obtained, and road network matching is performed on these trajectories to determine detailed information about the road segments where pavement defects occur. These determined road segment locations and defect types serve as the final data delivered to the business application.

[0044] In one implementation, before inputting road image data into a pre-trained road basic information recognition model, the process further includes training the road basic information recognition model and training the disease detection model.

[0045] In this embodiment, image data containing road damage is manually annotated to form an annotated dataset. The annotation content includes the effective boundary of the road, various road markings and their types. When the effective road area contains defects, the annotation content also includes the defect type and defect boundary. A YOLO neural network model is trained based on the road boundary and road marking data of the annotated dataset to obtain a trained road basic information recognition model. The road basic information recognition model is used to identify road boundaries and road markings.

[0046] Specifically, firstly, the RDD Road Damage Open Dataset is downloaded, and its images and labeled data are used to fine-tune a target detection model based on YOLOv11 or a later version to train a preliminary road damage identification model. Then, this model is used to batch screen all road images collected by vehicles equipped with front-view imaging devices within a set time period, identifying and recording images with potential damage. Next, these selected images are manually and meticulously labeled, covering the effective boundaries of the road surface, various road marking types, and, according to domestic road maintenance standards, the types and extent boundaries of existing defects are labeled. Finally, based on the labeled road extent and marking data, a high-precision road basic information identification model is trained using an instance segmentation model based on YOLOv11 or a later version.

[0047] In one implementation, before inputting road image data into a pre-trained disease detection model, the method further includes: training a YOLO neural network model based on disease types and disease range boundary data from a labeled dataset to obtain a trained disease detection model, which is used to identify disease types and disease range boundaries.

[0048] To facilitate understanding of the model training method in the embodiments of this application, the following is in conjunction with the appendix. Figure 2 Further description. For example... Figure 2 As shown, download the RDD Road Damage Open Dataset. Using the images and annotations provided in this dataset, fine-tune the YOLOv11 (or later) object detection model to train a model capable of initially identifying damaged roads. Select vehicles equipped with front-facing imaging devices and iterate through all the images acquired by these vehicles within a certain time frame. Using the model trained in the previous step, determine whether damage exists; if damage is found, record the image.

[0049] The images selected in the previous step are manually annotated. The annotations include the road surface boundaries, various road marking types, and if there is road damage, the types of damage and the boundaries of the road damage in the image are annotated according to the domestic transportation industry road maintenance standards.

[0050] Using labeled road extent data and road marking data, a road infrastructure information recognition model for identifying road boundaries and road markings is trained by fine-tuning the YOLOv11 instance segmentation model. Using labeled pavement distress data, a domestically standardized distress detection model is trained by fine-tuning the YOLOv11 (or a later version) instance segmentation model.

[0051] This application employs a collaborative processing flow of road infrastructure information identification, precise defect identification, spatial logic verification, and road network mapping positioning. Through a defect and marking spatial relationship filtering mechanism, it effectively suppresses false detections caused by marking interference and off-road misjudgments, thus improving the accuracy of defect detection. Simultaneously, by integrating vehicle trajectory with high-precision road network matching, it elevates defect positioning accuracy from the traditional meter level to the lane level, achieving precise binding of defects to road segments and lanes. It can directly interface with maintenance work order systems, significantly improving the automation level of road inspections and the accuracy of maintenance responses.

[0052] According to another aspect of the embodiments of this application, a road defect detection device for implementing the above-described road defect detection method is also provided. For example... Figure 5 As shown, the device includes: Image acquisition module 501 is used to traverse road image data captured by vehicles traveling on the road through vehicle-mounted imaging equipment within a preset time period; The road information recognition module 502 is used to input road image data into a pre-trained road basic information recognition model to identify road basic information, which includes road boundaries and road markings. The defect detection module 503 is used to input road image data into a pre-trained defect detection model in response to the road boundary containing the effective road surface area, and output the initial detection result of whether defects exist. The verification output module 504 is used to perform spatial logic verification based on the initial detection results and road basic information to obtain a valid detection result of whether there are defects, and output the defect detection result in combination with road network data.

[0053] It should be noted that the road defect detection device provided in the above embodiments is only illustrated by the division of the above functional modules when performing the road defect detection method. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the road defect detection device and the road defect detection method embodiments provided in the above embodiments belong to the same concept, and the implementation process is detailed in the method embodiments, which will not be repeated here.

[0054] According to another aspect of the present application, an electronic device corresponding to the road defect detection method provided in the foregoing embodiments is also provided to perform the road defect detection method described above.

[0055] Please refer to Figure 6 This illustrates a schematic diagram of an electronic device provided by some embodiments of this application. For example... Figure 6As shown, the electronic device includes: a processor 600, a memory 601, a bus 602, and a communication interface 603. The processor 600, the communication interface 603, and the memory 601 are connected via the bus 602. The memory 601 stores a computer program that can run on the processor 600. When the processor 600 runs the computer program, it executes the road defect detection method provided in any of the foregoing embodiments of this application.

[0056] The memory 601 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 603 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network.

[0057] Bus 602 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. The memory 601 is used to store programs. After receiving execution instructions, the processor 600 executes the program. The road defect detection method disclosed in any of the aforementioned embodiments of this application can be applied to the processor 600, or implemented by the processor 600.

[0058] The processor 600 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 600 or by instructions in software form. The processor 600 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 601. Processor 600 reads the information in memory 601 and, in conjunction with its hardware, completes the steps of the above method.

[0059] The electronic device provided in this application embodiment and the road defect detection method provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0060] According to another aspect of the present application, a computer-readable storage medium corresponding to the road defect detection method provided in the foregoing embodiments is also provided, wherein a computer program (i.e., a program product) is stored thereon, and the computer program, when run by a processor, executes the road defect detection method provided in any of the foregoing embodiments.

[0061] It should be noted that examples of computer-readable storage media may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.

[0062] The computer-readable storage medium provided in the above embodiments of this application and the road defect detection method provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0063] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0064] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for detecting road defects, characterized in that, include: Iterate through road image data captured by vehicles traveling on the road using in-vehicle imaging equipment within a preset time period; The road image data is input into a pre-trained road basic information recognition model to identify the road basic information, which includes road boundaries and road markings. In response to the fact that the road boundary includes the effective road surface area, the road image data is input into a pre-trained disease detection model, and an initial detection result of whether or not a disease exists is output. Based on the initial detection results and the road infrastructure information, spatial logic verification is performed to obtain valid detection results for the presence or absence of defects. The defect detection results are then output in conjunction with the road network data.

2. The method according to claim 1, characterized in that, Based on the initial detection results and the road infrastructure information, a spatial logic verification is performed to obtain a valid detection result indicating whether or not there are road defects, including: Obtain the candidate regions of diseases from the initial detection results; Determine whether the candidate area of ​​the defect is located within the effective road surface area defined by the road boundary; When the candidate area of ​​the defect is located within the effective road surface area, the intersection-combination ratio of the candidate area of ​​the defect and the road markings is calculated; When the intersection ratio of the candidate disease area and the road marking is less than a preset threshold, the candidate disease area is determined to be a valid disease.

3. The method according to claim 2, characterized in that, Also includes: When the candidate area of ​​a defect is located within the effective road surface area, and the candidate area of ​​a defect and the road markings do not intersect, the candidate area of ​​a defect is determined to be a valid defect.

4. The method according to claim 1, characterized in that, The output of defect detection results, based on road network data, includes: Based on the timestamp information in the road image data, the driving trajectory data of the photographed vehicle in the corresponding time period is obtained; The driving trajectory data is matched with the road network data to obtain information on road sections with effective defects; Associate the disease types corresponding to the effective diseases, generate and output disease detection results that include disease types and road segment information.

5. The method according to claim 4, characterized in that, Also includes: Based on the geometric characteristics and disease type of the effective diseases, calculate their severity index; Based on the road grade and current traffic flow characteristics in the road segment information of the effective defects, the risk level of the effective defects is obtained; The severity score of the effective disease is obtained by weighted summation based on the severity index and risk level, and the severity score is output.

6. The method according to claim 1, characterized in that, Before inputting the road image data into the pre-trained road basic information recognition model, the following steps are also included: The image data containing road damage is manually annotated to form an annotated dataset. The annotations include the effective boundary of the road, various road markings and their types. When the road has defects within its effective area, the labeling content also includes the defect type and the defect range boundary; A YOLO neural network model is trained based on the road boundary and road marking data in the labeled dataset to obtain a trained road basic information recognition model, which is used to identify the road boundary and road marking.

7. The method according to claim 6, characterized in that, Before inputting the road image data into the pre-trained disease detection model, the following steps are also included: A YOLO neural network model is trained based on the disease types and disease range boundary data of the labeled dataset to obtain a trained disease detection model. The disease detection model is used to identify disease types and disease extent boundaries.

8. A road defect detection device, characterized in that, include: The image acquisition module is used to traverse road image data captured by vehicles traveling on the road through in-vehicle imaging equipment within a preset time period; The road information recognition module is used to input the road image data into a pre-trained road basic information recognition model to identify the road basic information, which includes road boundaries and road markings. The defect detection module is used to respond to the fact that the road boundary includes the effective road surface area, input the road image data into the pre-trained defect detection model, and output the initial detection result of whether there is a defect; The verification output module is used to perform spatial logic verification based on the initial detection results and the road basic information to obtain a valid detection result of whether there are defects, and output the defect detection result in combination with the road network data.

9. An electronic device, characterized in that, It includes a processor and a memory storing program instructions, the processor being configured to perform the road defect detection method as described in any one of claims 1 to 7 when executing the program instructions.

10. A computer storage medium, characterized in that, It stores computer-readable instructions that are executed by a processor to implement the road defect detection method as described in any one of claims 1 to 7.

Images