Community road damage monitoring and maintenance method and device based on images and storage medium
By installing cameras and intelligent visual analysis equipment on community roads, and combining CNN and Transformer models for road damage identification, the problem of lack of scientific basis for road damage monitoring and maintenance in existing technology in community roads is solved, and efficient and accurate road maintenance and maintenance are achieved.
Patent Information
- Application Number
- CN202510292553.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-12
- Publication Date
- 2025-07-01
AI Technical Summary
In the prior art, the monitoring and maintenance of community roads lacks scientific basis, resulting in poor management results, and relying on manual inspections leads to large labor consumption, untimely maintenance and insignificant results.
The intelligent image-based monitoring system is adopted, and the road range and material information is obtained by installing cameras and intelligent visual analysis equipment, combining drone aerial photography or manual annotation, and feature extraction and damage identification of road images are used to achieve accurate identification and classification of road damage.
It improves the accuracy and efficiency of road damage monitoring, reduces manpower consumption, realizes efficient management and control of maintenance work, and improves the accuracy and timeliness of road maintenance.
Smart Images
Figure CN120235830A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image processing, and particularly relates to a method, device, and storage medium for monitoring and maintaining damaged community roads based on images. Background Art
[0002] Currently, the technology of smart communities is developing rapidly. The gradual maturity and formal commercialization of new-generation information technologies have also brought more possibilities to the construction of smart communities. With the gradual implementation of the renovation of old communities, the continuous integration of digital technologies and community grass-roots governance has gradually formed a smart community construction model of "Internet + community". Nowadays, the operation and maintenance by property companies and the supervision by owners' committees are still the main organizational forms for the community to carry out various tasks, and the efficiency requirements for the daily management of the community are also increasing day by day.
[0003] With the gradual aging of urban old communities, the problem of road damage is becoming increasingly serious. For road maintenance management departments, timely and accurately identifying road damage is of great significance for ensuring road smoothness and pedestrian safety. However, traditional road damage detection methods mainly rely on manual inspections. Most property communities lack the management and control work for the standardized development of road maintenance work in the community, lack scientific basis, and the management effect is poor.
[0004] To solve the above problems of the lack of quantitative analysis of road damage monitoring and maintenance work in the community, the inability to accurately maintain, and the large human consumption, untimely maintenance, and ineffective results caused by manual inspections, a precise road damage monitoring and maintenance system is realized through the present invention to guide property management in the community to achieve efficient management and control of road conditions.
[0005] Glossary of Terms:
[0006] CNN: Convolutional Neural Network (abbreviated as CNN) is a special artificial neural network structure that has been widely used in the fields of image recognition, speech recognition, natural language processing, etc. The characteristic of CNN is that it can automatically extract the features of input data, thereby realizing the efficient classification and recognition of input data.
[0007] Transformer: The Transformer model is a neural network model based on the self-attention mechanism proposed by Google for processing sequence data. Compared with traditional recurrent neural network models, the Transformer model has better parallel performance and shorter training time, so it has been widely used in the field of natural language processing. Summary of the Invention
[0008] A method, device, and storage medium for monitoring and maintaining damaged community roads based on images proposed by the present invention can at least solve one of the technical problems in the background art.
[0009] To achieve the above object, the present invention adopts the following technical solutions:
[0010] A method for monitoring and maintaining damaged community roads based on images includes the following steps:
[0011] Step 1: Install cameras and intelligent vision analysis devices;
[0012] Step 2: Obtain the edge space of the road ranges based on five types of roads, namely main roads, group roads, paths between houses, garden paths, and squares, through drone aerial photography or manual annotation; at the same time, annotate the materials of the roads in the community.
[0013] Step 3: Obtain the picture information of the specified road section through the camera, and annotate the corresponding pictures based on the data in Step 2.
[0014] Step 4: Process and analyze the data and pictures in Step 2 and Step 3 through intelligent algorithms to identify the degree of road damage.
[0015] Furthermore, the specific implementation process of the analysis algorithm in Step 4 is as follows:
[0016] Data acquisition module: Collect image data of community roads;
[0017] Image preprocessing module: Preprocess the collected road images, including denoising, enhancement, scaling, and cropping operations, to facilitate subsequent modules for feature extraction and damage recognition;
[0018] Feature extraction module: Use a convolutional neural network (CNN) to extract features from the preprocessed road images and obtain key road damage feature information from the images;
[0019] Damage recognition module: Use a Transformer model to process the extracted image features and parameters such as road type and road material to automatically identify different damage types in the road images;
[0020] Classification and evaluation module: According to the output results of the damage recognition module, use a classification algorithm to classify and evaluate the severity of road damage according to damage type, road type, and road material, providing a basis for subsequent maintenance decisions.
[0021] Furthermore, in Step 2, define the parameter annotation form for annotating road type, length, width, thickness, and road material.
[0022] Further, the feature extraction module uses a convolutional neural network (CNN) for feature extraction, including the following steps:
[0023] Use multiple convolutional layers for feature abstraction. Each convolutional layer extracts features within the local receptive field of the input image, capturing the local detailed information of road damage. Add an activation function after the convolutional layer to enhance the non-linear expression ability of the model. During the training process, adopt data augmentation and early stopping strategy techniques to improve the generalization performance of the model. Output the feature vector as the input to the damage recognition module.
[0024] Further, in the damage recognition module, use a Transformer model for damage type recognition and classification, including the following steps:
[0025] Take the feature vector output by the feature extraction module and the annotation parameters as the input of the Transformer model. Set up a multi-head self-attention mechanism to capture feature information at different levels. Use the cross-entropy loss function for parameter optimization during the training process;
[0026] Output the recognized road details and damage types in a one-dimensional vector structure as the input to the classification evaluation module.
[0027] Further, in the damage recognition module, referring to the "Highway Technical Condition Evaluation Standard", the output parameters include the following parts:
[0028] Damage type: This parameter represents the recognized road damage type, including cracks, potholes, abrasions, and subsidence;
[0029] Damage area: This parameter represents the area of the recognized damaged area.
[0030] On the other hand, the present invention also discloses a computer-readable storage medium storing a computer program, which when executed by a processor causes the processor to execute the steps of the above method.
[0031] On yet another aspect, the present invention also discloses a computer device including a memory and a processor, where the memory stores a computer program, and when the computer program is executed by the processor, it causes the processor to execute the steps of the above method.
[0032] As can be seen from the above technical solutions, a method and system for monitoring and maintaining damaged community roads based on image intelligence are proposed in the present invention. Since there are many parameters affecting the classification recognition rate of road damage, such as road materials, damage types, shooting angles, etc., it is difficult to ensure the recognition accuracy of mainstream image recognition algorithms. Therefore, before simply using image algorithms, the present invention proposes a parameter annotation method adapted to the transformer model, defines data such as road types, length, width, thickness, and road materials, and introduces the self-attention mechanism of the transformer model, which can consider the information of more parameters at the same time. At the same time, the present invention structures the road damage types, sizes, and evaluation annotations, so as to better use classification algorithms to classify the damage degree and determine the basis for the maintenance plan.
[0033] The purpose of the present invention is to provide a classification scheme for identifying damaged community roads, so as to improve the ability to accurately maintain roads, solve the problem that it is difficult to quantify the determination of road damage degree based on manual experience, reduce cost consumption, improve processing efficiency, and guide the property to efficiently manage the damaged roads in the community.
[0034] Specifically, the beneficial effects of the present invention are as follows:
[0035] 1. Standardized processing of road annotation parameters and damage parameters in the present invention.
[0036] 2. Parameter binding and dynamic update: Solve the problem of the disconnection between images and parameters in traditional methods and improve data consistency;
[0037] 3. Multimodal Transformer model: Significantly improve the damage classification accuracy by fusing image features and structured parameters (experimental data shows that the accuracy rate is increased by 12%);
[0038] 4. Rule-driven maintenance decision-making: Combine industry standards and custom rules to achieve accurate matching and efficient execution of maintenance plans. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Figure 1 is a flowchart of the road damage monitoring and maintenance system based on image intelligence of the present invention;
[0040] Figure 2 is a flowchart of the road damage identification and classification method. DETAILED DESCRIPTION OF THE INVENTION
[0041] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention.
[0042] Such asFigure 1 As shown in Figure 1 , the method for monitoring and maintaining damaged community roads based on images in this embodiment is implemented by a computer device through the following steps:
[0043] In the present invention, a method for monitoring and maintaining damaged community roads based on image intelligence is proposed. The specific method is as Figure 1 follows:
[0044] Step 1: Install cameras and intelligent vision analysis devices.
[0045] Step 2: Obtain the edge space of the road ranges of five types, namely main roads, group roads, paths between houses, garden roads, and squares, through aerial photography by drones or manual annotation; at the same time, annotate the materials of the roads in the community.
[0046] Step 3: Obtain picture information of specific road sections through cameras, and annotate the corresponding pictures based on the data in Step 3.
[0047] Step 4: Use intelligent algorithms to process and analyze the data and pictures in Steps 2 and 3 to identify the degree of road damage.
[0048] The following is a specific elaboration of each step.
[0049] In Step 2, as shown in Table 1, the roads can be divided into main roads (width greater than 9 meters), group roads (width 6 - 9 meters), paths between houses (width 3 - 6 meters), garden roads (width less than 3 meters), and squares (length and width each greater than 9 meters). Different road conditions are different, so the materials of the roads also need to be considered, such as asphalt, cement, and gravel; at the same time, due to the limitations of the picture shooting position and picture ratio, it is often difficult to determine the length, width, and thickness of the roads. Therefore, the length, width, and thickness of the roads should also be used as one of the parameters. Based on this idea, a parameter annotation form is defined to further annotate the road type, length, width, thickness, road material, etc.
[0050] Table 1 is the numbering rule for road annotation data
[0051]
[0052] In Step 3, through the camera devices corresponding to the roads, obtain high-definition captured pictures of each road section, then bind the annotation data of the corresponding roads obtained in Step 2, and send them to the visual intelligent analysis device for the next algorithm analysis.
[0053] Specifically, after obtaining high-definition captured pictures of specific road sections through cameras, the system will automatically call the structured data such as the road type, length, width, thickness, and material annotated in Step 2, and associate these parameters with the corresponding image files. The specific implementation method is as follows:
[0054] 1. Metadata Binding: Each captured image generates a unique identifier (such as UUID), which is associated with the road annotation parameters (type number, material code, etc.) in Step 2 through a database table to form an "image - parameter" mapping relationship.
[0055] 2. Spatial Coordinate Matching: For drone aerial images, combined with GPS coordinate information, spatially align the road areas in the images with the marked spatial boundary data to ensure the precise correspondence between the parameters and the image areas.
[0056] 3. Dynamic Update Mechanism: When the road material or type changes, the system supports updating the annotation parameters through the management interface and synchronizing them to the associated image dataset to ensure the accuracy of subsequent analysis.
[0057] In Step 4, using the high - definition images captured in Step 3, first identify the corresponding road area features, then add the annotation data obtained in Step 2 and the road image feature values to analyze the specific damage conditions of the road. Then, combined with the classification algorithm, classify the damage conditions as severe, moderate, mild, and normal.
[0058] The specific implementation process of the analysis algorithm in Step 4 is as follows:
[0059] 1. Data Acquisition Module: Collect image data of community roads.
[0060] 2. Image Pre - processing Module: Pre - process the collected road images, including operations such as denoising, enhancement, scaling, and cropping, to facilitate feature extraction and damage recognition in subsequent modules.
[0061] 3. Feature Extraction Module: Use a Convolutional Neural Network (CNN) to extract features from the pre - processed road images and obtain key road damage feature information from the images.
[0062] 4. Damage Recognition Module: Use a Transformer model to process the extracted image features and parameters such as road type and road material to automatically identify different damage types in the road images.
[0063] 5. Classification and Evaluation Module: According to the output results of the damage recognition module, use a classification algorithm to classify and evaluate the severity of road damage according to damage type, road type, road material, etc., providing a basis for subsequent maintenance decisions.
[0064] Step 5: Determine the maintenance plan based on the data in Step 2 and Step 4.
[0065] The specific implementation process of the road damage recognition and classification algorithm based on intelligent images is mainly as follows:
[0066] I. Road Damage Recognition Algorithm
[0067] 1. Image preprocessing module
[0068] In the image preprocessing module, the collected road image data needs to be preprocessed to facilitate subsequent feature extraction and damage recognition. The preprocessing process includes the following steps:
[0069] Noise removal: Perform noise removal processing on the image data, such as using median filtering, Gaussian filtering and other methods to eliminate the noise in the image;
[0070] Image enhancement: Enhance the contrast and dynamic range of the image through methods such as contrast stretching and histogram equalization to make the damage features more obvious;
[0071] Image cropping: Crop the image to reduce the interference of irrelevant regions;
[0072] Color space conversion: Convert the image from the RGB color space to the Lab color space according to needs to better extract road damage features;
[0073] Data augmentation: Augment the image data through methods such as rotation, flipping, translation, and scaling to expand the training dataset and improve the generalization ability of the model.
[0074] Through the above steps, the image preprocessing module processes the original road image data into a format suitable for feature extraction and damage recognition, providing high-quality input data for subsequent modules.
[0075] 2. Feature extraction module
[0076] In the feature extraction module, a convolutional neural network (CNN) is used for feature extraction, including the following steps:
[0077] Use multiple convolutional layers for feature abstraction. Each convolutional layer will extract features within the local receptive field of the input image, capturing the local detailed information of road damage; add activation functions after the convolutional layers to enhance the non-linear expression ability of the model; adopt data augmentation and early stopping strategy techniques during training to improve the generalization performance of the model; output feature vectors as the input for the damage recognition module.
[0078] The network structure uses ResNet-50 as the backbone network, optimizes the last fully connected layer, and outputs a 1024-dimensional feature vector. It should be noted that the output here uses feature map flattening as the output. The feature map can capture local patterns in the image, such as edges, textures, and other complex features, while flattening can flatten the feature map output into a one-dimensional vector for subsequent module processing.
[0079] 3. Damage recognition module
[0080] In the damage recognition module, a Transformer model is used for damage type recognition and classification, including the following steps:
[0081] The image feature vector (1024 - dimensional) output by the CNN is concatenated with the labeled structured parameters (such as road type, material, etc., encoded as 32 - dimensional vectors) to form a 1056 - dimensional input sequence.
[0082] A 4 - head attention mechanism is adopted to calculate the correlation weights between different parameters. For example, the model can automatically focus on the associated features between "asphalt material" and "crack damage" to improve the classification accuracy.
[0083] The final output layer uses a fully - connected layer + Softmax to generate the probability distribution of damage types (such as cracks, potholes, subsidence), and combines a classification threshold (such as probability > 0.8) to determine the existence of damage.
[0084] The key point is to output the recognized road details and damage types in a one - dimensional vector structure as the input to the classification evaluation module.
[0085] In the damage recognition module, referring to the "Highway Technical Condition Evaluation Standard", the output parameters should mainly include the following parts:
[0086] 1. Damage type: This parameter represents the recognized road damage types, such as cracks, potholes, abrasion, subsidence, etc. These types can be defined and extended according to specific requirements.
[0087] 2. Damage area: This parameter represents the area of the recognized damage area. It is usually composed of length, width, and height.
[0088] II. Road damage loss assessment and maintenance algorithm
[0089] Table 2 shows the parameter rules for damage situations
[0090]
[0091] Based on the output parameters of the damage recognition algorithm, further, the present invention can standardize the parameter rules of the output. As shown in Table 2, and combined with the labeled parameters in step two, the present invention can obtain structured parameters that include road type, road total length, width, thickness, road material, damage type, and damage area combined together. In the classification evaluation module, the present invention uses SVM to process this parameter to further evaluate the damage degree.
[0092] The road labeled parameters obtained in step two and the road damage parameters obtained by the recognition algorithm are extended to a multi - dimensional hyperspace, and the classification function is used to divide the feature points on the hyperplane. The parameters can be set as vectors. According to the "Highway Technical Condition Evaluation Standard", the present invention can obtain corresponding standards and set them as vectors. Thus, a classification evaluation function can be established. Among them, the vector is the trained parameter adjustment value, and this adjustment value can be adjusted according to the training requirements.
[0093] By training the classification function with different standards, the road damage degree can be classified into: severe, moderate, mild, and normal according to the needs.
[0094] In step five, based on different road types and damage severity, referring to the "Highway Maintenance Technical Specification", specific maintenance plans are selected.
[0095]
[0096]
[0097] Meanwhile, a dynamic decision-making mechanism for the maintenance plan can be established according to the specific situation.
[0098] 1. Multi-dimensional decision rule base:
[0099] Establish a rule base according to the "Highway Maintenance Technical Specification", for example:
[0100] For asphalt roads: If the crack width > 5mm and the area ratio > 10%, it is determined as "severe", and "local patching" is triggered;
[0101] For cement roads: If the pothole depth > 3cm, it is determined as "severe", and "demolition and re-pouring" is triggered.
[0102] The rules support dynamic expansion, and users can customize the thresholds and maintenance measures through the management interface.
[0103] 2. Parameter optimization of the SVM classifier:
[0104] In the classification evaluation module, the kernel function (RBF) parameters of the SVM are optimized by grid search (Grid Search) (penalty factor C = 10, kernel parameter γ = 0.1) to ensure the accuracy of the damage degree classification (severe / moderate / mild).
[0105] 3. Visualization and alarm linkage:
[0106] When the system generates a maintenance work order, it automatically associates the high-definition pictures and parameter annotation information of the damaged location and visualizes them through the map interface;
[0107] If "severe" damage is detected, a real-time alarm (text message / email) is triggered and pushed to the property maintenance system to shorten the response time.
[0108] The following is an example for illustration:
[0109] Example 1:
[0110] The community road damage monitoring and maintenance system based on image intelligence in this example can operate through the following steps:
[0111] 1. Refer to Figure 1 the system flow chart, and first install and debug the equipment.
[0112] 2. Based on step two, label and sort out the overall road conditions of the community, import the labeling information into the system, and technicians assist the community property to complete the road calibration and filing work. At the same time, adjust the installation position of the camera according to the specific situation to ensure that the pictures can capture as complete a road as possible.
[0113] 3. After completing the preparatory work, enable the relevant equipment and start collecting image information and environmental data to start algorithm analysis.
[0114] 4. After completing the damage situation assessment through step four, select different maintenance plans for different road damage situations and road types.
[0115] 5. Use the visualization module to visually display the results of road damage identification and classification for easy viewing and analysis by users.
[0116] Through the above steps, the old community road damage identification and classification system based on image intelligence of the present invention can efficiently and accurately identify and classify the damage of old community roads, providing strong support for road repair and maintenance.
[0117] Example 2:
[0118] On the basis of Example 1, the old community road damage identification and classification system based on image intelligence in this example can also add an alarm module for sending alarm information to relevant departments when serious damage is detected. The alarm module can set a threshold, and when the damage degree exceeds the threshold, the alarm is automatically triggered for timely repair.
[0119] Through the above various examples, the old community road damage identification and classification system based on image intelligence of the present invention can better meet the road damage detection tasks under different scenarios and requirements, and make greater contributions to improving the urban traffic environment and enhancing the quality of residents' lives.
[0120] On the other hand, the present invention also discloses a computer-readable storage medium storing a computer program, which when executed by a processor causes the processor to execute the steps of the above method.
[0121] In another aspect, the present invention also discloses a computer device, including a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor is caused to execute the steps of the above method.
[0122] In another embodiment provided by the present application, there is also provided a computer program product containing instructions. When it runs on a computer, the computer is caused to execute any one of the above-mentioned image-based community road damage monitoring and maintenance methods in the embodiments.
[0123] It can be understood that the system, device and storage medium provided by the embodiments of the present invention correspond to the method provided by the embodiments of the present invention. For the explanations, examples and beneficial effects of related content, reference can be made to the corresponding parts in the above method.
[0124] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wire (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that integrates one or more available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)).
[0125] It should be noted that in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the phrase "comprising a..." does not exclude the presence of additional identical elements in the process, method, article or device comprising said element.
[0126] Each embodiment in this specification is described in a related manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and reference can be made to the relevant parts of the method embodiments for the relevant content.
[0127] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. However, such modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A community road damage monitoring and maintenance method based on an image, characterized in that: The following steps are included: Step 1: Install cameras and intelligent visual analysis equipment; Step 2: Obtain the road range and space based on five types of roads, including main roads, group roads, residential paths, garden roads, and squares, through drone aerial photography or manual annotation; and also annotate the materials of the roads in the community; Step 3: Obtain image information of the specified road section through the camera, and annotate the corresponding image based on the data in step 2; Step 4: Use intelligent algorithms to process and analyze the data and images from steps 2 and 3 to identify the degree of road damage.
2. The image-based community road damage monitoring and maintenance method according to claim 1 is characterized by: The specific implementation process of the analysis algorithm in step 4 is as follows: Data collection module: collects image data of community roads; Image preprocessing module: preprocesses the collected road images, including denoising, enhancement, scaling, and cropping operations, so as to facilitate feature extraction and damage identification in subsequent modules; Feature extraction module: Convolutional neural network (CNN) is used to extract features from preprocessed road images to obtain key road damage feature information from the images; Damage recognition module: The Transformer model is used to process the extracted image features and parameters such as road type and road material to automatically identify different damage types in road images; Classification and assessment module: Based on the output results of the damage identification module, a classification algorithm is used to classify road damage and assess its severity according to damage type, road type, and road material, providing a basis for subsequent maintenance decisions.
3. The image-based community road damage monitoring and maintenance method according to claim 1, characterized in that: In step 2, define the parameter annotation format to mark the road type, length, width, thickness, and road material.
4. The image-based community road damage monitoring and maintenance method according to claim 2 is characterized in that: The feature extraction module uses convolutional neural network (CNN) for feature extraction, which includes the following steps: Multiple convolutional layers are used for feature abstraction. Each convolutional layer extracts features within the local receptive field of the input image to capture local details of road damage. An activation function is added after the convolutional layer to enhance the nonlinear expression ability of the model. Data enhancement and early stopping strategy techniques are used during training to improve the generalization performance of the model. The output feature vector is used as the input of the damage recognition module.
5. The image-based community road damage monitoring and maintenance method according to claim 2 is characterized by: In the damage identification module, the Transformer model is used to identify and classify damage types, including the following steps: The feature vector and annotation parameters output by the feature extraction module are used as the input of the Transformer model; a multi-head self-attention mechanism is set up to capture feature information at different levels; The cross entropy loss function is used for parameter optimization during training; The identified road details and damage types are output as a one-dimensional vector structure as the input of the classification evaluation module.
6. The image-based community road damage monitoring and maintenance method according to claim 5 is characterized by: In the damage identification module, referring to the Highway Technical Condition Assessment Standard, the output parameters include the following parts: Damage type: This parameter indicates the type of road damage identified, including cracks, potholes, wear and tear, and subsidence; Damaged area: This parameter indicates the area of the identified damaged area.
7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 6.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the computer program is executed by the processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 6.