Road disease self-updating patrol method, vehicle-mounted terminal, server and vehicle

Abstract

The invention discloses a road disease self-updating patrol method, a vehicle-mounted terminal, a server and a vehicle, and relates to the field of road intelligent patrol, and the method comprises the steps: collecting disease feature data and safety operation feature data through the vehicle-mounted terminal, generating disease information and construction safety early warning through a locally deployed disease recognition model, and generating a data packet based on the difference data and the priority information and sending the data packet to a server. After the server receives the data packet, confidence verification is carried out on the disease information, and high-confidence data is screened for automatic iterative updating of the disease recognition model. By dynamically adjusting identification parameters and optimizing data transmission, the data redundancy is effectively reduced in a complex environment, the identification precision is improved, and meanwhile, the safety guarantee is provided for constructors.

Description

Technical Field

[0001] The present application relates to the field of intelligent highway inspection, and in particular to a self-updating inspection method for highway defects, a vehicle-mounted terminal, a server and a vehicle. Background Art

[0002] Traditional highway disease inspections mainly rely on manual operations, with low inspection efficiency and difficulty in real-time monitoring and identification of large-scale highway diseases. In addition, manual identification methods are easily affected by environmental factors, personnel fatigue, etc., resulting in reduced identification accuracy, missed reports, and false reports. With the advancement of technology, more and more intelligent inspection systems have been proposed and implemented, including vehicle-mounted terminals, image recognition technology, Beidou positioning system, etc., which can improve inspection efficiency and identification accuracy to a certain extent. However, these existing systems still have some problems and limitations.

[0003] For example, the Chinese patent with publication number CN118426477A discloses a method and system for unmanned highway disease inspection based on an autonomous driving vehicle. The solution realizes automated highway disease inspection through autonomous driving vehicles, deep learning models, monitoring centers and disease database modules. Autonomous driving vehicles collect highway disease information through built-in GPS, information capture cameras and other equipment, and upload data to the monitoring center in real time through wireless communication modules.

[0004] The above scheme still has problems such as insufficient dynamic adaptability of the recognition model, efficiency problems in data processing and transmission, and lack of risk assessment associated with subsequent safety operations. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present application discloses a self-updating inspection method for highway hazards, a vehicle-mounted terminal, a server and a vehicle.

[0006] In a first aspect, the present application provides a self-updating inspection method for highway diseases, which is executed by a vehicle-mounted terminal and includes: For the target highway, disease characteristic data and safe operation characteristic data are obtained, wherein the safe operation characteristic data includes: environmental information and context information; wherein the safe operation characteristic data is used to perform a local assessment of safety risks at the construction site on the vehicle terminal; Using the disease recognition model deployed on the vehicle-mounted terminal, the disease feature data is identified to generate first disease information of the target highway, and safety risks are evaluated based on the first disease information to generate safety risk warning information; Based on a preset priority allocation rule, marking priority information for the disease characteristic data; Using the priority information, a first data packet is generated based on the disease feature data, the safe operation feature data, and the first disease information, and the first data packet is transmitted to a server; wherein the first data packet is used for iterative updating of the disease identification model.

[0007] As an optional implementation manner, the generating of security risk warning information includes: Based on the geographic information of the disease feature data and the safe operation feature data, matching the disease feature data with the safe operation feature data whose geographic information differs by less than a preset value; In response to the presence of the disease feature data matching the safety operation feature data, determining the disease type, severity, and location information based on the first disease information corresponding to the disease feature data, and generating safety risk warning information based on the disease type, severity, location information, and the safety operation feature data using a preset safety risk assessment model; In response to the absence of the disease feature data matching the safety operation feature data, a preset safety risk assessment model is used to generate safety risk warning information based on the safety operation feature data.

[0008] As an optional implementation manner, before marking the priority information for the disease feature data based on the preset priority allocation rule, the method further includes: Based on the type, severity, and location information of the first defect information, a comparison is performed in a first defect database locally deployed on the vehicle terminal to screen out differential data; wherein the data in the first defect database only includes the first defect information collected for the target highway in the most recent target inspection round; The difference data refers to the significant changes or newly added disease feature data between the disease feature data currently collected for the target highway inspection and the disease feature data recorded in the first disease database; The significant changes include: a change in the disease type, a disease severity exceeding a preset change threshold, and a disease location information offset exceeding a preset distance; The marking of priority information for the disease characteristic data includes: For the difference data, priority information is marked based on the preset priority allocation rule.

[0009] As an optional implementation manner, the obtaining of disease characteristic data and safe operation characteristic data includes: Using the camera deployed on the vehicle terminal, the disease characteristic data and the safe operation characteristic data are acquired at a collection frequency that matches the Beidou positioning data update frequency; The image data collected by the camera and the Beidou positioning data share a consistent timestamp; The positioning data of the target highway is obtained through the Beidou positioning module at an update frequency matching the camera acquisition frequency, as the position information corresponding to the disease feature data and the safe operation feature data.

[0010] As an optional implementation manner, generating a first data packet and transmitting the first data packet to a server includes: Generate a first data packet based on the priority information corresponding to each of the difference data and the safety operation characteristic data corresponding to each of the difference data; wherein the difference data and the safety operation characteristic data are matched based on the geographic information; Dynamically evaluate the current wireless network and generate network status evaluation information; based on the network status evaluation information, determine the size of the data packet that can be currently sent; Based on the data packet size and the priority sorting result, select the difference data with the highest priority to generate a first data packet; The first data packet is transmitted to the server via a wireless network.

[0011] And a self-updating inspection method for highway diseases, executed by a server, comprising: receiving a first data packet transmitted by the vehicle-mounted terminal, and based on the first data packet, performing confidence verification on the first disease information to generate verified feature data; Iteratively updating the disease recognition model based on the verified feature data; The iteratively updated disease identification model is sent to the vehicle-mounted terminal.

[0012] As an optional implementation manner, the step of performing confidence verification on the first disease information and generating verified feature data includes: Calculating a confidence value of a disease identification result based on the first disease information and safe operation feature data in the first data packet; the safe operation feature data includes: environmental information and context information; Comparing the confidence value with a preset confidence threshold; wherein the preset confidence threshold is dynamically adjusted based on the safe operation characteristic data; The dynamic adjustment includes: in response to the presence of bad weather information in the environmental information, lowering the confidence threshold; wherein the bad weather information includes: rainy days, snowy days, and haze; in response to the presence of traffic flow greater than the flow threshold in the context information, raising the confidence threshold; in response to the presence of traffic artery information or key construction area information in the context information, raising the confidence threshold; In response to the confidence value being less than the preset confidence threshold, marking the corresponding first disease information as having low credibility; In response to the confidence value being greater than or equal to the preset confidence threshold, marking the corresponding first disease information as high confidence, and using the first disease information marked as high confidence as verified feature data for iteratively updating the disease recognition model; The iterative updating of the disease identification model includes: Using the verified feature data and a preset machine learning algorithm, the disease recognition model is trained to update the model parameters; According to the preset model update strategy, when the verified feature data accumulates to a predetermined amount or a preset time interval has passed, the model update is executed.

[0013] In a second aspect, the present application also provides a vehicle-mounted terminal, including: A collection module, used for acquiring disease characteristic data and safe operation characteristic data for a target highway, wherein the safe operation characteristic data includes: environmental information and context information; wherein the safe operation characteristic data is used for local assessment of safety risks at the construction site on a vehicle-mounted terminal; A processing module, configured to identify the defect feature data using a defect recognition model deployed on the vehicle-mounted terminal, generate first defect information of the target highway, evaluate safety risks based on the first defect information, and generate safety risk warning information; An allocation module, used for marking priority information for the disease characteristic data based on a preset priority allocation rule; The first transmission module is used to use the priority information to generate a first data packet based on the disease characteristic data, the safe operation characteristic data, and the first disease information, and transmit the first data packet to the server; wherein the first data packet is used for iterative update of the disease recognition model.

[0014] In a third aspect, the present application further provides a server, including: A verification module, configured to receive a first data packet transmitted by the vehicle-mounted terminal, and based on the first data packet, perform confidence verification on the first disease information to generate verified feature data; A construction module is used to iteratively update the disease recognition model based on the verified feature data; The second transmission module is used to send the iteratively updated disease recognition model to the vehicle-mounted terminal.

[0015] In a fourth aspect, the present application further provides a vehicle, comprising a vehicle body, and a vehicle-mounted terminal as described in the second aspect and disposed inside the vehicle body. The vehicle-mounted terminal executes the steps of the first aspect or any possible implementation of the first aspect when it is operated.

[0016] Compared with the prior art, the beneficial effects of this application are: this application realizes an automated model update process based on high-confidence data through the collaborative work between the vehicle terminal and the server. After receiving the difference data and safety operation feature data, the server performs automated iterative updates to ensure that the disease identification model can always maintain an efficient and accurate state. The automatic update of the model enables the system to quickly adapt to new road conditions, significantly improving the accuracy of maintenance event identification and inspection efficiency.

[0017] In addition, a differential data screening mechanism based on geographic information matching is introduced, which can effectively eliminate redundant and duplicate data and retain only disease feature data with significant changes or new additions. This mechanism reduces the amount of data while ensuring that the data uploaded to the server has high value, which helps to accurately iterate and update the disease identification model. In addition, through the confidence verification and dynamic adjustment mechanism, the present invention can accurately mark and screen the identification results under different environmental conditions, further improving the reliability of disease identification.

[0018] In addition, safe operation characteristic data is also used to provide safety warnings to construction workers, automatically reminding the command center and on-site construction workers of safety risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flow chart of a self-updating inspection method for highway hazards performed by a vehicle-mounted terminal provided in an embodiment of the present application; Figure 2 A flow chart of a self-updating inspection method for highway hazards performed by a server according to an embodiment of the present application; Figure 3 A schematic diagram of a vehicle-mounted terminal for executing a self-updating inspection method for highway hazards provided by an embodiment of the present application; Figure 4 A schematic diagram of a server for executing a self-updating inspection method for highway hazards provided in an embodiment of the present application. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0021] The study found that the existing highway disease inspection system usually relies on a single disease recognition mode, which is difficult to dynamically adjust to different environmental conditions and road conditions. Most disease recognition models lack automated update and optimization mechanisms after initial deployment, making it difficult to adapt to dynamic changes in road conditions, resulting in a gradual decrease in recognition accuracy. With the use of patrol vehicle equipment, the amount of data collected in real time is huge. Without effective priority sorting and transmission strategies, it may lead to data redundancy and increased network load, making it difficult to ensure the timely processing of key data.

[0022] Therefore, the present application discloses a self-updating inspection method for highway diseases, which includes steps S101 to S104 executed by a vehicle-mounted terminal, and steps S105 to S107 executed by a server.

[0023] See also Figure 1 , Figure 1 This is a flow chart of a self-updating inspection method for highway hazards performed by a vehicle-mounted terminal according to an embodiment of the present application, wherein: S101: Acquire disease characteristic data and safe operation characteristic data for a target highway, wherein the safe operation characteristic data includes: environmental information and context information; wherein the safe operation characteristic data is used to perform a local assessment of safety risks at a construction site on a vehicle-mounted terminal; S102: using the damage recognition model deployed on the vehicle terminal, identifying the damage feature data, generating first damage information of the target highway, and assessing the safety risk based on the first damage information to generate safety risk warning information; S103: marking priority information for the disease feature data based on a preset priority allocation rule; S104: Generate a first data packet based on the disease feature data, the safe operation feature data, and the first disease information using the priority information, and transmit the first data packet to a server; wherein the first data packet is used for iterative updating of the disease recognition model.

[0024] Also, see Figure 2 , Figure 2 This is a flow chart of a self-updating inspection method for highway hazards performed by a server according to an embodiment of the present application, wherein: S105: receiving a first data packet transmitted by the vehicle-mounted terminal, and based on the first data packet, performing confidence verification on the first disease information to generate verified feature data; S106: Iteratively updating the disease recognition model based on the verified feature data; S107: Sending the iteratively updated disease identification model to the vehicle-mounted terminal.

[0025] Regarding S101~S104 above: In the process of highway disease inspection, the accuracy and real-time performance of data are very critical, especially in dynamic inspections in complex environments. In order to achieve the synchronous collection of disease feature data and safety operation feature data, this application uses the camera and Beidou positioning module deployed on the vehicle terminal to obtain image data and positioning information at a consistent collection frequency and ensure the temporal and spatial consistency of the data. In this way, each disease feature can be accurately located during the inspection process, as well as the environmental information and contextual information of the construction operation.

[0026] The vehicle terminal is equipped with a camera with high-speed capture capability and a Beidou positioning module. The camera is used to capture image data of the road surface, which contains information about road surface defects, such as cracks, potholes and other features. In addition, the camera is also used to capture safety operation feature data at the construction site, including environmental information (such as weather and lighting conditions) and contextual information (such as traffic flow, construction signs, helmet wearing, etc.). In order to ensure the temporal and spatial consistency of the data, the Beidou positioning module and the camera use the same acquisition frequency to update the positioning data.

[0027] The camera configured on the vehicle terminal can capture image data of the target highway at a high acquisition frequency. For example, when the vehicle-mounted inspection vehicle is traveling at a certain speed, the camera will capture multiple frames of images per second to ensure that no key disease information is missed during the inspection. The acquisition frequency will be adjusted according to the speed limit and inspection needs of the highway. For example, a higher acquisition frequency can be used on high-speed sections, while it can be appropriately reduced on low-speed sections.

[0028] The Beidou positioning module acquires positioning data at an update frequency that matches the camera acquisition frequency. For example, if the camera acquires 10 frames of image data per second, the Beidou positioning module will also update the positioning information 10 times per second. Each acquired image data will be attached with the corresponding Beidou positioning data to ensure the consistency of the acquired data in time and space.

[0029] In order to ensure the accurate match between the image data collected by the camera and the Beidou positioning data, the vehicle terminal adopts a timestamp synchronization mechanism. When each frame of image data is collected, a high-precision timestamp will be attached. The timestamp is provided by the Beidou positioning module to ensure data consistency.

[0030] For example, each time the camera captures image data, the system will synchronously obtain the current timestamp and attach it to the metadata of the image data. The timestamp is used to identify the acquisition time of the image data.

[0031] At the same time, the Beidou positioning module will generate the same timestamp every time the positioning information is updated. This ensures that each frame of image data and its corresponding geographic location information are accurately matched in time. Through the timestamp synchronization mechanism, the geographic location of each disease feature can be accurately determined in subsequent data processing and analysis, improving the accuracy of data analysis.

[0032] During the inspection process, the Beidou positioning module is not only used to obtain the location information of the disease characteristics, but also to determine the geographical location of the safety operation characteristic data. For example, if the camera captures the characteristics of the construction area (such as safety signs or traffic cones), the Beidou positioning module will simultaneously provide the latitude and longitude data of these characteristics. This can accurately locate the specific location of the construction site and provide basic data for subsequent safety risk assessments.

[0033] For example, every time the camera captures a frame of image data, the system will synchronously call the Beidou positioning module to obtain the current geographic location information. This information includes longitude, latitude and altitude data to ensure the accuracy of the location of the disease and safety operation feature data.

[0034] The image data collected by the camera will be bound to the corresponding Beidou positioning data to form a complete data record. For example, a complete disease feature data record includes: image data, timestamp, disease type information, and corresponding geographic location information. During the data collection process, the vehicle terminal will continuously update the image data and geographic information to ensure that the entire target highway section is covered without omission during the inspection.

[0035] Thus, through the above data collection and synchronization mechanism, the present application ensures the consistency of disease characteristic data and safe operation characteristic data in time and space. This consistency is crucial for subsequent data processing and analysis, which can improve the accuracy of disease identification and ensure the accuracy of safety assessment of construction personnel.

[0036] In the specific implementation, a local damage recognition model is deployed on the vehicle terminal, such as a convolutional neural network, which is responsible for identifying and analyzing the collected damage feature data to generate the recognition result of the highway damage, that is, the first damage information, such as: damage type, upper left corner coordinates of the damage, lower right corner coordinates, damage length, damage width, and geographical location of the damage, etc. The damage types may include: rutting, potholes, transverse cracks, longitudinal cracks, shedding, and repaired cracks, etc.

[0037] The safety operation feature data collected by the vehicle terminal is mainly used to evaluate the safety risks of the construction site locally and generate safety risk warning information to ensure the safety of the workers. In addition, this data will also be uploaded to the server for iterative updating of the disease identification model.

[0038] It should be noted that the server-side function is limited to automatically iterating and updating the damage recognition model based on the first data packet sent by the vehicle terminal, and sending the updated model to the vehicle terminal, without participating in the specific highway damage recognition process. In this way, the damage recognition effect based on the local model will not be affected by the network transmission quality.

[0039] In addition, the quality of network signal transmission in different sections of the highway may not be consistent, which poses challenges to the real-time uploading of data and real-time updating of models. In this application, the priority of each disease feature data will be marked according to factors such as the severity of the disease, location, and traffic flow. For example, high-risk diseases (such as severe cracks in tunnels) will be given a higher priority. After the marking is completed, the vehicle terminal generates a first data packet based on the priority information and uploads it to the server through the wireless module in the vehicle terminal.

[0040] Regarding S105~S107 above: The server receives a first data packet transmitted by the vehicle terminal. The data packet includes disease feature data, first disease information, and safe operation feature data. The confidence calculation module built into the server calculates the confidence value of the disease recognition result based on the information in the data packet. For example, if it is detected that the disease data uploaded by the vehicle terminal has a low recognition result credibility in a complex environment (such as bad weather or high-traffic road sections), the confidence calculation module will lower the confidence score of the data.

[0041] When the confidence verification is completed, the server will filter out high-confidence data for training the disease recognition model. The server uses a machine learning algorithm based on deep learning (such as a convolutional neural network or an integrated learning model) to adjust the model parameters using high-confidence data. When the verified data reaches a certain amount or a certain time interval has passed, the system will trigger a model update. The updated model is transmitted back to the vehicle terminal via the wireless network to ensure that the vehicle terminal always uses the latest disease recognition model.

[0042] The updated model file is transmitted to the vehicle terminal through a wireless channel. After receiving the new model file, the vehicle terminal will automatically replace the old model to ensure the accuracy and real-time performance of disease identification.

[0043] In addition, existing highway hazard identification systems usually only focus on the identification and classification of hazards, but lack effective risk assessment for the safety of construction workers. Therefore, in this application, the collected hazard feature data and safety operation feature data are used to generate safety risk warning information for the construction site based on geographic information matching and a preset safety risk assessment model. This method can effectively reduce the safety hazards of construction workers in complex environments.

[0044] As an optional implementation, the generation of safety risk warning information includes: based on the geographic information of the disease characteristic data and the safety operation characteristic data, matching the disease characteristic data whose geographic information differs by less than a preset value with the safety operation characteristic data; in response to the existence of matching disease characteristic data among the safety operation characteristic data, determining the disease type, severity, and location information based on the first disease information corresponding to the disease characteristic data, and generating safety risk warning information based on the disease type, severity, location information, and the safety operation characteristic data using a preset safety risk assessment model; in response to the absence of matching disease characteristic data among the safety operation characteristic data, generating safety risk warning information based on the safety operation characteristic data using a preset safety risk assessment model.

[0045] In the specific implementation, the disease characteristic data and safety operation characteristic data collected by the vehicle terminal include geographic location information, which can be obtained through Beidou positioning. In order to generate accurate safety risk warning information, it is necessary to match the geographic information of the disease characteristic data and the safety operation characteristic data.

[0046] For example, the geographical distance between each piece of disease feature data and safe operation feature data can be calculated. When the geographical distance is less than a preset matching threshold (such as 10 meters), the two types of data are considered to be matched. The preset matching threshold can be adjusted according to the complexity of the construction site and the actual situation of the highway. By matching geographic information, the disease situation near the location of the construction personnel can be accurately determined to provide basic data for risk assessment. When a matching relationship is detected between the safe operation feature data and the disease feature data, an in-depth analysis is performed based on the first disease information corresponding to the disease feature data.

[0047] In the specific implementation, after the vehicle terminal completes the highway disease identification, it generates the first disease information. This information includes the specific type of disease (such as cracks, potholes, falling off, etc.), severity (such as crack width and depth) and precise location information (longitude and latitude obtained by Beidou positioning vehicle terminal). By analyzing these characteristic parameters, a preliminary assessment of the potential dangers in the current construction area is carried out.

[0048] The safety risk assessment model can combine the type, severity, location information of the disease and the current environment and working conditions (i.e., safe operation characteristic data) to assess the safety risks of construction workers in the current construction environment. The safe operation characteristic data includes the current environmental information (such as weather conditions, visibility, light intensity) and contextual information (such as the type of construction activities, surrounding traffic flow, etc.). This information helps to comprehensively assess the risk factors of the construction site and ensure the safety of construction workers.

[0049] The vehicle terminal inputs the disease characteristic data and safe operation characteristic data into the safety risk assessment model. The model will comprehensively analyze the specific situation of the disease and the current construction environment to generate safety risk warning information.

[0050] For example, in heavy rain weather conditions, even if the severity of the damage itself is not high, the risk in the construction area will increase significantly due to low visibility and slippery roads.

[0051] In specific implementation, the safety risk assessment model can comprehensively evaluate the safety risks of the construction site based on multiple factors to ensure the safety of construction workers under different environmental conditions. The model can use fuzzy logic or machine learning algorithms (such as decision trees or random forests). The input variables of the model include disease type, disease severity, location information, environmental information, and context information.

[0052] For example, the type of damage may be cracks, potholes, or falling off. Each type of damage has different impacts on construction safety. For example, potholes may increase the instability of construction equipment, while cracks may increase the risk of slipping. The severity of the damage is assessed by measuring parameters such as the width, depth, and area of ​​the damage. More serious damage usually corresponds to a higher construction risk. In addition, location information can include the precise latitude and longitude of the construction site and its surrounding terrain features, whether it is close to a busy traffic area, etc. For example, when construction is carried out at a bend on a highway, safety hazards may be more prominent.

[0053] Environmental information can include weather conditions (such as sunny or rainy), visibility (day or night), wind speed and humidity, etc., which directly affect the difficulty and risk level of construction. Contextual information covers the type of construction activity (such as repair or maintenance), surrounding traffic flow and current operation status (such as whether certain road sections are closed), all of which determine the overall level of construction risk.

[0054] In the core algorithm part of the model, a weighted comprehensive scoring method can be used. By assigning appropriate weights to each input variable, the overall safety risk score can be calculated. For example, the severity of the disease can be given a higher weight because it has a greater direct impact on construction safety, while environmental information and location information may have a secondary weight. Finally, the weighted sum of each factor is calculated to determine the risk level. The risk level can be divided into different levels, such as low risk, medium risk, and high risk.

[0055] The construction of the model first requires a data collection phase. Through the camera, sensor and positioning module of the vehicle terminal, the disease characteristic data and safe operation characteristic data are collected, and these data are used as the input variables of the model. Subsequently, based on the historical construction data, the model is trained using machine learning algorithms (such as decision trees, random forests or neural networks) to determine the weights of each input variable and the calculation method of the risk score.

[0056] For example, during a specific highway inspection, the vehicle terminal detected a large crack at the entrance of a highway tunnel. The severity of the crack was high, the current weather condition was heavy rain, and the operation type was emergency repair. Based on these factors, the model will comprehensively analyze these input variables. Since the crack is located on a major traffic route and has a high severity, coupled with the background of bad weather and emergency operations, the vehicle terminal gave a risk score of 85 points and was judged as "high risk". The vehicle terminal then generates a red warning message. Based on this red warning message, construction personnel can take additional safety protection measures, such as adding warning signs and arranging special personnel to guide traffic, or suspending construction.

[0057] Among them, there are many methods to transmit the information to the construction workers on site, such as transmitting voice warning information through the smart safety helmets worn by the construction workers on site. The smart safety helmet has a built-in network module and sound unit that can be used to receive safety risk warning information issued by the vehicle-mounted terminal, and generate corresponding voice commands to be played through the smart safety helmet.

[0058] Exemplarily, there may be many forms of voice instructions, but considering the cost limit of the smart helmet, simplified voice instructions may be issued, such as preset risk 1, risk 2, risk 3, etc.

[0059] In addition, safety risk warning information can also be received through smart bracelets worn by on-site construction workers, and different risks can be conveyed by controlling the extinction of different indicator lights.

[0060] Exemplarily, the smart bracelet is provided with three indicator lights, namely the first indicator light, the second indicator light, and the third indicator light. When all three indicator lights are on, it can indicate no risk, and when each of the indicator lights flashes separately, it can indicate the corresponding type of risk. In addition, different risks can also be indicated by the flashing of multiple indicator lights or the combination of lit colors, which will not be elaborated here.

[0061] In addition, if no matching disease feature data is detected in the collected data, that is, there is no disease data near the geographical location information of the safe operation feature data, the vehicle terminal will still perform risk assessment. Based on the environmental information and context information in the safe operation feature data, the safety risk warning information is generated through the preset safety risk assessment model.

[0062] In this case, the main consideration is the impact of environmental factors on the construction site. For example, when the weather is bad (such as heavy fog, heavy rain) or the surrounding traffic volume is abnormally large, even if there is no clear disease data, the vehicle terminal will issue appropriate safety warning prompts. In this way, it can ensure that construction workers can be warned in various situations where there may be risks.

[0063] In this way, by combining the disease information and the safety operation characteristic data, the present invention can evaluate the safety risks of construction personnel in highway disease repair and maintenance operations in real time. The generated safety risk warning information can provide clear safety tips to construction personnel in a timely manner, help them avoid potential safety hazards, and improve the safety of construction operations. Through this safety risk assessment mechanism, the present invention not only improves the accuracy of disease inspections, but also significantly improves the level of safety protection during construction operations.

[0064] In the above, this application has collected a large amount of data to ensure the consistency of disease feature data and safe operation feature data in time and space, and these data will be used for subsequent model iteration updates. However, due to the large amount of data, if there is no reasonable screening and priority sorting mechanism, it may lead to data redundancy and accumulation of invalid data, increasing the network load. Therefore, in this application, by comparing the disease feature data collected by the current inspection with the first disease database local to the vehicle terminal, it is possible to effectively screen out differential data with significant changes, thereby reducing the amount of data and ensuring that the data uploaded to the server is targeted for the iterative update of the disease recognition model.

[0065] As an optional implementation, before marking the priority information for the disease feature data based on the preset priority allocation rule, it also includes: based on the disease type, severity, and location information in the first disease information, comparing in a first disease database locally deployed on the vehicle terminal to screen out differential data; wherein the data in the first disease database only includes the first disease information collected for the target highway in the most recent target patrol round; the differential data is significant changes or new disease feature data between the disease feature data currently collected for the target highway patrol and the disease feature data recorded in the first disease database; the significant changes include: changes in disease type, disease severity exceeding a preset degree change threshold, and disease location information offset exceeding a preset distance; marking the priority information for the disease feature data includes: marking the priority information for the differential data based on the preset priority allocation rule.

[0066] In a specific implementation, the vehicle terminal will compare the disease feature data collected during the current inspection with the historical data in the first disease database. The data in the first disease database only includes the first disease information collected in the most recent inspection round to ensure the freshness and validity of the data.

[0067] For example, before each inspection mission begins, the vehicle terminal will load data from the first disease database, which records the disease characteristic data of the target highway in the most recent inspection round, including the type, severity, geographic location and other information of the disease.

[0068] During the current inspection process, the vehicle-mounted terminal will collect new disease characteristic data in real time and compare it with the data in the database one by one.

[0069] The comparison criteria include: whether the type of damage has changed, whether the severity of the damage exceeds the preset change threshold, and whether the location of the damage has shifted significantly. The preset threshold can be adjusted according to the characteristics of different highway environments and construction requirements.

[0070] Among them, differential data refers to data that has significant changes or new additions between the disease characteristic data collected during the current inspection and the data recorded in the first disease database. Specifically, the screening criteria for differential data can include the following aspects: Change of disease type: If the disease type collected during the current inspection is inconsistent with the records in the database, it is considered as differential data. For example, if the disease type previously recorded was "longitudinal cracks" and is currently identified as "ponds", the disease type is considered to have changed.

[0071] Severity change: If the severity of the currently identified disease exceeds the preset change threshold (such as an increase or decrease of more than 20%) compared to the historical data, it is considered to have changed significantly. For example, if the crack width was previously recorded as 10 cm and is currently identified as 12 cm (an increase of 20%), the severity is considered to have changed.

[0072] Location deviation: If the current location of the disease exceeds the preset distance threshold (such as location deviation of more than 10 meters) compared with the historical record, it is considered a significant change. This is particularly important in certain sections that are susceptible to environmental changes, such as bridges and tunnels.

[0073] After screening out the differential data, the vehicle terminal will prioritize these data according to the preset priority allocation rules, which is different from prioritizing all data.

[0074] As an optional implementation, generating a first data packet and transmitting the first data packet to a server includes: generating a first data packet based on priority information corresponding to each of the difference data and safety operation characteristic data corresponding to each of the difference data; wherein the difference data and the safety operation characteristic data are matched based on the geographic information; dynamically evaluating the current wireless network to generate network status evaluation information; determining the size of the data packet that can currently be sent based on the network status evaluation information; selecting the difference data with the highest priority based on the data packet size and the priority sorting result to generate a first data packet; and transmitting the first data packet to a server via a wireless network.

[0075] In the specific implementation, in the vehicle terminal, based on the disease feature data and safe operation feature data, the difference data will be marked with priority information. The priority information is determined by factors such as the severity of the disease, the type of disease, and the complexity of the construction environment. In order to ensure the accuracy and effectiveness of the data, the difference data and safe operation feature data will be matched based on geographic information.

[0076] For example, after completing the differential screening of the disease feature data, the vehicle terminal will mark the priority of each differential data according to the severity and risk level of the disease. The differential data will be matched with the corresponding safe operation feature data by geographic information to ensure that the disease information corresponds to the construction site information.

[0077] Before generating a data packet, the vehicle terminal will dynamically evaluate the current wireless network status to ensure that even when the network status is unstable, critical data can be transmitted first to avoid data upload failure or delay due to network fluctuations.

[0078] For example, the vehicle terminal has a built-in network status assessment module, which regularly detects the current wireless network signal strength, transmission rate, packet loss rate and other parameters. Network status assessment information is generated based on these parameters to determine the transmission capacity of the current network. Based on the network status assessment information, the size of the currently sendable data packet is dynamically adjusted. For example, when the network condition is good, a larger data packet can be generated to increase the data transmission volume; when the network condition is poor, a smaller data packet will be generated to ensure that the data can be uploaded stably.

[0079] In a specific implementation, after differential data screening and priority marking are completed, high-priority data can be selected to generate the first data packet. The size of the data packet will be dynamically adjusted according to the network status evaluation result so that it can be stably transmitted under different network conditions.

[0080] According to the priority sorting results, the data with the highest priority is selected and the total size of the data is calculated. According to the current sendable data packet size, the system will continue to add data until the preset size limit is reached. For example, if the current network status allows the size of the data packet to be sent to be 500KB, the system will select the highest priority data for packaging until the total amount of data reaches or approaches 500KB. The generated first data packet may include differential data and its corresponding safety operation feature data.

[0081] This data packet generation and transmission mechanism significantly optimizes data transmission efficiency and increases the update speed of disease identification models.

[0082] In the highway disease inspection system, the accuracy of disease identification directly affects the decision-making and subsequent maintenance measures at the construction site. Therefore, in order to improve the reliability of disease identification, the present invention introduces a confidence verification mechanism. By calculating the confidence of the disease identification results and dynamically adjusting the confidence threshold based on the safety operation feature data, the validity of the identification results can be ensured, and the most reliable data can be screened out for model updating.

[0083] As an optional implementation, the confidence verification of the first disease information and the generation of verified characteristic data include: calculating the confidence value of the disease identification result based on the first disease information and the safe operation characteristic data in the first data packet; the safe operation characteristic data include: environmental information and context information; comparing the confidence value with a preset confidence threshold; wherein the preset confidence threshold is dynamically adjusted based on the safe operation characteristic data; the dynamic adjustment includes: in response to the presence of severe weather information in the environmental information, lowering the confidence threshold; wherein the severe weather information The information includes: rainy days, snowy days, and haze; in response to the traffic flow in the context information being greater than the flow threshold, the confidence threshold is increased; in response to the traffic artery information or key construction area information in the context information, the confidence threshold is increased; in response to the confidence value being less than the preset confidence threshold, the corresponding first disease information is marked as low confidence; in response to the confidence value being greater than or equal to the preset confidence threshold, the corresponding first disease information is marked as high confidence, and the first disease information marked as high confidence is used as the verified feature data for iteratively updating the disease recognition model; The iterative update of the disease recognition model includes: using the verified feature data and a preset machine learning algorithm to train the disease recognition model and update the model parameters; according to a preset model update strategy, when the verified feature data accumulates to a predetermined number or a preset time interval has passed, executing the model update.

[0084] In a specific implementation, after receiving the first data packet transmitted by the vehicle-mounted terminal, the server extracts the first disease information and safety operation feature data contained therein. Based on these data, the confidence value of each disease identification result is calculated. The confidence value is an important parameter for measuring the accuracy of the disease identification result. The specific calculation method may include the confidence score of the machine learning algorithm, the rule-based evaluation, etc. This application does not limit this.

[0085] Exemplarily, the server extracts the first disease information from the first data packet, including the type, severity, location information, etc. of the disease. At the same time, the corresponding safety operation feature data is also extracted, which includes environmental information (such as weather and lighting conditions) and contextual information (such as traffic flow, construction signs, etc.). The confidence value of each disease identification result is calculated by combining the disease feature data and the safety operation feature data through a machine learning algorithm or a rule-based analysis method. For example, a deep learning model outputs a confidence score (a value between 0 and 1) to evaluate the reliability of the identification.

[0086] In order to adapt to the complexity of different construction environments and road conditions, this application proposes a mechanism for dynamically adjusting the confidence threshold. The confidence threshold will vary according to different environmental information and context information to ensure the accuracy of disease identification.

[0087] In specific implementation, when bad weather (such as rain, snow, and haze) is detected in the environmental information, the confidence threshold will be lowered. This is because the difficulty of identifying disease characteristics increases under bad weather conditions, and lowering the threshold can improve the sensitivity of the warning.

[0088] When the traffic flow in the context information is greater than the set flow threshold, the confidence threshold will be increased to reduce the possibility of false alarms. In addition, when the context information contains information about traffic arteries or key construction areas, the threshold will also be appropriately increased to ensure that the identification results of high-risk areas are more accurate.

[0089] After the confidence value is calculated and dynamically adjusted, each disease identification result is verified. The confidence value is compared with the dynamically adjusted confidence threshold to determine the credibility status of the disease information. Low-confidence data can trigger manual review or special processing, while high-confidence data will be used to iteratively update the disease identification model.

[0090] Among them, high-credibility disease information will be used to iteratively update the disease recognition model to improve the overall recognition accuracy of the system. The iterative update of the model is an ongoing process. With the continuous accumulation of high-credibility data, the model retraining and parameter adjustment are regularly triggered.

[0091] In the specific implementation, the disease recognition model is trained using high-confidence verified feature data and a preset machine learning algorithm (such as convolutional neural network in deep learning).

[0092] According to the preset model update strategy, the accumulated amount and time interval of the verified feature data are checked regularly. When the verified feature data reaches the predetermined amount or a specific time interval has passed, the server will perform a model update to ensure that the model can adapt to the dynamic changes in road conditions.

[0093] After the model update is completed, the server will send the latest disease recognition model to the vehicle terminal to replace the old model. After receiving the new model, the vehicle terminal will automatically deploy it to ensure that the latest recognition algorithm can be used in subsequent inspection tasks.

[0094] In this way, the above verification and update mechanism can significantly improve the accuracy of disease identification and ensure that the model can always maintain efficient performance in complex construction environments.

[0095] Those skilled in the art will appreciate that, in the above method of specific implementation, the order in which the steps are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of the steps should be determined by their functions and possible internal logic.

[0096] Based on the same inventive concept, the embodiments of the present application also provide a vehicle-mounted terminal and a server for executing a self-update inspection method for highway defects. Since the principles of solving the problems by the vehicle-mounted terminal and the server in the embodiments of the present application are similar to the above-mentioned self-update inspection method for executing highway defects in the embodiments of the present disclosure, the implementation of the vehicle-mounted terminal and the server can refer to the implementation of the method, and the repeated parts will not be repeated.

[0097] See also Figure 3 , Figure 3 The schematic diagram of the vehicle terminal for executing the self-updating inspection method for highway diseases provided in the embodiment of the present application is as follows. The vehicle terminal includes: a collection module 10, a processing module 20, a distribution module 30, and a first transmission module 40; wherein: The acquisition module 10 is used to obtain disease characteristic data and safe operation characteristic data for the target highway, wherein the safe operation characteristic data includes: environmental information and context information; wherein the safe operation characteristic data is used to perform a local assessment of safety risks at the construction site on the vehicle terminal; The processing module 20 is used to identify the disease feature data using the disease recognition model deployed on the vehicle terminal, generate first disease information of the target highway, evaluate the safety risk based on the first disease information, and generate safety risk warning information; An allocation module 30, configured to mark priority information for the disease characteristic data based on a preset priority allocation rule; The first transmission module 40 is used to use the priority information to generate a first data packet based on the disease characteristic data, the safe operation characteristic data, and the first disease information, and transmit the first data packet to the server; wherein the first data packet is used for iterative update of the disease recognition model.

[0098] See also Figure 4 , Figure 4 A schematic diagram of a server for executing a self-updating inspection method for highway diseases provided in an embodiment of the present application, wherein the server comprises: a verification module 50, a construction module 60, and a second transmission module 70; wherein: The verification module 50 is used to receive the first data packet transmitted by the vehicle-mounted terminal, and based on the first data packet, perform confidence verification on the first disease information to generate verified feature data; A construction module 60 is used to iteratively update the disease recognition model based on the verified feature data; The second transmission module 70 is used to send the iteratively updated disease recognition model to the vehicle terminal.

[0099] The embodiment of the present application further provides a vehicle, comprising a vehicle body and the above-mentioned vehicle-mounted terminal configured inside the vehicle body; wherein the vehicle-mounted terminal is used to execute the above-mentioned self-updating inspection method for highway diseases, comprising: For the target highway, disease characteristic data and safe operation characteristic data are obtained, wherein the safe operation characteristic data includes: environmental information and context information; wherein the safe operation characteristic data is used to perform a local assessment of safety risks at the construction site on the vehicle terminal; Using the disease recognition model deployed on the vehicle-mounted terminal, the disease feature data is identified to generate first disease information of the target highway, and safety risks are evaluated based on the first disease information to generate safety risk warning information; Based on a preset priority allocation rule, marking priority information for the disease characteristic data; Using the priority information, a first data packet is generated based on the disease feature data, the safe operation feature data, and the first disease information, and the first data packet is transmitted to a server; wherein the first data packet is used for iterative updating of the disease identification model.

[0100] The execution method of the above-mentioned self-updating inspection method for highway defects can refer to the steps of the self-updating inspection method for highway defects recorded above, which will not be repeated here.

[0101] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0102] It should be understood that determining B based on A does not mean determining B only based on A. B can also be determined based on A and / or other information.

[0103] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0104] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0105] The preferred embodiments of the present application disclosed above are only used to help explain the present application. The preferred embodiments do not describe all the details in detail, nor do they limit the present application to specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present application, so that those skilled in the art can understand and use the present application well. The present application is only limited by the claims and their full scope and equivalents.

Claims

1. A self-updating inspection method for highway diseases, characterized in that: Executed by the vehicle-mounted terminal, the self-update inspection method includes: For the target highway, disease characteristic data and safe operation characteristic data are obtained, wherein the safe operation characteristic data includes: environmental information and context information; wherein the safe operation characteristic data is used to perform a local assessment of safety risks at the construction site on the vehicle terminal; Using the disease recognition model deployed on the vehicle-mounted terminal, the disease feature data is identified to generate first disease information of the target highway, and safety risks are evaluated based on the first disease information to generate safety risk warning information; Based on a preset priority allocation rule, marking priority information for the disease characteristic data; Using the priority information, a first data packet is generated based on the disease feature data, the safe operation feature data, and the first disease information, and the first data packet is transmitted to a server; wherein the first data packet is used for iterative updating of the disease identification model.

2. The method according to claim 1, characterized in that: The generating of security risk warning information includes: Based on the geographic information of the disease feature data and the safe operation feature data, matching the disease feature data with the safe operation feature data whose geographic information differs by less than a preset value; In response to the presence of the disease feature data matching the safety operation feature data, determining the disease type, severity, and location information based on the first disease information corresponding to the disease feature data, and generating safety risk warning information based on the disease type, severity, location information, and the safety operation feature data using a preset safety risk assessment model; In response to the absence of the disease feature data matching the safety operation feature data, a preset safety risk assessment model is used to generate safety risk warning information based on the safety operation feature data.

3. The method according to claim 2, characterized in that Before marking the priority information for the disease characteristic data based on the preset priority allocation rule, the method further includes: Based on the type, severity, and location information of the first defect information, a comparison is performed in a first defect database locally deployed on the vehicle terminal to screen out differential data; wherein the data in the first defect database only includes the first defect information collected for the target highway in the most recent target inspection round; The difference data refers to the significant changes or newly added disease feature data between the disease feature data currently collected for the target highway inspection and the disease feature data recorded in the first disease database; The significant changes include: a change in the disease type, a disease severity exceeding a preset change threshold, and a disease location information offset exceeding a preset distance; The marking of priority information for the disease characteristic data includes: For the difference data, priority information is marked based on the preset priority allocation rule.

4. The method according to claim 3, characterized in that The acquisition of disease characteristic data and safe operation characteristic data includes: Using the camera deployed on the vehicle terminal, the disease characteristic data and the safe operation characteristic data are acquired at a collection frequency that matches the Beidou positioning data update frequency; The image data collected by the camera and the Beidou positioning data share a consistent timestamp; The positioning data of the target highway is obtained through the Beidou positioning module at an update frequency matching the camera acquisition frequency, as the position information corresponding to the disease feature data and the safe operation feature data.

5. The method according to claim 4, characterized in that The generating a first data packet and transmitting the first data packet to the server comprises: Generate a first data packet based on the priority information corresponding to each of the difference data and the safety operation characteristic data corresponding to each of the difference data; wherein the difference data and the safety operation characteristic data are matched based on the geographic information; Dynamically evaluate the current wireless network and generate network status evaluation information; based on the network status evaluation information, determine the size of the data packet that can be currently sent; Based on the data packet size and the priority sorting result, select the difference data with the highest priority to generate a first data packet; The first data packet is transmitted to the server via a wireless network.

6. A self-renewal inspection method for highway diseases, characterized in that: Executed by the server, the self-update inspection method includes: receiving a first data packet transmitted by the vehicle-mounted terminal, and based on the first data packet, performing confidence verification on the first disease information to generate verified feature data; Iteratively updating the disease recognition model based on the verified feature data; The iteratively updated disease identification model is sent to the vehicle-mounted terminal.

7. The method according to claim 6, characterized in that The step of performing confidence verification on the first disease information and generating verified feature data comprises: Calculating a confidence value of a disease identification result based on the first disease information and safe operation feature data in the first data packet; the safe operation feature data includes: environmental information and context information; Comparing the confidence value with a preset confidence threshold; wherein the preset confidence threshold is dynamically adjusted based on the safe operation characteristic data; The dynamic adjustment includes: in response to the presence of bad weather information in the environmental information, lowering the confidence threshold; wherein the bad weather information includes: rainy days, snowy days, and haze; in response to the presence of traffic flow greater than the flow threshold in the context information, raising the confidence threshold; in response to the presence of traffic artery information or key construction area information in the context information, raising the confidence threshold; In response to the confidence value being less than the preset confidence threshold, marking the corresponding first disease information as having low credibility; In response to the confidence value being greater than or equal to the preset confidence threshold, marking the corresponding first disease information as high confidence, and using the first disease information marked as high confidence as verified feature data for iteratively updating the disease recognition model; The iterative updating of the disease identification model includes: Using the verified feature data and a preset machine learning algorithm, the disease recognition model is trained to update the model parameters; According to the preset model update strategy, when the verified feature data accumulates to a predetermined amount or a preset time interval has passed, the model update is executed.

8. A vehicle-mounted terminal, used to execute the method according to any one of claims 1 to 5, characterized in that: include: A collection module, used for acquiring disease characteristic data and safe operation characteristic data for a target highway, wherein the safe operation characteristic data includes: environmental information and context information; wherein the safe operation characteristic data is used for local assessment of safety risks at the construction site on a vehicle-mounted terminal; A processing module, configured to identify the defect feature data using a defect recognition model deployed on the vehicle-mounted terminal, generate first defect information of the target highway, evaluate safety risks based on the first defect information, and generate safety risk warning information; An allocation module, used for marking priority information for the disease characteristic data based on a preset priority allocation rule; The first transmission module is used to use the priority information to generate a first data packet based on the disease characteristic data, the safe operation characteristic data, and the first disease information, and transmit the first data packet to the server; wherein the first data packet is used for iterative update of the disease recognition model.

9. A server, used to execute the method according to any one of claims 6 to 7, characterized in that: include: A verification module, configured to receive a first data packet transmitted by the vehicle-mounted terminal, and based on the first data packet, perform confidence verification on the first disease information to generate verified feature data; A construction module is used to iteratively update the disease recognition model based on the verified feature data; The second transmission module is used to send the iteratively updated disease recognition model to the vehicle-mounted terminal.

10. A vehicle, comprising a vehicle body, characterized in that: Also includes: The vehicle-mounted terminal according to claim 8 is arranged inside the vehicle body.

Images