Road surface state cooperative intelligent analysis method and system

Abstract

The invention discloses a road surface state cooperative intelligent analysis method and system, and relates to the technical field of road surface analysis. A multi-source sensor is deployed, real-time data of a road surface is collected, and road surface multi-source data is generated; carrying out fusion processing on different types of data collected by multiple sensors to generate a comprehensive pavement state data vector; analyzing the comprehensive road surface state data vector by using a machine learning model, determining a specific position of an abnormal state on the road surface, and generating road surface abnormal state reminding information; generating a pavement maintenance control instruction; and according to the road surface maintenance control instruction, matching corresponding prompt content, and sending prompt information to related management departments and maintenance personnel. According to the invention, the technical problem of low road maintenance and management efficiency caused by untimely road surface state monitoring and inaccurate abnormity judgment in the prior art is solved, and the technical effects of real-time and accurate road surface state monitoring and improvement of the road maintenance and management efficiency are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of pavement analysis, and particularly to a method and system for collaborative intelligent analysis of pavement conditions. Background Art

[0002] In today's transportation infrastructure field, roads, as the key link connecting various regions and supporting economic activities and social operations, their good operating conditions are crucial. However, traditional pavement condition monitoring and maintenance technologies face many challenges. On the one hand, early pavement monitoring mainly relied on manual inspections. This method not only consumed a large amount of manpower, material resources and time, but also was greatly affected by the subjective factors of inspectors, making it difficult to achieve comprehensive and real-time monitoring, and it was extremely easy to miss some hidden pavement diseases or early minor abnormal conditions. On the other hand, even when using some sensors for monitoring, there were often problems such as single data type and lack of effective data fusion means, and it was impossible to accurately reflect the complex actual pavement conditions. In the maintenance link, due to the lack of accurate positioning of pavement anomalies and an efficient maintenance strategy matching mechanism, the maintenance work was often carried out blindly, not only consuming a large amount of resources, but also unable to solve pavement problems in a timely and effective manner.

[0003] The prior art has technical problems such as untimely monitoring of pavement conditions and inaccurate abnormal judgment, resulting in low efficiency of road maintenance management. Summary of the Invention

[0004] This application provides a method and system for collaborative intelligent analysis of pavement conditions, which are used to solve the technical problems in the prior art of untimely monitoring of pavement conditions and inaccurate abnormal judgment, resulting in low efficiency of road maintenance management.

[0005] In view of the above problems, this application provides a method for collaborative intelligent analysis of pavement conditions.

[0006] In the first aspect of this application, a method for collaborative intelligent analysis of pavement conditions is provided. The method includes:

[0007] Deploy multi-source sensors, and based on a preset acquisition period, collect real-time data of the pavement to generate multi-source pavement data; based on a preset data fusion algorithm, fuse and process different types of data collected by multiple sensors to generate a comprehensive pavement condition data vector; use a machine learning model to analyze the comprehensive pavement condition data vector, judge whether the pavement has an abnormal condition, and determine the location of the abnormal condition according to the time and location information of the collected data, and generate a pavement abnormal condition reminder message; obtain the type and location of the pavement abnormal condition, match the abnormal condition with a preset maintenance strategy library to generate a pavement maintenance control instruction; according to the pavement maintenance control instruction, match the corresponding prompt content and send the prompt message to relevant management departments and maintenance personnel.

[0008] In a second aspect of the present application, a collaborative intelligent analysis system for road surface conditions is provided. The system includes:

[0009] A multi-source data generation module for road surface, which is used to deploy multi-source sensors, collect real-time data of the road surface based on a preset collection period, and generate multi-source data of the road surface; a comprehensive road surface condition data vector generation module, which is used to fuse and process different types of data collected by multiple sensors based on a preset data fusion algorithm to generate a comprehensive road surface condition data vector; a reminder information generation module, which is used to analyze the comprehensive road surface condition data vector by using a machine learning model, judge whether there is an abnormal condition on the road surface, and determine the location of the abnormal condition according to the time and location information of the collected data, and generate a reminder information for the abnormal condition of the road surface; a road surface maintenance control instruction generation module, which is used to obtain the type and location of the abnormal condition of the road surface, match the abnormal condition with a preset maintenance strategy library, and generate a road surface maintenance control instruction; a prompt information sending module, which is used to match the corresponding prompt content according to the road surface maintenance control instruction and send the prompt information to relevant management departments and maintenance personnel.

[0010] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0011] Deploy multi-source sensors, collect real-time data of the road surface based on a preset collection period, and generate multi-source data of the road surface; fuse and process different types of data collected by multiple sensors based on a preset data fusion algorithm to generate a comprehensive road surface condition data vector; analyze the comprehensive road surface condition data vector by using a machine learning model, judge whether there is an abnormal condition on the road surface, and determine the location of the abnormal condition according to the time and location information of the collected data, and generate a reminder information for the abnormal condition of the road surface; obtain the type and location of the abnormal condition of the road surface, match the abnormal condition with a preset maintenance strategy library, and generate a road surface maintenance control instruction; match the corresponding prompt content according to the road surface maintenance control instruction and send the prompt information to relevant management departments and maintenance personnel. It achieves the technical effect of real-time and accurate monitoring of the road surface conditions and improves the efficiency of road maintenance management. Description of the Drawings

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0013] Figure 1Schematic flowchart of a method for collaborative intelligent analysis of road surface conditions provided by an embodiment of the present application;

[0014] Figure 2 Schematic structural diagram of a system for collaborative intelligent analysis of road surface conditions provided by an embodiment of the present application.

[0015] Explanation of reference numerals: Multi-source data generation module 10 for road surface, Comprehensive road surface condition data vector generation module 20, Reminder information generation module 30, Road surface maintenance control instruction generation module 40, Prompt information sending module 50. Detailed implementation manners

[0016] The present application provides a method and a system for collaborative intelligent analysis of road surface conditions, which are used to solve the technical problems in the prior art that the monitoring of road surface conditions is not timely and the abnormal judgment is inaccurate, resulting in low efficiency of road maintenance management.

[0017] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0018] Embodiment 1, as Figure 1 shown, the present application provides a method for collaborative intelligent analysis of road surface conditions, and the method includes:

[0019] Step S100: Deploy multi-source sensors, and based on a preset acquisition period, collect real-time data of the road surface to generate multi-source data of the road surface.

[0020] Specifically, multi-source sensors are deployed on the road surface. These sensors are of various types. Among them, image sensors are used to capture real-time images of the road surface, which can intuitively show whether there are surface conditions such as cracks and potholes on the road surface; humidity sensors can accurately measure the humidity of the road surface, providing data for judging whether problems such as changes in friction caused by dampness occur on the road surface; temperature sensors are responsible for monitoring the road surface temperature, which helps to understand the impact of temperature changes on the performance of road surface materials; stress and strain sensors are used to sense the stress and strain borne by the road surface and analyze the stability of the road surface structure. According to the preset acquisition period, these sensors continuously and automatically collect various real-time data of the road surface. The setting of the acquisition period comprehensively considers the frequency of road surface condition changes and the efficiency of data processing, ensuring that dynamic information of the road surface can be obtained in a timely manner without causing data redundancy and excessive processing burden due to overly frequent acquisition. As the sensors continue to work, road surface image data, humidity data, temperature data, and stress and strain data are collected, and finally integrated to generate multi-source road surface data containing rich information, laying a foundation for subsequent accurate analysis of the road surface state.

[0021] Step S200: Based on a preset data fusion algorithm, fuse and process different types of data collected by multiple sensors to generate a comprehensive road surface state data vector.

[0022] Specifically, for road surface image data, using image feature extraction technology, based on the method of convolutional neural network, accurately extract road surface texture features from the image and convert them into representative road surface texture feature vectors, which can effectively reflect the surface microstructure and damage signs of the road surface. At the same time, normalize the road surface humidity data, temperature data, and stress and strain data, unify these data into the same numerical interval, eliminate the influence brought by the differences in dimension and numerical range of different data types, so that they can play a fair role in the subsequent fusion process. Finally, adopt a weighted fusion algorithm, assign weights to each data according to its importance in reflecting the road surface state, and fuse the processed road surface texture feature vectors, normalized humidity data, temperature data, and stress and strain data, thus generating a comprehensive road surface state data vector. This vector comprehensively and orderly integrates the key information of multi-source data, provides a high-quality and more valuable data basis for subsequent analysis of the road surface state using machine learning models, and greatly improves the accuracy and reliability of the analysis results.

[0023] Step S300: Use a machine learning model to analyze the comprehensive road surface state data vector, judge whether there is an abnormal state on the road surface, and determine the location of the abnormal state according to the time and location information of the collected data, and generate a road surface abnormal state reminder message.

[0024] Specifically, first construct a comprehensive and representative training dataset that covers a large number of sample data in normal road surface states and various abnormal states. With the help of these rich samples, use deep learning algorithms to repeatedly train the machine learning model, enabling it to continuously learn and master the characteristic manifestations of different road surface states in data vectors, thus obtaining a trained road surface state analysis model. When the comprehensive road surface state data vector is input into the trained model, the model will quickly and accurately process the input data based on the knowledge and patterns it has learned, and output the judgment result of the road surface state to determine whether there is an abnormal state on the road surface. Once the model determines that there is an abnormality on the road surface, immediately start the positioning program. By using the time and location information synchronously recorded during data collection and combining with the positioning algorithm, it is possible to quickly determine the specific location of the abnormal state on the road surface. For example, through the layout information of the sensors and the time sequence of data collection, use triangulation or positioning technology based on geographic information system (GIS) to accurately lock the abnormal location. After determining the location of the abnormal state, generate a road surface abnormal state reminder message containing detailed information. These reminder messages not only include the types of abnormal states, such as cracks, potholes, subsidence, etc., but also clearly indicate the specific location where the abnormality appears, accurate to the specific section of the road. In addition, the reminder messages will also attach an estimated severity level of the abnormal state and recommended treatment measures, so that relevant management departments and maintenance personnel can obtain key information in a timely manner and take effective measures to repair and maintain the abnormal road surface in a timely manner to ensure the safe passage of the road.

[0025] Step S400: Obtain the type and location of the road surface abnormal state, match the abnormal state with a preset maintenance strategy library, and generate a road surface maintenance control instruction.

[0026] Specifically, first, obtain the specific types of road surface abnormal states, such as cracks, potholes, bumps, etc., as well as accurate location information. Subsequently, use this information as the retrieval basis to perform a matching operation with a pre-constructed maintenance strategy library. The maintenance strategy library stores a large amount of data, including historical maintenance records, the mapping relationships between various abnormal types and corresponding maintenance measures, and the distribution locations of maintenance resources, etc. According to the obtained abnormal type, screen out the corresponding candidate maintenance measures from the mapping relationships of maintenance measures. For example, if the abnormal type is a crack, the candidate maintenance measures include crack filling treatment, patch tape repair, etc. Then, combine the location of the abnormal state and the distribution location of maintenance resources to calculate the response priority of each candidate maintenance measure. For example, by calculating the distance between the abnormal location and the locations of various maintenance resources, and referring to the road traffic condition data, use a path planning algorithm to estimate the actual travel time required to reach the abnormal location; then set weights according to the urgent demand levels of different types of maintenance resources, and use a weighted average algorithm to obtain the response priority of each candidate maintenance measure. Finally, based on this priority information, generate detailed road surface maintenance control instructions. The instruction content includes the type of maintenance measure, specifying exactly how to perform the repair; execute path planning to plan the best route from the location of the maintenance resource to the abnormal location; and the required resource list, listing the resources such as manpower, equipment, and materials required to complete the maintenance work, ensuring that the maintenance work can be carried out efficiently and orderly.

[0027] Step S500: According to the road surface maintenance control instruction, match the corresponding prompt content and send the prompt information to the relevant management departments and maintenance personnel.

[0028] Specifically, according to the specific content in the road surface maintenance control instruction, the corresponding information is matched from a pre-set prompt content library. This prompt content library contains standardized prompt texts for different maintenance measure types, execution path planning, and required resource lists. For example, if the maintenance control instruction requires pothole repair, and the execution path involves a busy main road, and the required resources include asphalt, road roller equipment, and several maintenance workers, the corresponding prompt content is matched, which details that the specific task of the maintenance work is pothole repair, and attention should be paid to the traffic conditions on the main road during the execution process, and the construction time should be reasonably arranged to reduce the impact on traffic. At the same time, key information such as the quantity of asphalt required, the model of road roller equipment, and the number of workers is listed. After the matching is completed, these prompt messages are accurately sent to relevant management departments and maintenance personnel through various communication methods, such as text messages, instant messaging software, professional management platforms, etc. For the management department, the prompt information can help them comprehensively understand the situation of road surface maintenance work, so as to carry out resource coordination, progress monitoring, and decision-making. For maintenance personnel, the prompt information is a specific work guide. Based on this information, they can prepare the required equipment and materials in advance, quickly reach the maintenance site according to the planned execution path, and accurately carry out the maintenance work, so as to ensure that road surface anomalies can be solved in a timely and effective manner, and the normal use of the road and traffic safety can be guaranteed.

[0029] In a possible implementation manner, step S100 further includes:

[0030] Step S110: Use multi-source sensors to collect and obtain road surface image data, road surface humidity data, road surface temperature data, and stress and strain data.

[0031] Step S120: Perform time synchronization and coordinate calibration processing on the road surface image data, road surface humidity data, road surface temperature data, and stress and strain data.

[0032] Specifically, relevant road surface data is comprehensively collected by means of deployed multi-source sensors. Among them, the high-definition camera, as a key sensor for collecting road surface image data, can continuously capture real-time images of the road surface, record the texture and color changes of the road surface, and whether there are signs of diseases such as cracks and potholes. The humidity sensor uses a sensing element to monitor the water content of the road surface in real time, obtain road surface humidity data, and provide a basis for judging the wetness degree of the road surface and whether there is a hidden danger of water accumulation. The temperature sensor uses a thermosensitive element to accurately measure the real-time temperature of the road surface, which is crucial for analyzing the impact of temperature changes on the performance of road surface materials, such as road surface softening caused by high temperature and frost heaving caused by low temperature. The stress and strain sensor is installed inside the road surface structure to sense the stress and strain conditions borne by the road surface, reflect the stress state of the road surface structure, and help judge whether there is potential structural damage to the road surface. These sensors work together to collect road surface image data, road surface humidity data, road surface temperature data, and stress and strain data from different dimensions, providing rich original information for subsequent road surface state analysis.

[0033] Since there are differences in the data acquisition time and reference coordinate system when different types of sensors collect data, in order to ensure the accuracy and consistency of the data, it is necessary to perform time synchronization and coordinate calibration processing on the collected road surface image data, road surface humidity data, road surface temperature data, and stress and strain data. In terms of time synchronization, high-precision clock synchronization technology, the clock synchronization method based on the Global Positioning System (GPS), or the Network Time Protocol (NTP) is adopted to accurately align the acquisition times of all sensors, ensuring the comparability of different types of data in the time dimension and facilitating the subsequent analysis of the change trend of the road surface state over time. Coordinate calibration is to adjust the coordinate information when each sensor collects data according to a unified geographic coordinate system or a specific reference coordinate system of the road surface. For road surface image data, through image matching and geographic information annotation technology, the position information in the image is corresponded to the actual geographic coordinates; for the data collected by humidity, temperature, and stress and strain sensors, based on the accurate measurement results of their installation positions, combined with the measurement principle and installation angle of the sensors, the coordinate information in the data is calibrated to ensure that all data can be accurately associated in terms of spatial position, providing a reliable data basis for subsequent comprehensive analysis and road surface anomaly positioning.

[0034] In a possible implementation manner, step S200 further includes:

[0035] Step S210: Extract features from the road surface image data to obtain a road surface texture feature vector.

[0036] Step S220: Normalize the road surface humidity data, road surface temperature data, and stress and strain data.

[0037] Step S230: Use the weighted fusion algorithm to fuse the processed various types of data to generate a comprehensive road surface state data vector.

[0038] Specifically, for road surface image data, image processing and machine learning techniques are used for feature extraction. The convolutional neural network (CNN) algorithm is applied to perform multi-layer convolution and pooling operations on the image, automatically identifying and extracting the texture features of the road surface, such as the thickness and direction of cracks, the size and shape of potholes, etc. These extracted texture features are quantified and encoded, and finally a road surface texture feature vector is formed. This vector can accurately describe the surface microstructure and damage characteristics of the road surface in numerical form, providing key image information for subsequent comprehensive analysis.

[0039] Since there are differences in the dimensions, numerical ranges, and distributions of road surface humidity data, road surface temperature data, and stress-strain data, in order to eliminate the influence of these differences on subsequent fusion analysis, it is necessary to normalize them. For road surface humidity data, by calculating its minimum and maximum values, the humidity data is mapped to the interval [0, 1], making the humidity data collected at different times and locations comparable. Similar normalization methods are also used for road surface temperature data and stress-strain data, unifying their numerical ranges to the same scale. After such processing, all data has the same importance and weight numerically, facilitating subsequent fusion operations.

[0040] Use the weighted fusion algorithm to organically fuse the processed various types of data. According to the importance and contribution degree of each type of data in reflecting the road surface state, different weights are assigned to the road surface texture feature vector, the normalized road surface humidity data, the road surface temperature data, and the stress-strain data respectively. For example, when judging whether there are cracks on the road surface, the weight of the road surface texture feature vector may be relatively high; while when analyzing the deformation of the road surface caused by temperature changes, the weight of the road surface temperature data will increase accordingly. Through weighted summation, these data are combined according to their respective weights to generate a comprehensive road surface state data vector. This vector synthesizes the information of multiple types of data and can more comprehensively and accurately reflect the actual state of the road surface, providing a high-quality data basis for subsequent road surface state analysis using machine learning models.

[0041] In a possible implementation manner, step S300 further includes:

[0042] Step S310: Construct a training data set containing various normal and abnormal state samples of the road surface, train the deep learning model, and obtain a trained road surface state analysis model.

[0043] Step S320: Input the comprehensive road surface state data vector into the road surface state analysis model, output the road surface state judgment result, and judge whether there is an abnormal state.

[0044] Step S330: If there is an abnormal state, use the positioning algorithm and combine it with the position information of the sensor to determine the location of the abnormal state.

[0045] Specifically, collect a large amount of road surface data from different sections and at different times, including road surface images, humidity, temperature, stress and strain data under normal conditions, as well as data containing various abnormal states such as cracks, potholes, and bumps. For road surface image data, use the feature extraction method in the convolutional neural network (CNN), utilize the VGG16 network, and extract features such as texture and edges in the image through multi-layer convolution and pooling operations. For humidity, temperature, and stress and strain data, perform normalization processing, map them to the [0, 1] interval to make them comparable. Divide the processed various data into a training set and a test set according to a certain ratio, for example, 70% as the training set and 30% as the test set. Select a suitable deep learning model, such as ResNet, build the model structure, set the input layer to receive the processed feature data, the middle layer contains multiple residual blocks for extracting deep features, the output layer sets nodes according to the number of road surface state classifications, and use the cross-entropy loss function and the Adam optimizer for training. During the training process, continuously adjust the model parameters until the accuracy of the model on the test set reaches a stable and high level, thereby obtaining a trained road surface state analysis model.

[0046] Input the obtained comprehensive road surface state data vector into the trained road surface state analysis model. The data enters from the input layer of the model, passes through each residual block in the middle in turn for feature extraction and calculation, and finally reaches the output layer. What the output layer outputs is a probability vector, and each element in the vector represents the probability that the road surface belongs to different states (normal, crack, pothole, etc.). Use the Softmax function to process the output vector to obtain the probability distribution of each state, and select the state with the highest probability as the road surface state judgment result. If the state corresponding to the highest probability is normal, then it is judged that there is no abnormal state on the road surface; if it is other abnormal states, then it is judged that there is an abnormality on the road surface.

[0047] Build a machine learning model dedicated to positioning, using the Support Vector Regression (SVR) model. In the model training stage, collect a large amount of historical abnormal state data, which includes data collected by multiple sensors when different abnormalities occur and the corresponding accurate abnormal location information. Use data such as signal strength and time of arrival of signals collected by sensors as input features, and the coordinates of the abnormal location as output labels to train the SVR model. During the training process, continuously adjust the kernel function parameters, penalty factors, etc. of the model to optimize the model so that it can accurately learn the complex relationship between the input features and the abnormal location. When a new abnormal state appears, extract information such as the signal strength and time of arrival of signals of each sensor that has collected abnormal data currently, form a feature vector and input it into the trained SVR model. The model will make a prediction according to the learned pattern and output a coordinate value, which is a preliminary estimate of the abnormal state location. Considering that there may be noise interference and data errors in the actual situation, then use an optimized clustering algorithm to optimize the preliminary estimation result. Conduct a clustering analysis on the location points obtained from multiple predictions, group the points with close distances into one category, and select the cluster center as the finally determined abnormal state location, so as to improve the positioning accuracy and provide an accurate location basis for subsequent road surface maintenance work.

[0048] In a possible implementation manner, step S400 further includes:

[0049] Step S410: Build the maintenance strategy library, where the maintenance strategy library includes the mapping relationships of historical maintenance records, abnormal types, maintenance measures, and the distribution locations of maintenance resources.

[0050] Step S420: According to the type of the abnormal state, match the corresponding candidate maintenance measures from the mapping relationships of the maintenance measures.

[0051] Step S430: Combine the location of the abnormal state with the distribution locations of the maintenance resources to calculate the response priority of the candidate maintenance measures.

[0052] Step S440: Generate the road surface maintenance control instruction based on the response priority, where the road surface maintenance control instruction includes the type of maintenance measure, execution path planning, and a list of required resources.

[0053] Specifically, a maintenance strategy library is constructed to collect extensive historical maintenance records, which include detailed information on each maintenance action, such as the time when maintenance was carried out, the specific road surface location involved, the type of abnormality presented by the road surface at that time, and the maintenance measures finally taken. At the same time, various possible road surface abnormalities are sorted out and classified to form a comprehensive list of abnormality types, such as cracks, potholes, ridges, and road surface peeling. For each type of abnormality, in-depth research is conducted and a corresponding maintenance measure mapping relationship is established. For example, cracks correspond to maintenance methods such as grouting and taping, and potholes are associated with measures such as filling asphalt materials. In addition, the distribution of maintenance resources is recorded in detail, including the sites where maintenance personnel are located, the storage locations of various types of construction equipment, and the locations of storage warehouses for different maintenance materials, so as to integrate and form a complete and well-organized maintenance strategy library, providing solid data support and decision-making basis for subsequent maintenance decisions.

[0054] After the maintenance strategy library is built, once the abnormal state of the road surface is determined and its abnormal type is identified, the matching work will be quickly carried out according to the maintenance measures mapping relationship established in the maintenance strategy library. For example, if the abnormal type detected is cracks on the road surface, the maintenance measures related to the cracks will be accurately retrieved in the maintenance strategy library immediately, and crack filling, seam tape repair, etc. will be quickly selected as candidate maintenance measures, providing a variety of options for the subsequent determination of specific maintenance plans.

[0055] The precise location information of the abnormal state on the road surface and the distribution location data of maintenance resources are fully combined. First, with the help of the path planning algorithm, the actual travel time from the location of the maintenance resource to the abnormal location is accurately calculated by comprehensively considering dynamic factors such as road traffic conditions and real-time traffic flow. For example, if the abnormal location is on a main road with heavy traffic and easy congestion, and there are complex situations such as road construction in the surrounding area, the actual travel time will increase significantly. At the same time, according to the degree of emergency demand for different types of maintenance resources in dealing with various types of abnormalities, corresponding weights are set for them. For example, for abnormal situations that may seriously affect traffic safety, the weights of professional materials and repair personnel used for emergency repairs will be set higher. Finally, through the weighted average algorithm, the actual travel time and the maintenance resource weight are comprehensively calculated to accurately calculate the response priority of each candidate maintenance measure. Candidate maintenance measures with higher response priorities mean that under the current circumstances, road abnormalities can be solved more quickly and efficiently to ensure the normal use of roads.

[0056] Generate detailed and practical pavement maintenance control instructions based on the calculated response priorities. The content of these instructions is rich and specific. First, clearly define the type of maintenance measures selected, that is, among numerous candidate maintenance measures, select the measure with the highest response priority as the final implementation plan. For example, determine to use crack filling to handle crack problems. Secondly, the instructions include detailed execution path planning. According to the abnormal location, the distribution location of maintenance resources, and real-time traffic information, plan the optimal driving route from the departure location of maintenance resources to the abnormal location to ensure that maintenance personnel and equipment can reach the maintenance site quickly and smoothly. Finally, the instructions will also generate a detailed list of required resources, clearly listing various types of resources required to complete this maintenance work, including the number of repair workers required in terms of manpower, the type of construction equipment required in terms of equipment, and the specific quantity of repair materials required in terms of materials. By generating such comprehensive and detailed pavement maintenance control instructions, provide clear and accurate action guidelines for subsequent pavement maintenance work, effectively ensuring that the maintenance work can be carried out in an orderly and efficient manner and quickly restore the normal use state of the pavement.

[0057] In a possible implementation manner, step S430 further includes:

[0058] Step S431: Coordinate the location of the abnormal state and the distribution location of the maintenance resources.

[0059] Step S432: Calculate the distance between the location of the abnormal state and the distribution location of the maintenance resources, and combine the road traffic condition data. Use the path planning algorithm to generate the actual travel time.

[0060] Step S433: Combine the actual travel time, set weights according to the urgent demand levels of different types of maintenance resources, and use the weighted average algorithm to generate the response priority.

[0061] Specifically, in order to more accurately analyze the relationship between the abnormal state location and the distribution location of maintenance resources, coordinate these two types of location information. Whether it is the specific location of the abnormal state on the pavement or the distribution locations of various maintenance resources (such as repair personnel stations, construction equipment storage locations, material reserve warehouses, etc.), they are all converted into coordinate forms in a unified coordinate system. The advantage of doing this is to digitize the location information, which is convenient for subsequent accurate distance calculation and path planning.

[0062] Based on the position information after coordinate transformation, calculate the distances between the positions of abnormal states and the distribution positions of various maintenance resources. However, it is not enough to only consider the straight-line distance because in the actual road environment, the driving of vehicles is also affected by the road traffic conditions. Therefore, combine real-time road traffic condition data, such as whether there is traffic congestion, road construction, traffic accidents, etc. Then use path planning algorithms, such as Dijkstra algorithm or A* algorithm, etc., to comprehensively consider the distance and road traffic conditions to generate the actual travel time from the location of the maintenance resource to the abnormal location. For example, if the road near the abnormal location is under construction, resulting in slow traffic on that section, then the actual travel time will be longer than the time calculated solely based on the distance.

[0063] Combined with the actual travel time calculated previously, set corresponding weights according to the emergency demand levels of different types of maintenance resources in dealing with abnormal states. For example, for some abnormal situations that seriously affect traffic safety, such as large areas of potholes or severe cracks, the emergency demand levels of the materials for emergency repair and professional emergency repair personnel are relatively high, and their corresponding weights will be set larger; while for some relatively less urgent maintenance resources, the weights will be set smaller. Then, use the weighted average algorithm to comprehensively calculate the actual travel time and the weights of each maintenance resource, and finally generate the response priority of each candidate maintenance measure. A candidate maintenance measure with a high response priority means that in the current actual situation, resources can be allocated more quickly and effectively to solve the road surface abnormality problem, providing a key basis for generating reasonable road surface maintenance control instructions in the future.

[0064] In a possible implementation manner, step S500 further includes:

[0065] Step S510: After the maintenance personnel perform the maintenance, collect the road surface state verification data through the multi-source sensors.

[0066] Step S520: Compare the road surface state verification data with the historical data of the abnormal state and calculate the maintenance effect score.

[0067] Step S530: If the score is lower than the preset threshold, trigger the update instruction of the maintenance strategy library and correct the maintenance measure type in the mapping relationship according to the verification result.

[0068] Specifically, after the maintenance personnel complete the maintenance operations according to the established maintenance plan, in order to evaluate the effectiveness of the maintenance work, they use the multi-source sensors deployed previously to collect the pavement state verification data after maintenance again. These sensors continue to function, collecting the image data of the pavement to check whether there are surface defects such as cracks and potholes on the pavement after maintenance; collecting the pavement humidity data to judge whether the maintenance measures have an impact on the waterproof performance of the pavement; collecting the pavement temperature data to analyze the stability of the pavement materials after maintenance under temperature changes; and collecting the stress-strain data to evaluate whether the bearing capacity of the pavement structure has returned to normal. Through these multi-source data, the actual state of the pavement after maintenance is comprehensively reflected.

[0069] To calculate the maintenance effect score, first, for the pavement image data, the feature matching algorithm in image recognition technology is used. The pavement image features after maintenance, such as key indicators like crack width and pothole area, are compared with the corresponding features in the historical image data of abnormal states to calculate the feature difference degree score. For the pavement humidity data, the difference analysis method is adopted to calculate the difference between the humidity data after maintenance and the average humidity value in the abnormal state, and according to the size of the difference, scores are given according to the pre-set humidity scoring standard. In terms of the pavement temperature data, by comparing the temperature change curves at different times before and after maintenance and using curve similarity algorithms such as the dynamic time warping algorithm, the similarity degree score of the temperature change is obtained. For the stress-strain data, a mechanical model is constructed, and the stress-strain data after maintenance is substituted into the model and compared with the model results constructed based on the historical abnormal state data, and the stress-strain score is calculated according to the difference output by the model. Finally, weights are assigned to each item of data according to its importance, and the scores of image, humidity, temperature, and stress-strain are combined by weighted summation to generate the final maintenance effect score, so as to quantitatively evaluate the actual effectiveness of the maintenance work in improving the abnormal state of the pavement.

[0070] The calculated maintenance effect score is compared with the pre-set threshold. If the score is lower than the pre-set threshold, it indicates that the current maintenance work fails to achieve the expected effect, and there may be room for optimization in the original maintenance strategy. At this time, an update instruction for the maintenance strategy library is automatically triggered. According to the comparison results between the pavement state verification data after maintenance and the historical abnormal state data, the mapping relationship between the abnormal types and maintenance measures in the maintenance strategy library is corrected. For example, if for the maintenance of a certain type of crack, the original caulking measure has a low score, and it is found through analyzing the verification data that it is due to the compatibility problem between the caulking material and the pavement material resulting in poor maintenance effect, then in the mapping relationship, the type of caulking material or the caulking process can be adjusted, and more appropriate maintenance measure options can be added, so as to continuously improve the maintenance strategy library and provide more effective reference for subsequent similar pavement abnormal maintenance work.

[0071] Embodiment 2. Based on the same inventive concept as the pavement condition collaborative intelligent analysis method in the foregoing embodiment, as Figure 2 shown, the present application provides a pavement condition collaborative intelligent analysis system. The system in the embodiment of the present application and the method embodiment are based on the same inventive concept. Among them, the system includes:

[0072] A pavement multi-source data generation module 10, configured to deploy multi-source sensors, and based on a preset acquisition period, collect real-time data of the pavement to generate pavement multi-source data.

[0073] A comprehensive pavement condition data vector generation module 20, configured to perform fusion processing on different types of data collected by multiple sensors based on a preset data fusion algorithm to generate a comprehensive pavement condition data vector.

[0074] A reminder information generation module 30, configured to analyze the comprehensive pavement condition data vector by using a machine learning model, determine whether there is an abnormal condition on the pavement, and determine the location of the abnormal condition according to the time and location information of the collected data, and generate a pavement abnormal condition reminder information.

[0075] A pavement maintenance control instruction generation module 40, configured to obtain the type and location of the pavement abnormal condition, and match the abnormal condition with a preset maintenance strategy library to generate a pavement maintenance control instruction.

[0076] A prompt information sending module 50, configured to match corresponding prompt content according to the pavement maintenance control instruction and send the prompt information to relevant management departments and maintenance personnel.

[0077] Further, the system is also used to implement the following functions:

[0078] Use the multi-source sensors to collect and obtain pavement image data, pavement humidity data, pavement temperature data, and stress and strain data; perform time synchronization and coordinate calibration processing on the pavement image data, pavement humidity data, pavement temperature data, and stress and strain data.

[0079] Further, the system is also used to implement the following functions:

[0080] Extract features from the pavement image data to obtain a pavement texture feature vector; perform normalization processing on the pavement humidity data, pavement temperature data, and stress and strain data; use a weighted fusion algorithm to fuse the processed various types of data to generate a comprehensive pavement condition data vector.

[0081] Further, the system is also used to implement the following functions:

[0082] Construct a training data set containing samples of various normal and abnormal road surface states, train a deep learning model, and obtain a trained road surface state analysis model; input the comprehensive road surface state data vector into the road surface state analysis model, output a road surface state judgment result, and determine whether there is an abnormal state; if there is an abnormal state, use a positioning algorithm and combine the position information of the sensor to determine the position of the abnormal state.

[0083] Furthermore, the system is also used to implement the following functions:

[0084] Construct the maintenance strategy library, which includes the mapping relationship of historical maintenance records, abnormal types, maintenance measures, and the distribution location of maintenance resources; according to the type of the abnormal state, match the corresponding candidate maintenance measures from the mapping relationship of the maintenance measures; combine the position of the abnormal state and the distribution location of the maintenance resources, and calculate the response priority of the candidate maintenance measures; generate the road surface maintenance control instruction based on the response priority, and the road surface maintenance control instruction includes the type of maintenance measures, execution path planning, and the list of required resources.

[0085] Furthermore, the system is also used to implement the following functions:

[0086] Coordinate the position of the abnormal state and the distribution location of the maintenance resources; calculate the distance between the position of the abnormal state and the distribution location of the maintenance resources, combine the road traffic condition data, and use a path planning algorithm to generate the actual travel time; combine the actual travel time, set weights according to the urgent demand degree of different types of maintenance resources, and use a weighted average algorithm to generate the response priority.

[0087] Furthermore, the system is also used to implement the following functions:

[0088] After the maintenance personnel perform the maintenance, collect the road surface state verification data after the maintenance through the multi-source sensor; compare the road surface state verification data with the historical data of the abnormal state, and calculate the maintenance effect score; if the score is lower than the preset threshold, trigger an update instruction for the maintenance strategy library, and correct the type of maintenance measures in the mapping relationship according to the verification result.

[0089] It should be noted that the above order of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above describes specific embodiments of this specification. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0090] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

[0091] This specification and the drawings are merely exemplary descriptions of the present application and are considered to have covered any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A road surface condition collaborative intelligent analysis method, characterized in that: The method comprises: Deploy multi-source sensors to collect real-time data on the road surface based on a preset collection cycle and generate multi-source data on the road surface; Based on the preset data fusion algorithm, different types of data collected by multiple sensors are fused and processed to generate a comprehensive road state data vector; Using a machine learning model to analyze the comprehensive road surface state data vector, determine whether there is an abnormal state of the road surface, and determine the location of the abnormal state based on the time and location information of the collected data, and generate road surface abnormal state reminder information; Obtain the type and location of the abnormal state of the road surface, match the abnormal state with a preset maintenance strategy library, and generate a road surface maintenance control instruction; According to the road maintenance control instruction, the corresponding prompt content is matched and the prompt information is sent to the relevant management department and maintenance personnel.

2. A road surface condition collaborative intelligent analysis method as claimed in claim 1, characterized in that: Deploy multi-source sensors to collect real-time data of the road surface based on a preset collection cycle, the method comprising: Utilize the multi-source sensors to collect and acquire road surface image data, road surface humidity data, road surface temperature data, and stress and strain data; The road surface image data, road surface humidity data, road surface temperature data and stress-strain data are subjected to time synchronization and coordinate calibration processing.

3. A road surface condition collaborative intelligent analysis method as claimed in claim 2, characterized in that: Based on a preset data fusion algorithm, different types of data collected by multiple sensors are fused and processed, and the method includes: Performing feature extraction on the road surface image data to obtain a road surface texture feature vector; Normalizing the road surface humidity data, road surface temperature data, and stress-strain data; The weighted fusion algorithm is used to fuse various processed data to generate a comprehensive road surface status data vector.

4. A road surface condition collaborative intelligent analysis method as claimed in claim 3, characterized in that: The comprehensive road surface state data vector is analyzed using a machine learning model, the method comprising: Construct a training data set containing a variety of normal and abnormal road state samples, train the deep learning model, and obtain a trained road state analysis model; Inputting the comprehensive road surface state data vector into the road surface state analysis model, outputting a road surface state judgment result, and judging whether there is an abnormal state; If an abnormal state exists, the location of the abnormal state is determined by using a positioning algorithm combined with the location information of the sensor.

5. A road surface condition collaborative intelligent analysis method as claimed in claim 4, characterized in that: According to the abnormal state, a preset maintenance strategy library is matched to generate a road maintenance control instruction, and the method includes: Constructing the maintenance strategy library, which includes historical maintenance records, abnormality types, mapping relationships of maintenance measures, and distribution locations of maintenance resources; According to the type of the abnormal state, matching corresponding candidate maintenance measures from the mapping relationship of the maintenance measures; Calculate the response priority of the candidate maintenance measures by combining the location of the abnormal state and the distribution location of the maintenance resources; The road surface maintenance control instruction is generated based on the response priority, and the road surface maintenance control instruction includes a maintenance measure type, an execution path planning, and a required resource list.

6. A road surface condition collaborative intelligent analysis method as claimed in claim 5, characterized in that: Combining the location of the abnormal state with the distribution location of the maintenance resources, calculating the response priority of the candidate maintenance measures, the method includes: Coordinate processing is performed on the location of the abnormal state and the distribution location of the maintenance resources; Calculate the distance between the location of the abnormal state and the location of the maintenance resource distribution, combine the road traffic condition data, and use the path planning algorithm to generate the actual travel time; In combination with the actual travel time, weights are set according to the emergency demand levels of different types of maintenance resources, and a weighted average algorithm is used to generate the response priority.

7. A road surface condition collaborative intelligent analysis method as claimed in claim 6, characterized in that: Also includes: After the maintenance personnel perform maintenance, the multi-source sensor is used to collect the road surface status verification data after maintenance; Comparing the road surface status verification data with the historical data of the abnormal status, and calculating a maintenance effect score; If the score is lower than a preset threshold, an update instruction of the maintenance strategy library is triggered, and the maintenance measure type in the mapping relationship is corrected according to the verification result.

8. A road surface condition collaborative intelligent analysis system, characterized in that: The system is used to implement a road surface state collaborative intelligent analysis method according to any one of claims 1 to 7, and the system comprises: The road surface multi-source data generation module is used to deploy multi-source sensors, collect real-time road surface data based on a preset collection cycle, and generate road surface multi-source data; A comprehensive road surface state data vector generation module is used to fuse different types of data collected by multiple sensors based on a preset data fusion algorithm to generate a comprehensive road surface state data vector; A reminder information generation module is used to analyze the comprehensive road surface state data vector using a machine learning model to determine whether there is an abnormal state of the road surface, and determine the location of the abnormal state according to the time and location information of the collected data, and generate road surface abnormal state reminder information; A road maintenance control instruction generation module is used to obtain the type and location of the abnormal state of the road surface, match the abnormal state with a preset maintenance strategy library, and generate a road maintenance control instruction; The prompt information sending module is used to match the corresponding prompt content according to the road maintenance control instruction and send the prompt information to the relevant management department and maintenance personnel.

Images