Routing inspection method and system for road operation and inspection integration

By collecting and integrating the processing of road information on floating vehicles in real time and processing data centers, using Bayesian and deep learning algorithms to improve detection accuracy, solving the problems of high cost and low efficiency of road detection in the existing technology, and achieving high frequency and accuracy The road monitoring improves the efficiency and environmental protection of road management.

CN119066616APending Publication Date: 2024-12-03SHANGHAI RESEARCH INSTITUTE OF BUILDING SCIENCES CO LTD +1
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411203972.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-30
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing road detection methods are costly, rely on professionals, and are difficult to meet daily operation and maintenance needs. The efficiency and accuracy are insufficient, making it difficult to achieve timely and accurate monitoring of roads.

Method used

Design a patrol method for road operation and inspection. By collecting road information in real time on floating vehicles, pre-processing and transmission using vehicle-mounted data terminals, the data center performs multi-vehicle and multi-type data fusion processing, adopts Bayesian Fusion algorithms and deep learning models improve detection accuracy and formulate road maintenance plans.

Benefits of technology

It reduces the detection cost, improves the detection frequency and accuracy, realizes timely identification and treatment of road diseases, improves the safety and durability of roads, reduces carbon emissions, and promotes green and low-carbon development.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066616A_ABST
    Figure CN119066616A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of road disease detection, in particular to a routing inspection method and system for road operation and inspection integration. The method comprises the following steps: S1, collecting road condition information in real time in the daily running process of the floating car; s2, processing the collected information data at a vehicle-mounted data terminal, and transmitting the information data to a data center in a wired or wireless transmission mode; s3, performing fusion processing on the multi-vehicle and multi-type data in the data center to obtain structured road condition data, the fusion processing including a method of performing fusion processing on the low-precision and high-frequency positioning data to improve the positioning precision; and S4, formulating a road maintenance scheme according to the acquired road technical condition data. The method has the advantages that the detection cost is reduced, the detection coverage range is expanded, high-precision road detection is realized, and the discovery and processing speed of road disease problems is increased.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of road detection, and in particular to a patrol inspection method and system for integrating road operation and inspection. Background Art

[0002] With the continuous optimization of my country's transportation network, the construction strategy of transportation infrastructure has gradually shifted from a single construction stage to a new model of parallel construction and maintenance. In this transformation, road detection, as a key link in road maintenance, has become increasingly important. Faced with the growing traffic volume, the frequency of road damage has increased. Therefore, it is crucial for road management departments to be able to grasp the real-time technical status of roads in a timely and accurate manner. This not only helps them respond quickly, take necessary maintenance and repair measures, and prevent the spread of damage, but also effectively improves the driving comfort of the road and extends its service life.

[0003] Currently, road inspection is mainly carried out in two ways: professional inspection vehicles and manual inspections. Professional inspection vehicles are equipped with advanced high-precision sensors that can regularly collect detailed data on road conditions. Although this method can quickly obtain structured road information, its high cost, high requirements for equipment accuracy, and reliance on professional operators have led to an economic burden for inspection. In addition, due to the limitations of inspection frequency and cycle, this method is difficult to meet the daily operation and maintenance needs of different levels of roads in my country.

[0004] On the other hand, manual inspections rely on on-site inspections by professionals to record road damage. Although this method can provide intuitive damage information, it is inefficient, has limited coverage, and the accuracy, authenticity, and integrity of the data are often questioned, which to some extent limits the refinement and systematization of road maintenance work. Summary of the invention

[0005] The purpose of the present invention is to overcome the deficiencies of the prior art and to provide a patrol method and system for integrated road operation and inspection, so as to reduce the additional labor and cost for inspection, significantly increase the inspection frequency, and further improve the inspection accuracy.

[0006] In order to achieve the above-mentioned purpose, a patrol method for integrated road operation and inspection is designed, and the method steps are as follows: S1. Through sensors, the technical condition information of the road is collected in real time during the daily operation of the floating vehicle; S2. The collected information data is processed on the on-board data terminal, and transmitted to the data center by wired or wireless transmission; S3. Multi-vehicle and multi-type data are fused in the data center to obtain structured road condition data, and the fusion processing includes a method for fusing low-precision and high-frequency positioning data to improve positioning accuracy, which is as follows: S31. Set the preset radius of the hexagon, divide the map into cells, and obtain the longitude and latitude coordinates and number of the center point of each cell; S32. Calculate the number of measured GPS data points falling within the hexagon, draw a cellular heat map, and sort them according to the total number of data points in each cell; S33. Set The segmentation threshold is α%, and the cells with the top α% of points are taken as the target area; S34. The GPS data points falling in the target area are retained, and the data falling outside the target area are eliminated, thereby obtaining the positioning data after noise removal; the fusion processing also includes a method for improving the evaluation accuracy using multi-frequency detection results, as follows: S35. According to the method for improving positioning accuracy, data located outside the road are eliminated; S36. A preset section length is set, and the road is divided according to the section length; S37. Data statistics are performed on the evaluation results of different floating vehicles in the same section; S38. For different recognition results in the same section, the Bayesian fusion algorithm is used to calculate the posterior probability of each evaluation result, and the result with the largest posterior probability is selected as the final evaluation result of the road condition of this section; S4. A road maintenance and repair plan is formulated based on the collected road condition data.

[0007] The inspection method for integrated road operation and inspection provided by the present invention also has other technical features, wherein the inspection method is used for road surface disease investigation, road flatness detection, road covering detection or road ancillary facilities inventory and evaluation.

[0008] The inspection method for integrated road operation and inspection provided by the present invention also has other technical features, wherein the floating vehicle is a bus, a taxi, an online car-hailing vehicle or a logistics vehicle.

[0009] The inspection method for integrated road operation and inspection provided by the present invention also has other technical features, wherein step S1 includes data collected by the sensor, collection time and real-time location information.

[0010] The inspection method for integrated road operation and inspection provided by the present invention also has other technical features, wherein the vehicle-mounted data terminal in step S2 is used to preprocess the collected data and store historical data, the data preprocessing includes data cleaning, time and space alignment, and automatic recognition of video image data, and the data center is used to store, manage, analyze and publish the collected data.

[0011] The inspection method for integrated road operation and inspection provided by the present invention also has other technical features, among which the fusion processing in step S3 includes data fusion of sensors on the same vehicle and spatiotemporal fusion of sensors on different vehicles, calculation of road condition evaluation indicators, improvement of the accuracy of detection results, and spatiotemporal correlation of detection data.

[0012] The inspection method for integrated road operation and inspection provided by the present invention has other technical features, wherein the calculation of the evaluation index of the road condition in step S3 includes: a. For the investigation of road surface defects, the road pavement image data obtained by the sensor is used to identify the type, size and location of the defects using a deep learning model, and the algorithms used include the YOLO series model, Faster R-CNN model, Mask After obtaining the above-mentioned disease information, the R-CNN model calculates the road technical condition evaluation index PCI according to the "Highway Technical Condition Assessment Standard" (JTG5210) or the "Technical Specifications for Urban Road Maintenance" (CJJ36); b. For road flatness detection, the road technical condition indicators IRI and RQI are calculated using the flatness analysis algorithm based on the vehicle vibration response, vertical acceleration, and road surface elevation information obtained by the sensor; c. For road cover detection, the road technical condition index friction coefficient is calculated using the road friction coefficient analysis algorithm based on the cover type and thickness information obtained by the sensor; d. For the inventory and assessment of road ancillary facilities, the road ancillary facilities image data obtained by the sensor is used to identify the type, location, and damage type of the ancillary facilities using a deep learning model. The algorithms used include the YOLO series models, the Faster R-CNN model, and the Mask R-CNN model.

[0013] The inspection method for integrated road transportation and inspection provided by the present invention has other technical features, wherein the calculation steps of the Bayesian fusion algorithm in step S3 are: S4.1. Define the problem and symbols; S4.1.1. Input data: X: recognition results of different train numbers, denoted as X 1 ,X 2 ,…,X n , each recognition result contains the number and type combination of diseases, denoted as X i ={(quantity i, type i)}; S4.1.2. Assumption: The correct disease identification result H includes the correct combination of disease quantity and type; S4.2 Calculate the likelihood function P(X i |H); S4.2.1. Assumption: According to the calculation results of the disease identification algorithm, determine the accuracy of the algorithm identification β Here, the correct identification of the algorithm is defined as the number and type of the identified diseases are accurate. If there is an error in the number or type, it is an identification error. From this, the probability of correct identification of each vehicle can be obtained as β , the probability of misidentification is (1- β ); S4.2.2. Likelihood function: For the hypothesis H j (hypothetical disease combination), i Vehicle recognition results X i The likelihood of is set to: ; S4.3. Set the prior probability P(H j ): In the absence of other information, the initial assumption is that the prior probabilities of all possible combinations of disease quantity and type are equal, that is: , where M is the number of all possible disease combinations; if there is other prior knowledge (for example, some disease combinations are more common), adjust the prior probability based on this knowledge; S4.4. Calculate the posterior probability P(H j |X): For every possible disease combination H j , its posterior probability is: , where the total likelihood function ,in X is the set of all train identification results, X= { X 1 , X 2 ,…, X n}, the posterior probability is: , where k is the combination of all possible diseases traversed; S4.5. Select the hypothesis with the maximum posterior probability: Finally, the disease combination with the largest posterior probability is selected H j As the final recognition result of this segment: The mean of the longitude and latitude coordinates of each disease location is taken as the final coordinates of each disease.

[0014] The inspection method for integrated road operation and inspection provided by the present invention also has other technical features. In the maintenance and repair plan of step S4, road maintenance decisions are made based on the technical condition data of the road throughout its life cycle obtained by the floating vehicle through machine learning and deep learning analysis algorithms. The analysis algorithms used include decision trees, support vector machines, multi-layer perceptrons, and reinforcement learning.

[0015] The present invention also provides an inspection system for integrated road operation and inspection that adopts the above-mentioned inspection method, including: a detection equipment module, including sensors that meet different detection needs, installed on a floating vehicle, and used to collect road condition information; an on-board data terminal, responsible for preprocessing and storing and updating historical data of sensor data; a data transmission module, responsible for transmitting the collected data to a data center; a data center, including a data receiving unit, a data storage unit, a data processing unit, and a data output unit, responsible for storing, managing, analyzing, and publishing the collected data.

[0016] Compared with the prior art, the present invention has the following advantages: 1. Cost-effectiveness: The present invention eliminates the reliance on professional inspectors by utilizing existing floating vehicles for data collection, thereby significantly reducing inspection costs.

[0017] 2. Detection frequency: Since the driving route and frequency of floating vehicles are relatively fixed, the present invention can achieve high-frequency detection of road information, effectively improving the coverage of detection.

[0018] 3. Detection accuracy: By collecting data from the same road section multiple times at different time points, the present invention uses data fusion technology to achieve high-precision road detection even when sensor performance is limited, reducing the reliance on single detection accuracy.

[0019] 4. Maintenance efficiency: The high-frequency detection capability enables the present invention to identify road diseases and related facility problems in a timely manner, thereby speeding up the discovery and handling of problems and ensuring the safety, comfort and durability of the road.

[0020] 5. Environmentally friendly: The detection method adopted by the present invention avoids the on-site operation of professional detection vehicles, reduces fuel consumption and carbon emissions, and promotes the green and low-carbon development of the road detection industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 , is a flow chart of the method of the present invention; Figure 2 , is a system structure diagram of the present invention; Figure 3 , is the original GPS distribution map of Example 1; Figure 4 , is a GPS cellular heat map of embodiment 1; Figure 5 , is a diagram of GPS positioning data denoising results in Example 1. DETAILED DESCRIPTION

[0022] In order to make the purpose, principle and structure of the present invention more clear, it is further described below in conjunction with the drawings and specific embodiments.

[0023] <Example 1> like Figure 1 As shown, the present invention provides a patrol inspection method for integrated road operation and inspection, and the specific process is as follows: Step 1: Install relevant sensors on the floating vehicle according to different detection needs or use existing sensors on the floating vehicle to collect technical condition information of the road in real time during the daily operation of the floating vehicle.

[0024] Step 2: Process the collected data on the vehicle data terminal and transmit it to the data center via wired or wireless transmission.

[0025] Step 3: In the data center, multi-vehicle and multi-type data are integrated to obtain structured road condition data.

[0026] Step 4: Develop a road maintenance plan based on the collected road condition data.

[0027] The inspection requirements in Step 1 include but are not limited to road surface disease investigation, road flatness inspection, road cover inspection, inventory and evaluation of road ancillary facilities, etc.

[0028] For the floating vehicles in Step 1, they include but are not limited to buses, taxis, online ride-hailing vehicles, logistics vehicles, etc.

[0029] For the sensors in Step 1, the corresponding relationship between them and the detection requirements is shown in Table 1. In the actual implementation process, any sensor in the table can be used, or a combination of multiple sensors can be used, or any other sensor that can meet the corresponding detection requirements can be used.

[0030] Table 1 - Correspondence between detection requirements and sensors Testing requirements sensor Survey on apparent road damage Driving recorder, HD camera, etc. Road roughness detection Accelerometer, vibration sensor, six-component force meter, etc. Road Covering Detection Road condition sensors, water accumulation sensors, ice / snow sensors, etc. Inventory and evaluation of road ancillary facilities Driving recorder, HD camera, etc. For the technical status information in Step 1, in addition to the data collected by the sensor, it should also include information such as collection time and real-time location.

[0031] The vehicle-mounted data terminal in Step 2 is used to pre-process the collected data and store historical data. Data pre-processing includes data cleaning, spatiotemporal registration, automatic recognition of video image data, etc. The vehicle-mounted data terminal should be able to store at least seven days of historical data.

[0032] The data center in Step 2 is used to store, manage, analyze and publish the collected data.

[0033] For the data transmission method in Step 2, it is recommended to transmit the pre-processed data to the data center by wireless transmission, and for the raw data of the sensor, it is recommended to transmit it by wired copy according to actual needs.

[0034] The fusion processing in Step 3 includes but is not limited to the data fusion of sensors on the same vehicle and the spatiotemporal fusion of sensors on different vehicles. The functions implemented include but are not limited to: calculation of road condition evaluation indicators, improvement of the accuracy of detection results, spatiotemporal correlation of detection data, etc.

[0035] For the calculation of road condition evaluation index in Step 3, for the investigation of road surface defects, the road pavement image data obtained by relevant sensors can be used to identify the type, size and location of the defects using a deep learning model. The algorithms that can be used include but are not limited to the YOLO series models, Faster R-CNN model, Mask R-CNN model, etc. After obtaining the above-mentioned defect information, the road technical condition evaluation index PCI can be calculated according to the "Highway Technical Condition Assessment Standard" (JTG5210) or the "Technical Specifications for Urban Road Maintenance" (CJJ36).

[0036] For the calculation of road condition evaluation indexes in Step 3, for road flatness detection, the road condition indexes IRI, RQI, etc. can be calculated using a flatness analysis algorithm based on the vehicle vibration response, vertical acceleration, road surface elevation and other information obtained by relevant sensors.

[0037] For the calculation of the road condition evaluation index in Step 3, for the detection of road surface covering, the road condition index friction coefficient can be calculated using the road surface friction coefficient analytical algorithm based on the covering type, thickness and other information obtained by relevant sensors.

[0038] For the calculation of road condition evaluation index in Step 3, for the inventory and assessment of road ancillary facilities, the image data of road ancillary facilities obtained by relevant sensors can be used to identify the type, location, and damage type of the ancillary facilities using a deep learning model. The algorithms that can be used include but are not limited to the YOLO series models, Faster R-CNN model, Mask R-CNN model, etc.

[0039] The accuracy improvement of the detection results in Step 3 is mainly divided into positioning accuracy improvement and evaluation accuracy improvement.

[0040] The reason for the improvement of positioning accuracy in Step 3 is that in the integrated operation mode of transportation and inspection, the inspection operation is mostly carried out by reusing the equipment on the floating vehicle. The GPS positioning accuracy is relatively low compared to the equipment with RTK correction, which may be only 5-10 meters. The positioning accuracy is even lower in areas with high-rise buildings or high-rise buildings. Therefore, it is necessary to fuse the low-precision and high-frequency positioning data to improve the positioning accuracy. The accuracy is improved based on the cellular heat map. The specific optimization steps are as follows: (1) Set the preset radius of the hexagon, divide the map into cellular (hexagon), and obtain the longitude and latitude coordinates and number of the center point of each cellular (hexagon); (2) Calculate the number of measured GPS data points falling within the hexagon, draw a cellular heat map, and sort them according to the total number of data points in each cellular (hexagon); (3) Set the segmentation threshold α%, and take the α% of cells (hexagons) with the top α% of points as the target area; (4) Keep the GPS data points that fall within the target area, and remove the data (GPS data and the evaluation index data associated with it) that fall outside the target area, so as to obtain the positioning data after noise removal.

[0041] The improvement of evaluation accuracy in Step 4 is mainly aimed at the investigation of road surface defects. Due to the resolution limitation of the video acquisition equipment of the floating vehicle and the interference of factors such as lighting and occlusion during vehicle driving, the accuracy of defect recognition is low. Therefore, it is necessary to use multi-frequency detection results to improve the recognition accuracy. The optimization steps for improving accuracy are: (1) Eliminate data located outside the road according to the method of improving positioning accuracy; (2) Set the preset segment length and divide the road according to the segment length; (3) Perform data statistics on the recognition results of different floating vehicles in the same segment; (4) For different recognition results in the same segment, use the Bayesian fusion algorithm to calculate the posterior probability of each recognition result, and select the result with the largest posterior probability as the final recognition result of the segment disease.

[0042] The calculation steps of the Bayesian fusion algorithm in Step 5 are as follows: S4.1. Define the problem and notation; S4.1.1. Input data: X: represents the recognition results of different train numbers, denoted by X 1 ,X 2 ,…,X n , each recognition result contains the number and type combination of diseases, denoted as X i ={(quantity i, type i )}; S4.1.2. Assumption: The correct disease identification result is H, which includes the correct combination of disease quantity and type; S4.2 Calculate the likelihood function P(X i∣H); S4.2.1. Assumption: Based on the calculation results of the disease identification algorithm, determine the accuracy of the algorithm identification β Here, the correct identification of the algorithm is defined as the number and type of the identified diseases are accurate. If there is an error in the number or type, it is an identification error. Therefore, the probability of correct identification of each vehicle can be obtained as β , the probability of misidentification is (1- β ); S4.2.2. Likelihood function: For the hypothesis H j (hypothetical disease combination), i Vehicle recognition results X i The likelihood of is set to: ; S4.3. Set the prior probability P(H j ): In the absence of other information, the initial assumption is that the prior probabilities of all possible combinations of disease quantity and type are equal, that is: , where M is the number of all possible disease combinations; if there is other prior knowledge (for example, some disease combinations are more common), adjust the prior probability based on this knowledge; S4.4. Calculate the posterior probability P(H j |X): For every possible disease combination H j , its posterior probability is: , where the total likelihood function ,in X is the set of all train identification results, X= { X 1 , X 2 ,…, X n}, the posterior probability is: , where k is a traversal of all possible disease combinations; S4.5. Select the hypothesis with the largest posterior probability: Finally, select the disease combination with the largest posterior probability H j As the final recognition result of this segment: The mean of the longitude and latitude coordinates of each disease location is taken as the final coordinates of each disease.

[0043] For the maintenance decision in Step 4, the road maintenance decision is made based on the technical condition data of the road throughout its life cycle obtained by the floating vehicle through machine learning, deep learning and other analysis algorithms. The analysis algorithms used include but are not limited to: decision tree, support vector machine, multi-layer perceptron, reinforcement learning, etc.

[0044] (1) Example of improving positioning accuracy: A plurality of buses are used as floating vehicles to collect data for 14 days as an example.

[0045] The scatter plot of GPS data collected from multiple buses over 14 days is shown in Figure 3 Take the preset radius as 10 meters and draw the cellular heat map. Figure 4 The segmentation threshold α% is set to 5%, and the denoised GPS positioning data is obtained, as shown in Figure 5 Indicated by the white line in the middle.

[0046] (2) Example of improving recognition accuracy: Calculated through actual training data, the accuracy of algorithm recognition β= In the embodiment, 20 floating vehicles perform integrated operation of transportation and inspection on the same section with a length of 100 meters. The recognition results of 20 times are shown in Table 2.

[0047] Table 2--20 floating car apparent disease identification results Train Identification results 1 2 potholes, 1 crack 2 3 potholes 3 2 potholes 4 2 potholes, 1 crack 5 2 potholes 6 2 potholes, 1 crack 7 2 potholes, 1 crack 8 3 potholes 9 2 potholes, 1 crack 10 2 potholes 11 2 potholes, 1 crack 12 3 potholes 13 2 potholes, 1 crack 14 2 potholes, 1 crack 15 2 potholes 16 2 potholes, 1 crack 17 2 potholes, 1 crack 18 3 potholes 19 2 potholes, 1 cable break 20 2 potholes, 1 crack It can be concluded that the recognition results of the 12 vehicles meet the assumption H 1 (2 potholes, 1 crack), the recognition results of 4 vehicles meet the hypothesis H 2 (3 potholes), the recognition results of 4 vehicles meet the hypothesis H 3 (2 potholes).

[0048] The overall calculation process is as follows: 1. Calculate the likelihood function P ( X i ∣ H ): Given the probability of correct recognition β = 0.6, the probability of error is 1− β =0.4.

[0049] 2. Set prior probability P ( H j ): Assume that the prior probabilities of all possible disease combinations are equal: P( H 1 )=P( H 2 )=P( H 3 )=1 / 3.

[0050] 3. Calculate the total likelihood function for each hypothesis P ( Xi ∣ H j ): Assumptions H 1 : 2 potholes, 1 crack: The recognition results of 12 vehicles are consistent with H1, and the recognition results of 8 vehicles are inconsistent with H1: .

[0051] Assumptions H 2 : 3 potholes: The recognition results of 4 vehicles are consistent with H2, and the recognition results of 16 vehicles are inconsistent with H2: .

[0052] Assumptions H 3 : 2 potholes: The recognition results of 4 vehicles are consistent with H3, and the recognition results of 16 vehicles are inconsistent with H3: .

[0053] 4. Calculate the posterior probability P ( H j ∣ X ): Use Bayes' formula to calculate the posterior probability of each hypothesis: ; From this we can get the posterior probability of each hypothesis: for ,for ,for .

[0054] 5. Choose the hypothesis with the largest a posteriori probability: According to the calculation results, H 1 The posterior probability of is the largest, which is 0.893, so 2 potholes and 1 crack are finally selected as the final evaluation results.

[0055] <Example 2> A patrol inspection system for integrated road operation and inspection, comprising: Detection equipment module: includes sensors that meet different detection needs, installed on the floating vehicle, and used to collect road condition information.

[0056] On-board data terminal: responsible for the pre-processing of sensor data and the storage and update of historical data.

[0057] Data transmission module: responsible for transmitting the collected data to the data center, which can adopt wireless network, mobile communication network or other suitable transmission methods.

[0058] Data Center: It includes data receiving unit, data storage unit, data processing unit and data output unit, and is responsible for storing, managing, analyzing and publishing the collected data.

[0059] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent substitutions or changes according to the technical solutions and novel concepts of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A patrol inspection method for integrated road operation and inspection. Features The method steps are as follows: S1. Through sensors, the technical condition information of the road is collected in real time during the daily operation of the floating vehicle; S2. Process the collected information data on the vehicle data terminal and transmit it to the data center via wired or wireless transmission; S3. In the data center, multi-vehicle and multi-type data are integrated to obtain structured road condition data. The fusion processing includes a method for fusing low-precision and high-frequency positioning data to improve positioning accuracy, which is specifically as follows: S31. Set the preset radius of the hexagon, divide the map into cells, and obtain the latitude and longitude coordinates and number of the center point of each cell; S32. Calculate the number of measured GPS data points that fall within the hexagon, draw a cellular heat map, and sort them according to the total number of data points in each cell; S33 sets the segmentation threshold α%, and takes the top α% of the cells as the target area; S34. Keep the GPS data points that fall in the target area, and remove the data that fall outside the target area, thereby obtaining the positioning data after removing the noise; The fusion processing also includes a method for improving the evaluation accuracy by using the detection results of multiple frequencies, as follows: S35. Eliminate data located outside the road according to a method for improving positioning accuracy; S36. Set a preset section length and divide the road according to the section length; S37. Performing statistical analysis on the evaluation results of different floating vehicles in the same section; S38. For different evaluation results in the same section, the Bayesian fusion algorithm is used to calculate the posterior probability of each recognition result, and the result with the largest posterior probability is selected as the final evaluation result of the road condition of this section; S4. Develop a road maintenance and repair plan based on the collected road condition data.

2. A road inspection method for integrated road operation and inspection as claimed in claim 1, Features The inspection method is used for road surface disease investigation, road flatness detection, road covering detection or road ancillary facilities inventory and evaluation.

3. A road inspection method for integrated road operation and inspection as claimed in claim 1, Features The floating vehicle is a bus, a taxi, an online car-hailing vehicle or a logistics vehicle.

4. A road inspection method for integrated road operation and inspection as claimed in claim 1, Features The step S1 includes the data collected by the sensor, the collection time and the real-time location information.

5. A road inspection method for integrated road operation and inspection as claimed in claim 1, Features The vehicle-mounted data terminal in step S2 is used to preprocess the collected data and store historical data. The data preprocessing includes data cleaning, spatiotemporal alignment, and automatic recognition of video image data. The data center is used to store, manage, analyze and publish the collected data.

6. A road inspection method for integrated road operation and inspection as claimed in claim 1, Features The fusion processing in step S3 includes data fusion of sensors on the same vehicle and spatiotemporal fusion of sensors on different vehicles, calculation of road condition evaluation indicators, improvement of the accuracy of detection results, and spatiotemporal correlation of detection data.

7. A road inspection method for integrated road operation and inspection as claimed in claim 1, Features The calculation of the road conditions and their evaluation indexes in step S3 includes: a. For the investigation of road surface defects, the road pavement image data obtained by sensors is used to identify the type, size and location of defects using a deep learning model. The algorithms used include the YOLO series model, Faster R-CNN model and Mask R-CNN model. After obtaining the above defect information, the PCI of the road technical condition evaluation index is calculated according to the Highway Technical Condition Assessment Standard (JTG5210) or the Technical Specification for Urban Road Maintenance (CJJ36); b. For road roughness detection, the road technical condition indicators IRI and RQI are calculated using a roughness analysis algorithm based on the vehicle vibration response, vertical acceleration, and road surface elevation information obtained by the sensor; c. For road surface cover detection, the friction coefficient of the road technical condition indicator is calculated using the road surface friction coefficient analytical algorithm based on the cover type and thickness information obtained by the sensor; d. For the inventory and assessment of road ancillary facilities, the image data of road ancillary facilities obtained by sensors is used to identify the type, location, and damage type of the ancillary facilities using a deep learning model. The algorithms used include the YOLO series model, Faster R-CNN model, and Mask R-CNN model.

8. A road inspection method for integrated road operation and inspection as claimed in claim 1, It is characterized in that The calculation steps of the Bayesian fusion algorithm in step S3 are: S4.

1. Define the problem and notation; S4.1.

1. Input data: X: represents the recognition results of different train numbers, denoted by X 1 ,X 2 ,…,X n , each recognition result contains the number and type combination of diseases, denoted as X i ={(quantity i, type i )}; S4.1.

2. Assumption: The correct disease identification result is H, which includes the correct combination of disease quantity and type; S4.2 Calculate the likelihood function P(X i ∣H); S4.2.

1. Assumption: Based on the calculation results of the disease identification algorithm, determine the accuracy of the algorithm identification β Here, the correct identification of the algorithm is defined as the number and type of the identified diseases are accurate. If there is an error in the number or type, it is an identification error. Therefore, the probability of correct identification of each vehicle can be obtained as β , the probability of misidentification is (1- β ); S4.2.

2. Likelihood function: For the hypothesis H j (hypothetical disease combination), i Vehicle recognition results X i The likelihood of is set to: ; S4.

3. Set the prior probability P(H j ): In the absence of other information, the initial assumption is that the prior probabilities of all possible combinations of disease quantity and type are equal, that is: , where M is the number of all possible disease combinations; if there is other prior knowledge, adjust the prior probability according to this knowledge; S4.

4. Calculate the posterior probability P(H j |X): For every possible disease combination H j , and its posterior probability is: , where the total likelihood function ,in X is the set of all train identification results, X= { X 1 , X 2 ,…, X n }, the posterior probability is: , where k is a traversal of all possible disease combinations; S4.

5. Select the hypothesis with the largest posterior probability: Finally, select the disease combination with the largest posterior probability H j As the final recognition result of this segment: The mean of the longitude and latitude coordinates of each disease location is taken as the final coordinates of each disease.

9. A road inspection method for integrated road operation and inspection as claimed in claim 1, Features In the maintenance and repair plan of step S4, road maintenance decisions are made based on the technical condition data of the road throughout its life cycle obtained by the floating vehicle through machine learning and deep learning analysis algorithms. The analysis algorithms used include decision trees, support vector machines, multi-layer perceptrons, and reinforcement learning.

10. An inspection system for integrated road operation and inspection using the inspection method according to any one of claims 1 to 9, Features include: The detection equipment module, including sensors that meet different detection requirements, is installed on the floating vehicle to collect road condition information; The vehicle-mounted data terminal is responsible for preprocessing sensor data and storing and updating historical data; The data transmission module is responsible for transmitting the collected data to the data center; The data center, including a data receiving unit, a data storage unit, a data processing unit and a data output unit, is responsible for storing, managing, analyzing and publishing the collected data.

Citation Information

Cited By

  • Intelligent road inspection method and device based on multi-source perception

    CN121475309A