Real-time road safety assessment method based on driving vehicle vibration sensing, medium and device

Abstract

The application provides a real-time road safety evaluation method based on vibration sensing of a driving vehicle, a medium and equipment, first, the road to be monitored is divided into regions according to a specified size, and the road to be monitored is divided into intervals according to region attribution and road type; then a neural network regression model is established, which is used for predicting vibration difference values according to road type, temperature data, weather conditions and road surface obstacles; finally, the vibration sensing values generated by the automobile in the driving process are compared with the historical standard vibration sensing values of the automobile after a series of standardization treatments, to help judge whether abnormal conditions such as cracks, depressions and changes in the structure below the road occur. The application can effectively reduce the cost of artificial detection, discover risks in time, take countermeasures and avoid the occurrence of road accidents.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of road traffic safety, and particularly relates to a real-time road safety evaluation method based on driving vehicle vibration sensing, a medium and equipment. BACKGROUND

[0002] With the continuous expansion of modern transportation system and large-scale vehicle use, road safety and maintenance have become an important part of public management. Road damage and potential collapse risk not only threaten the personal safety of drivers and passengers, but also can cause serious traffic delays and economic losses. Traditional road detection methods usually rely on regular manual patrol and expensive professional equipment, which not only consumes time and effort, but also often fails to monitor road conditions in real time, making it difficult to discover potential safety hazards in time and take early warning measures to avoid accidents. SUMMARY

[0003] The present application aims at the deficiencies in the prior art, and provides a real-time road safety evaluation method based on driving vehicle vibration sensing, a medium and equipment.

[0004] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0005] In a first aspect, the present application provides a real-time road safety evaluation method based on driving vehicle vibration sensing, characterized in that it comprises:

[0006] S1: dividing the road to be monitored into areas according to a specified size, the divided areas being referred to as points, and recording the latitude and longitude range contained in each point; dividing the road to be monitored into intervals according to area ownership and road type, and recording the point information contained in each interval and the road type to which it belongs;

[0007] S2: in the normal interval road, real-time collecting latitude, longitude, vibration sensing value and temperature data of a plurality of vehicles traveling at a specified range of speed, real-time shooting photos of the road conditions in front of the vehicle, and identifying weather conditions and road obstacles according to the photos of the road conditions in front of the vehicle;

[0008] S3: taking each latitude and longitude and its corresponding road type, temperature data, weather conditions and road obstacles as a group of input data, inputting into a neural network regression model for training, and the label of the neural network regression model being the absolute value of the difference between the average value of the historical vibration sensing value of the latitude and longitude corresponding point and the measured vibration sensing value, denoted as vibration difference value ΔF;

[0009] S4: Collect the time, latitude and longitude GPS, vibration sensor value F, temperature data T, front road condition photo and vehicle speed uploaded by the vehicle driving on the road to be evaluated in real time, exclude the uploaded data whose vehicle speed is not within the specified range, and after the vehicle finishes running an interval, correspond the latitude and longitude uploaded by the vehicle in the interval to the point, and determine the data corresponding to the point;

[0010] S5: Input the data corresponding to each point into the trained neural network regression model, and predict the vibration difference value AF to supplement the vibration sensor value F to obtain the standard vibration sensor value F';

[0011] S6: Compare the standard vibration sensor value F' with the N historical vibration sensor values of the point to determine whether the vibration sensor value of the point is abnormal, and handle the abnormality.

[0012] Optionally, in step S1, a unique id is used to identify the point, and the point set is represented as Pid=(pid1, pid2,..., pid k ), k is the number of points; a unique id is used to identify the interval, and the interval set is represented as Aid=(aid1, aid2,..., aid l ), l is the number of intervals; record the road type Rid=(rid1, rid2,..., rid n ), n is the number of road types; record the belonging relationship between Pid and Aid.

[0013] Optionally, in step S2, the vehicle is equipped with a vibration sensor, a temperature sensor and a front camera, which are used to collect vibration sensor values, temperature data and vehicle front road condition photos, respectively.

[0014] Optionally, in step S3, the weather condition and road surface obstacle are one-hot coded before being input into the neural network regression model; according to each latitude and longitude, the corresponding road type is found, and each road type, temperature data, weather code and road surface obstacle code are taken as a group of input data and input into the neural network regression model for training.

[0015] Optionally, in step S4, after the vehicle finishes running an interval, all the uploaded data of the vehicle in the interval are extracted, the time, latitude and longitude, vibration sensor value, temperature data, vehicle speed, weather code and road surface obstacle code uploaded per second are taken as a group of data set, the valid data set satisfying the vehicle speed requirement is screened, the latitude and longitude are extracted from the valid data set, the corresponding road type and point are found according to the latitude and longitude, and then the group of data set belongs to the data set in the point; if there are multiple groups of data sets in the point, the average value of each type of data is taken to represent the data set corresponding to the point.

[0016] Optionally, in step S5, the vibration sensing values ​​F = (F1, F2, ..., F) are extracted from all valid datasets uploaded by the car within the interval that day. m And the corresponding point Pid′=(pid1, pid2, ..., pid) m ), where m is the number of data points; simultaneously, extract the N most recent Pid′ corresponding standard vibration sensor values ​​from the vehicle's historical data. i∈[1,N]; Input the road type, temperature data, weather code, and road obstacle code corresponding to Pid′ for that day into the trained neural network regression model to predict a set of vibration difference values ​​ΔF=(ΔF1, ΔF2, ..., ΔF m This set of ΔF values ​​is used as supplementary values ​​to the vibration sensing value F for the day. These are then added together to obtain the standard vibration sensing value F′ = (F1 + ΔF1, F2 + ΔF2, ..., F′) after eliminating the influence of external environmental factors. m +ΔF m )=(F′1, F′2,...,F′ m ).

[0017] Optionally, in step S6, the isolated forest algorithm is used to analyze F′ and F′. i Data analysis is performed to determine if any vibration sensor values ​​at any location are abnormal on that day; if abnormal locations are found, they are marked; if multiple vehicles show abnormalities at a certain location, an alarm is issued for that location.

[0018] Optionally, after step S6, the effective dataset obtained that day is used as the training set and then input into the neural network regression model for training again.

[0019] In a second aspect, the present invention provides a computer-readable storage medium storing a computer program, characterized in that the computer program causes a computer to execute the real-time road safety assessment method based on vehicle vibration sensing as described in the first aspect.

[0020] Thirdly, the present invention provides an electronic device, characterized in that it includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the real-time road safety assessment method based on vehicle vibration sensing as described in the first aspect.

[0021] The beneficial effects of the present application are: the present application utilizes the vibration sensing value monitored during the driving process of the vehicle, eliminates the influence of the automobile itself and external environmental factors on the automobile vibration sensing value by standardizing the vibration sensing value, and evaluates the structural integrity and potential collapse risk of the road according to the comparison between the real-time standard vibration sensing value and the past standard vibration sensing value. The present application has the advantages of strong real-time performance, low labor cost, wide coverage, etc., and can effectively predict the occurrence of road accidents, facilitate timely response, and improve the safety of road use. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 It is a data processing flowchart of a real-time road safety evaluation method based on vehicle vibration sensing during driving.

[0023] Figure 2 It is a whole method flowchart of a real-time road safety evaluation method based on vehicle vibration sensing during driving.

[0024] Figure 3 It is a flowchart for training a prediction model using a single car as an example. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application.

[0026] In an embodiment, as shown in Figure 1 and Figure 2 , the present application proposes a real-time road safety evaluation method based on vehicle vibration sensing during driving, which specifically includes the following steps.

[0027] Step 1: mounting vibration sensors, temperature sensors and front-end cameras on the car.

[0028] Step 2: dividing the road to be monitored into multiple areas with 10m x 10m as a unit, which are called points and identified by unique ids, and the point set can be expressed as Pid=(pid1,pid2,...,pid k ), and at the same time, the latitude and longitude range contained by each point is recorded. In addition, the road to be monitored is divided into intervals according to the area ownership, road type (asphalt pavement, cement concrete pavement, etc.) and other factors, and the intervals are identified by unique ids, and the interval set can be expressed as Aid=(aid1,aid2,...,aid l ), and at the same time, the road type Rid=(rid1,rid2,...,rid n ) of each interval is recorded. Finally, the ownership relationship between Pid and Aid is recorded, for example, pid1,pid2...,pid ubelonging to aid1, where u e (1, k). The above data is saved in a record table.

[0029] Step 3: During the pre-training process, a large number of cars are allowed to drive on the road in the normal situation (no cracks, no obvious concave-convex, etc.) for several days in a row within the specified speed range. During the driving process, the real-time latitude and longitude GPS, vibration sensor value F, temperature data T, and real-time photos of the road ahead taken by the front camera of the car are uploaded every second. At the same time, the back end will use artificial intelligence algorithms to identify weather and obstacles to process the uploaded photos, identify the weather and road obstacles on the current road. Among them, the weather can be divided into eight situations: sunny, light rain, moderate rain, heavy rain, light snow, heavy snow, gale, and sandstorm. These situations are based on the degree to which weather affects vehicle vibration. Road obstacles are considered to be the types that cars usually roll over when driving, so they are divided into foam board, speed bump, small stone, small branch, mud and sand, and plastic bag types.

[0030] Step 4: The weather and road obstacle conditions are respectively one-hot encoded, and the encoding details are shown in Table 1 and Table 2. According to the actual recognized weather type and road obstacle type, the corresponding encoding value is input. Then according to each GPS in the record table, find the corresponding road type rid, take each rid, T, photo recognition result corresponding weather encoding and road obstacle encoding as a group of input data, input into the neural network regression model for training, and the relevant vibration data of the GPS corresponding point after processing is taken as the model label. The label is Each time a group of data is input into the regression model, all the collected data is input into the model, and after training, the model can predict the vibration difference value AF of the same car at the same point caused by external environmental factors such as road type, temperature, weather and road obstacles. The training process is as shown in Figure 3 .

[0031] Table 1 Weather Encoding

[0032] Sunny 1 0 0 0 0 0 0 0 Light rain 0 1 0 0 0 0 0 0 Moderate rain 0 0 1 0 0 0 0 0 Heavy rain 0 0 0 1 0 0 0 0 Light snow 0 0 0 0 1 0 0 0 Heavy snow 0 0 0 0 0 1 0 0 Strong wind 0 0 0 0 0 0 1 0 Sandstorm 0 0 0 0 0 0 0 1

[0033] Table 2 Road Obstacle Encoding

[0034] Foam board 1 0 0 0 0 0 Speed bump 0 1 0 0 0 0 Small stones 0 0 1 0 0 0 Small branches 0 0 0 1 0 0 Mud and sand 0 0 0 0 1 0 Plastic bag 0 0 0 0 0 1

[0035] Step 5: In practical applications, every day, cars driving on the road to be evaluated continuously upload their real-time time (S), GPS coordinates (latitude and longitude), vibration sensor value (F), temperature data (T), real-time road condition photos ahead, and vehicle speed (v) at a rate of one second. The vehicle speed (V) is used to check if the car's speed is within a specified range. Only when the car's speed is within the specified range will its data be used for subsequent analysis of road anomalies. This method effectively eliminates the phenomenon of large differences in vibration sensor values ​​caused by factors such as speeding and traffic jams that result in large differences in car speed. After a car completes a section, all the data uploaded by the car within that section are extracted. The S, GPS, F, T, V, weather code, and road obstacle code uploaded per second are compiled into a dataset. If V is not within the specified range, the dataset is discarded. Simultaneously, GPS values ​​that meet the v requirement are extracted. Based on the GPS information, the corresponding RID and PID are found in the record table. This dataset belongs to the dataset corresponding to the PID. If there are multiple datasets with very similar times within a certain point, the average value of the data is used to represent the value of that data. For example... Where v is the number of data sets.

[0036] Step 6: Extract the vibration sensing values ​​F = (F1, F2, ..., F) from all valid datasets uploaded by the car within this interval on that day. m And the corresponding point Pid′=(pid1, pid2, ..., pid) m ), where m represents the number of valid datasets for the vehicle within this interval. Simultaneously, the standard vibration sensor values ​​corresponding to the 10 most recent Pid′ values ​​from the vehicle's historical data are extracted. i∈[1,10]. If some points have fewer than N data points in the historical data, the actual amount of data available will be used; if some points have no historical data before the current day, these points will not be included in the subsequent analysis, but the processed standard vibration sensing value of these points on the current day will be stored in the historical data of these points. The day's rid, T, weather code corresponding to the photo recognition result, and road obstacle code corresponding to Pid′ are input into the neural network regression model, and finally a set of vibration difference values ​​ΔF=(ΔF1, ΔF2, ..., ΔF m Meanwhile, this set of ΔF values ​​is used as a supplementary value to the vibration sensing value F for that day. The two are added together to obtain the standard vibration sensing value F′ = (F1 + ΔF1, F2 + ΔF2, ..., F′) after eliminating the influence of external environmental factors. m +ΔF m )=(F′1, F′2,...,F′ m The predicted ΔF is added to the F value of a point based on the data input into the regression model.

[0037] Step 7: Finally, use the Isolation Forest algorithm to process F′=(F′1,F′2,...,F′ m )and Data analysis is performed to determine if any vibration sensor values ​​at any location show abnormalities on that day, i.e., values ​​that are too high or too low. If abnormal locations are found, they are marked. The core idea of ​​the Isolation Forest algorithm is that in a normal data distribution, data points tend to cluster together, while outliers are relatively isolated. Therefore, by randomly partitioning the data space, outliers are isolated into separate subspaces earlier. During the algorithm analysis, the current day's and historical data for the same location are analyzed as a single set of values. There are a total of m sets of values. It's worth noting that the value for the current day is always placed last in a set of data. The process of analyzing anomalies is as follows:

[0038] 1. The sample set input to the algorithm each time is: Since the sample set is small, the number of trees to be built should not be set too high. The reference range is [5, 20] for training.

[0039] 2. Calculate the F' of the day. j The input is fed into the trained Isolation Forest model, and its anomaly score is calculated using the following formula:

[0040]

[0041] Among them, E(h(F′) j )) is the data point F′ j The average path length across all trees is given by c(t), where t is the size of the training set and c(t) is the average path length of the tree. The outlier score ranges from [0, 1]. A score closer to 1 indicates a sample is more likely to be isolated and is more probable as an outlier.

[0042] 3. If the abnormal score is greater than 0.5, the value is judged as an abnormal value, and the point corresponding to the value is recorded.

[0043] Step 8: All data uploaded by vehicles are processed according to the methods in Steps 5, 6, and 7, segmented by vehicle and interval. The standard vibration sensing values ​​obtained by different vehicles in different intervals are only recorded in their own databases. It should be noted that each vehicle has its own dedicated database, which stores the standard vibration sensing values ​​F′ of that vehicle at different times, intervals, and locations. Each F′ corresponds to an interval (aid), location (pid), and time (S).

[0044] Step 9: If multiple vehicles have abnormal values ​​at a certain point, then issue an alarm for that point.

[0045] Step 10: The newly obtained data set is fed back into the neural network regression model, and the model is continuously trained and adjusted for accuracy.

[0046] In another embodiment, the present application provides a computer readable storage medium, storing a computer program, which causes a computer to execute the real-time road safety assessment method based on driving vehicle vibration sensing as described in Embodiment One.

[0047] In another embodiment, the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the real-time road safety assessment method based on driving vehicle vibration sensing as described in Embodiment One when executing the computer program.

[0048] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium, which can contain or store programs for use by or in connection with an instruction execution system, apparatus or device. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus or devices, or any suitable combination of the above. More specific examples of computer storage media can include one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0049] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed in the present application can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those of ordinary skill in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0050] The above is only the preferred embodiment of the present application, and the protection scope of the present application is not limited to the above-mentioned embodiments. Any technical solution falling within the concept of the present application shall fall within the protection scope of the present application. It should be noted that, for those of ordinary skill in the art, some improvements and refinements without departing from the principles of the present application shall be considered within the protection scope of the present application.

Claims

1. A method for real-time road safety assessment based on driving vehicle vibration sensing, characterized in that, The method comprises the following steps: S1: dividing the road to be monitored into regions according to a specified size, the divided region being referred to as a point, and recording the latitude and longitude range contained in each point; dividing the road to be monitored into intervals according to region ownership and road type, and recording the point information contained in each interval and the road type to which the interval belongs; S2: in the interval road under normal conditions, collecting the latitude and longitude, vibration sensor value and temperature data of a plurality of vehicles traveling at a speed within a specified range for multiple days in real time, taking photos of the road conditions in front of the vehicle in real time, and identifying the weather conditions and road surface obstacles according to the photos of the road conditions in front of the vehicle; S3: inputting each latitude and longitude and its corresponding road type, temperature data, weather conditions and road surface obstacles as a group of input data into a neural network regression model for training, and taking the difference between the average value of the historical vibration sensor value of the latitude and longitude corresponding point and the measured vibration sensor value as the label of the neural network regression model, which is referred to as vibration difference value ΔF; S4: collecting the time, latitude and longitude GPS, vibration sensor value F, temperature data T, front road condition photo and vehicle speed uploaded by the vehicle traveling on the road to be evaluated in real time, excluding the uploaded data whose vehicle speed is not within the specified range, and corresponding the latitude and longitude uploaded by the vehicle in the interval to the point after the vehicle has run through the interval, to determine the data corresponding to the point; S5: extract the vibration sensor value F=(F1, F2, …, F m ) and the corresponding point position Pid'=(pid1, pid2, …, pid m ), m is the number of data; at the same time, extract the standard vibration sensor value corresponding to the N Pid' from the past history data of the vehicle closest to the current day Extract the data corresponding to each point position and input it into the trained neural network regression model to predict the vibration difference value AF. Use the vibration difference value AF to supplement the vibration sensor value F to obtain the standard vibration sensor value F'. S6: using the isolation forest algorithm to F' and F' i Data analysis is performed to determine whether the vibration sensor value of the point position on the same day is abnormal. If there is an abnormal point position, the point position is marked. If multiple vehicles have abnormal point positions, the point position is warned and predicted.

2. The real-time road safety assessment method based on the vibration sensing of the traveling vehicle according to claim 1, wherein: In step S1, a unique id is used to identify a point, and a point set is represented as Pid=(pid1, pid2, …, pid k ), k being the number of points; a unique id is used to identify an interval, and an interval set is represented as Aid=(aid1, aid2, …, aid l ), l being the number of intervals; a road type Rid=(rid1, rid2, …, rid n ) is recorded for each interval, n being the number of road types; and a belonging relationship between Pid and Aid is recorded.

3. The real-time road safety assessment method based on the vibration sensing of the traveling vehicle according to claim 1, wherein: In step S2, the vehicle is equipped with a vibration sensor, a temperature sensor and a front camera, which are used to collect vibration sensor values, temperature data and photos of the road conditions in front of the vehicle, respectively.

4. The real-time road safety assessment method based on the vibration sensing of the traveling vehicle according to claim 1, wherein: In step S3, the weather conditions and road surface obstacles are one-hot encoded before being input into the neural network regression model; according to each latitude and longitude, the corresponding road type is found, and each road type, temperature data, weather encoding and road surface obstacle encoding are input as a group of input data into the neural network regression model for training.

5. The real-time road safety assessment method based on the vibration sensing of the moving vehicle as claimed in claim 1, wherein: In step S4, after the vehicle has run through an interval, all the data uploaded by the vehicle in the interval are extracted, and the time, latitude and longitude, vibration sensor value, temperature data, vehicle speed, weather encoding and road surface obstacle encoding uploaded every second are taken as a group of data sets, the effective data sets whose vehicle speed meet the requirements are screened, the latitude and longitude are extracted from the effective data sets, the corresponding road type and point are found according to the latitude and longitude, and then the group of data sets belongs to the data sets in the point; if there are multiple groups of data sets in the point, the average value of each type of data is taken to represent the data set corresponding to the point.

6. A real-time road safety assessment method based on driving vehicle vibration sensing as claimed in claim 5, wherein: In step S5, the road type, temperature data, weather code and road obstacle code of the day corresponding to Pid' are input into the trained neural network regression model to predict a set of vibration difference values ΔF = (ΔF1, ΔF2, …, ΔF m ) of the day; the set of ΔF is added as a supplementary value of the vibration sensor value F of the day to obtain the standard vibration sensor value F' = (F1+ΔF1, F2+ΔF2, …, F m +ΔF m ) = (F'1, F'2, …, F' m ) which eliminates the influence of external environmental factors.

7. A real-time road safety assessment method based on driving vehicle vibration sensing as claimed in claim 5, wherein: After step S6, the method further comprises inputting the effective data sets obtained on the same day into the neural network regression model again for training.

8. A computer readable storage medium storing a computer program, characterized in that, The computer program enables the computer to execute the real-time road safety evaluation method based on vibration sensing of a traveling vehicle according to any one of claims 1-7.

9. An electronic device, comprising: The computer program product comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the real-time road safety evaluation method based on vibration sensing of a traveling vehicle according to any one of claims 1-7 when executing the computer program. The computer program product comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the real-time road safety evaluation method based on vibration sensing of a traveling vehicle according to any one of claims 1-7 when executing the computer program.

Images