CN112153574B

CN112153574B - Method and system for checking accuracy of roadside device clock based on floating vehicle

Info

Publication number: CN112153574B
Application number: CN202010987499.5A
Authority: CN
Inventors: 朱磊; 齐家; 卞加佳
Original assignee: Nanjing Microvideo Technology Co ltd
Current assignee: Nanjing Microvideo Technology Co ltd
Priority date: 2020-09-18
Filing date: 2020-09-18
Publication date: 2022-09-20
Anticipated expiration: 2040-09-18
Also published as: CN112153574A

Abstract

The invention relates to a method and a system for checking the accuracy of a roadside device clock based on a floating vehicle, which comprises the steps of retrieving and recording GPS records near the roadside device, including vehicle GPS positioning information and time information; forming a passing track record of the vehicle passing by the roadside device based on the acquired GPS record; establishing a fitting model to predict the real GPS time of the vehicle passing through the roadside equipment based on the GPS positioning information of the vehicle passing through the roadside equipment each time; and matching data records of the same vehicle passing through the same roadside device, and checking whether the observation time of the roadside device is consistent or not by using the real GPS time when the event occurs. The method and the system have the advantages of low cost, easy realization, and good accuracy and stability.

Description

Method and system for checking accuracy of roadside device clock based on floating vehicle

Technical Field

The invention belongs to the field of highway operation management informatization, and relates to a method for realizing roadside device clock calibration by using floating car data and an application system thereof.

Background

The wide construction of highway roadside equipment (RSU) greatly enhances the outfield awareness of highway operation management. As an end-aware infrastructure, RSU devices operate with time information provided by their system-internal clock. In practical engineering, a clock error of the RSU device sometimes occurs, which causes the device to provide a timestamp carrying an error in data. This phenomenon occurs for a variety of reasons, for example: 1) the internal clock of the device is not corrected (timed) or set wrong; 2) the built-in clock of the equipment runs for a long time to generate errors; 3) clock errors are caused by external factors such as power failure and the like in the working process of equipment; 4) some devices generate errors during time conversion (e.g., from UTC to local time); 5) although the device clock is calibrated by the time service server (the time service server can usually obtain accurate time through satellite time service), errors such as delay are caused by network transmission and the like in the process of finally reaching the device through the time service server.

Due to clock errors, the RSU device cannot know the real time, so that the generated data affects the accuracy and quality of service of the upper layer service, for example: vehicle path restoration, charging audit, dynamic flow monitoring, interval speed measurement and the like. To solve this problem, a method of analyzing the time service server log to check the clock abnormality of the device is widely adopted. However, this method can only carry out limited auxiliary checking on clock errors caused by some of the above factors, and it is more desirable for a third party to adopt an accurate clock and device clock pair at each device in view of clock anomalies caused by such factors as time service network delay or internal calculation errors. However, from a cost and feasibility perspective, it is less practical to arrange for personnel to go back and forth regularly. The invention provides a roadside device clock checking method which is low in cost and easy to implement by referring to the thought.

Disclosure of Invention

The invention aims to overcome the defects of the prior art and provides a low-cost and easy-to-implement road side equipment clock abnormity detection method and an online real-time clock verification system for realizing road network road side equipment based on the method.

The core thought of the scheme of the invention is that the accuracy of the clock of the road side equipment is tested by comparing the time-space consistency of the vehicle track and the time-space consistency of the road side equipment through a series of calculation and analysis by using data from different sources, namely the time-space data collected by the GPS equipment of the floating car and the observation data of the road side equipment to the vehicle, and by judging the time stamp consistency of different equipment to the same time-space occurrence event and superposing statistical analysis.

In order to achieve the technical purpose, the invention adopts the following technical scheme:

a method for checking the clock accuracy of road side equipment based on a floating vehicle comprises the following steps:

retrieving and recording GPS records near the roadside equipment, including vehicle GPS positioning information and time information;

forming a passing track record of the vehicle passing by the roadside device based on the acquired GPS record;

establishing a fitting model to predict the real GPS time of the vehicle passing through the roadside equipment based on the GPS record of the vehicle passing through the roadside equipment each time;

and matching data records of vehicles passing through the same road side equipment, and checking whether the observation time of the road side equipment is consistent or not by using the real GPS time when the event occurs.

As a further refinement of the invention, GPS records that fall in the vicinity of the roadside device are retrieved with a radius r, and the distance of the vehicle from the roadside device is calculated based on the haversine formula. The method can realize accurate distance calculation; or creating a rectangle based on the longitude and latitude coordinates of the road side equipment, and acquiring the GPS record of the floating vehicle with the longitude and latitude coordinates falling into the rectangular range. Under the scene of large data volume, the method can realize quick calculation so as to improve the calculation efficiency. Preferably, when the data scale is large, a quick search method is firstly adopted for preliminary screening, and the preliminary screening result adopts the above hemiversive formula to re-screen the accurate distance.

As a further improvement of the invention, when the GPS records are retrieved, the value of the search radius is determined by combining the vehicle speed and the vehicle GPS acquisition frequency, the search radius range is calculated by the average speed, the acquisition frequency and the required sample number, the preliminary screening value is selected, different preliminary screening values are selected for testing based on small sample data, and the search radius is determined.

As a further improvement of the invention, for any roadside device, the GPS records of the same vehicle passing through the roadside device are sorted according to the GPS time, and when the vehicle running distance in the time interval of the two GPS records is larger than the search range, the time interval is marked to be used for distinguishing the dividing points of the vehicle passing through the roadside device for multiple times, so that the records of multiple passes of the same roadside device are distinguished.

Further, vehicle running direction judgment is carried out on the traffic track record, a running azimuth angle of the traffic record is calculated, the calculated running azimuth angle is compared with the running direction of the route where the corresponding road side equipment is located, and the running record on the route where the road side equipment is located in the opposite direction is removed.

As a further improvement of the present invention, the real GPS time is obtained based on a linear fit, including: assuming that the vehicle motion track is approximately in uniform linear motion in the formed passing track recording range, establishing a fitting equation by adopting a general linear model, and calculating the real GPS time of the vehicle passing through the road side equipment based on the GPS record of the vehicle passing through the road side equipment each time.

As a further improvement of the present invention, the method further includes obtaining the average clock error of the road side device based on the single train number:

for any combination (p, g) of the floating vehicle and the road side equipment, acquiring the real GPS time of the vehicle p passing the road side equipment g each time, the time of the corresponding road side equipment observing the vehicle passing and the record writing time;

calculating absolute value of difference between true GPS time and record write time

Will be provided with

Sorting in ascending order, and taking the first kappa bit; taking the smaller value of the record number of the real GPS time and the record number of the record writing time;

calculating the difference value between the real GPS time and the observation time of the roadside equipment of the vehicle, and obtaining a result comparison set of k times that the vehicle p passes through equipment g and the real GPS time inspection equipment clock;

for all (p, g) groups, the time difference is calculated separately, and each group of time difference means, i.e. the average clock error of the roadside device g obtained based on the vehicle p.

The method is further improved by counting the clock average error of each roadside device and the fluctuation of error calculation of each floating vehicle; the volatility is characterized by a standard deviation of the mean error of the roadside device clock.

Further, different group thresholds are set according to the clock average error of each road side device and the standard deviation of the clock average error, and different handling measures are taken for the road side devices of different groups according to the error range and the fluctuation.

The invention also provides a system for checking the accuracy of the clock of the road side equipment based on the floating vehicle, which comprises the following steps:

the coordinate record retrieval unit is used for retrieving and recording GPS records near the roadside equipment, including vehicle GPS positioning information and time information;

the passing track record generating unit is used for forming a passing track record of the vehicle passing through the vicinity of the roadside equipment based on the GPS record acquired by the coordinate record retrieval unit;

the GPS time fitting unit is used for establishing a fitting model to predict the real GPS time of the vehicle passing through the roadside equipment based on the GPS record of the vehicle passing through the roadside equipment each time;

and the roadside device clock error checking unit is used for matching data records of the same vehicle passing through the same roadside device and checking whether the observation time of the roadside device is consistent or not by using the real GPS time when the event occurs.

Further, the system also comprises a checking unit, wherein the checking unit is used for counting the clock average error of each roadside device and the fluctuation of each floating vehicle in error calculation; the volatility is characterized by a standard deviation of the mean error of the roadside device clock.

The invention utilizes GPS data of the floating vehicle to check the clock accuracy of the road side equipment. The roadside equipment in the road network is passed by a large number of different vehicles (carrying GPS) in a reciprocating way, and the time (real GPS) when the floating vehicle passes through the roadside equipment can be effectively extracted by the method provided by the invention; then, the difference between the time based on the roadside device clock and the real GPS time can be effectively found through the record matching and comparing algorithm provided by the invention; and finally, obtaining an evaluation result of the accuracy of the road side equipment clock based on the traffic observation and statistical analysis of the large data volume. The invention has the following beneficial effects:

(1) compared with other modes, the thought and the method for checking the road side equipment clock are more practical. Compared with indirect checking of the time service log, the method is more direct, and is equivalent to direct detection of the road side device clock in the same time space by using a GPS (global positioning system) -based clock.

(2) The technical scheme of the invention has good accuracy and stability. Because the floating car which provides space-time information based on the GPS equipment is used, the time acquisition is based on the GPS satellite atomic clock, and the accuracy guarantee is provided for a reference standard system for detecting the roadside equipment clock. Meanwhile, the non-identical vehicle/equipment is used for carrying out repeated reciprocating inspection on the same roadside equipment in the inspection process, and the error influence of the inspection of a single vehicle is eliminated by using weighted average and standard deviation description.

(3) The cost for realizing the technical scheme of the invention is lower. The scheme has a plurality of implementation methods and ways, and the required raw materials (three source data) are all data of the existing business system. On the basis of realizing data sharing exchange, the implementation scheme does not need additional hardware and engineering investment.

(4) The technical scheme of the invention has high implementation and operation speed and high efficiency. The technical scheme can realize real-time rolling detection on road side equipment in the road network, and after systematization of the technical scheme, the required manual intervention in the execution process is very little, and the result can be automatically pushed regularly. Because the road side equipment is widely distributed in the road network, and the equipment which really needs to be checked or maintained by personnel only accounts for a small proportion, the achievement provided by the technical scheme can greatly reduce the equipment list which needs to be checked by personnel on the spot, and the maintenance efficiency is greatly improved while the manual checking/maintenance cost is saved.

(5) In the intelligent process of highway operation, the all-round perception to the road network depends on various more widely deployed roadside devices, and the effective work of the road network requires the running accuracy of the clocks of the devices and the continuous monitoring and maintenance of the devices. Under the same business and scenes of dynamic flow monitoring, interval speed measurement and vehicle-road coordination, the accuracy of the clock of the road side equipment plays a decisive important role. Although the server can ensure the accuracy of clocks of a plurality of devices on the whole, because the external field devices are large in quantity and wide in deployment and distribution, due to the fact that single-node accidental or periodic faults caused by various factors need to be periodically checked by an efficient detection means, and the scheme of the invention fills the gap of the application scene in technical implementation.

Drawings

FIG. 1 is a flow chart of a method according to an embodiment of the present invention.

Fig. 2 is a schematic position diagram of equipment on the highway network side of Jiangsu in embodiment 2.

Fig. 3 is a histogram of the average vehicle speed in a non-stationary state of the vehicle.

Fig. 4 is a schematic diagram of the vehicle trajectory determination split.

FIG. 5 is a schematic diagram of the roadside apparatus time-fitting calculation.

Fig. 6 distribution of gantry clock errors (absolute values).

FIG. 7 is a standard difference distribution scatter plot of gantry clock error estimation and estimation samples.

Detailed Description

The technical scheme of the invention is further explained by combining the description of the attached drawings and the detailed description.

Example 1

The method of the invention has the flow shown in figure 1, and comprises the following steps:

1. input data for implementing the technical scheme is prepared.

The scheme needs to utilize input data from three sources: 1) basic data, 2) space-time data (data set V) collected by floating car GPS equipment, and 3) observation data (data set G) of a road side device to a vehicle.

The basic data includes basic information of the roadside apparatus and the vehicle, such as:

the roadside device number is used for identifying the unique number of the roadside device;

the spatial position information of the road side equipment at least comprises longitude and latitude information;

vehicle basic information, such as: vehicle identification information (e.g., license plate), vehicle type, color, etc.

The floating car data (denoted as data set V) describes the vehicle spatial position records collected by the vehicle GPS device at a fixed frequency over a given time range d, each record containing at least the following information:

vehicle identification information, which is information for identifying a unique identity of a vehicle, such as a license plate, a unique pass code, or ETC _ ID;

the vehicle positioning information at least comprises longitude, latitude and azimuth based on vehicle position information obtained by a GPS;

vehicle time information, typically coordinated Universal Time (UTC) to local (beijing) time, based on time obtained by the GPS device;

logging time-to-garage information describing the time the log was written to the database, which lags behind the vehicle acquisition location (GPS) time, which was acquired from the server clock when the GPS data was warehoused.

Roadside device data (denoted as data set G) describes observations of vehicles by roadside devices over a time range of d days, each record containing at least the following information:

vehicle identification information, such as license plate information recognized by a camera, or ETC _ ID information in a running water record in which interaction with an ETC device occurs;

observation time information, a record of time that the vehicle was observed by the device, obtained based on the device clock, which may be expressed as an unreal time for device reasons;

and recording warehousing time information, describing and recording the time written into the database, and lagging the real observation time, wherein the time is acquired from the server clock when the data is warehoused by the road side equipment.

2. And cleaning and sorting the original data.

Step 2 is used for removing invalid, error and repeated data in the input data. Where the data source is a structured database, it is contemplated that the data cleansing may be accomplished directly when the data is acquired.

Invalid data

Data that cannot be utilized in subsequent analysis due to the absence of critical information should be removed, such as: 1) v contains the record of the invalid license plate, 2) G has the record of the mistake in the license plate recognition.

Abnormal data

Based on the data sets and records themselves, it can be directly determined that records with data errors should be cleaned. For example, the timestamp caused by the conversion error in the time conversion (UTC time to local time) process can be distinguished by comparing the record warehousing time.

Repeating data

Duplicate data, typically timestamps that occur for a vehicle due to multiple transmissions by the device or other reasons, may exist in both data sets V and G.

3. The vehicle GPS data is spatially searched based on the roadside device location, i.e., GPS records that fall near the roadside device are retrieved at a certain radius r.

Step 3, recording that the vehicle passes through any equipment in G in the space is picked in V, and the purpose is 1) judging whether the vehicle has the behavior of passing through roadside equipment by using GPS position information and keeping the space-time coordinate record of passing through any equipment; 2) the dataset to be examined V is scaled down for subsequent analytical calculations.

Accurate distance calculation search

For any of V records V _i For any roadside apparatus G of G _j Calculating v _i To g _j Is a spatial distance d _ij (ii) a If d is _ij <r then retains v _i Otherwise, discarding.

d _ij The hemiversine formula (Haversine formula) can be taken for calculation, namely:

wherein (phi) _i ,λ _i ) Is v is _i Latitude and longitude coordinates of (phi) _j ,λ _j ) Is g _j And R is the radius of the earth.

Simplified fast search

Because (1) adopts sin ² Sum-squared calculation, calculating distance d _ij The occupied resources are high, and the search can be simplified for searching the vehicle GPS coordinate points which are close to the road side equipment. For any road side equipment G in G _j With latitude and longitude coordinates (phi) _j ,λ _j ) Creating a rectangle using θ centered at the point; theta refers to radian; for any of V records V _i Latitude and longitude coordinate (phi) _i ,λ _i ) And judging whether the rectangular image falls into the rectangle or not, namely simultaneously satisfying the following conditions:

then v is reserved _i Otherwise, discarding.

Compared with the hemiversine distance judgment, the simplified method reduces the retrieval precision within the allowable range, can greatly improve the calculation efficiency, and is suitable for scenes with large data volume. In the practical implementation process, two-step retrieval can be carried out under the scene of large data scale, firstly, a rapid method is used for carrying out primary screening by using a slightly larger theta value (an approximate arc value which is slightly larger than r can be used for representing a value), and then, an accurate distance method is used for carrying out secondary screening on the passed result.

Determining search radius values

The value of the search radius theta needs to be comprehensively considered by combining factors such as vehicle speed and vehicle GPS acquisition frequency, and different values can be selected for trial based on small sample data (for example, 100-. The effective GPS coordinate point of the vehicle passing through the roadside equipment is lost due to the fact that the radius is too small, and particularly for vehicle GPS records with low sampling frequency and high speed; too large radius can lead to too many GPS coordinate points to be selected, increases the difficulty and performance of post processing, and is particularly easy to select the GPS data on the lane where the non-roadside equipment is located for the roadside equipment close to the interworking junction.

4. And analyzing and obtaining a passing track record of the vehicle passing through the roadside equipment based on the GPS record of the vehicle near the roadside equipment obtained in the step 3.

And 4, restoring the GPS recording points of the vehicles passing by the roadside equipment, which are obtained in the step 3, into the track records of the vehicles passing by through calculation and analysis. This process will be handled on a per-road-side-device basis and per-vehicle basis, where two types of problems are encountered: 1) distinguishing multiple-pass records of the same roadside device; 2) the record of the vehicle passing the roadside apparatus from the reverse lane is removed. The following are described separately:

segmenting GPS path points to obtain a passing record

For any roadside apparatus g _j The GPS location record of the vehicle passing by its vicinity, denoted v, is obtained by step 3 _pi Wherein p is the unique identification information (such as license plate) of the vehicle, i is the passing g of the vehicle p _j Vicinity records sorted by GPS timeThe latter index. For any p e k, i e n, the time interval between every two GPS records is calculated:

Δt _pi ＝t _p(i+1) -t _pi (3) wherein t is _pi Is v is _pi GPS time stamp in (1). If present:

Δt _pi ·s _pi >2r (4) judging that the vehicle running distance in the time interval is larger than the search range (s in the formula) _pi Is v is _pi Medium vehicle running speed) during which the vehicle has run out g _i The search space of, then marks the Δ t _pi Device g for distinguishing multiple-pass road side of vehicle _j The point of tangency of (a).

By vehicle direction of travel determination

After the splitting, the formed vehicle p passes through the roadside apparatus g _j Pass record r of _pjl Wherein l is that the vehicle p passes through the road side equipment g _j Is composed of a set of GPS coordinates, denoted v, ordered by GPS time _pji ,v _pji+1 ,…,v _pjn . For r _pjl Calculating the driving azimuth angle of the traffic record:

wherein (phi) ₁ ,λ ₁ ) Is v is _pji Latitude and longitude coordinates of (phi) ₂ ,λ ₂ ) Is v is _pjn The longitude and latitude coordinates of (c).

For roadside equipment g _j Having a of ^j The driving direction of the route can be obtained from the basic data. Comparison of alpha _pjl And alpha ^j The included angle between the two is used for judging the vehicle p is at the roadside equipment g _j Whether the first pass of (1) belongs to the group g _j And recording the driving on the lane in the direction, if:

|α _pjl -α ^j |<90 (6) determination of α _pjl Corresponding roadside apparatus g for vehicle p _j The first pass record of (c) belongs to g _j In the lane direction, the relevant passage record r _pjl To be reserved for later use,otherwise, it is the road side equipment g _j And the driving record on the reverse route is discarded.

5. And fitting the real GPS time of the vehicle passing through the roadside equipment.

And 5, predicting and obtaining the real GPS time of the vehicle passing through the roadside equipment by adopting a linear model (fitting) based on the GPS coordinate point of the roadside equipment each time the vehicle passes through. The vehicle passage record obtained in step 4 can be represented as r _pjl And describing the GPS space coordinates and the corresponding GPS time information acquired when the p vehicles pass through the j road side equipment for l times. For any r _pjl Let its GPS record number be m, r _pjl Can be expressed as:

{(φ ₁ ,λ ₁ ,t ₁ ),(φ ₂ ,λ ₂ ,t ₂ ),…,(φ _m ,λ _m ,t _m )}. For any r _pjl Assuming that the vehicle motion track is similar to uniform linear motion in the formed traffic track recording range, establishing a fitting equation by using a general linear model based on the recorded data:

t＝aφ+bλ+c (7)

λ＝a′φ+b′ (8)

the coordinates of the existing roadside apparatus are expressed as (phi) _g ,λ _g ) If its position does not fall on (8) (i.e., (phi)) _g ,λ _g ) The equation of (8) does not hold), needs to be calculated _g ,λ _g ) Drop foot coordinate on 7 (phi' _g ,λ′ _g ) Calculated in the following way:

otherwise it is used directly (phi' _g ,λ′ _g )＝(φ _g ,λ _g ) Carrying in (7), calculating and obtaining the GPS time t when the vehicle passes through the road side equipment _g 。

t _g ＝aφ′ _g +bλ′ _g +c (10)

6. And (4) collating the data obtained at present.

And 6, sorting the data results obtained in the preorder step for performing subsequent matching and time comparison analysis among the data sets.

By using the data set V, through the processing of the steps 1-5, which floating vehicles pass by which roadside devices at which time can be known, and the data and the calculation result are arranged for standby. The sorted data set V contains the following information: roadside device numbers, vehicle identification information (license plates), real GPS time of vehicle passing.

Meanwhile, the data set G is arranged in the following order: the method comprises the steps of numbering road side equipment, identifying information (license plate) of vehicles, observing time of the road side equipment and recording warehousing time.

7. Record matches versus time.

The purpose of step 7 is to compare the data sets R and G, match records of different data sources corresponding to the same event (namely, the vehicle p passes through the road side equipment G), and then check whether observation time of the road side equipment is consistent by using GPS time when the event occurs, so as to obtain single-vehicle time clock detection of the road side equipment G. The specific process is as follows:

for records within data sets R and G, grouped by (p, G), respectively; for any (p, g) combination, if there is a record in the data set R, it is denoted t ₁ ,t ₂ ,…,t _m Corresponding to the real GPS time of each time the vehicle passes by the roadside device G, there is a correlation record in G, denoted as(s) ₁ ,t′ ₁ ),(s ₂ ,t′ ₂ ),…,(s _n ,t′ _n ) Time (t ') that the corresponding roadside device observes the vehicle to pass' _n ) And recording the write time(s) _n )，

Calculating t _i And s _j Time length between

T can be represented as an m n matrix:

will be provided with

Sorting in ascending order, taking the first k bits, denoted as

Calculating the time t for the vehicle to pass through the road side equipment _i T 'observed by road side equipment of vehicle' _j Difference of (a), Δ t _pg ＝t _i -t′ _j The set of comparisons of the results of k times the vehicle p passes the device g to check the device clock with the on-board GPS time, denoted Δ T, is obtained _pg ＝Δt _pg1 ,Δt _pg2 ,…,Δt _pgκ . For all (p, g) packets, their respective sets of clock comparisons Δ T are calculated _pg And the mean value of each group

I.e., the average clock error of the roadside apparatus g obtained based on the vehicle p.

8. Statistical description of roadside device clock accuracy.

The purpose of step 8 is to sum and count the average clock error of each roadside device and the fluctuation of error calculation of each floating vehicle pair.

Firstly, summarizing the result of the step 7 by adopting a weighted average method to obtain the average error delta t of the road side equipment g _g ：

Wherein the content of the first and second substances,

mean inspection error, ω, for vehicle p at roadside unit g _p Weight for vehicle p, expressed as:

in the formula, n ^pg The total number of times the vehicle p passes by the roadside apparatus g.

Next, the error standard deviation Δ δ of the roadside apparatus g is calculated _g ：

9. And outputting the final result.

The result obtained in the step 8 is processed to obtain the clock calibration evaluation value delta t of each path side device _g And evaluation value stability Δ δ _g And can be represented as data set C:

C＝{(Δt ₁ ,Δδ ₁ ),(Δt ₂ ,Δδ ₂ ),…,(Δt _g ,Δδ _g )} (15)

in the formula,. DELTA.t _g The method is used for describing the average clock error of the roadside device g in the observation time, and the smaller the absolute value of the average clock error is, the smaller the error is; delta delta _g The standard deviation is used for describing the fluctuation of multiple observation errors of the device g, and the smaller the value of the standard deviation, the higher the accuracy of error judgment.

For the final result, different group thresholds can be set according to the two values, the clock accuracy of the road side equipment is discussed respectively, and different handling measures are taken for different groups according to the error range and the fluctuation.

Example 2

The case is developed based on a highway network in Jiangsu and highway side equipment in the Jiangsu, the total mileage of a case research area is 4084 kilometers, and 1316 roadside equipment are arranged in the case research area. The floating cars are expressway wreckers, cleaning cars and patrol police cars, all of which are provided with GPS positioning systems and report the position and motion information to the system in real time.

The implementation of the method is based on the development and implementation of a private cloud environment, and the hardware environment of the host computer is as follows: 1) intel (R) Xeon (R) Gold5218R x 2 CPU; 2)256GB memory; 3)1Tb SSD +16Tb HDD. The virtual host used for executing data processing and analysis is configured to be a 40-core 128Gb memory, the system platform is Ubuntu 18.04LTS, the analysis software platform is JupyterLab, the analysis uses python 3.7, and the third-party tool comprises: numpy, pandas, sklern, matplotlib, and the like.

The implementation process is as follows:

1. all data required for the embodiment were collected and prepared.

In example 2, basic data is basic information of roadside equipment (portal frames) of the highway network of Jiangsu, and the total network is provided with portal frames 2234. A case uses only portals (1316 seats) with type attributes as road classes. Based on the raw data table (25 columns total of fields), the following information was extracted for subsequent analysis:

each device unique number (f _ vc _ gaming _ id), as: g250332001000310010, the code is composed of city code, road section code, stake number and other information;

spatial location information (longitude, latitude) of each device.

The floating car data is stored in a cloud database and directly acquired by using SQL sentences, and the total 7 days from 7 months and 6 days to 12 days in 2020 is selected as a research window in the example. Obtaining original floating car data about 2.42X 10 ⁷ The strip records, about 3.6 Gb. The following information was extracted from the 20 columns of raw fields for case analysis:

a license plate number (F _ license plate) for identifying information of the vehicle identity;

vehicle GPS positioning data comprising longitude, latitude, azimuth, speed; (F _ length, F _ position, F _ order, F _ SPEED), wherein the lowest precision of the coordinate information is 6 bits after the decimal point (8 bits for some vehicle devices) for analysis in step 3-5;

time information (F _ GPSTIME) with accuracy of second based on the time obtained by the GPS equipment;

time information written to the database is recorded (F INSERTTIME).

The observation records of the road side equipment are the license plate identification data of the portal snapshot camera, the source of the license plate identification data is the cloud database, and the total number of the license plate identification data is about 1.32 multiplied by 10 in 7 days after preliminary retrieval ⁸ Strip recording, subsequently (step 6) according toAnd (5) inquiring the conditions (license plate and roadside equipment number) by using SQL sentences according to the results of the steps 1 to 5. The following information in each record was extracted for analysis:

a device number (GANTRYID), a device number that generates a snapshot record, corresponding to a base data device number;

license plate information (VEHICLEPLATE), wherein the average recognition accuracy of the license plate information of the passing vehicle recognized by the camera is 92%;

snapshot time (PICTIME), a record of the time of observation of the vehicle obtained based on the camera clock, this time being assumed as information to be calibrated;

and (4) warehousing time (AUDCREATETIME), the time for uploading the vehicle-passing snapshot data to a system warehousing device, and the delay of about 10 minutes exists on average.

2. Original data are loaded by using a Jupyter platform, and cleaning data are developed according to the step 2 in the embodiment 1, wherein the aim is to remove three types of data, namely invalid data, abnormal data and error data.

And screening the license plate numbers in the GPS data by using a regular expression, and rejecting records containing invalid license plates. The license plate information consists of province names, license plate issuing place codes (A-Z), license plate serial numbers (numbers and letters) and license plate colors. The license plate matching mode is written by using a regular expression ('A-Z0-9 ] {5,6 }'), so that the purpose is achieved.

And comparing the GPS time in the floating car data with the warehousing time, and screening time abnormity records. Statistics shows that the GPS data entry time (F _ INSERTTIME) is later than the GPS acquisition time (F _ GPS) 330.04 seconds on average, which is caused by data reporting processing, network delay, and other factors. The method comprises the following steps that a small part of data is subjected to data conversion error to cause that the GPS acquisition time is greatly different from the GPS data storage time (the original time acquired by GPS equipment is coordinated Universal Time (UTC), local conversion can be carried out after the GPS data is uploaded to the cloud, occasional faults exist in the process, the GPS data recording time is wrong), the GPS data storage time and the GPS data acquisition time are used for carrying out difference value calculation, and the difference value is analyzed and judged through a histogram: 1) and (3) the warehousing time is earlier than the acquisition time, or 2) the warehousing time is later than the acquisition time by more than 4, and the GPS time is recorded as the GPS time abnormity, and the GPS time is eliminated.

And removing repeated data caused by multiple retransmission or other reasons of the equipment, and respectively carrying out deduplication operation on the floating car data and the roadside equipment observation data.

Cleaning the three types of data according to the method, and eliminating the floating car data by 2.46 multiplied by 10 ⁵ The bars are recorded to account for about 1% of the total.

3. The floating car data is screened and filtered by the roadside device position by using the space retrieval method in the step 3 of the embodiment 1. The purpose of space retrieval is to extract the driving record of the floating car in a certain range near the equipment through the position information of the roadside equipment and the floating car, so that the GPS data volume of the floating car is further greatly reduced, and the method is used for path segmentation and fitting in the steps 4 and 5. The process first obtains an optimal search radius through sampling analysis, and then performs spatial retrieval on the full amount of data based on the obtained parameters.

(1) Determining an optimal search radius for a spatial search

For one roadside device, too large radius of spatial search may cause too many GPS data points to be selected, increasing the fitting calculation of the vehicle passing through the roadside device (step 5 in embodiment 1) but not beneficial to improving the accuracy; an excessively small radius setting may result in omission of the fitted GPS points (step 5 in example 1) and increase difficulty in determining the driving trajectory (step 4 in example 1), and selecting reasonable parameters is beneficial to controlling the performance and result quality of subsequent analysis.

And (3) performing description statistical analysis on the floating car GPS data obtained in the step (2), wherein the average speed of the reference car in the non-static state of the vehicle is 62.29km/h in a histogram (figure 3) of the average speed, and the average GPS acquisition frequency is 3.99 s. Calculating by using average speed and acquisition frequency, if more than 5 samples are selected near roadside equipment, selecting at least 173.05m of search radius, taking the search radius as reference, selecting 100, 200, 300 and 400 to-be-selected space search radii for the cleaned floating car data, and respectively executing quick search according to a simple calculation mode formula (2) (the space ranges are N30.89502400-N34.90732000 and E116.57126564-E121.70781180, and the central positions of research areas are N32.90117200 and E119.13953872). The longitude and latitude are calculated to be about (0.001141,0.000899) in terms of 100 meters in the area, so that the integral multiple is taken and substituted into the formula (2) for searching. And performing statistics on the obtained results according to the road side equipment and the vehicle groups to obtain the average GPS record number of the vehicles passing through the road side equipment as follows: 2.12, 4.37, 6.84, 12.61. Ideally, 3-5 records are selected when the expected vehicle passes by the roadside device every time, so that 200 meters are selected as the optimal search radius, namely recording points within 200 meters of any roadside device are selected from the cleaned floating vehicle data.

(2) And after 200 meters are selected as the optimal retrieval radius, floating car data are respectively used for spatial retrieval filtering by using a formula (2) and a formula (1). Firstly, a simple method of formula (2) is used for combining a 300 m distance (corresponding to a longitude and latitude distance of 0.003) to carry out rapid primary screening to obtain a record of 9.66 multiplied by 10 ⁵ Bar, then actual review using equation (1) in combination with 200 meter distance parameters to obtain record 6.12 x 10 ⁵ And a bar, searching the final result for the space.

4. And (4) restoring the GPS record obtained after the space search into the track of the vehicle passing through the road side equipment according to the step 4 of the embodiment 1.

Firstly, grouping GPS data records according to road side equipment and vehicles, wherein the data screened by taking 200 meters as a retrieval radius relates to 979 vehicles with 1257 road side equipment, the road side equipment and the license plate count 13994 groups in total, and the average number of the GPS records in each group is 43.74. And sequencing the GPS data in each group according to the GPS time, calculating the GPS time difference between records in each group by using a formula (3), and judging the dividing points in each group by using a formula (4), thereby obtaining the path point set of each vehicle passing through each road side device. The 1257 roadside device 979 vehicles add up to 95196 GPS path sets.

Next, for each GPS waypoint set obtained above, the travel direction angle α of each group is calculated using formula (5), and whether or not the waypoint set (fig. 4) belongs to the roadside apparatus in the group is determined in combination with formula (6). It is calculated that 79833 the set of waypoints are eligible for reservation.

This step finally yields: 979 vehicles accumulated 79833 times through 1257 roadside devices within 7 days; in the observation time period, the average 63.51 vehicle passes by each roadside device; the average roadside apparatus passed by each vehicle was 81.54 seats.

5. According to the step 5 of the embodiment 1, the driving track of the vehicle passing through the roadside device is restored by 79833 groups of GPS point sets, and the real GPS time of the vehicle passing through the roadside device at the accurate position is extracted by utilizing a linear fitting model. In the embodiment, LinerRegistration provided by a sk-leann tool set is used for linear fitting, and equations are respectively carried out according to formulas (7-8) to establish a linear fitting model. For each set, equation (7) establishes a trend surface model for inverting possible GPS time for the vehicle at different locations; equation (8) establishes a linear model of the trajectory. And (3) obtaining a projection point of the road side equipment on the driving track (namely, the accurate coordinates of the vehicle passing through the road side equipment) by using a formula (9) in combination with the coordinates of the road side equipment corresponding to each set, and calculating the accurate GPS time of the vehicle passing through the road side equipment in a trend surface model by using the coordinate point (figure 5).

6. And (4) based on floating car data, obtaining the local track reduction of the car passing through the portal frame in the steps 1-5, and calculating the GPS time when each car passes through the portal frame. The intermediate results, collated to obtain data set R with reference to step 6 of example 1, contain the following fields: the total of 79833 door numbers, license plate numbers, GPS time passed by the door, and other fields (as auxiliary information, including vehicle original location and GPS time information). Meanwhile, in order to improve the result quality, data with lower vehicle passing number and gantry record number in R are removed, and 200 vehicles (corresponding to 1089 gantries) with the most passing times are selected for data record in combination with histogram distribution. And (5) inquiring the observation records of the vehicle and the roadside device by using an SQL statement to obtain a data set G, wherein 51448 pieces of data are obtained.

7. In step 7 of embodiment 1, two data sets (R and G) obtained at present are matched, so as to find out the corresponding reflection (corresponding record) of the roadside device events of all vehicles in the two data sets, and then the device internal clock in G is successively calibrated by the GPS time in R. Because the data sets R and G are incomplete in practical implementation, the matching records are subsets of the intersection of the two data sets.

For the records in data sets R and G, the records are grouped by (p, G), respectively, resulting in 12428 sets of records. For each (p, G) subgroup, records in R and G are extracted, the difference between GPS time and warehousing time between R and G records in the subgroup is calculated according to formula (11), and a matrix T is listed. And (4) establishing a matching relation (the row record and the column record, corresponding to the GPS track and the matching record) by taking the minimum value of each row in the matrix, and establishing the matching relation by only taking the minimum value if the minimum values of a plurality of rows of records in the same row correspond to each other. This example was conducted in 12428(p, g) subgroups, and each match was completed to obtain 30911 record pairs.

In order to ensure that the mismatching record pairs are eliminated, the embodiment performs secondary threshold filtering on the minimum time difference for establishing matching, and performs subsequent clock calibration analysis only by using the matching pairs with the minimum time difference for establishing matching relation smaller than the threshold. The threshold is determined by the statistical result of the difference between the acquisition of the observation data of the reference full-scale road side equipment and the write-in delay, and in this example, 1000 seconds is used as the secondary screening threshold. After the second screening, 29424 pairs of valid records were obtained.

And for each matched record pair in the (p, g) subgroups, comparing the GPS time with the equipment snapshot time to obtain the time error of the equipment built-in clock compared with the GPS time of the vehicle every time the vehicle passes through a certain roadside equipment, wherein if the equipment clock is normal, the error value is 0.

8. According to step 8 of embodiment 1, the obtained GPS timing results are grouped by the roadside apparatus numbers, the average error (fig. 6) and the standard deviation of the clocks of the respective roadside apparatuses are calculated using the formulas (12 to 14), respectively, and the final calculation results are collated in step 9. The results of the examples are described below:

the GPS data of highway sweeper, wrecker and police car in Jiangsu province and in a week from 6 days to 12 days in 7 months in 2020 and the snapshot data of the portal frame of the road network are extracted from the original data for matching and clock verification. The raw data totals 1225 vehicles, seated in portal 1316. Through the steps of cleaning, analyzing, track restoring and calculating, 200 vehicles which are accumulated most frequently pass through the door frame are finally selected for snapshot observation data matching, 29424 times of (sample) passing are successfully matched, equipment clock inspection is carried out, 1089 door frames are involved, and the inspection is carried out 27.01 times on average on each piece of equipment. The weighted average was used to calculate the clock error for each portal, and the standard deviation of the clock-corrected samples, presented in scattered-point form (fig. 7). The portal clock error condition and the percentage are counted according to different levels of error and the reliability (standard deviation) of the error test (table 1). From table 1, it can be seen: the number of the time-error-free portal frames is 149, and the time-error-free portal frames account for about 13.7%; 773 total portal frames with the error within 1-60 seconds account for about 70.9%; 164 seats of portal frame with error more than 1 minute (60-600 seconds), accounting for about 24.4%; the error of the gantry with 2 seats is more than 10 minutes. From the perspective of the check reliability, the check reliability of about 67.6% (737) of the gantries is high, the check reliability of about 32.1% (348) of the gantries is medium, and the gantry number with low check reliability accounts for a very low percentage (0.3%). From the perspective of guiding actual inspection, manual on-site review can be preferentially carried out from the group with large error and high reliability.

Table 1 real-time example roadside device clock error grouping statistics.

Claims

1. A method for checking the clock accuracy of road side equipment based on a floating vehicle is characterized by comprising the following steps:

matching data records of vehicles passing through the same road side equipment, and checking whether the observation time of the road side equipment is consistent by using the real GPS time when an event occurs, wherein the method comprises the following steps: acquiring the average clock error of the road side equipment based on the single train number:

calculating absolute value of difference between real GPS time and recorded write time

Will be provided with

Sorting in ascending order, and taking the first k bits; k, taking the smaller value of the number of records of the real GPS time and the number of records of the record writing time;

calculating the difference value between the real GPS time and the observation time of the vehicle by the road side equipment to obtain a set of k times of comparison of the results of the vehicle p passing through the equipment g and the real GPS time inspection equipment clock;

for all the (p, g) groups, respectively calculating time difference and the average value of each group of time difference, namely the average clock error of the roadside device g obtained based on the vehicle p;

counting the clock average error of each roadside device and the fluctuation of each floating vehicle in error calculation; the volatility is characterized by a standard deviation of the mean error of the roadside device clock.

2. The method of claim 1, wherein the GPS records that fall near the roadside device are retrieved at a radius r, and the distance of the vehicle from the roadside device is calculated based on the haversine formula; or creating a rectangle based on the longitude and latitude coordinates of the road side equipment, and acquiring a GPS record of the floating vehicle with the longitude and latitude coordinates falling into the rectangular range; or the GPS records are searched by combining the two modes, a rectangular preliminary screening is firstly established, and then the accurate distance re-screening calculated based on the hemiversine formula is adopted.

3. The method of claim 1, wherein the search radius is determined by taking into account vehicle speed, vehicle GPS acquisition frequency, calculating the search radius range from the average speed, acquisition frequency and required sample number, selecting prescreening values, selecting different prescreening values for testing based on small sample data, and determining the search radius r when retrieving GPS records.

4. The method of claim 1, wherein for any roadside device, the GPS records of the same vehicle passing the roadside device are sorted by GPS time, and when the vehicle travel distance within two GPS record time intervals is greater than the search range, the time interval is marked for distinguishing the cut points of the vehicle passing the roadside device multiple times.

5. The method according to claim 1 or 4, characterized in that the vehicle driving direction is judged for the traffic track record, the driving azimuth angle of the traffic record is calculated, the calculated driving azimuth angle is compared with the driving direction of the route of the corresponding road side equipment, and the driving record on the route of the road side equipment in the opposite direction is removed.

6. The method of claim 1, wherein the true GPS time is obtained based on a linear fit, comprising: assuming that the vehicle motion track is approximately in uniform linear motion in the formed traffic track recording range, establishing a fitting equation by adopting a general linear model, and calculating the real GPS time of the vehicle passing through the road side equipment based on the GPS record of the vehicle passing through the road side equipment.

7. A roadside apparatus clock accuracy verification system based on a floating vehicle, characterized by comprising:

roadside device clock error check unit matches the data record of same car through same roadside device, and the real GPS time when utilizing the incident to take place checks whether the observation time of roadside device is unanimous, includes: acquiring the average clock error of the road side equipment based on the single train number:

Will be provided with

8. The system of claim 7, further comprising a verification unit for counting the average clock error of each roadside device and the fluctuation of error calculation of each floating vehicle; the volatility is characterized by a standard deviation of the mean error of the roadside device clock.