CN114663992B - Multisource data fusion expressway portal positioning method - Google Patents

Abstract

The invention discloses a multi-source data fusion highway portal positioning method, which comprises the following steps: step 1, preprocessing transaction data of an ETC portal of the expressway to ensure the accuracy of transaction information; and 2, obtaining a potential portal position set by using a DR algorithm based on ETC portal transaction data and two-passenger one-risk vehicle track data. And step 3, obtaining a central point as a potential portal position by using a median central algorithm based on the potential portal position set. And 4, carrying out data fusion on the original portal position and the potential portal position by using a Kalman data fusion algorithm based on the selection strategy data, so as to obtain a more accurate portal position.

Description

Multisource data fusion expressway portal positioning method

Technical Field

The invention relates to the technical field of traffic facilities, in particular to a multi-source data fusion highway portal positioning method.

Background

With the rapid development of the expressway, an expressway Electronic Toll Collection (ETC) system is used as one of important subsystems of an intelligent traffic system, the traffic capacity of the expressway toll collection station is effectively improved by applying a combined networking electronic toll collection technology to the expressway toll collection station, queuing congestion caused by manual toll collection is improved, the problems of excessive consumption of energy sources, environmental pollution and the like are reduced until 31 months in 2019, the traffic transportation department of China completes the construction of 29 ETC portal systems with networking provinces, 487 expressway province toll collection stations are canceled, 24588 ETC portal systems are built, and 48211 ETC lanes are modified. The ETC portal system realizes the functions of real-time supervision and recording of vehicle running information, vehicle path identification, charging data fitting and the like, and plays a significant role in the data convergence system. The information accuracy of the ETC portal system is critical to the overall highway informatization process.

However, the system has some problems at present, such as a certain lack or deviation of the geographical position information of the portal frame. It is therefore a problem how to effectively detect whether the position information of the gantry is erroneous and correct after the deviation.

Disclosure of Invention

The invention aims to provide a multi-source data fusion highway portal positioning method.

The technical scheme adopted by the invention is as follows:

a multi-source data fusion highway portal positioning method comprises the following steps:

step 1, ETC transaction records of two-passenger one-risk vehicles passing through a portal and track data of the vehicles are extracted;

step 2, comparing and extracting transaction points of all vehicles according to the transaction time of ETC transaction data and track points in vehicle track data;

step 3, calculating a portal potential position set based on the transaction points of the vehicle,

step 4, clustering a portal potential position set by using a median center clustering algorithm to obtain portal potential positions;

and 5, inputting the potential positions of the portal into a Kalman filtering algorithm based on a selection strategy, and fusing the original portal positions and the potential positions of the portal to obtain final portal positions.

Further, the two-passenger one-risk vehicles comprise travel package vehicles, more than three classes of on-duty buses and road special vehicles for transporting dangerous chemicals, fireworks and crackers and civil explosive substances; attributes of the two-passenger one-risk trajectory data include positioning time, longitude, latitude, and instantaneous speed.

Further, the portal transaction record data is a series of time sequence data formed by the vehicle in the high-speed driving process, and the portal transaction record data comprises a transaction identifier TRADEID, a portal number FLAGID, a vehicle license plate OBUPLATE, a traffic identifier PASSID, a vehicle type VEHCLASS and a transaction time TRADETIME.

Further, the specific steps of step 2 are as follows:

step 2-1, acquiring the transaction time TRADETIME of ETC transaction data of the current vehicle;

step 2-2, judging whether track points which are equal to the transaction time TRADETIME of ETC transaction data exist in the vehicle track data; if yes, extracting a corresponding track point as a time coincidence point, taking the time coincidence point as a transaction point of a corresponding vehicle, and executing the steps 2-4; otherwise, executing the step 2-3;

step 2-3, taking one track point with the smallest time difference between the vehicle track data acquisition transaction times TRADETIME as a time front point, and taking one track point with the smallest time difference between the vehicle track data acquisition transaction times TRADETIME as a time rear point; and takes the time point before and the time point after as the transaction points of the corresponding vehicles

And 2-4, ending the extraction of the transaction points of the current vehicle, acquiring the track data of the next vehicle, and executing the step 2-1 until all vehicles finish the extraction of the transaction points.

If there is no track point equal to the store time, a front point and a rear point are extracted, the front point is the point with the track positioning time before the transaction time and the minimum time difference, and the rear point is the point with the track positioning time after the transaction time and the minimum time difference.

Further, the specific steps of step 3 are as follows:

step 3-1, judging whether the transaction point of the vehicle is a time coincidence point; if yes, adding the corresponding time coincidence point into the portal potential position set and executing the step 3-3; otherwise, executing the step 3-2;

step 3-2, calculating potential portal positions corresponding to the current vehicle by using a dead reckoning algorithm based on the front point, the rear point and the transaction time, and adding the potential portal positions into a portal potential position set;

and 3-3, acquiring transaction point data of the next vehicle, and executing the step 3-1 until all vehicles finish calculation to form a final portal potential position set.

Further, the specific calculation step of step 3-2 is as follows;

step 3-2-1, A is the point P before time _a B is the time point P _b N is the north pole, O is the sphere center of the earth, N is the radian corresponding to the arc AB, and the cosine value of N is calculated by using the spherical cosine theorem and the Pythagorean theorem, and the calculation formula is as follows:

cos(n)＝cos(90-B _Lat )×cos(90-A _Lat )+sin(90-B _Lat )×sin(90-A _Lat )×cos(B _Lng - A _Lng ) (3-6)

wherein n is the radian corresponding to the arc AB, B _Lng Longitude of point B, B _Lat Is the latitude of point B, A _Lat Longitude of point A, A _Lat The latitude of the point A;

and 3-2-2, calculating a dihedral angle NAB between the NOA and the AOB based on the cosine value, wherein the calculation formula is as follows:

wherein ,B_Lng Longitude of point B, B _Lat Is the latitude of point B, A _Lat Longitude of point A, A _Lat The latitude of the point A;

step 3-2-3, determining the value of the direction angle eta based on the angle NAB;

setting the definition field of the inverse cosine to be

When the point B is in the first quadrant, the direction angle eta= NAB;

when the point B is at the second quadrant, the direction angle eta=360+ & NAB;

when the point B is at the third or fourth quadrant, the direction angle eta=180- & lt NAB;

wherein the direction angle eta is 0 DEG in true north, and is clockwise to 360 DEG;

step 3-2-4, calculating a time difference between a time front point and the transaction time as a front time difference, and simultaneously calculating a time difference between a time rear point and the transaction time as a rear time difference;

step 3-2-5, judging whether the absolute value of the front time difference value is smaller than the absolute value of the rear time difference value; if yes, selecting a time point before the time to carry out accumulation calculation and executing the steps 3-2-6; otherwise, selecting a time later point to carry out accumulation calculation and executing the steps 3-2-6;

step 3-2-6, substituting the selected time point or the time point into a corresponding formula to calculate the position of the portal:

when the accumulation calculation is performed at the time point before the selection, the calculation formula of the portal position is as follows:

Node _Lng ＝A _Lng +V _drive ×t _trade ×sinη/[ARC×cosA _Lat ×2π/360] (3-8)

Node _Lat ＝A _Lat +V _drive ×t _trade ×cosη/[ARC×2π/360] (3-9)

wherein ,t_trade ＝|t _A -t _trade Absolute value of i, i.e. the previous time difference, ARC is the earth radius, here 6371393m, η is the time-previous point P _a To P _b The direction angle of travel;

when the accumulation calculation is performed at the selected time, the calculation formula of the portal position is as follows:

Node _Lng ＝B _Lng +V _drive ×t _trade ×sinη/[ARC×cosB _Lat ×2π/360] (3-10)

Node _Lat ＝B _Lat +V _drive ×t _trade ×cosη/[ARC×2π/360] (3-11)

wherein ,t_trade ＝|t _B -t _trade Absolute value of i, i.e. the difference in time after, ARC is the radius of the earth, here 6371393m, η is the point before time P _b To P _a Steering angle of travel.

Further, in the step 4, a method of searching a center point by adopting a median center is adopted to determine the potential position of the portal, and the specific steps are as follows:

step 4-1, establishing a median center for initializing a dataset LocaAll by utilizing all portal potential positions

Wherein, localall= { location ₁ ,locations ₂ ,…,location _j }，location _j ＝{x _j ,y _j }，x _j Is the j potential portal position longitude, y _j Is the j-th potential portal location longitude;

longitude and latitude representing the initial median center;

step 4-2, extracting portal information from the track data and calculating the distance from the potential portal position to the median center;

wherein ,s_pj Representing the distance between the point and the center of the median;

step 4-3, judging s _pj Whether the value of (a) is less than a threshold value alpha; if yes, stopping the iterative process to obtain the final result

Otherwise, executing the step 4-4;

and 4-4, updating the unit distance weight from the point to the median center according to the change of the median center, wherein the updating formula is as follows:

wherein ,w_j Is the initial unit weight, s _pj Representative of the distance between the point and the center of the median of the present invention,

the unit distance weight after updating;

step 4-5, iterating the center of the median by using the updated unit weight,

step 4-6, judging whether all portal information in the track data is extracted; if yes, stopping the iterative process to obtain the final result

Otherwise, step 4-2 is performed.

Further, the existing raw gantry position data is regarded as a measurement vector in step 5

The potential portal position obtained by the central clustering method is used as a system vector x _k-1 More accurate portal position information is fused through data.

Further, the specific steps of the Kalman filtering algorithm based on the step 5 are as follows:

step 5-1, initializing x _k-1 ,Q,R,P _k-1 ,F,H，

wherein ,x_k-1 The longitude and latitude of the potential portal position obtained by the central clustering method are obtained, Q is an error matrix of the potential portal position obtained by the central clustering method and the true value, and R is the longitude and latitude of the original portal

Error matrix with true value, P _k-1 To indicate x _k-1 And->

Covariance matrix of errors between the two, H is X _k and />

The method comprises the steps of converting into a conversion matrix with a unified unit, wherein F is a state transition matrix, and F is a state conversion process for an object;

step 5-2, carrying out state prediction, wherein the state prediction formula is as follows

wherein ,x_k Represents x _k-1 A new position obtained through iteration is used for indicating the drift of the position of the portal; b is a control matrix, and the control matrix is a control matrix,

for controlling vector +.>

P is the noise of covariance Q in the system _k Representing P _k-1 The error covariance matrix obtained through iteration, namely, representing x _k And->

Errors between;

step 5-3, obtaining a measurement of time k

S _k ＝(HP _k H ^T +R)

wherein ,

is the actual measurement +.>

And a predicted measurement Hx _k Difference vector between S _k To represent the residual covariance matrix, express +.>

And the difference between the true values;

and 5-4, correcting the state based on the measured value, wherein the specific formula is as follows:

K _k ＝P _k H ^T S _k ^-1

P′ _k ＝(I-K _k H)P _k

wherein ,K_k For Kalman gain, x' _k Longitude and latitude of potential portal position obtained by central clustering method and original portal data by Kalman filtering

A fused result; p'. _k Represents x _k-1 And->

The covariance matrix of the errors between the two is P _k-1 The method comprises the steps of carrying out a first treatment on the surface of the I is an identity matrix.

Further, in step 5, when the original gantry is located

With potential gantry position x _k When the difference delta is smaller than or equal to 2000m, the final gantry position is considered as Kalman filtering result x' _k ；

When delta original portal position

With potential gantry position x _k If the difference delta is larger than 2000m kilometers, the position of the portal frame is considered as x _k The potential gantry position is taken as the final gantry position,

the specific expression formula of the final portal position is as follows:

wherein ,x_est Representing final gantry position, x' _k In order to obtain the result of the Kalman filtering,

x is the original gantry position _k Is a potential portal location.

By adopting the technical scheme, the invention effectively obtains the information of the position of the portal through ETC portal transaction records and track data of two-passenger one-risk vehicles by using a DR algorithm and a median center algorithm, and calculates the potential position of the portal by using the position information; and inputting the potential portal position and the portal original position into a Kalman filtering model based on a selection strategy to obtain the portal position.

Drawings

The invention is described in further detail below with reference to the drawings and detailed description;

FIG. 1 is a schematic diagram of a multi-source data fusion highway portal positioning method according to the present invention;

FIG. 2 is a schematic diagram of a direction angle algorithm according to the present invention;

FIG. 3 is a schematic diagram of one of the pre-time and post-time points of the present invention through a portal;

fig. 4 is a diagram showing a second point before and a second point after a certain portal according to the present invention.

Detailed Description

For the purposes, technical solutions and advantages of the embodiments of the present application, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application.

The geographical position of the ETC portal has very important significance for evaluating the driving speed of the vehicle and the subsequent driving behavior. The present portal geographic position is manually collected during construction, but because of various reasons, the portal position can have errors or deletions, and aiming at the problem, the invention provides an ETC portal geographic position positioning algorithm based on two-passenger one-risk vehicle track data and ETC transaction data.

As shown in one of fig. 1 to 4, the invention discloses a multi-source data fusion highway portal positioning method, which comprises the following steps:

and step 1, extracting ETC transaction records of two-passenger one-risk vehicles passing through the portal and track data of the vehicles.

Specifically, portal topology data, original portal position data, two-passenger one-risk track data and portal transaction record data are acquired from a data center; and acquiring position data and direction angle data of the vehicle by a wireless signal transmission technology in the running process of the vehicle. Attributes of the two-passenger one-risk trajectory data include positioning time, longitude, latitude, and instantaneous speed.

The two-passenger one-danger vehicles comprise travel package vehicles, more than three classes of bus buses and special road vehicles for transporting dangerous chemicals, fireworks and crackers and civil explosive substances.

Specifically, two-passenger one-risk data attributes: the invention uses the track data of two-passenger one-risk vehicles running on the expressway from No. 3 of 9 months in 2020 to No. 5 of 9 months in 2020, and the vehicles automatically acquire the running information data under the running behaviors at the time and upload the running information data to the data center in the running process of the vehicles. The two-passenger one-risk track data is that the position data, the direction angle and other data of the vehicle are acquired through a wireless signal transmission technology in the running process of the vehicle. The attributes mainly used in the present invention include positioning time, longitude, latitude, instantaneous speed, etc. The acquisition frequency is 10s-1min, and the detailed data attribute is shown in table 1.

TABLE 1 two-passenger one-risk vehicle track data attributes

Sequence number	Field name	Description of the invention	Example
				1	Longitude	Longitude and latitude	119.392654
2	Latitude	Latitude of latitude	26.463268
				3	Speed	Speed of speed	30.22
4	Direction	Direction	0
				5	Time	Time stamp	2018-6-14 11:50:26

Further, the portal transaction record data is a series of time sequence data formed by the vehicle in the high-speed driving process, and the portal transaction record data comprises a transaction identifier TRADEID, a portal number FLAGID, a vehicle license plate OBUPLATE, a traffic identifier PASSID, a vehicle type VEHCLASS and a transaction time TRADETIME.

ETC transaction record data belongs to: ETC portal frame transaction data is a series of time sequence data formed in the high-speed running process of the vehicle, the time sequence data can form a complete track of the vehicle at a high speed, and extraction of dynamic data of the vehicle is achieved. The data attributes mainly include transaction identifier TRADEID, portal frame number FLAGID, vehicle license plate obupelate, pass identifier passsid, vehicle type VEHCLASS, transaction time TRADETIME, and the like. The detailed data attributes are shown in table 2.

Table 2 ETC portal frame transaction data attributes

Sequence number	Field name	Alias name	Description of the invention
				1	TRADEID	Transaction identification ID	Transaction record unique identification
2	FLAGID	Portal frame numbering	ETC portal frame whole network unique number
				3	TRANSTIME	Transaction time	Transaction recordTime of generation
4	PASSID	Pass identifier ID	Vehicle passing unique mark
				5	OBUPLATE	License plate number	License plate number
6	VEHCLASS	Vehicle type	Comprises six kinds of passenger cars, six kinds of trucks and six kinds of special working vehicles
				7	ENSTAION	Import toll station	Initial toll station for expressway on vehicle

Step 2, comparing and extracting a time coincidence point, a time front point and a time rear point according to the transaction time of ETC transaction data and track points in vehicle track data;

extracting track points which are equal to the transaction time TRADETIME of ETC transaction data in vehicle track data, and adding the data of the track points into a potential portal position set; if there is no track point equal to the store time, a front point and a rear point are extracted, the front point is the point with the track positioning time before the transaction time and the minimum time difference, and the rear point is the point with the track positioning time after the transaction time and the minimum time difference.

Carrying out topology correction on two-passenger one-risk portal data according to standard portal topology data; and identifying the existing portal frames without position information and the quantity thereof through ETC transaction data and position data of two-passenger one-risk vehicles, filling the missing position information of the portal frames by using a supplement algorithm, verifying whether the position information is wrong, correcting the position information, and finally generating a section set of the expressway according to the supplemented and corrected portal frames.

Specifically, in the intelligent high speed, the ETC portal is used as road side unit equipment, plays a vital role in the implementation of vehicle-road cooperation, but the erection of the ETC portal is primarily completed, basic data still has problems, such as the position information of partial portal is missing and inaccurate when the information of the ETC portal is recorded. The imperfect portal position information and inaccuracy have important influence on the development and research of vehicle-road coordination. Therefore, in order to solve the problem of data of ETC (electronic toll collection) portal frames constructed and erected, how to quickly and effectively verify the positions of the ETC portal frames, the invention provides a method for identifying the portal frames without position information and the number thereof through ETC transaction data and position data of two-passenger one-risk vehicles, filling the missing position information of the portal frames by a supplement algorithm, verifying whether the position information is wrong or not, correcting the position information, and finally generating a section set of the expressway according to the supplemented and corrected portal frames. For the algorithm described above, the definition of the correlation is given first as follows:

definition 1 (highway segment QD): the portal frame and the entrance and exit (including the entrance and exit of the cross province) of the expressway are commonly called as nodes, and two adjacent nodes form an expressway section QD, which is simply called as a section, and the expression form is shown in a formula 3-2:

QD＝<Node ₁ ,Node ₂ > (3-1)

wherein Node1 is a segment start point, and Node2 is a segment end point; portal nodes are called class-one nodes, and toll gate import and export nodes are called class-two nodes.

Definition 2 (gantry track Ntraj): the portal track formed by the vehicle through the segment QD is called Ntraj as shown in equation 3-2.

Ntraj＝<Node ₁ ,…,Node _n > (3-2)

wherein Node₁ Represents the starting point of the gantry track Ntraj, node _n Representing the end of the gantry track Ntraj.

Because of the problem of portal equipment or other reasons, the track information may have error information, such as false detection, missing detection and repeated detection, so the invention classifies completely correct data into one type of track, has error classified into two types of tracks, and the tracks obtained by the track cleaning algorithm are three types of tracks.

Definition 3 (travel NTrav): the track formed by a vehicle from an entrance toll station to an exit toll station of an expressway is called the journey NTrav of the vehicle, which represents travel as shown in formulas 3 to 3.

Ntrav＝{Ntraj ₁ ,…,Ntraj _n } (3-3)

Wherein Ntraj ₁ Starting portal representing trajectory Ntrav, ntraj _n Represents the end portal of the track Ntrav.

Definition 4 (two-passenger one-danger vehicle GPS track Gtraj) _i GPS point of two-passenger one-danger vehicle _i ): the two-passenger one-danger vehicles comprise travel package vehicles, more than three classes of bus buses and special road vehicles for transporting dangerous chemicals, fireworks and crackers and civil explosive substances. These vehicles collect information on the position, time, speed, and direction angle of the vehicles through GPS devices mounted on the vehicles. The data are arranged according to time sequence, gtraj _i By Gpoint _i The composition is represented by the following formulas 3-4 and 3-5.

Gtraj _i ＝{Gpoint ₁ ,...,Gpoint _n } (3-4)

Gpoint _i ＝{＜lon _i ,lat _i >,t _i ,v _i ,d _i } (3-5)

Wherein Gpoint _i The track point in the two-passenger one-risk track comprises attribute information such as time, longitude, latitude, running speed and direction angle of the vehicle at the moment.

Definition 5 (time coincidence point P) _c ): two-passenger one-danger vehicleTime t in vehicle GPS track point _location Transaction time t with ETC portal _trade Identical, i.e. t _location ＝t _trade The locus point Gpoint _i Referred to as a time coincidence point.

Definition 6 (point before time P _a ): when there is no time coincidence point in the vehicle, it is necessary to find the positioning time t _location At transaction time t _trade Previously, the track point P with the smallest time difference _a I.e. P _a ＝min(t _trade -t _location )

Definition 7 (time point P _b ): when there is no time coincidence point in the vehicle, it is necessary to find the positioning time t _location At transaction time t _trade Then, the track point P with the smallest time difference _b I.e. P _b ＝min(t _location -t _trade )

And 3, calculating a potential portal position set by using a dead reckoning algorithm based on the front point, the rear point and the transaction time.

Point of use P _a And P after time _b Estimating a forward direction eta of the vehicle, and estimating a forward mileage by using speed information and time in a track of the vehicle to obtain a position of the portal

Specifically, based on the data and definitions after the preprocessing, it is known that when two-passenger one-risk vehicles pass through the portal, the position of the portal can be considered as the position of the portal, but since the two-passenger one-risk vehicles sample track points at a certain frequency and the sampling frequency is relatively large, many portals do not have time coincidence points P _c The position of the portal cannot be accurately positioned by simply using the time coincidence point. The present invention thus uses the dead reckoning algorithm (DR algorithm), utilizing the point in time P _a And P after time _b And the speed of the vehicle, the position of the gantry is presumed.

The DR algorithm is a method of estimating the position of the next point by using the direction angle and the distance, and is often used in places where GPS such as ships and airplanes cannot be used effectively. In order to ensure the safety of the offshore navigation, it is important that the navigation personnel must know their own ship position at any time and under any circumstances. Only then can the navigation conditions around the vessel be known on the sea chart according to the ship position, so that necessary navigation measures are taken. The DR algorithm is essentially a process of accumulating information, accumulating displacement vectors at an initial position, and calculating a current position.

In the present invention, the sampling frequency is not dense enough, so the present invention uses the point P before the time _a And P after time _b The forward direction η of the vehicle is estimated, and the forward mileage is estimated by using the speed information and time in the track of the vehicle, so as to obtain the position of the portal. FIG. 2 is a schematic diagram of a DR algorithm, A is a point P before time _a B is the time point P _b N is the north pole and O is the center of the sphere of the earth. L (L) _AB The arc length of the inferior arc corresponding to the point A and the point B is that a is the radian corresponding to the arc NB, B is the radian corresponding to the arc NA, n is the radian corresponding to the arc AB, and the direction angle eta is 0 DEG in true north and is clockwise to 360 deg.

According to definition, the direction angle is NAB, and NAB can be obtained by calculating the spherical triangle sine theorem 3-1:

wherein NAB is the dihedral angle between NOA and AOB, ABN is the dihedral angle between AOB and BON, and ANB is the dihedral angle between AON and NOB.

Bringing the latitude and longitude of A, B points into formula 3-1 yields formula 3-2:

wherein B_Lng Longitude of point B, B _Lat Is the latitude of point B, A _Lat Longitude of point A, A _Lat Is the latitude of point a.

Performing an arcsine transformation on the formula 3-2 to obtain the formula 3-3:

from equation 3-3, when sin (n) is determined, then NAB is known.

sin (n) can be known from the spherical cosine theorem and the Pythagorean theorem, and the calculation formulas are shown in formulas 3-4 and 3-5.

Cos(n)＝cos(a)×cos(b)+sin(a)×sin(b)×cos(∠ANB) (3-4)

Bringing the latitude and longitude of the point A and the point B into the formula 3-4 can obtain the formula 3-6.

cos(n)＝cos(90-B _Lat )×cos(90-A _Lat )+sin(90-B _Lat )×sin(90-A _Lat )×cos(B _Lng -A _Lng ) (3-6)

The invention sets the definition domain of the inverse cosine as

However, since the direction angle η is defined as 0 in the north direction, the clockwise rotation increases, so the ++nab needs to be converted: 1) When the point B is in the first quadrant, the direction angle eta= NAB; 2) The direction angle η=360++nab when the point B is at the second quadrant, 3) the direction angle η=180++nab when the point B is at the third or fourth quadrant.

In order to estimate the portal position, in addition to the forward direction η of the vehicle, a forward distance of the vehicle is required, which may be determined by the travel speed V of the vehicle _drive Travel time t _driveo For the convenience of calculation, the invention uses the vehicle from the time point P _a To the point P after the time _b The running process is assumed to be uniform linear motion, and the speed is the time front point P in the track of the two-passenger one-dangerous vehicle _a Instantaneous speed of moment V _A And a point P after the time _b Instantaneous speed of moment V _B Mean value of (2), calculation formulaAs shown in equations 3-7.

Since the invention assumes that the vehicle is in uniform linear motion at a speed V _drive The direction is eta, but the situations of acceleration, deceleration, lane change, turning and the like can occur in the actual running process of the vehicle, and the DR algorithm deduces the next track point according to the accumulation result of the last track point, so that the shorter the running time is, the smaller the accumulated error is, and the more accurate the positioning portal position is. At this time, a judgment is required to be made on the running time difference between the time front point and the time rear point and the portal, and the portal position is judged by selecting the point with shorter running time. If |t _A -t _trade |<|t _B -t _trade Selecting a time front point to carry out accumulation calculation to obtain a portal position; otherwise, |t _A -t _trade |>|t _B -t _trade And selecting a time later point to carry out accumulation calculation to obtain the position of the portal. The calculation formulas for calculating the position of the gantry by using the time front point and the time rear point are shown as formulas 3-8, 3-9, 3-10 and 3-11 respectively.

Node _Lng ＝A _Lng +V _drive ×t _trade ×sinη/[ARC×cosA _Lat ×2π/360] (3-8)

Node _Lat ＝A _Lat +V _drive ×t _trade ×cosη/[ARC×2π/360] (3-9)

wherein t_trade ＝|t _A -t _trade The ARC is the radius of the earth, 6371393m is taken here and eta is the point P before time _a To P _b Steering angle of travel.

Node _Lng ＝B _Lng +V _drive ×t _trade ×sinη/[ARC×cosB _Lat ×2π/360] (3-10)

Node _Lat ＝B _Lat +V _drive ×t _trade ×cosη/[ARC×2π/360] (3-11)

wherein t_trade ＝|t _B -t _trade I, ARC is groundSphere radius, here 6371393m, η is the point before time P _b To P _a Steering angle of travel.

Step 4, clustering a portal potential position set by using a median center clustering algorithm to obtain portal potential positions;

specifically, each vehicle passing through the portal will obtain a potential portal position, as shown in FIG. 3, where the deep color point is the time point P in the different vehicle trajectories _a The light color point is the time back point P in different vehicle tracks _b Each set of front and rear points can obtain a set of potential portal positions, as shown in fig. 4, gray points are potential portal position sets obtained through a DR algorithm, diamond points are positions of real portal positions, most of the potential portal positions are distributed around the real portal positions, and the few potential portal positions are abnormal points, so that the potential portal positions are far away from roads and portals. It is therefore desirable to extract a set of potential gantry positions from which one position represents all potential gantry positions and to effectively remove outliers.

This is considered a generalized Weber problem in the present invention and uses a median center [6 ]]To find the center point. The algorithm has good robustness, and can effectively identify the center point P of the potential portal position, wherein the P point is closest to all points of the potential portal, and the formula is expressed as formula 3-12. The algorithm calculates the median center P (x _p ，y _p ). The median center to other point distance calculation formulas are shown in formulas 3-13.

wherein s_pj Is the distance from the center of the median to the point elsewhere, w _j Represents the unit distance weights of other points to the median center, and phi is the sum of all the point to median center distances. X is x _j Longitude, y representing other points _j Weft threads representing other pointsDegree.

Based on the theory, the algorithm firstly establishes a data set LocaAll= { location for all portal potential positions ₁ ,locations ₂ ,…,locations _j Location in which _j ＝{x _j ,y _j }，x _j Is the potential portal position longitude, y _j Is a potential portal location longitude. Bringing the formula LocaAll to 3-14 and 3-15 yields an initialized median center

wherein

Representing the longitude and latitude of the initial median center.

With the change of the median center, the unit distance weight between the points is updated using equations 3-16.

wherein w_j Is the initial unit weight, s _pj Representative of the distance between the point and the center of the median of the present invention,

is the unit distance weight after updating.

And carrying out iteration on the center of the median by taking the updated unit weight into formulas 3-14 and 3-15, calculating the distance difference after each iteration, and stopping the iteration process if the distance difference is smaller than a certain threshold value alpha to obtain a final result, wherein the specific algorithm flow is shown in the table 3.

Table 3: transaction data cleaning algorithm based on Dijkstra algorithm

And 5, inputting the potential positions of the portal into a Kalman filtering algorithm based on a selection strategy, and fusing the original portal positions and the potential positions of the portal to obtain the final portal positions.

Specifically, the DR algorithm is relatively dependent on a time front point and a time rear point in a vehicle track, and the motion state between the two points is assumed to be uniform linear motion, so that the change of the motion state of the vehicle and the change of road conditions are not considered, and the positioning offset of a vehicle positioning system is not considered, therefore, the invention uses a Kalman filtering algorithm to fuse the median center with the original portal data, and obtains more accurate longitude and latitude of the ETC portal of the expressway.

The kalman filter mainly includes two models: a predictive model and a follow-up model. The prediction model is mainly used for describing the transition between the states of the system; the new model is responsible for describing the relationship between the state quantity and the observed quantity, and specific formulas are shown as 3-17 and 3-18.

wherein x_k A state vector for the k time period including position information and state information; f is a change matrix, which is a state transition process for the object. B is a control matrix, and the control matrix is a control matrix,

for the control vector, both represent the influence of external behavior on the movement of the object, such as sudden acceleration and deceleration of the vehicle or a change in direction. />

Is the noise of the covariance Q in the system. />

Represents the measurement vector at time k, +.>

Is the measurement noise matrix of covariance R. The specific algorithm steps are shown in Table 4 below, where P _k Represents the state covariance matrix at the moment k, H is x _k and />

Conversion matrix between z _k Is the actual measurement +.>

And a predicted measurement Hx _k The difference vector between K _k The matrix is a posterior covariance matrix for use in state prediction, which is the Kalman gain. />

TABLE 4 Kalman Filter rationale

Algorithm 1: kalman filtering

In the present invention, the existing raw gantry position data is considered as a measurement vector

The potential portal position obtained by the central clustering method is used as a system vector x _k More accurate portal position information is fused through data.

Because of the existing high speed mast position

There are extreme outliers, which cannot be effectively handled using only the kalman data fusion algorithm,because of the extreme outliers present in the raw data, the final result of the data fusion will be greatly affected. The present invention therefore proposes a portal position selection strategy: when the original portal is located->

With potential gantry position x _k When the difference delta is smaller than 2000m, the gantry position is considered as Kalman filtering result x' _k The method comprises the steps of carrying out a first treatment on the surface of the If delta is greater than 2000m km, then the gantry position is considered to be x _k The potential gantry position is taken as the gantry position. The formulas are shown in 3-17.

According to the technical scheme, as shown in table 5, the portal positioning algorithm extracts ETC transaction records of two-passenger one-risk vehicles passing through the portal and the track data of the vehicles, and performs topology correction on the two-passenger one-risk portal data according to standard portal topology data; inputting the front and rear positions of a two-passenger one-risk vehicle through a portal into a Dead Reckoning (DR) algorithm to obtain the position of the vehicle when the vehicle passes through the portal, and considering the position as a portal potential position set; clustering the portal potential position set by using a median center clustering algorithm to obtain portal potential positions; inputting the potential positions of the portal into a Kalman filtering algorithm based on a selection strategy, and fusing the original portal positions and the potential positions of the portal to obtain final portal positions.

Table 5: portal positioning algorithm

Algorithm 3: portal positioning algorithm

/>

It will be apparent that the embodiments described are some, but not all, of the embodiments of the present application. Embodiments and features of embodiments in this application may be combined with each other without conflict. The components of the embodiments of the present application, which are generally described and illustrated in the figures herein, may be arranged and designed in a wide variety of different configurations. Thus, the detailed description of the embodiments of the present application is not intended to limit the scope of the application, as claimed, but is merely representative of selected embodiments of the application. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, are intended to be within the scope of the present application.

Claims (8)

1. A multi-source data fusion highway portal positioning method is characterized in that: which comprises the following steps:

step 1, ETC transaction records of two-passenger one-risk vehicles passing through a portal and track data of the vehicles are extracted;

step 2, comparing and extracting transaction points of all vehicles according to the transaction time of ETC transaction data and track points in vehicle track data;

step 3, calculating a portal potential position set based on the transaction point of the vehicle, wherein the specific steps of the step 3 are as follows:

step 3-1, judging whether the transaction point of the vehicle is a time coincidence point; if yes, adding the corresponding time coincidence point into the portal potential position set and executing the step 3-3; otherwise, executing the step 3-2;

step 3-2, calculating potential portal positions corresponding to the current vehicle by using a dead reckoning algorithm based on the front point, the rear point and the transaction time, and adding the potential portal positions into a portal potential position set; the specific calculation step of the step 3-2 is as follows;

step 3-2-1, A is the point P before time _a B is the time point P _b N is the north pole, O is the sphere center of the earth, N is the radian corresponding to the arc AB, and the cosine value of N is calculated by using the spherical cosine theorem and the Pythagorean theorem, and the calculation formula is as follows:

cos(n)＝cos(90-B _Lat )×cos(90-A _Lat )+sin(90-B _Lat )×sin(90-A _Lat )×cos(B _Lng -A _Lng ) (3-6)

wherein n is the radian corresponding to the arc AB, B _Lng Longitude of point B, B _Lat Is the latitude of point B, A _Lat Longitude of point A, A _Lat The latitude of the point A;

and 3-2-2, calculating a dihedral angle NAB between the NOA and the AOB based on the cosine value, wherein the calculation formula is as follows:

wherein ,B_Lng Longitude of point B, B _Lat Is the latitude of point B, A _Lat Longitude of point A, A _Lat The latitude of the point A;

step 3-2-3, determining the value of the direction angle eta based on the angle NAB;

setting the definition field of the inverse cosine to be

When the point B is in the first quadrant, the direction angle eta= NAB;

when the point B is at the second quadrant, the direction angle eta=360+ & NAB;

when the point B is at the third or fourth quadrant, the direction angle eta=180- & lt NAB;

wherein the direction angle eta is 0 DEG in true north, and is clockwise to 360 DEG;

step 3-2-4, calculating a time difference between a time front point and the transaction time as a front time difference, and simultaneously calculating a time difference between a time rear point and the transaction time as a rear time difference;

step 3-2-5, judging whether the absolute value of the front time difference value is smaller than the absolute value of the rear time difference value; if yes, selecting a time point before the time to carry out accumulation calculation and executing the steps 3-2-6; otherwise, selecting a time later point to carry out accumulation calculation and executing the steps 3-2-6;

step 3-2-6, substituting the selected time point or the time point into a corresponding formula to calculate the position of the portal:

when the accumulation calculation is performed at the time point before the selection, the calculation formula of the portal position is as follows:

Node _Lng ＝A _Lng +V _drive ×t _trade ×sinη/[ARC×cosA _Lat ×2π/360] (3-8)

Node _Lat ＝A _Lat +V _drive ×t _trade ×cosη/[ARC×2π/360] (3-9)

wherein ,t_trade ＝|t _A -t _trade Absolute value of i, i.e. the previous time difference, ARC is the earth radius and η is the time previous point P _a To P _b The direction angle of travel;

when the accumulation calculation is performed at the selected time, the calculation formula of the portal position is as follows:

Node _Lng ＝B _Lng +V _drive ×t _trade ×sinη/[ARC×cosB _Lat ×2π/360] (3-10)

Node _Lat ＝B _Lat +V _drive ×t _trade ×cosη/[ARC×2π/360] (3-11)

wherein ,t_trade ＝|t _B -t _trade Absolute value of i, i.e. the difference in time after, ARC is the earth radius and η is the point before time P _b To P _a The direction angle of travel;

step 3-3, acquiring transaction point data of the next vehicle and executing step 3-1 until all vehicles finish calculation to form a final portal potential position set;

step 4, clustering a portal potential position set by using a median center clustering algorithm to obtain portal potential positions;

and 5, inputting the potential positions of the portal into a Kalman filtering algorithm based on a selection strategy, and fusing the original portal positions and the potential positions of the portal to obtain final portal positions.

2. The multi-source data fusion highway portal positioning method according to claim 1, wherein the method comprises the following steps: the two-passenger one-danger vehicles comprise travel package vehicles, more than three classes of bus buses and road special vehicles for transporting dangerous chemicals, fireworks and crackers and civil explosive articles; attributes of the two-passenger one-risk trajectory data include positioning time, longitude, latitude, and instantaneous speed.

3. The multi-source data fusion highway portal positioning method according to claim 1, wherein the method comprises the following steps: the portal transaction record data is a series of time sequence data formed by the vehicle in the high-speed driving process, and comprises a transaction identifier TRADEID, a portal number FLAGID, a vehicle license plate OBUPLATE, a traffic identifier PASSID, a vehicle type VEHCLASS and a transaction time TRADETIME.

4. A multi-source data fusion highway portal positioning method according to claim 3, wherein: the specific steps of the step 2 are as follows:

step 2-1, acquiring the transaction time TRADETIME of ETC transaction data of the current vehicle;

step 2-2, judging whether track points which are equal to the transaction time TRADETIME of ETC transaction data exist in the vehicle track data; if yes, extracting a corresponding track point as a time coincidence point, taking the time coincidence point as a transaction point of a corresponding vehicle, and executing the steps 2-4; otherwise, executing the step 2-3;

step 2-3, taking one track point with the smallest time difference between the vehicle track data acquisition transaction times TRADETIME as a time front point, and taking one track point with the smallest time difference between the vehicle track data acquisition transaction times TRADETIME as a time rear point; and takes the time point before and the time point after as the transaction points of the corresponding vehicles

And 2-4, ending the extraction of the transaction points of the current vehicle, acquiring the track data of the next vehicle, and executing the step 2-1 until all vehicles finish the extraction of the transaction points.

5. The multi-source data fusion highway portal positioning method according to claim 1, wherein the method comprises the following steps: in the step 4, a method for searching a center point by adopting a median center is adopted to determine the potential position of the portal, and the specific steps are as follows:

step 4-1, utilizing all portal potential positionsCenter of median for initialization of setup dataset localall

Wherein, localall= { location ₁ ,locations ₂ ,…,locations _j }，location _j ＝{x _j ,y _j }，x _j Is the j potential portal position longitude, y _j Is the j-th potential portal location longitude;

longitude and latitude representing the initial median center;

step 4-2, extracting portal information from the track data and calculating the distance from the potential portal position to the median center;

/>

wherein ,s_pj Representing the distance between the point and the center of the median;

step 4-3, judging s _pj Whether the value of (a) is less than a threshold value alpha; if yes, stopping the iterative process to obtain the final result

Otherwise, executing the step 4-4;

and 4-4, updating the unit distance weight from the point to the median center according to the change of the median center, wherein the updating formula is as follows:

wherein ,w_j Is the initial unit weight, s _pj Representing the distance between the point and the center of the median,

the unit distance weight after updating;

step 4-5, iterating the center of the median by using the updated unit weight,

step 4-6, judging whether all portal information in the track data is extracted; if yes, stopping the iterative process to obtain the final result

Otherwise, step 4-2 is performed.

6. The multi-source data fusion highway portal positioning method according to claim 5, wherein the positioning method comprises the following steps: in step 5, the existing raw gantry position data is regarded as a measurement vector

The potential portal position obtained by the central clustering method is used as a system vector x _k-1 More accurate portal position information is fused through data.

7. The multi-source data fusion highway portal positioning method according to claim 5, wherein the positioning method comprises the following steps: the specific steps of the Kalman filtering algorithm based on the step 5 are as follows:

step 5-1, initializing x _k-1 ,Q,R,P _k-1 ,F,H，

wherein ,x_k-1 The longitude and latitude of the potential portal position obtained by the central clustering method are obtained, Q is an error matrix of the potential portal position obtained by the central clustering method and the true value, and R is the longitude and latitude of the original portal

Error matrix with true value, P _k-1 To indicate x _k-1 And (3) with

Covariance matrix of errors between the two, H is X _k and />

The method comprises the steps of converting into a conversion matrix with a unified unit, wherein F is a state transition matrix, and F is a state conversion process for an object;

step 5-2, carrying out state prediction, wherein the state prediction formula is as follows

wherein ,x_k Represents x _k-1 A new position obtained through iteration is used for indicating the drift of the position of the portal; b is a control matrix, and the control matrix is a control matrix,

for controlling vector +.>

To coordinate in the systemNoise of variance Q, P _k Representing P _k-1 The error covariance matrix obtained through iteration, namely, representing x _k And->

Errors between;

step 5-3, obtaining a measurement of time k

S _k ＝(HP _k H ^T +R)

wherein ,

is the actual measurement +.>

And a predicted measurement Hx _k Difference vector between S _k To represent the residual covariance matrix, express +.>

And the difference between the true values;

and 5-4, correcting the state based on the measured value, wherein the specific formula is as follows:

K _k ＝P _k H ^T S _k ^-1

P _k ′＝(I-K _k H)P _k

wherein ,K_k For Kalman gain, x' _k To use Kalman filteringLongitude and latitude of potential portal position and original portal data obtained by wave-pair central clustering method

A fused result; p'. _k Represents x _k-1 And->

The covariance matrix of the errors between the two is P _k-1 The method comprises the steps of carrying out a first treatment on the surface of the I is an identity matrix.

8. The multi-source data fusion highway portal positioning method according to claim 5, wherein the positioning method comprises the following steps: in step 5, when the original portal frame is positioned

With potential gantry position x _k When the difference delta is smaller than or equal to 2000m, the final gantry position is considered as Kalman filtering result x' _k ；

When delta original portal position

With potential gantry position x _k If the difference delta is larger than 2000m kilometers, the position of the portal frame is considered as x _k The potential gantry position is taken as the final gantry position,

the specific expression formula of the final portal position is as follows:

wherein ,x_est Representing final gantry position, x' _k In order to obtain the result of the Kalman filtering,

x is the original gantry position _k Is a potential portal location. />

Images