CN109686090A - A kind of virtual traffic method of calculating flux based on multisource data fusion - Google Patents
A kind of virtual traffic method of calculating flux based on multisource data fusion Download PDFInfo
- Publication number
- CN109686090A CN109686090A CN201910042487.2A CN201910042487A CN109686090A CN 109686090 A CN109686090 A CN 109686090A CN 201910042487 A CN201910042487 A CN 201910042487A CN 109686090 A CN109686090 A CN 109686090A
- Authority
- CN
- China
- Prior art keywords
- track
- individual
- section
- time window
- population
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- 230000004927 fusion Effects 0.000 title claims abstract description 13
- 230000004907 flux Effects 0.000 title claims abstract description 12
- 238000005457 optimization Methods 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims abstract description 10
- 230000000903 blocking effect Effects 0.000 claims abstract description 8
- 230000006978 adaptation Effects 0.000 claims description 18
- 230000002068 genetic effect Effects 0.000 claims description 17
- 230000035772 mutation Effects 0.000 claims description 12
- 230000002159 abnormal effect Effects 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 4
- 230000008569 process Effects 0.000 description 6
- 230000008859 change Effects 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 241001269238 Data Species 0.000 description 3
- 238000001514 detection method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 230000001174 ascending effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0125—Traffic data processing
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/065—Traffic control systems for road vehicles by counting the vehicles in a section of the road or in a parking area, i.e. comparing incoming count with outgoing count
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/123—Traffic control systems for road vehicles indicating the position of vehicles, e.g. scheduled vehicles; Managing passenger vehicles circulating according to a fixed timetable, e.g. buses, trains, trams
- G08G1/127—Traffic control systems for road vehicles indicating the position of vehicles, e.g. scheduled vehicles; Managing passenger vehicles circulating according to a fixed timetable, e.g. buses, trains, trams to a central station ; Indicators in a central station
- G08G1/13—Traffic control systems for road vehicles indicating the position of vehicles, e.g. scheduled vehicles; Managing passenger vehicles circulating according to a fixed timetable, e.g. buses, trains, trams to a central station ; Indicators in a central station the indicator being in the form of a map
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a kind of virtual traffic method of calculating flux based on multisource data fusion, comprising the following steps: Step 1: obtaining floating vehicle travelling track；Step 2: clustering after carrying out blocking processing to floating vehicle travelling track, and obtain the hot spot track in every class；Step 3: recording to obtain vehicle observed volume data of the bayonet corresponding road section in time windows using bayonet；Step 4:, based on the section in the time window with vehicle observed volume data, optimization objective function is established, to solve the expansion spline coefficient of each hot spot track in the time window for any time window T；Step 5: method solves optimization objective function；Step 6: the link flow that tracks therethrough all in the time window generate is superimposed using corresponding expand after spline coefficient carries out expansion sample, the virtual traffic flow as the section for the section for not having vehicle observed volume data in time window T.The present invention can obtain the traffic flow data with precision and range.
Description
Technical field
The present invention relates to a kind of virtual traffic method of calculating flux based on multisource data fusion.
Background technique
In urban highway traffic status analysis, road traffic flow is an important index, can be urban congestion etc.
Analysis, the early warning of situation provide important data basis, while can also provide for the planning, control, induction of urban highway traffic
Data are supported.The acquisition of traditional cities road traffic flow is mainly by ring coil detector or video detector come real
It is existing, have the advantages that precision is higher, but due to input cost is high, approval process is complicated, use environment is severe etc., cause
The data volume of acquisition is smaller, and many roads can not cover in city road network.Floating Car GPS data is that a kind of coverage area is wider
City dweller's crawler behavior data, can preferably represent the feature of Urban Residential Trip, have the advantages that range is higher, but
Since Floating Car is only a part in motor vehicle, thus its flow cannot be carried out directly as traffic flow data using.It is comprehensive
Upper described, the acquisition methods of Forecast of Urban Traffic Flow exist at present or data coverage is smaller, it is difficult to the high data of range are obtained,
And it is limited simultaneously by input cost, hardware facility etc. condition；Or data integrity degree itself is inadequate, it is difficult to as height
The defect that accuracy data uses.
There are certain requirement, magnitude of traffic flow number to the precision and range of traffic flow data in urban highway traffic research
According to precision and range affect the result of traffic decision-making.Therefore, urgent need provides one kind and has both higher precision and range
Traffic flow data acquisition methods.
Summary of the invention
Technical problem solved by the invention is, for the deficiency of available data, to propose a kind of based on multisource data fusion
Virtual traffic method of calculating flux, by fusion Floating Car GPS data and bayonet data, to calculate in city road network without vehicle
The virtual traffic flow in the section of observed volume data, data acquisition is convenient, has real-time and higher precision and range, fits
Formulation for urban highway traffic research and successive policy.
A kind of virtual traffic method of calculating flux based on multisource data fusion, comprising the following steps:
Step 1: Floating Car GPS data is distributed by map-matching method to city road network, the row of Floating Car is obtained
Sail trace information；
Step 2: carrying out blocking processing to floating vehicle travelling track, to guarantee the track between two bayonets without it
His bayonet；Divide time window to carry out the cluster based on time-space matrix the track obtained after cell processing, and obtains in every class
Hot spot track；
Step 3: vehicle of the bayonet corresponding road section in time windows is obtained using the traffic flow data that bayonet records
Observed volume data；
Step 4: definition expands spline coefficient and carries out expansion sample come the link flow generated to known trajectory；For each time window
T constructs an expansion spline coefficient set A={ α respectivelyk| k=1,2 ..., K }, size K, K are full track data in the time
Hot spot track number in window.Any time window T is built based on the section in the time window with vehicle observed volume data
Vertical optimization objective function, for solving the corresponding expansion spline coefficient set A of the time window:
F=Minimize Z
VE(k)=αkVT(k)
Wherein, f is target function value, and Z is fitness, and i and j are the roads in time window T with vehicle observed volume data
Section, N are the sum in the section in time window T with vehicle observed volume data, VR(i) be section i in time window T vehicle see
Measurement of discharge, PiJ is the hot spot track set in time window T from section i to section j, VTIt (k) is that hot spot track k is produced in time window T
Raw link flow, value are equal to the number that hot spot track k occurs in time window T, VEIt (k) is VT(k) expand the flow after sample, αk
It is the expansion spline coefficient of hot spot track k；
Step 5: solving optimization objective function using genetic algorithm, obtains the corresponding expansion spline coefficient set of time window T
Global optimum is as a result, be denoted as A*；
Step 6: for not having the section of vehicle observed volume data in time window T, by all processes in the time window
The link flow that its track generates is superimposed using corresponding expand after spline coefficient carries out expansion sample, the virtual traffic stream as the section
Amount；Specifically, for the hot spot track by section l, expand the link flow that spline coefficient generates it using it and carry out expansion sample；It is right
In the non-thermal locus of points by section l, it is first determined with its maximally related hot spot track, determine method are as follows: it is non-hot to obtain this
Common segment of each hot spot track, calculates separately the length of each common segment, by longest common segment pair in track and cluster belonging to it
The hot spot track answered is considered as its maximally related hot spot track；Then it utilizes with the expansion spline coefficient of its maximally related hot spot track to it
The link flow of generation carries out expansion sample.
Further, in the step one, to guarantee that the trip of Floating Car has Urban Residential Trip feature
Trip, filters out the Floating Car GPS data of passenger carrying status, and wherein each GPS point is a three-dimensional vector, and the 1st component is
World concordant time, the 2nd component are the longitude of Floating Car position, and the 3rd component is the latitude of Floating Car position；
GPS point is matched on city road network using ST-Matching map-matching algorithm, obtains the driving trace information of Floating Car,
Wherein each tracing point is a three-dimensional vector, and the 1st component is that section is passed through the time, and the 2nd component is track number, the 3rd
A component is to number by section；
Further, in the step two, Floating Car GPS data is after the processing by map-matching algorithm, GPS
Point is matched on section, and the expression-form of track is changed by tracing point Point-A, Point-B...Point-C, Point-D
For section Road-U, Road-V ..., Road-Y, Road-Z；In order to guarantee the track between two bayonets without other bayonets,
In order to during subsequent assignment of traffic will not duplicate allocation flow, to track carry out blocking processing, specific steps are such as
Under:
All tracks are numbered, item identifies the bayonet section number that every track is passed through one by one, if certain track is passed through
Bayonet section number M≤1, then abandon the track；If the bayonet section number M=2 that certain track is passed through, which is passed through
Track between two bayonet sections retains by way of section as new track；If bayonet section number M >=3 that certain track is passed through,
Then this track is split, since first bayonet section that the track is passed through, every two phase which is passed through
Track between adjacent bayonet section retains by way of the section track new as one.
Further, in the step two, carrying out the cluster based on time-space matrix to track is that similarity is high
Track integration, and using the hot spot track in a kind of track as such representative track, to subsequent calculating step
Simplified.Specific step is as follows:
2.1) untreated state is set by all tracks, and sets hot spot track collection and is combined into empty set；
2.2) a track Traj is randomly selected, track collection identical as its origin and destination and in same time window is obtained
It closes, i.e., the space-time of track Traj is neighbouring set C (Traj)；
2.3) classify to the track in set C (Traj), wherein kth class track is denoted as CTk, CTkIn any two rails
The overlap length ratio q of mark is greater than the minimum overlay length ratio Minq (empirical value) of setting, and adds the hot spot in every one kind
Track to hot spot track set；Wherein overlap length ratio q possesses road divided by this two tracks for two tracks overlapping section numbers
The mean value of number of segment, hot spot track (HT, Hot Trajectory) are that all tracks in a kind of track are arranged by frequency of occurrence descending
After column, take account for n% before the sum of all track frequency of occurrence in such track (value range of n be [0,100], be experience
Value)；Tag set C (TrajiIn track be processed state；
2.4) judge whether that all tracks are all marked as processed state, if so, terminating, otherwise with untreated state
Track based on, return step 2.2).
Further, in the step three, using DBSCAN clustering algorithm to the same card of time window same in more days
The traffic flow data of mouth record is clustered, and removes exception stream magnitude, wherein cluster the point for including in the maximum cluster of acquisition
It is considered as normal stream magnitude, remaining point is considered as exception stream magnitude；Meanwhile if being less than record flow comprising point number in maximum cluster
The 50% of number of days is then considered as the bayonet hardware (bayonet test device) and is abnormal, and removes the traffic flow data of bayonet record；
Bayonet is matched on the section of its detection, each bayonet match information is a bivector, and the 1st component is bayonet volume
Number, the 2nd component is section number；The traffic flow data that remaining bayonet is recorded is as the bayonet corresponding road section (with this
The matched section of bayonet) vehicle observed volume data.
Further, in the step five, the solution of optimization objective function is carried out by genetic algorithm, it is specific to walk
It is rapid as follows:
5.1) initialize: setting evolutionary generation counter g=0, n individual of random generation are used as initial population, each
Individual is an expansion spline coefficient set A={ αk| k=1,2 ..., K }, wherein element αkValue given birth at random in the range of [0,1]
At；Setting genetic algebra is Ng, convergence precision Q；
5.2) individual choice: the probability that q-th of individual is selected in population in g generation is calculatedWherein Zg(q)Indicate that g, i.e., will be in the individual for the fitness of q-th of individual in population
Element substitutes into optimization objective function, obtained fitness value；The probability being selected according to Different Individual is continuously to g for population
In individual repeat n wheel and select, obtain the new individuals of n；
5.3) individual intersection and variation:
The n individual that step 5.2) is obtained carries out random pair two-by-two, takes fixed crossover probability pcCarry out crossover operation；
If certain group individual need is intersected, a crosspoint is randomly generated in all elements of group individual, by group individual
Element after crosspoint is exchanged with each other, and generates two new individuals；If certain group individual does not need to be intersected, group individual
It remains unchanged；
Take fixed mutation probability pmMutation operation is carried out, n individual after successively selecting crossover operation, to each individual
All elements traversed, if some element needs to make a variation, change the value of the element at random in the range of [0,1]；
It results in through n individual after heredity, constitutes g+1 for population；
Wherein crossover probability pcWith mutation probability pm, it is calculated respectively according to following formula:
Wherein, pc1It is the crossover probability for the individual that g is less than average fitness value for fitness value in population, pc2It indicates
Crossover probability of the g for the individual in population with maximum adaptation angle value, FmaxIt is g for the maximum adaptation of all individuals of population
Angle value, FaveIt is g for the average fitness value of all individuals of population, F*It is biggish adaptation in two individuals intersected
Angle value；pm1It is the mutation probability for the individual that g is less than average fitness value for fitness value in population, pm2It is g in population
The mutation probability of individual with maximum adaptation angle value, F ' are the fitness values of the individual to make a variation；pc1、pc2、pm1、pm2For
Empirical parameter (can be adjusted by many experiments and be obtained)；
5.4) g+1 is calculated for the fitness of individual each in population, and the maximum for obtaining g+1 for all individuals of population is suitable
Response, i.e. g+1 are for the corresponding target function value f of population；
5.5) algorithm termination condition judges；
Convergence judgement: in g+1 generation and g, are compared for target function value f corresponding to population, if difference is less than
Precision Q, then it is assumed that meet convergence, algorithm terminates, using g+1 for the individual in population with maximum adaptation degree as the complete of A
Office's optimal result；
Genetic algebra judgement: if current genetic algebra g=Ng, then algorithm terminates, by this NgFor in population have maximum adaptation
Global optimum result of the individual of degree as A；
If being unsatisfactory for above-mentioned algorithm termination condition, genetic algebra g=g+1 is enabled, is grasped back to step 5.2) circulation
Make.
Since hot spot track is the representative of full track mark, and the lower section of some category of roads is not by hot spot track institute
Covering, but covered by the remaining non-thermal locus of points is covered and do not covered by hot spot track by the non-thermal locus of points to fill up these
The flow in the section of lid, the invention proposes a kind of calculations based on hot spot track and non-hot Track association in trajectory clustering
Method, the supplement of Lai Jinhang data on flows.According to trajectory clustering in step 3, similar track space-time with higher is similar
Degree, hot spot track therein and the non-thermal locus of points have certain relevance, therefore utilize relevance by the expansion sample system of hot spot track
Number is transmitted on the similar non-thermal locus of points, the specific steps are as follows:
6.1) all non-thermal locus of points in acquisition time window T, and be marked as untreated；
6.2) optional one untreated non-thermal locus of points non-HT determines its affiliated class in trajectory clustering, obtain its with
The common segment of each hot spot track in class, calculates separately the length of each common segment, and the corresponding hot spot track of longest common segment is regarded
For its maximally related hot spot track；
6.3) using the expansion spline coefficient with its maximally related hot spot track as the expansion spline coefficient of the non-thermal locus of points；Label should
The non-thermal locus of points is processed；It repeats 6.2).
Further, in the step six, by the expansion spline coefficient acquired be applied to whole track by way of section
On, can calculate according to the following formula single track Traj by way of the expansion sample stream that is generated by the single track of i-th section
Amount:
tReal, i=tTraj, i*αi
I=1,2 ..., u
Wherein, tTraj, iFor track Traj by way of the link flow that is generated by track Traj of i-th section, αiFor track
The expansion spline coefficient of Traj, tReal, iFor tTraj, iCarry out the flow value obtained after expansion sample, u be track Traj by way of section quantity.
After obtaining the expansion sample flow that each section is generated by single track therethrough, to any section, by institute therethrough
The expansion sample flow for having track to generate is overlapped, and can acquire the virtual traffic flow in the section:
Wherein VSum, rFor the flow total amount that section r is obtained after expanding sample, that is, the virtual traffic flow of the section r required, s
For by the tracking quantity of section r.
The utility model has the advantages that
The present invention provides a kind of combination multi-source data, the characteristic of different data is made full use of to obtain traffic flow data
Method, the traffic flow data of acquisition has precision and range simultaneously, and the present invention is to the traffic programme in intelligent transportation, traffic
Tissue, traffic administration and control, traffic safety all have very high real value.
Detailed description of the invention
Fig. 1 is flow diagram of the present invention
Fig. 2 is trajectory unit schematic diagram
Fig. 3 is the virtual flow result figure of the 8:00-9:00 on the 1st of September in 2016
Specific embodiment
Further detailed number of seconds is made to the present invention with specific implementation example with reference to the accompanying drawing, but not as limit of the invention
It is fixed.
Floating Car GPS data used in the present embodiment is China Shenzhen Luohu District and Futian District on September 1, to 30 2016
Day data, record include a record point longitude coordinate, latitude coordinate and time tag information, and data record summary is 202,391,
745, Floating Car sum is 8,168；Bayonet data are 1 to 14 data of China Shenzhen September in 2016, are amounted to 14 days.
The present invention specific implementation the following steps are included:
Step 1: Floating Car GPS data record sum is 202,391,745, the vehicle 8,168 for having passenger carrying status to record
?.
Step 1: by continuous carrying record as primary trip, decision condition are as follows: continuous 3 or more record, is adjacent
The time difference is recorded less than 35 seconds, adjacent record linear distance less than 1500m.
Step 2: filtering out effective trip, all trips are screened with the following conditions: trip radius distance 0.5km <
Distance < 32km；Travel time 3min < Time < 60min；Go on a journey radius speed 0.5km/h < Speed < 120km/h.
Obtain 420,129 trips altogether by screening.
Step 3:ST-Matching map-matching algorithm is a kind of map-matching algorithm suitable for low sampling rate, is utilized
Effective trip GPS point is matched on city road network by the map-matching algorithm, and obtains corresponding trace information；Algorithm specifically walks
It is rapid as follows:
1) track trajectory and Shenzhen's road network information that input taxi is once gone on a journey；
2) Shenzhen's road network is divided into the regional scope of 10*10；
3) after road network being divided into zonule, the trip trace information in same zonule is subjected to section matching, it will
Tracing point, according to being matched on candidate side, obtains the path of path matching using distance；
Step 2: after the processing by map-matching algorithm, tracing point is matched on section Floating Car GPS data,
The expression-form of track is changed into section Road-U, Road- by tracing point Point-A, Point-B ... Point-C, Point-D
V ..., Road-Y, Road-Z；In order to guarantee the track between two bayonets without other bayonets, in order in subsequent flow
In assigning process will not duplicate allocation flow, to track carry out blocking processing, the specific steps are as follows:
420,129 tracks are numbered, respectively 1,2 ... ..., 420,129, track carries out identification process one by one
Bayonet quantity, if bayonet section number M≤1 that certain track is passed through, abandons the track；If the bayonet section that certain track is passed through
Number M=2 then retains the track between two bayonet sections by way of section, retains in this, as new track；If certain track is passed through
Bayonet section number M >=3, then the track is split, since the track pass through first bayonet section, by the track
Track between the adjacent bayonet section of the every two of process retains by way of the section track new as one.
As shown in Fig. 2, A, B, C are respectively 3 bayonets, then it is 2 (AB by the track after blocking track, obtained
With BC).
Trajectory clustering based on space-time is the track integration that similarity is high, and utilizes the hot spot in a kind of track
Track further simplifies subsequent calculating step as such representative track.The specific steps of trajectory clustering
It is as follows:
2.1) untreated state is set by all tracks, and sets hot spot track collection and is combined into empty set；
2.2) a track Traj is randomly selected, track collection identical as its origin and destination and in same time window is obtained
It closes, i.e., the space-time of track Traj is neighbouring set C (Traj)；
2.3) classify to the track in set C (Traj), wherein kth class track is denoted as CTk, CTkIn any two rails
The overlap length ratio q of mark is greater than the minimum overlay length ratio Minq (0.7 is set as in the present embodiment) of setting, and adds every
Hot spot track to hot spot track set in one kind；Wherein overlap length ratio q be two tracks be overlapped section numbers divided by this two
Track possesses the mean value of section number, and hot spot track (HT, Hot Trajectory) is that all tracks in a kind of track press out
Occurrence number descending arrangement after, take account for n% before the sum of all track frequency of occurrence in such track (value range of n be [0,
100], it is empirical value, is set as 90) in the present embodiment；Tag set C (Traji) in track be processed state；
2.4) judge whether that all tracks are all marked as processed state, if so, terminating, otherwise with untreated state
Track based on, return step 2.2)
9,132 track classes and 54,796 hot spot tracks are obtained altogether by step 2.
Step 3: bayonet data have 5,287,649 enumeration datas altogether in detection in 14 days, and definition number of days occurs and is greater than
Vehicle equal to 2 days is to commonly use vehicle, totally 287 ten thousand；Bayonet test device is matched with record section, including bayonet test
Crossing and direction, in this, as the observed volume of the section in this direction.
Due to bayonet probably due to hardware or other external factor cause failure, thus some abnormal datas are generated, in order to
These abnormal datas are removed, DBSCAN method used herein is clustered, to remove abnormal data；Cluster the Clique obtained
The point for including in cluster is considered as normal value, remaining point is considered as exceptional value；If being less than record flow comprising point number in maximum cluster
The 50% of number of days is then considered as the bayonet test device hardware and is abnormal, the magnitude of traffic flow for not using the bayonet test device to record.
There are two main major parameters needed for DBSCAN clustering algorithm: a parameter is radius (Eps), is indicated with given
The range of circle shaped neighborhood region centered on point A (in the present invention, A indicates flow)；Another parameter is the circle centered on point A
The quantity (MinAts) at least put in neighborhood.If meet: centered on point A, radius for the point in the neighborhood of Eps number not
Less than MinAts, then point P is referred to as core point.
By bayonet, at more days, the data on flows of window record was denoted as data set A={ a (i) i=1 ... n } at the same time,
Middle p (i) indicates the bayonet in the flow of i-th day time window；For each point A (i), the son that point A (i) arrives set A is calculated
Collect the distance between all the points in B={ a (1), a (2) ..., a (i-1), a (i+1) ..., a (n) }, distance according to from it is small to
Big sequence sequence, the distance set after being sorted are D={ d (1), d (2) ..., d (k-1), d (k), d (k+1) ..., d
(n) }, wherein d (k) is known as k- distance (it is close apart from kth between all the points other than a (i) point that k- distance is that point a (i) is arrived
Distance)；
The value of k in empirically determined k- distance, so that it is determined that the quantity MinAts at least put；K=is taken in the present embodiment
4, then MinAts=4；
It treats each point a (i) in cluster set and calculates k- distance, finally obtain the k- distance set E={ e of all the points
(1),e(2),…,e(n)}。
According to the k- distance set E of obtained all the points, k- distance set E ' is obtained after carrying out ascending sort to set E,
It is fitted the change curve of k- distance in E ' set, in change curve, x-axis coordinate point is directly using incremental natural number
Sequence, each pair of point answer a natural number, and y-axis coordinate point is k- distance in E ' set；Select the most express delivery in change curve
Increase point and be used as flow radius Eps, the corresponding k- distance of the two o'clock of maximum slope is averagely steepest incremented point；
According to the value of given MinAts and the value of radius Eps, all core points are calculated；According to obtained core point set
The value of conjunction and radius Eps calculate the core point that can be connected to；The each group of core point that will be connected to, and arrive core point
Distance is less than the point of radius Eps, all puts together, forms a cluster；Thus cluster obtains one group of cluster；
The present invention utilizes rejecting outliers method of this kind based on density, can effectively resist exceptional value (" noise ")
Interference.In the present embodiment, the effective percentage for obtaining bayonet record data is 72%.
Step 4: definition expands spline coefficient and carries out expansion sample come the link flow generated to known trajectory；For each time window
T constructs an expansion spline coefficient set A={ α respectivelyk| k=1,2 ..., K }, size K, K are full track data in the time
Hot spot track number in window.1 hour is used in this example as a time window, the expansion spline coefficient collective number of acquisition is 24.
Optimization objective function is established based on the section in the time window with vehicle observed volume data for any time window T,
For solving the corresponding expansion spline coefficient set A of the time window:
F=Minimize Z
VE(k)=αkVT(k)
Wherein, f is target function value, and Z is fitness, and i and j are the roads in time window T with vehicle observed volume data
Section, N are the sum in the section in time window T with vehicle observed volume data, VR(i) be section i in time window T vehicle see
Measurement of discharge, PijIt is the hot spot track set in time window T from section i to section j, VTIt (k) is that hot spot track k is produced in time window T
Raw link flow, value are equal to the number that hot spot track k occurs in time window T, VEIt (k) is VT(k) expand the flow after sample, αk
It is the expansion spline coefficient of hot spot track k.
Optimization objective function is solved using genetic algorithm, obtains the corresponding overall situation for expanding spline coefficient set of time window T most
Excellent result, the specific steps are as follows:
5.1) initialize: setting evolutionary generation counter g=0, n individual of random generation are used as initial population, each
Individual is an expansion spline coefficient set A={ αk| k=1,2 ..., K }, wherein element αkValue given birth at random in the range of [0,1]
At；Setting genetic algebra is Ng, convergence precision Q；In this example, Ng=1500；
5.2) individual choice: the probability that q-th of individual is selected in population in g generation is calculatedWherein Zg(q)Indicate that g, i.e., will be in the individual for the fitness of q-th of individual in population
Element substitutes into optimization objective function, obtained fitness value；The probability being selected according to Different Individual is continuously to g for population
In individual repeat n wheel and select, obtain the new individuals of n；
5.3) individual intersection and variation:
The n individual that step 5.2) is obtained carries out random pair two-by-two, takes fixed crossover probability pcCarry out crossover operation；
If certain group individual need is intersected, a crosspoint is randomly generated in all elements of group individual, by group individual
Element after crosspoint is exchanged with each other, and generates two new individuals；If certain group individual does not need to be intersected, group individual
It remains unchanged；
Take fixed mutation probability pmMutation operation is carried out, n individual after successively selecting crossover operation, to each individual
All elements traversed, if some element needs to make a variation, change the value of the element at random in the range of [0,1]；
It results in through n individual after heredity, constitutes g+1 for population；
Wherein crossover probability pcWith mutation probability pm, it is calculated respectively according to following formula:
Wherein, pc1It is the crossover probability for the individual that g is less than average fitness value for fitness value in population, pc2It indicates
Crossover probability of the g for the individual in population with maximum adaptation angle value, FmaxIt is g for the maximum adaptation of all individuals of population
Angle value, FaveIt is g for the average fitness value of all individuals of population, F*It is biggish adaptation in two individuals intersected
Angle value；pmIt is the mutation probability for the individual that g is less than average fitness value for fitness value in population, pm2It is g in population
The mutation probability of individual with maximum adaptation angle value, F ' are the fitness values of the individual to make a variation；pc1、pc2、pm1、pm2For
Empirical parameter；In the present embodiment, pc1Take 0.9, pm1Take 0.1, pc2And pm2Two class values, p are taken respectivelyc2=0.6 and pm2=0.001 or
pc2=0.5 and pm2=0.05；
5.4) g+1 is calculated for the fitness of individual each in population, and the maximum for obtaining g+1 for all individuals of population is suitable
Response, i.e. g+1 are for the corresponding target function value f of population；
5.5) algorithm termination condition judges；
Convergence judgement: in g+1 generation and g, are compared for target function value f corresponding to population, if difference is less than
Precision Q, then it is assumed that meet convergence, algorithm terminates, using g+1 for the individual in population with maximum adaptation degree as the complete of A
Office's optimal result；
Genetic algebra judgement: if current genetic algebra g=Ng, then algorithm terminates, by this NgFor in population have maximum adaptation
Global optimum result of the individual of degree as A；
If being unsatisfactory for above-mentioned algorithm termination condition, genetic algebra g=g+1 is enabled, is grasped back to step 5.2) circulation
Make.
To every each time window, calculating through the above steps, can obtain the corresponding 24 expansions sample of 24 time windows respectively
Coefficient sets.
Step 6: for not having the section of vehicle observed volume data in time window T, by all processes in the time window
The link flow that its track generates is superimposed using corresponding expand after spline coefficient carries out expansion sample, the virtual traffic stream as the section
Amount；Specifically, for the hot spot track by section l, expand the link flow that spline coefficient generates it using it and carry out expansion sample；It is right
In the non-thermal locus of points by section l, it is first determined with its maximally related hot spot track, determine method are as follows: it is non-hot to obtain this
Common segment of each hot spot track, calculates separately the length of each common segment, by longest common segment pair in track and cluster belonging to it
The hot spot track answered is considered as its maximally related hot spot track；Then it utilizes with the expansion spline coefficient of its maximally related hot spot track to it
The link flow of generation carries out expansion sample.
The virtual traffic flow value passed through in certain time window T obtained after calculating in this example is as shown in Figure 3.
It is different with traditional magnitude of traffic flow acquisition modes, will there are bayonet data section and no bayonet by GPS data
The association of data section, it is abundant in the present invention by the virtual flow in expansion spline coefficient calculating acquisition Floating Car GPS data covering section
The precision of bayonet flow metering number evidence and the high coverage property of Floating Car GPS data is utilized, has the characteristics that implement simple, tool
There is real-time, the virtual flow in no bayonet data on flows section can be directly acquired, be conducive to carry out traffic in city road network
The work that flow is filled up.
Claims (6)
1. a kind of virtual traffic method of calculating flux based on multisource data fusion, which comprises the following steps:
Step 1: obtaining Floating Car GPS data, Floating Car GPS data distributed by map-matching method to city road network,
Obtain floating vehicle travelling trace information；
Step 2: carrying out blocking processing to floating vehicle travelling track, to guarantee the track between two bayonets without other cards
Mouthful；To all tracks obtained after blocking processing, temporally window carries out the cluster based on time-space matrix, and obtains in every class
Hot spot track；
Step 3: vehicle of the bayonet corresponding road section in time windows is obtained using the traffic flow data that bayonet records and is observed
Data on flows；
Step 4: being directed to each time window T, constructs an expansion spline coefficient set A={ α respectivelyk| k=1,2 ..., K }, size
It is hot spot track number of the full track data in the time window for K, K；For any time window T, based on having in the time window
The section of vehicle observed volume data, establishes optimization objective function, for solving the corresponding expansion spline coefficient set of the time window
A:
F=Minimize Z
VE(k)=αkVT(k)
Wherein, f is target function value, and Z is fitness, and i and j are the section in time window T with vehicle observed volume data, N
It is the sum in the section in time window T with vehicle observed volume data, VR(i) be section i in time window T vehicle observation stream
Amount, PijIt is the hot spot track set in time window T from section i to section j, VTIt (k) is that hot spot track k is generated in time window T
Link flow, value are equal to the number that hot spot track k occurs in time window T, VEIt (k) is VT(k) expand the flow after sample, αkIt is heat
The expansion spline coefficient of locus of points k；
Step 5: solving optimization objective function using genetic algorithm, obtains the corresponding overall situation for expanding spline coefficient set of time window T
Optimal result；
Step 6:, will be all therethrough in the time window for not having the section of vehicle observed volume data in time window T
The link flow that track generates is superimposed using corresponding expand after spline coefficient carries out expansion sample, the virtual traffic flow as the section；
Specifically, for the hot spot track by section l, expand the link flow that spline coefficient generates it using it and carry out expansion sample；For
By the non-thermal locus of points of section l, it is first determined with its maximally related hot spot track, determine method are as follows: obtain the non-hot rail
Common segment of each hot spot track, calculates separately the length of each common segment in mark and cluster belonging to it, and longest common segment is corresponding
Hot spot track be considered as its maximally related hot spot track；Then it utilizes and it is produced with the expansion spline coefficient of its maximally related hot spot track
Raw link flow carries out expansion sample.
2. the virtual traffic method of calculating flux according to claim 1 based on multisource data fusion, which is characterized in that institute
In the step of stating one, the Floating Car GPS data of passenger carrying status is filtered out first, recycles ST-Matching map-matching algorithm
GPS point in Floating Car GPS data is matched on city road network, floating vehicle travelling trace information is obtained.
3. the virtual traffic method of calculating flux according to claim 1 based on multisource data fusion, which is characterized in that institute
It states in step 2, the method that blocking processing is carried out to floating vehicle travelling track are as follows:
All tracks are numbered, item identifies the bayonet section number that every track is passed through one by one, if the card that certain track is passed through
Mouth section number M≤1, then abandon the track；If the bayonet section number M=2 that certain track is passed through, two which is passed through
Track between bayonet section retains by way of section as new track；If bayonet section number M >=3 that certain track is passed through, right
This track is split, since first bayonet section that the track is passed through, the adjacent card of the every two which is passed through
Track between mouth section retains by way of the section track new as one.
4. the virtual traffic method of calculating flux according to claim 1 based on multisource data fusion, which is characterized in that institute
It states in step 2, to all tracks, temporally window carries out the cluster based on time-space matrix specific step is as follows:
2.1) untreated state is set by all tracks in sometime window, and sets hot spot track collection and is combined into empty set；
2.2) a track Traj is randomly selected, track set identical as its origin and destination and in same time window is obtained,
I.e. the space-time of track Traj is neighbouring set C (Traj)；
2.3) classify to the track in set C (Traj), wherein kth class track is denoted as CTk, CTkIn any two tracks
Overlap length ratio q is greater than the minimum overlay length ratio Minq of setting, and adds the hot spot track in every one kind to hot spot rail
Trace set；Wherein overlap length ratio q is the mean value that two tracks are overlapped that section number possesses section number divided by this two tracks, heat
The locus of points be all tracks in a kind of track by the arrangement of frequency of occurrence descending after, take account in such all track frequency of occurrence it
With the track of preceding n%；Tag set C (Traji) in track be processed state；
2.4) judge whether all tracks in the time window are all marked as processed state, if so, terminating the time window
The cluster of interior track, otherwise based on the track of untreated state, return step 2.2).
5. the virtual traffic method of calculating flux according to claim 1 based on multisource data fusion, which is characterized in that
In the step two, the same bayonet of time window same in more days is recorded using DBSCAN clustering algorithm traffic flow data
It is clustered, removes exception stream magnitude, wherein it clusters the point for including in the maximum cluster of acquisition and is considered as normal stream magnitude,
Remaining point is considered as exception stream magnitude；Meanwhile if being less than the 50% of record flow number of days comprising point number in maximum cluster, it is considered as this
Bayonet hardware is abnormal, and removes the traffic flow data of bayonet record；Using remaining traffic flow data as bayonet pair
Answer the vehicle observed volume data in section.
6. the virtual traffic method of calculating flux according to claim 1 based on multisource data fusion, which is characterized in that institute
It states in step 5, solves optimization objective function using genetic algorithm, the specific steps are as follows:
5.1) initialize: setting evolutionary generation counter g=0, n individual of random generation are used as initial population, each individual
For an expansion spline coefficient set A={ αk| k=1,2 ..., K }, wherein element αkValue generated at random in the range of [0,1]；If
Setting genetic algebra is Ng, convergence precision Q；
5.2) individual choice: the probability that q-th of individual is selected in population in g generation is calculatedWherein Zg(q)Indicate that g, i.e., will be in the individual for the fitness of q-th of individual in population
Element substitutes into optimization objective function, obtained fitness value；The probability being selected according to Different Individual is continuously to g for population
In individual repeat n wheel and select, obtain the new individuals of n；
5.3) individual intersection and variation:
The n individual that step 5.2) is obtained carries out random pair two-by-two, takes fixed crossover probability pcCarry out crossover operation；If certain
Group individual need is intersected, then a crosspoint is randomly generated in all elements of group individual, by this group of individual intersection
Element after point is exchanged with each other, and generates two new individuals；If certain group individual does not need to be intersected, which is kept
It is constant；
Take fixed mutation probability pmMutation operation is carried out, n individual after successively selecting crossover operation, to the institute of each individual
There is element to be traversed, if some element needs to make a variation, changes the value of the element at random in the range of [0,1]；
It results in through n individual after heredity, constitutes g+1 for population；
Wherein crossover probability pcWith mutation probability pm, it is calculated respectively according to following formula:
Wherein, pc1It is the crossover probability for the individual that g is less than average fitness value for fitness value in population, pc2Indicate g generation
The crossover probability of individual in population with maximum adaptation angle value, FmaxIt is g for the maximum adaptation angle value of all individuals of population,
FaveIt is g for the average fitness value of all individuals of population, F*It is biggish fitness value in two individuals intersected；
pm1It is the mutation probability for the individual that g is less than average fitness value for fitness value in population, pm2It is g for having in population
The mutation probability of the individual of maximum adaptation angle value, F ' are the fitness values of the individual to make a variation；pc1、pc2、pm1、pm2For experience
Parameter；
5.4) g+1 is calculated for the fitness of individual each in population, obtains g+1 for the maximum adaptation of all individuals of population
Degree, i.e. g+1 is for the corresponding target function value f of population；
5.5) algorithm termination condition judges；
Convergence judgement: in g+1 generation and g, are compared for target function value f corresponding to population, if difference is less than precision
Q, then it is assumed that meet convergence, algorithm terminates, using g+1 for the individual in population with maximum adaptation degree as the overall situation of A most
Excellent result；
Genetic algebra judgement: if current genetic algebra g=Ng, then algorithm terminates, by this NgFor in population with maximum adaptation degree
Global optimum result of the individual as A；
If being unsatisfactory for above-mentioned algorithm termination condition, genetic algebra g=g+1 is enabled, is operated back to step 5.2) circulation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910042487.2A CN109686090B (en) | 2019-01-17 | 2019-01-17 | Virtual traffic flow calculation method based on multi-source data fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910042487.2A CN109686090B (en) | 2019-01-17 | 2019-01-17 | Virtual traffic flow calculation method based on multi-source data fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109686090A true CN109686090A (en) | 2019-04-26 |
| CN109686090B CN109686090B (en) | 2020-03-27 |
Family
ID=66193374
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910042487.2A Expired - Fee Related CN109686090B (en) | 2019-01-17 | 2019-01-17 | Virtual traffic flow calculation method based on multi-source data fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109686090B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110827540A (en) * | 2019-11-04 | 2020-02-21 | 黄传明 | Motor vehicle movement mode recognition method and system based on multi-mode data fusion |
| CN111063189A (en) * | 2019-11-05 | 2020-04-24 | 阿里巴巴集团控股有限公司 | Traffic flow processing method and device and electronic equipment |
| CN111123333A (en) * | 2019-12-30 | 2020-05-08 | 公安部交通管理科学研究所 | Vehicle track positioning method fusing bayonet and GPS data |
| CN111414719A (en) * | 2020-04-28 | 2020-07-14 | 中南大学 | Method and device for extracting peripheral features of subway station and estimating traffic demand |
| CN112394380A (en) * | 2019-08-16 | 2021-02-23 | 阿里巴巴集团控股有限公司 | Data processing method, device and system |
| CN113706875A (en) * | 2021-10-29 | 2021-11-26 | 深圳市城市交通规划设计研究中心股份有限公司 | Road function studying and judging method |
| CN113990075A (en) * | 2021-12-30 | 2022-01-28 | 广州市交通规划研究院 | Traffic survey data and track data fused flow distribution method and system |
| CN115641721A (en) * | 2022-12-06 | 2023-01-24 | 南京感动科技有限公司 | Multi-source traffic flow index fusion and road condition calculation method based on fusion index |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009259158A (en) * | 2008-04-21 | 2009-11-05 | Toyota Central R&D Labs Inc | Traffic state simulation device, and program |
| CN105374210A (en) * | 2015-12-03 | 2016-03-02 | 中山大学 | Gate detection rate obtaining method based on floating vehicle data |
| CN106530704A (en) * | 2016-11-25 | 2017-03-22 | 杭州电子科技大学 | Floating car aggregation detection method based on multivariate data fusion |
| JP2018077756A (en) * | 2016-11-11 | 2018-05-17 | Kddi株式会社 | Device, program, and method for estimating traffic volume on the basis of moving position range group including target route |
| CN108877227A (en) * | 2018-08-30 | 2018-11-23 | 中南大学 | A kind of global dynamic trip requirements estimation method based on multi-source traffic data |
-
2019
- 2019-01-17 CN CN201910042487.2A patent/CN109686090B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009259158A (en) * | 2008-04-21 | 2009-11-05 | Toyota Central R&D Labs Inc | Traffic state simulation device, and program |
| CN105374210A (en) * | 2015-12-03 | 2016-03-02 | 中山大学 | Gate detection rate obtaining method based on floating vehicle data |
| JP2018077756A (en) * | 2016-11-11 | 2018-05-17 | Kddi株式会社 | Device, program, and method for estimating traffic volume on the basis of moving position range group including target route |
| CN106530704A (en) * | 2016-11-25 | 2017-03-22 | 杭州电子科技大学 | Floating car aggregation detection method based on multivariate data fusion |
| CN108877227A (en) * | 2018-08-30 | 2018-11-23 | 中南大学 | A kind of global dynamic trip requirements estimation method based on multi-source traffic data |
Non-Patent Citations (1)
| Title |
|---|
| 孙超 等: "基于多源浮动车数据融合的道路交通运行评估", 《同济大学学报（自然科学版）》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112394380A (en) * | 2019-08-16 | 2021-02-23 | 阿里巴巴集团控股有限公司 | Data processing method, device and system |
| CN112394380B (en) * | 2019-08-16 | 2024-04-02 | 阿里巴巴集团控股有限公司 | Data processing method, device and system |
| CN110827540A (en) * | 2019-11-04 | 2020-02-21 | 黄传明 | Motor vehicle movement mode recognition method and system based on multi-mode data fusion |
| CN110827540B (en) * | 2019-11-04 | 2021-03-12 | 黄传明 | Motor vehicle movement mode recognition method and system based on multi-mode data fusion |
| CN111063189A (en) * | 2019-11-05 | 2020-04-24 | 阿里巴巴集团控股有限公司 | Traffic flow processing method and device and electronic equipment |
| CN111063189B (en) * | 2019-11-05 | 2022-05-03 | 阿里巴巴集团控股有限公司 | Traffic flow processing method and device and electronic equipment |
| CN111123333B (en) * | 2019-12-30 | 2022-05-03 | 公安部交通管理科学研究所 | Vehicle track positioning method fusing bayonet and GPS data |
| CN111123333A (en) * | 2019-12-30 | 2020-05-08 | 公安部交通管理科学研究所 | Vehicle track positioning method fusing bayonet and GPS data |
| CN111414719A (en) * | 2020-04-28 | 2020-07-14 | 中南大学 | Method and device for extracting peripheral features of subway station and estimating traffic demand |
| CN113706875A (en) * | 2021-10-29 | 2021-11-26 | 深圳市城市交通规划设计研究中心股份有限公司 | Road function studying and judging method |
| CN113990075B (en) * | 2021-12-30 | 2022-03-18 | 广州市交通规划研究院 | Traffic survey data and track data fused flow distribution method and system |
| CN113990075A (en) * | 2021-12-30 | 2022-01-28 | 广州市交通规划研究院 | Traffic survey data and track data fused flow distribution method and system |
| CN115641721A (en) * | 2022-12-06 | 2023-01-24 | 南京感动科技有限公司 | Multi-source traffic flow index fusion and road condition calculation method based on fusion index |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109686090B (en) | 2020-03-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109686090A (en) | A kind of virtual traffic method of calculating flux based on multisource data fusion | |
| CN104318324B (en) | Shuttle Bus website and route planning method based on taxi GPS records | |
| CN106548645B (en) | Vehicle route optimization method and system based on deep learning | |
| CN102592447B (en) | Method for judging road traffic state of regional road network based on fuzzy c means (FCM) | |
| CN101599217B (en) | Method for rapidly judging traffic state | |
| CN109686091B (en) | Traffic flow filling algorithm based on multi-source data fusion | |
| CN104809112B (en) | A kind of city bus development level integrated evaluating method based on multi-source data | |
| CN105355049B (en) | A kind of highway evaluation of running status method based on macroscopical parent map | |
| CN108447256A (en) | Trunk road vehicle trajectory reconstruction method based on electric police and fixed point detector data fusion | |
| CN112365711A (en) | Vehicle track reconstruction method based on license plate recognition data | |
| CN109886474B (en) | Closed test field planning method for automatic driving vehicle test | |
| CN108961804A (en) | Method is determined based on the alternative set of public bus network adjustment of multi objective classification intersection | |
| CN107195180A (en) | A kind of traffic trip track extraction method and device based on the alert data of electricity | |
| CN102521965A (en) | Effect evaluation method of traffic demand management measures based on identification data of license plates | |
| CN110188797A (en) | A kind of intelligent automobile method for rapidly testing based on Bayes's optimization | |
| CN106856049A (en) | Crucial intersection demand clustering analysis method based on bayonet socket number plate identification data | |
| CN110149593A (en) | Road network passenger flow state identification method based on Mobile Phone Signalling | |
| CN112784000B (en) | Passenger searching method based on taxi track data | |
| CN110413855A (en) | A kind of region entrance Dynamic Extraction method based on taxi drop-off point | |
| CN114139251B (en) | Integral layout method for land ports of border regions | |
| CN109489679A (en) | A kind of arrival time calculation method in guidance path | |
| CN106327867A (en) | Bus punctuality prediction method based on GPS data | |
| CN117238126A (en) | Traffic accident risk assessment method under continuous flow road scene | |
| CN114091581A (en) | Vehicle operation behavior type identification method based on sparse track | |
| CN110211380B (en) | Multisource traffic data fusion expressway congestion interval detection method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200327 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |