CN109426791A - A kind of polynary vehicle match method of multi-site, server and system - Google Patents
A kind of polynary vehicle match method of multi-site, server and system Download PDFInfo
- Publication number
- CN109426791A CN109426791A CN201710779960.6A CN201710779960A CN109426791A CN 109426791 A CN109426791 A CN 109426791A CN 201710779960 A CN201710779960 A CN 201710779960A CN 109426791 A CN109426791 A CN 109426791A
- Authority
- CN
- China
- Prior art keywords
- information
- vehicles
- vehicle
- collection
- website
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
- G08G1/017—Detecting movement of traffic to be counted or controlled identifying vehicles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Traffic Control Systems (AREA)
Abstract
The present invention relates to a kind of polynary vehicle match methods of multi-site, comprising: obtains the road image data of the first website and the second website acquired respectively；The road image data of first website acquisition are pre-processed to obtain the first information of vehicles collection, the road image data of second website acquisition are pre-processed to obtain the second information of vehicles collection, the first information of vehicles collection and the second information of vehicles collection respectively include multiple information of vehicles, and the information of vehicles includes at least one vehicle characteristics；Fusion Features are carried out to the first information of vehicles collection and the second information of vehicles collection, generate fusion feature data set；The corresponding information of vehicles of the first information of vehicles collection and the corresponding information of vehicles of the second information of vehicles collection are matched according to the fusion feature data set, and obtain matching result.This method uses multivariate information fusion scheme, precisely matches so as to carry out more vehicles of vehicle between website.
Description
Technical field
The present invention relates to the fields intelligent transportation (Intelligent Transportation System, abbreviation ITS), especially
It is related to a kind of polynary vehicle match method of multi-site, server and system.
Background technique
Traditional acquisition telecommunication flow information method is the rough traffic flow data of section for acquiring point, this traffic fluxion
According to including information such as vehicle number, speed, occupancy, license plate number, type of vehicle, vehicle flowrates.These information pass through traditional
Traffic flow model can estimate the operating condition of section or road network, but it is rough, such as road not accurately that this estimation, which is,
The speed of section, running time.
By the vehicle match between website, available accurately road section information.Although Car license recognition, RFID technique can be with
The unique match vehicle between website, but RFID label tag installation amount is limited, and there are license plates to be stained, false license plate etc. is asked for Car license recognition
Topic.The especially identical license plate of highway same type vehicle, intercourses password card and carrys out fee evasion, these are all RFID and vehicle
Board identification can't resolve.Vehicle pictures search technique at present can scan for similar vehicle, but it is only capable of providing a system
Column cannot carry out unique match to vehicle by the picture of similarity arrangement.
Summary of the invention
To solve the above problems, the present invention provides a kind of polynary vehicle match method of multi-site, comprising:
The road image data of the first website and the second website acquired respectively are obtained, wherein second station point is located at first stop
The front of point in the direction of travel；
The road image data of first website acquisition are pre-processed to obtain the first information of vehicles collection, to described the
The road image data of two websites acquisition are pre-processed to obtain the second information of vehicles collection, the first information of vehicles collection and second
Information of vehicles collection respectively includes multiple information of vehicles, and the information of vehicles includes at least one vehicle characteristics；
Fusion Features are carried out to the first information of vehicles collection and the second information of vehicles collection, generate fusion feature data
Collection；
According to the fusion feature data set to the corresponding information of vehicles of the first information of vehicles collection and second vehicle
The corresponding information of vehicles of information collection is matched, and obtains matching result.
Further, the vehicle characteristics are vehicle color, vehicle shape, Vehicle length or vehicle angle point.
Further, the road image data to first website acquisition are pre-processed to obtain the first vehicle letter
Breath collection is pre-processed to obtain the second information of vehicles collection step to the road image data of second website acquisition specifically:
The road image data for choosing the first website acquisition in first time period are pre-processed to obtain the first vehicle
Information collection；
Second time period is chosen according to the first time period, obtains the road of the second website acquisition in second time period
Road image data is pre-processed to obtain the second information of vehicles collection.
Further, described that Fusion Features are carried out to the first information of vehicles collection and the second information of vehicles collection, it is raw
At fusion feature data set step specifically:
Calculate each vehicle letter in each information of vehicles and the second information of vehicles collection in the first information of vehicles collection
Breath feature group between any two, the feature group include between two information of vehicles the characteristic distance of at least two vehicle characteristics or
The correspondence probability of at least two vehicle characteristics respectively；
Fusion feature data, fusion feature data structure corresponding to a plurality of feature groups are calculated according to the feature group
At the fusion feature data set.
Further, when the characteristic distance that the feature group includes at least two vehicle characteristics between two information of vehicles
When, it is described that fusion feature data step is calculated according to the feature group specifically:
Fusion feature data are calculated using the direct blending algorithm of characteristic distance according to the feature group；Or,
It is described when at least two vehicle characteristics correspondence probability respectively between the feature group includes two information of vehicles
Fusion feature data step is calculated according to the feature group specifically:
Fusion feature data are calculated using probability metrics blending algorithm according to the feature group.
Further, described that fusion feature data tool is calculated using the direct blending algorithm of characteristic distance according to the feature group
Body are as follows:
The i-th information of vehicles in the first information of vehicles collection is selected, jth vehicle letter is chosen out of described second information of vehicles collection
Breath；
Calculate the characteristic distance d of each vehicle characteristics between i-th information of vehicles and jth information of vehicles1(i,j)、d2
(i,j)、……dk(i,j)、……dK(i,j)；
The then fusion feature data D=w1*d1(i,j)+w2*d2(i,j)+……+wk*dk(i,j)+……+wK*dK(i,
J), wherein weighted value w1+w2+……+wk+……+wK=1.
Further, described specific using probability metrics blending algorithm calculating fusion feature data according to the feature group
Are as follows:
The i-th information of vehicles in the first information of vehicles collection is selected, jth vehicle letter is chosen out of described second information of vehicles collection
Breath, wherein the first information of vehicles collection includes N number of information of vehicles, the second information of vehicles collection includes each information of vehicles of M；
Calculate the characteristic distance d of each vehicle characteristics between i-th information of vehicles and jth information of vehicles1(i,j)、d2
(i,j)、……dk(i,j)……dK(i, j), and obtaining in the first information of vehicles collection and/or the second information of vehicles collection includes plural number
When vehicle, i-th information of vehicles is judged using corresponding vehicle characteristics and the jth information of vehicles is the probability of same vehicle
p1(d1(i,j))、p1(d2(i,j))、……p1(dk(i,j))、……p1(dK(i, j)), institute is judged using corresponding vehicle characteristics
It states the i-th information of vehicles and the jth information of vehicles is not the Probability p of same vehicle2(d1(i,j))、p2(d2(i,j))、……p2
(dk(i,j))、……p2(dK(i,j))；
Then the fusion feature data lnP (D) can be acquired by following formula:
Wherein:
λ (i, j) be the i-th information of vehicles and jth information of vehicles be same vehicle Fusion Features after probability, λ (i, j) by
ρ (the d of at least two characteristic distances between i-th information of vehicles and jth information of vehiclesk(i, j)) weighting multiplication acquisition；λ(i,τ)
It is the probability that the i-th information of vehicles does not have any vehicle matched；λ (i, τ) is ξ (dk(i, j)) multiple weightings are multiplied, ξ (dk(i,
It j) is) that the i-th information of vehicles uses probability of the same vehicle characteristics without vehicle match；P is multiple vehicle characteristics in a feature group
Matched overall probability.
Further, it is described according to the fusion feature data set to the corresponding information of vehicles of the first information of vehicles collection
Information of vehicles corresponding with the second information of vehicles collection is matched, and obtains matching result step specifically:
Determine each vehicle letter in each information of vehicles and the second information of vehicles collection in the first information of vehicles collection
The relative weighting of breath between any two, the relative weighting are the corresponding fusion feature data；
It according to relative weighting between any two, is matched using shortest path first, and obtains matching result.
Another aspect, the invention also discloses a kind of polynary vehicle match servers of multi-site, comprising:
Module is obtained, for obtaining the road image data of the first website and the second website acquired respectively, wherein second
Website is located at the front of the first website in the direction of travel；
Processing module, the road image data for acquiring to first website are pre-processed to obtain the first vehicle letter
Breath collection is pre-processed to obtain the second information of vehicles collection, first vehicle to the road image data of second website acquisition
Information collection and the second information of vehicles collection respectively include multiple information of vehicles, and the information of vehicles includes that at least one vehicle is special
Sign；
Fusion Module, it is raw for carrying out Fusion Features to the first information of vehicles collection and the second information of vehicles collection
At fusion feature data set；
Matching module is used for according to the fusion feature data set to the corresponding information of vehicles of the first information of vehicles collection
Information of vehicles corresponding with the second information of vehicles collection is matched, and obtains matching result.
Another aspect, the present invention disclose a kind of polynary vehicle match system of multi-site, including at least two for adopting again
Collect the website of road image data, the connected polynary vehicle match server of multi-site as described above of the website.
Implement technical solution of the present invention, using multivariate information fusion method, carries out accurate, more vehicles of vehicle between website
Match, according to the accurately vehicle match, track vehicle, and then solves the problems, such as fee evasion, vacation license plate and unlicensed vehicle problem.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.In attached drawing:
Fig. 1 is a kind of flow chart of the polynary vehicle match method of multi-site of the present invention；
Fig. 2 is of the invention a kind of based on the matched method flow diagram of shortest path first progress.
Specific embodiment
To solve the above problems, the present invention provides a kind of polynary vehicle match method of multi-site, this method is applied at least
System including two websites.
Embodiment one, as shown in Figure 1, the polynary vehicle match method of multi-site includes:
S101: the road image data of the first website and the second website acquired respectively are obtained, wherein second station point is located at
The front of first website in the direction of travel；It should be understood that target vehicle first passes through when vehicle normally travel on road
There are multiple vehicles including target vehicle in the road image data of first website acquisition at this time in the first website.Work as vehicle
When normally travel to the second website, there is also more including target vehicle in the road image data of the second website acquisition
A vehicle.For the accurate control to road conditions, it need to determine target vehicle in the road image data that the first website acquires
With the corresponding relationship in the road image data of the second website acquisition.
S102: being pre-processed to obtain the first information of vehicles collection to the road image data of first website acquisition, right
The road image data of the second website acquisition are pre-processed to obtain the second information of vehicles collection, the first information of vehicles collection
Multiple information of vehicles are respectively included with the second information of vehicles collection, the information of vehicles includes at least one vehicle characteristics；Wherein it is
More accurate matching vehicle, information of vehicles preferably include at least two vehicle characteristics.Wherein vehicle characteristics refer to vehicle color,
The non-unique features such as vehicle shape, Vehicle length and vehicle angle point, such as the feature of Car license recognition uniqueness are within the rule, are easy
Understand, it is unique although the feature identification of the uniqueness such as Car license recognition can also be completed to carry out vehicle matched purpose
Property feature obtain it is extremely difficult, be relatively easy to be disturbed.
The road image data of website acquisition need to be pre-processed for convenience of subsequent application, which can be based on
Artificial intelligence carries out deep learning, continues to optimize.By deep learning, system can be made to pass through the feature of single nonuniqueness
Accurately matching vehicle.By pretreatment, the vehicle that can extract each vehicle in the road image data of website acquisition is special
Sign.Wherein vehicle characteristics may include the global characteristics and notable feature of vehicle.It, can be by a website for the same vehicle
The time unification that the vehicle characteristics and vehicle of acquisition reach saves.In the present embodiment, vehicle characteristics include vehicle color, vehicle
Shape, Vehicle length and vehicle angle point.Vehicle color is indicated that vehicle shape is obtained by template matching by color histogram, vehicle
Angle point is extracted by SURF (Speeded-Up Robust Features accelerates robust features) algorithm.
Since vehicle arrival time is can to determine the running time of vehicle it is believed that simultaneously by historical data
Distribution, therefore for higher matching efficiency.Choose the road image data of first website acquisition in first time period into
Row pretreatment obtains the first information of vehicles collection；Second time period is chosen according to the first time period, is obtained in second time period
The road image data of the second website acquisition are pre-processed to obtain the second information of vehicles collection.If vehicle is average from first stop
The time that point drives to the second website is 30 minutes, is needed to the N number of vehicle for passing through the first website in the 12:00-12:01 period
It is matched, can be matched to guarantee N number of vehicle in the period, a biggish range can be chosen, because
It is matched in this selection 12:20-12:31 period by M vehicle of the second website.The selection of certain second time period can
To be chosen according to the actual situation.By choosing to the period, matching efficiency can be effectively improved.
S103: Fusion Features are carried out to the first information of vehicles collection and the second information of vehicles collection, it is special to generate fusion
Levy data set；
S104: according to the fusion feature data set to the corresponding information of vehicles of the first information of vehicles collection and described
The corresponding information of vehicles of two information of vehicles collection is matched, and obtains matching result.
The M vehicle that the N number of vehicle and the second information of vehicles collection for include to the first information of vehicles collection contain matches, wherein
Multiple vehicle characteristics of the first information of vehicles collection and all vehicles in the second information of vehicles collection carry out whole examine in matching process
Amount, therefore Optimum Matching scheme on the whole can be obtained, to reach accurate matched purpose, or can also be by list
One non-unique features deep learning is matched, to reach accurate matched purpose.It should be understood that in practical application
In, which can also be used in combination with RFID technique or license plate recognition technology, reach more accurately identification and require.
Embodiment two, it is described to the first information of vehicles collection and second information of vehicles on the basis of example 1
Collection carries out Fusion Features, generates fusion feature data set step specifically:
Calculate each vehicle letter in each information of vehicles and the second information of vehicles collection in the first information of vehicles collection
Breath feature group between any two, the feature group include between two information of vehicles the characteristic distance of at least two vehicle characteristics or
The correspondence probability of at least two vehicle characteristics respectively；
Fusion feature data, fusion feature data structure corresponding to a plurality of feature groups are calculated according to the feature group
At the fusion feature data set.
Specifically, being melted according to the feature group using the direct blending algorithm of characteristic distance or the calculating of probability metrics blending algorithm
Characteristic is closed, repeats the step for different feature groups, that is, for the vehicle in the first different information of vehicles collection
Fusion feature data, a plurality of fusion feature data structures are calculated with feature group corresponding to the vehicle in the second information of vehicles collection
At fusion feature data set.In preferred embodiment, fusion feature data set include arbitrary first information of vehicles collection in vehicle and
Fusion feature data corresponding to the feature group between vehicle in second information of vehicles collection.
Wherein when the characteristic distance of at least two vehicle characteristics between the feature group includes two information of vehicles, according to
The feature group calculates fusion feature data using the direct blending algorithm of characteristic distance specifically:
The i-th information of vehicles in the first information of vehicles collection is selected, jth vehicle letter is chosen out of described second information of vehicles collection
Breath；
Calculate the characteristic distance d of each vehicle characteristics between i-th information of vehicles and jth information of vehicles1(i,j)、d2
(i,j)、……dk(i,j)、……dK(i,j)；Wherein it should be understood that information of vehicles includes K kind vehicle characteristics, variety classes
Vehicle characteristics characteristic distance calculation it is not fully identical, chosen according to vehicle characteristics such as normalized cross-
correlation(NCC)、sum of absolute difference(SAD)、sum of squared difference
(SSD), Euclidean distance, COS distance, mahalanobis distance scheduling algorithm, it is not limited here.
The then fusion feature data D=w1*d1(i,j)+w2*d2(i,j)+……+wk*dk(i,j)+……+wK*dK(i,
J), wherein weighted value w1+w2+……+wk+……+wK=1.Wherein weighted value wkMost common calculation is least square method,
The matched vehicle of one group of historical data such as 300 is chosen, makes the D of same vehicle minimum by least square method, different vehicles
D it is maximum, to find out wk.It, can also be using in one d of exclusive use in addition to least square methodkIn the case where (i, j), it is seen
Success rate situation in the historical data, wk=ck/sum(c1:ck), ckIt is dkCorresponding success rate, exactly this feature distance
Weight is equal to its contribution degree situation that success rate is summed in all feature success rates in the historical data.
It repeats to calculate fusion feature data using the direct blending algorithm of characteristic distance, until every in the first information of vehicles collection
Corresponding fusion feature data are generated between each vehicle in one vehicle and the second information of vehicles collection.
In other embodiments, when the feature group includes at least two vehicle characteristics difference between two information of vehicles
When corresponding probability, fusion feature data can also be calculated according to the feature group using probability metrics blending algorithm, wherein vehicle
The correspondence probability of feature refers to the probability according to current vehicle characteristic matching for same vehicle, specific:
The i-th information of vehicles in the first information of vehicles collection is selected, jth vehicle letter is chosen out of described second information of vehicles collection
Breath；
Calculate the characteristic distance d of each vehicle characteristics between i-th information of vehicles and jth information of vehicles1(i,j)、d2
(i,j)、……dk(i,j)、……dKIt includes plural vehicle that (i, j), which is obtained in the first information of vehicles collection and/or the second information of vehicles collection,
When, judge i-th information of vehicles using corresponding vehicle characteristics and the jth information of vehicles be the Probability p of same vehicle1
(d1(i,j))、p1(d2(i,j))、……p1(dk(i,j))、……p1(dK(i, j)), institute is judged using corresponding vehicle characteristics
It states the i-th information of vehicles and the jth information of vehicles is not the Probability p of same vehicle2(d1(i,j))、p2(d2(i,j))、……p2
(dk(i,j))、……p2(dK(i,j))；
Then the fusion feature data lnP (D) can be acquired by following formula:
Wherein:
λ (i, j) be the i-th information of vehicles and jth information of vehicles be same vehicle Fusion Features after probability, λ (i, j) by
ρ (the d of at least two characteristic distances between i-th information of vehicles and jth information of vehiclesk(i, j)) weighting multiplication acquisition；λ(i,τ)
It is the probability that the i-th information of vehicles does not have any vehicle matched；λ (i, j) is ξ (dk(i, j)) multiple weightings are multiplied, ξ (dk(i,
It j) is) that the i-th information of vehicles uses probability of the same vehicle characteristics without vehicle match.
It is identical with the direct blending algorithm of characteristic distance, it repeats to calculate fusion feature data using probability metrics blending algorithm,
Up to generating corresponding melt between each vehicle in each vehicle and the second information of vehicles collection in the first information of vehicles collection
Close characteristic.
Obtain the fusion between each vehicle in each vehicle and the second information of vehicles collection in the first information of vehicles collection
After characteristic, according to the fusion feature data set to the corresponding information of vehicles of the first information of vehicles collection and described
The corresponding information of vehicles of two information of vehicles collection is matched, and obtains matching result step specifically:
Determine each vehicle letter in each information of vehicles and the second information of vehicles collection in the first information of vehicles collection
The relative weighting of breath between any two, the relative weighting are the corresponding fusion feature data；
It according to relative weighting between any two, is matched using shortest path first, and obtains matching result.
Specifically, as shown in Fig. 2, left side i=1, i=2 are the vehicle in the first information of vehicles collection, right side j=1, j in figure
=2, j=3, j=4 are the vehicle in the second information of vehicles collection.Using the fusion feature data being calculated in step before as
Relative weighting assignment to weight between any two, ask by the matching that matching problem becomes minimum value and maximum value in graph theory at this time
Topic is solved using shortest path first such as Dijkstra (Di Jiesitela) algorithm.
Specifically, assuming relative weighting S between any two11=0.2, S12=0.3, S13=0.3, S14=0.2, S21=
0.1,S22=0.5, S23=0.4, S24=0.3, wherein S be probability metrics blending algorithm in fusion feature data lnP (D) or
Fusion feature data D, P in the direct blending algorithm of characteristic distance are that multiple matched totality of vehicle characteristics are general in a feature group
Rate, then specific calculation are as follows:
Firstly, from the smallest S11Start to search for, Pmin1=S11.I=1 is matched to j=1, and no longer matches；Then secondary small
Sij, then Pmin1=S11+S24=0.5.First result obtains.
Secondly, from the smallest S12, start to search for, Pmin2=S12.I=1 is matched to j=2, and no longer matches；Then secondary small
Sij, then Pmin2=S12+S21=0.4.Second result obtains.
Again, from the smallest S13Start to search for, Pmin3=S13.I=1 is matched to j=3, and no longer matches；Then secondary small
Sij, then Pmin1=Pmin3=S13+S21=0.4.Third result obtains.
Finally, from the smallest S14Start to search for, Pmin4=S14.I=1 is matched to j=4, and no longer matches；Then secondary small
Sij, then Pmin4=S14+S21=0.3.4th result obtains.
All possibility are all searched, acquire the 4th result minimum, then have i=1 to be matched to j=4, and i=2 is matched
To j=1, then vehicle i=1, i=2 all unique match.
Another aspect, the invention also discloses a kind of polynary vehicle match servers of multi-site, comprising:
Module is obtained, for obtaining the road image data of the first website and the second website acquired respectively, wherein second
Website is located at the front of the first website in the direction of travel；
Processing module, the road image data for acquiring to first website are pre-processed to obtain the first vehicle letter
Breath collection is pre-processed to obtain the second information of vehicles collection, first vehicle to the road image data of second website acquisition
Information collection and the second information of vehicles collection respectively include multiple information of vehicles, and the information of vehicles includes that at least one vehicle is special
Sign；
Fusion Module, it is raw for carrying out Fusion Features to the first information of vehicles collection and the second information of vehicles collection
At fusion feature data set；
Matching module is used for according to the fusion feature data set to the corresponding information of vehicles of the first information of vehicles collection
Information of vehicles corresponding with the second information of vehicles collection is matched, and obtains matching result.
Another aspect, the present invention disclose a kind of polynary vehicle match system of multi-site, including at least two for adopting again
Collect the website of road image data, the connected polynary vehicle match server of multi-site as described above of the website.
The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, for the skill of this field
For art personnel, the invention may be variously modified and varied.All within the spirits and principles of the present invention, made any bun
Change, equivalent replacement, improvement etc., should be included within scope of the presently claimed invention.
Claims (10)
1. a kind of polynary vehicle match method of multi-site characterized by comprising
The road image data of the first website and the second website acquired respectively are obtained, wherein second station point is located at first stop point and exists
Front on direction of traffic；
The road image data of first website acquisition are pre-processed to obtain the first information of vehicles collection, to the second station
The road image data of point acquisition are pre-processed to obtain the second information of vehicles collection, the first information of vehicles collection and the second vehicle
Information collection respectively includes multiple information of vehicles, and the information of vehicles includes at least one vehicle characteristics；
Fusion Features are carried out to the first information of vehicles collection and the second information of vehicles collection, generate fusion feature data set；
The corresponding information of vehicles of the first information of vehicles collection and second vehicle are believed according to the fusion feature data set
Breath collects corresponding information of vehicles and is matched, and obtains matching result.
2. matching process according to claim 1, which is characterized in that the vehicle characteristics be vehicle color, vehicle shape,
Vehicle length or vehicle angle point.
3. matching process according to claim 1, which is characterized in that the road image to first website acquisition
Data are pre-processed to obtain the first information of vehicles collection, pre-process to the road image data of second website acquisition
To the second information of vehicles collection step specifically:
The road image data for choosing the first website acquisition in first time period are pre-processed to obtain the first information of vehicles
Collection；
Second time period is chosen according to the first time period, obtains the mileage chart of the second website acquisition in second time period
As data are pre-processed to obtain the second information of vehicles collection.
4. matching process according to claim 1, which is characterized in that described to the first information of vehicles collection and described
Two information of vehicles collection carry out Fusion Features, generate fusion feature data set step specifically:
Calculate each information of vehicles two in each information of vehicles and the second information of vehicles collection in the first information of vehicles collection
Feature group between two, the feature group include the characteristic distance of at least two vehicle characteristics or at least between two information of vehicles
The correspondence probability of two vehicle characteristics respectively；
Fusion feature data are calculated according to the feature group, fusion feature data corresponding to a plurality of feature groups constitute institute
State fusion feature data set.
5. matching process according to claim 4, which is characterized in that when the feature group includes between two information of vehicles
It is described that fusion feature data step is calculated according to the feature group when characteristic distance of at least two vehicle characteristics specifically:
Fusion feature data are calculated using the direct blending algorithm of characteristic distance according to the feature group；Or,
When at least two vehicle characteristics correspondence probability respectively between the feature group includes two information of vehicles, the basis
The feature group calculates fusion feature data step specifically:
Fusion feature data are calculated using probability metrics blending algorithm according to the feature group.
6. matching process according to claim 5, which is characterized in that described straight using characteristic distance according to the feature group
It connects blending algorithm and calculates fusion feature data specifically:
The i-th information of vehicles in the first information of vehicles collection is selected, jth information of vehicles is chosen out of described second information of vehicles collection；
Calculate the characteristic distance d of each vehicle characteristics between i-th information of vehicles and jth information of vehicles1(i,j)、d2(i,
j)、……dk(i,j)、……dK(i,j)；
The then fusion feature data D=w1*d1(i,j)+w2*d2(i,j)+……+wk*dk(i,j)+……+wK*dK(i, j),
Wherein weighted value w1+w2+……+wk+……+wK=1.
7. matching process according to claim 5, which is characterized in that described to be melted according to the feature group using probability metrics
Hop algorithm calculates fusion feature data specifically:
The i-th information of vehicles in the first information of vehicles collection is selected, jth vehicle letter is chosen out of described second information of vehicles collection
Breath, wherein the first information of vehicles collection includes N number of information of vehicles, the second information of vehicles collection includes each information of vehicles of M；
It obtains when the first information of vehicles collection and/or the second information of vehicles collection include plural vehicle, using corresponding
Vehicle characteristics judge that i-th information of vehicles and the jth information of vehicles are the Probability ps of same vehicle1(d1(i,j))、p1(d2
(i,j))、……p1(dk(i,j))、……p1(dK(i, j)), using corresponding vehicle characteristics judge i-th information of vehicles with
The jth information of vehicles is not the Probability p of same vehicle2(d1(i,j))、p2(d2(i,j))、……p2(dk(i,j))、……p2
(dK(i,j))；
Then the fusion feature data lnP (D) can be acquired by following formula:
Wherein:
λ (i, j) be the i-th information of vehicles and jth information of vehicles be same vehicle Fusion Features after probability, λ (i, j) is by i-th
ρ (the d of at least two characteristic distances between information of vehicles and jth information of vehiclesk(i, j)) weighting multiplication acquisition；λ (i, τ) is
The probability that i-th information of vehicles does not have any vehicle matched；λ (i, τ) is ξ (dk(i, j)) multiple weightings are multiplied, ξ (dk(i,
It j) is) that the i-th information of vehicles uses probability of the same vehicle characteristics without vehicle match.
8. according to matching process described in claim 1, which is characterized in that it is described according to the fusion feature data set to described
The corresponding information of vehicles of one information of vehicles collection and the corresponding information of vehicles of the second information of vehicles collection are matched, and are obtained
With result step specifically:
Determine each information of vehicles two in each information of vehicles and the second information of vehicles collection in the first information of vehicles collection
Relative weighting between two, the relative weighting are the corresponding fusion feature data；
It according to relative weighting between any two, is matched using shortest path first, and obtains matching result.
9. a kind of polynary vehicle match server of multi-site characterized by comprising
Module is obtained, for obtaining the road image data of the first website and the second website acquired respectively, wherein the second website
Positioned at the front of the first website in the direction of travel；
Processing module, the road image data for acquiring to first website are pre-processed to obtain the first information of vehicles
Collection is pre-processed to obtain the second information of vehicles collection, first vehicle to the road image data of second website acquisition
Information collection and the second information of vehicles collection respectively include multiple information of vehicles, and the information of vehicles includes at least one vehicle characteristics；
Fusion Module, for carrying out Fusion Features to the first information of vehicles collection and the second information of vehicles collection, generation is melted
Close characteristic data set；
Matching module is used for according to the fusion feature data set to the corresponding information of vehicles of the first information of vehicles collection and institute
It states the corresponding information of vehicles of the second information of vehicles collection to be matched, and obtains matching result.
10. a kind of polynary vehicle match system of multi-site, which is characterized in that including at least two for acquiring road image data
Website, the connected polynary vehicle match server of multi-site as claimed in claim 9 of the website.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710779960.6A CN109426791B (en) | 2017-09-01 | 2017-09-01 | Multi-site and multi-vehicle matching method, server and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710779960.6A CN109426791B (en) | 2017-09-01 | 2017-09-01 | Multi-site and multi-vehicle matching method, server and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109426791A true CN109426791A (en) | 2019-03-05 |
| CN109426791B CN109426791B (en) | 2022-09-16 |
Family
ID=65513019
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710779960.6A Active CN109426791B (en) | 2017-09-01 | 2017-09-01 | Multi-site and multi-vehicle matching method, server and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109426791B (en) |
Citations (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001344689A (en) * | 2000-06-02 | 2001-12-14 | Hiroshi Imai | Vehicle correspondence device and method |
| JP2002342872A (en) * | 2001-05-11 | 2002-11-29 | Sumitomo Electric Ind Ltd | Device and method for detecting abnormality of traffic flow |
| CN1464487A (en) * | 2002-06-03 | 2003-12-31 | 昆明利普机器视觉工程有限公司 | A traffic flow detection system based on visual vehicle optical characteristic recognition and matching |
| US20060245617A1 (en) * | 2005-03-30 | 2006-11-02 | Ying Shan | Object identification between non-overlapping cameras without direct feature matching |
| CN101916383A (en) * | 2010-08-25 | 2010-12-15 | 浙江师范大学 | Vehicle detecting, tracking and identifying system based on multi-camera |
| CN102081846A (en) * | 2011-02-22 | 2011-06-01 | 交通运输部公路科学研究所 | Expressway charge data track matching based traffic state recognition method |
| CN102200999A (en) * | 2011-04-27 | 2011-09-28 | 华中科技大学 | Method for retrieving similarity shape |
| CN102201165A (en) * | 2010-03-25 | 2011-09-28 | 北京汉王智通科技有限公司 | Monitoring system of vehicle traffic violation at crossing and method thereof |
| CN102289948A (en) * | 2011-09-02 | 2011-12-21 | 浙江大学 | Multi-characteristic fusion multi-vehicle video tracking method under highway scene |
| CN102354389A (en) * | 2011-09-23 | 2012-02-15 | 河海大学 | Visual-saliency-based image non-watermark algorithm and image copyright authentication method |
| CN103020989A (en) * | 2012-12-05 | 2013-04-03 | 河海大学 | Multi-view target tracking method based on on-line scene feature clustering |
| CN103729892A (en) * | 2013-06-20 | 2014-04-16 | 深圳市金溢科技有限公司 | Vehicle positioning method and device and processor |
| CN104298990A (en) * | 2014-09-15 | 2015-01-21 | 西安电子科技大学 | Rapid graph matching and recognition method based on skeleton graphs |
| US20150163638A1 (en) * | 2013-12-05 | 2015-06-11 | Deutsche Telekom Ag | Method and system for tracking the whereabouts of people in urban settings |
| CN104732485A (en) * | 2015-04-21 | 2015-06-24 | 深圳市深图医学影像设备有限公司 | Method and system for splicing digital X-ray images |
| CN104794731A (en) * | 2015-05-12 | 2015-07-22 | 成都新舟锐视科技有限公司 | Multi-target detection and tracking method for speed dome camera control strategy |
| US20150332588A1 (en) * | 2014-05-15 | 2015-11-19 | Xerox Corporation | Short-time stopping detection from red light camera evidentiary photos |
| CN105096590A (en) * | 2014-04-23 | 2015-11-25 | 株式会社日立制作所 | Traffic information generation method and device |
| CN105894542A (en) * | 2016-04-26 | 2016-08-24 | 深圳大学 | Online target tracking method and apparatus |
| CN106960182A (en) * | 2017-03-02 | 2017-07-18 | 云南大学 | A kind of pedestrian integrated based on multiple features recognition methods again |
-
2017
- 2017-09-01 CN CN201710779960.6A patent/CN109426791B/en active Active
Patent Citations (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001344689A (en) * | 2000-06-02 | 2001-12-14 | Hiroshi Imai | Vehicle correspondence device and method |
| JP2002342872A (en) * | 2001-05-11 | 2002-11-29 | Sumitomo Electric Ind Ltd | Device and method for detecting abnormality of traffic flow |
| CN1464487A (en) * | 2002-06-03 | 2003-12-31 | 昆明利普机器视觉工程有限公司 | A traffic flow detection system based on visual vehicle optical characteristic recognition and matching |
| US20060245617A1 (en) * | 2005-03-30 | 2006-11-02 | Ying Shan | Object identification between non-overlapping cameras without direct feature matching |
| CN102201165A (en) * | 2010-03-25 | 2011-09-28 | 北京汉王智通科技有限公司 | Monitoring system of vehicle traffic violation at crossing and method thereof |
| CN101916383A (en) * | 2010-08-25 | 2010-12-15 | 浙江师范大学 | Vehicle detecting, tracking and identifying system based on multi-camera |
| CN102081846A (en) * | 2011-02-22 | 2011-06-01 | 交通运输部公路科学研究所 | Expressway charge data track matching based traffic state recognition method |
| CN102200999A (en) * | 2011-04-27 | 2011-09-28 | 华中科技大学 | Method for retrieving similarity shape |
| CN102289948A (en) * | 2011-09-02 | 2011-12-21 | 浙江大学 | Multi-characteristic fusion multi-vehicle video tracking method under highway scene |
| CN102354389A (en) * | 2011-09-23 | 2012-02-15 | 河海大学 | Visual-saliency-based image non-watermark algorithm and image copyright authentication method |
| CN103020989A (en) * | 2012-12-05 | 2013-04-03 | 河海大学 | Multi-view target tracking method based on on-line scene feature clustering |
| CN103729892A (en) * | 2013-06-20 | 2014-04-16 | 深圳市金溢科技有限公司 | Vehicle positioning method and device and processor |
| US20150163638A1 (en) * | 2013-12-05 | 2015-06-11 | Deutsche Telekom Ag | Method and system for tracking the whereabouts of people in urban settings |
| CN105096590A (en) * | 2014-04-23 | 2015-11-25 | 株式会社日立制作所 | Traffic information generation method and device |
| US20150332588A1 (en) * | 2014-05-15 | 2015-11-19 | Xerox Corporation | Short-time stopping detection from red light camera evidentiary photos |
| CN104298990A (en) * | 2014-09-15 | 2015-01-21 | 西安电子科技大学 | Rapid graph matching and recognition method based on skeleton graphs |
| CN104732485A (en) * | 2015-04-21 | 2015-06-24 | 深圳市深图医学影像设备有限公司 | Method and system for splicing digital X-ray images |
| CN104794731A (en) * | 2015-05-12 | 2015-07-22 | 成都新舟锐视科技有限公司 | Multi-target detection and tracking method for speed dome camera control strategy |
| CN105894542A (en) * | 2016-04-26 | 2016-08-24 | 深圳大学 | Online target tracking method and apparatus |
| CN106960182A (en) * | 2017-03-02 | 2017-07-18 | 云南大学 | A kind of pedestrian integrated based on multiple features recognition methods again |
Non-Patent Citations (7)
| Title |
|---|
| JINWOO CHOI 等: "Data association using relative compatibility of multiple observations for EKF-SLAM", 《INTEL SERV ROBOTICS》 * |
| XINCHEN LIU 等: "A Deep Learning-Based Approach to Progressive Vehicle Re-identification for Urban Surveillance", 《ECCV 2016》 * |
| ZHENGHAO XI 等: "Multiple object tracking using A* association algorithm with dynamic weights", 《JOURNAL OF INTELLIGENT & FUZZY SYSTEMS》 * |
| 刘加运: "一种多维特征融合的车辆对象同一性匹配方法", 《计算机技术与发展》 * |
| 单玉刚 等: "基于运动检测的多车辆跟踪方法研究", 《计算机测量与控制》 * |
| 李贤慧 等: "基于概率距离及融合时空特征的镜头相似性度量", 《计算机应用研究》 * |
| 王先彬 等: "跨摄像头目标跟踪综述", 《图形图像》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109426791B (en) | 2022-09-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11586992B2 (en) | Travel plan recommendation method, apparatus, device and computer readable storage medium | |
| Liu et al. | Urban traffic prediction from mobility data using deep learning | |
| Shen et al. | Learning deep neural networks for vehicle re-id with visual-spatio-temporal path proposals | |
| Huang et al. | Object identification: A Bayesian analysis with application to traffic surveillance | |
| CN111832514B (en) | Unsupervised pedestrian re-identification method and unsupervised pedestrian re-identification device based on soft multiple labels | |
| Meuter et al. | A decision fusion and reasoning module for a traffic sign recognition system | |
| CN107451619A (en) | A kind of small target detecting method that confrontation network is generated based on perception | |
| Oliveira-Neto et al. | Online license plate matching procedures using license-plate recognition machines and new weighted edit distance | |
| CN103778441B (en) | A kind of sequence Aircraft Target Recognition based on DSmT and HMM | |
| WO2020000191A1 (en) | Method for driver identification based on car following modeling | |
| Martinsson et al. | Clustering vehicle maneuver trajectories using mixtures of hidden markov models | |
| CN111126327B (en) | Lane line detection method and system, vehicle-mounted system and vehicle | |
| Benton et al. | Nearly d-linear convergence bounds for diffusion models via stochastic localization | |
| CN110363238A (en) | Intelligent vehicle damage identification method, system, electronic equipment and storage medium | |
| Krishnakumari et al. | Traffic congestion pattern classification using multiclass active shape models | |
| CN108280481A (en) | A kind of joint objective classification and 3 d pose method of estimation based on residual error network | |
| CN102693258A (en) | High-accuracy similarity search system | |
| Zhang et al. | C 3-GAN: Complex-Condition-Controlled Urban Traffic Estimation through Generative Adversarial Networks | |
| CN113553975A (en) | Pedestrian re-identification method, system, equipment and medium based on sample pair relation distillation | |
| Agarwal et al. | Attention guided cosine margin to overcome class-imbalance in few-shot road object detection | |
| CN109426791A (en) | A kind of polynary vehicle match method of multi-site, server and system | |
| Silva et al. | A multi-layer k-means approach for multi-sensor data pattern recognition in multi-target localization | |
| Kwon et al. | Face friend-safe adversarial example on face recognition system | |
| Lidman et al. | Clustering, shape extraction and velocity estimation applied to radar detections | |
| CN110263196B (en) | Image retrieval method, image retrieval device, electronic equipment and storage medium |
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 |