CN106022296B - A kind of fake-licensed car detection method of the probability polymerization based on vehicle hot spot region - Google Patents
A kind of fake-licensed car detection method of the probability polymerization based on vehicle hot spot region Download PDFInfo
- Publication number
- CN106022296B CN106022296B CN201610380318.6A CN201610380318A CN106022296B CN 106022296 B CN106022296 B CN 106022296B CN 201610380318 A CN201610380318 A CN 201610380318A CN 106022296 B CN106022296 B CN 106022296B
- Authority
- CN
- China
- Prior art keywords
- bayonet
- vehicle
- probability
- hot spot
- fake
- 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.)
- Active
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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
-
- 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
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/625—License plates
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Traffic Control Systems (AREA)
Abstract
A kind of fake-licensed car detection method of the probability polymerization based on vehicle hot spot region, comprising the following steps: the extraction of S1. bayonet data；S2. the cleaning of bayonet data；S3. probability is flowed between calculating bayonet；S4. it calculates abnormal behavior vehicle inventory: according to the bayonet pair of each vehicle extracted in S2, calculating bayonet to probability, be lower than given threshold T according to probabilityiNumber, descending arrangement, as needed choose before N row as abnormal behavior vehicle inventory；S5. vehicle hot spot region bayonet point is calculated；S6. vehicle probability current between 2 hot spot region bayonet points is calculated；S7. according to probability between bayonet obtained by S6, all bayonets are polymerize；S8. situation is polymerize according to vehicle bayonet and calculates vehicle fake-license probability Z.The present invention overcomes the bayonet number plate of randomness to identify mistake, using the relatively-stationary characteristic in movable vehicle region, filters out that the biggish vehicle of deck suspicion degree, to be substantially reduced suspicion range, practicability good.
Description
Technical field
The present invention relates to intelligent transportation fields, and in particular to a kind of fake-licensed car detection method.
Background technique
Fake-licensed car is commonly called as clone's vehicle, refers to by forging or illegally extracting in the formalities such as other number plate of vehicle and driving license
The vehicle of road traveling.With the development of economy, vehicle population is more and more, and fake-licensed car is also continuously increased.Fake-licensed car can be upset
Control of the public security organ to public safety, since fake-licensed car does not have legal procedure and insurance, once traffic accident, driver occurs
Easily escape.Fake-licensed car also provides tool used in crime for criminal activity, considerably increases difficulty of solving a case.Fake-licensed car can escape various taxes
Take, cause being largely lost for the national expenses of taxation, and upsets the order of transport market.Fake-licensed car can damage the legitimate rights and interests of true car owner,
Vehicular traffic violation, in terms of, true car owner will often serve as " person who spends money wastefully and foolishly ".
It is difficult that the identification of this act of violating regulations of fake-licensed car is investigated and prosecuted, and people's police are difficult the short time according to vehicle during on duty
Board and external appearance characteristic artificial judgment go out whether vehicle is fake-licensed car.There are some fake-licensed cars based on bayonet data automatic now
Change recognition methods.Bayonet is referred to using advanced photoelectricity, computer, image procossing, pattern-recognition, remote data access etc.
Technology carries out round-the-clock real time monitoring to car lane, the non-motorized lane in monitoring section and records dependent image data, and from
Dynamic acquisition vehicle passes through the data such as time, place, driving direction, brand number, number plate color, body color.
In published patent and patent in examining, it is largely divided into three categories, 1, be based on image recognition, it is special to pass through vehicle
Sign is judged.2, be based on auxiliary device: such as installation electronic license plate reserves security password etc..3, bayonet data are based on, it is right
Track of vehicle is analyzed.
Patent based on first kind method has:
One: what application number [201510102368.3] proposed, the deck of license plate number and brand message based on identification vehicle
Vehicle detection method.1, identify that the license plate number and brand message of vehicle add the vehicle if license plate number and brand message do not correspond to
Enter fake-licensed car suspicion library；2, the vehicle that fake-licensed car suspicion library is added accurately is differentiated, removes the vehicle of wrong report；3) vehicle is counted
The number occurred in fake-licensed car suspicion library, sounds an alarm when the frequency of occurrence of certain vehicle is more than setting value K.
Two: what application number [201410333789.2] proposed, a kind of side of the fake-licensed car identification based on testing vehicle register identification
Method and device.1, vehicle is extracted from picture or video, identifies the vehicle characteristics.It 2, is to be compared with traffic administration institute's database
It is right, when finding that feature mismatches in the certain features of vehicle and database, alarm.2, inspection cloth is carried out in urban district inner bayonet
Control then carries out set alarm when finding that different location of the identical vehicle of feature in the same time occurs.3, historical data carries out
It compares, when discovery license plate number identical vehicle is alarmed when other features of vehicle are not identical.
In the above patented method, the factors such as weather, light can interfere the identification of image, partial domestic vehicle and imported car shape
It is more similar, also difficulty can be caused to identification.The defects of identical can not then being detected there are former vehicle with fake-licensed car brand, model.
Patent based on the second class method has:
One: what application number [201320365827.3] proposed, a kind of vacation based on super high frequency radio frequency identification technology, fake-licensed car
Identification device.Vacation, fake license plate vehicle identification device based on super high frequency radio frequency identification technology, pass through radio frequency identification read-write equipment pair
Electronic tag on automobile is detected, and is judged the information in electronic tag, then acquires equipment with license plate image
The license plate number obtained is compared, and carries out fake-licensed car differentiation.
Two: what application number [200910107671.7] proposed, a kind of motor vehicle deck false-trademark based on security password is illegal
Behavioral value method.Traffic police's interior vehicle management system platform reserves vehicle safety and detects password；The hand-held long-range control of the traffic police that enforces the law
Terminal processed, which inputs, is examined vehicle region code and license plate number；It is compared with registration information；It is close that vehicle safety detection is looked into input
Code；Input password compares automatically with reserving cipher, judges whether deck false-trademark.
The above patent needs to be related to additional platform construction and vehicle remoulding.
Patent based on third class method has:
One: what application number [201210438702.9] proposed, a kind of fake-licensed car detection method based on time intersection.It obtains
Residence time section at the time of vehicle enters and leaves the position and in the position, is matched, two-by-two if when corresponding stop
Between section have intersection, then there is fake-licensed car, if without intersection, according between two o'clock the theoretical shortest time and real time difference come
Judge whether there is fake-licensed car.In the program, the theoretical shortest time calculates complexity, is related to road network structure, road section length, traffic
All multi-parameters such as congestion degree.
Two: what application number [201310730531.1] proposed, a kind of region automatic capture method of fake-licensed car vehicle.It is sealing
The zone boundary setting monitoring point closed, if vehicle is driven out to from the region, but does not drive into information, then determines the vehicle for set
Board vehicle.Closed area in the program is difficult to realize in practice.
Three: what application number [201310034242.8] proposed, a kind of capture for the fake-licensed car vehicle formatted based on road network
Method.Urban road area is subjected to gridding according to grid segmentation principle, examines same license plate number vehicle in temporal sequence
Whether grid track is continuous, if there is discontinuous grid track, using the corresponding license plate number in grid track as doubtful set
Board vehicle license plate number.The program needs detailed road network structure.
Four: what application number [201410094882.2] proposed, a kind of side of the deck analysis based on bayonet connectivity.1, root
It is whether consistent with the bayonet communication information pre-saved by the sequence of bayonet according to vehicle, if inconsistent, protected further according to preparatory
The bayonet communication information deposited determines that any inconsistent bayonet to all reachable paths of its adjacent bayonet, judges the reachable road
It whether there is a paths in diameter, the bayonet quantity which passes through is less than or equal to preset value N, if it is, the license plate
Non- deck, otherwise, there are decks for the license plate.
Existing method is all based on single vehicle driving trace and is judged.In actual conditions, since bayonet number plate is known
Not wrong presence, the fake-licensed car suspicion inventory that algorithm above obtains often reach tens of thousands of or even hundreds of thousands.
The bayonet number plate identification mistake of randomness refers to due to factors such as illumination, angles, license plate in part bayonet or
Period can correctly identify that part bayonet or period are identified as other license plates.Such case accounts for number plate identification mistake
The overwhelming majority.Caused by nonrandomness number plate identification mistake is mainly stained due to number plate of vehicle, which can be by all bayonets
It is identified as fixed wrong number plate.Example license plate number " Zhejiang A****R ", wherein the lower right corner character R is blocked, but rest part is equal
It is apparent, all bayonets are by the Car license recognition at " Zhejiang A****P ".
Summary of the invention
In order to overcome existing fake license plate vehicle detection method due to there are the bayonet number plate of randomness identification mistake, suspicion model
Deficiency excessive, that practicability is poor is enclosed, the present invention provides a kind of bayonet number plate identification mistake for overcoming randomness, living using vehicle
The dynamic relatively-stationary characteristic in region, filters out that the biggish vehicle of deck suspicion degree, to be substantially reduced suspicion range, practicability good
The fake-licensed car detection method of probability polymerization based on movable vehicle hot spot region.
The technical solution adopted by the present invention to solve the technical problems is:
A kind of fake-licensed car detection method of the probability polymerization based on vehicle hot spot region, the fake license plate vehicle detection method packet
Include following steps:
S1. it the extraction of bayonet data: obtains bayonet and crosses vehicle record data, retain the dimension of needs, including bayonet number, card
Mouth direction, brand number, number plate type, spend the vehicle time；
S2. the cleaning of bayonet data；
S3. probability is flowed between calculating bayonet, process is as follows:
3.1, according to wheelpath, extract bayonet vector: by bayonet data according to brand number, bayonet crosses the progress of vehicle time
Sequence, obtains the wheelpath of each vehicle, with the bayonet sequence (K passed througha,Kb,Kc·Ki·Kn) indicate, it extracts each
Adjacent bayonet pair in wheelpath, forms bayonet sequence vector (Ka,Kb)·(Ki, Kj)·。Ka, indicate vehicle passes through the
A bayonet, vector (Ka,Kb) indicate that sequence of cars passes through two bayonet KaAnd Kb, and KaAnd KbIt is adjacent in wheelpath；
3.2, probability is flowed between calculating bayonet: to bayonet KiIt counts, obtains KiBayonet flows out vehicle summation, with count (Ki)
It indicates.To vector (Ki,Kj) count, obtain bayonet KiTo bayonet KjPassing vehicle summation, with count (Ki,Kj) indicate.Bayonet Ki
To bayonet KjFlow to probability be Pij=count (Ki,Kj)/count(Ki)；
S4. it calculates abnormal behavior vehicle inventory: according to the bayonet pair of each vehicle extracted in S2, calculating bayonet pair
Probability is lower than given threshold T according to probabilityiNumber, descending arrangement, N row before choosing as needed, as abnormal behavior vehicle
Inventory, N choose according to actual needs；
S5. calculate vehicle hot spot region bayonet point: thermal point structure region bayonet point refers to vehicle often captured card
Mouthful, the number that calculating vehicle occurs in each bayonet point, if it is greater than the threshold value M of settingv, then it is assumed that the bayonet is the heat of vehicle
Point zone of action bayonet point；
S6. vehicle probability current between 2 hot spot region bayonet points: the heat calculated in S5 each vehicle is calculated
The probability that point bayonet does cartesian product, and calculated according to S2 carries out assignment to every a pair of of bayonet, and choose in forward and reverse compared with
Big probability；If the probability P (Kj, Ki) of probability P (Ki, Kj) <bayonet Kj to bayonet Ki of bayonet Ki to bayonet Kj, then will
P (Kj, Ki) is assigned to bayonet to (Ki, Kj)；
S7. according to probability between bayonet obtained by S6, all bayonets are polymerize, it is ensured that each packets inner, for any
Bayonet is to (Ki, Kj), there are a paths from KiTo Kj, the probability of all adjacent bayonets pair is not less than given threshold on path
Ti, between each grouping, for any bayonet Ki, there is no bayonet K for other groupsj, meet P (Ki,Kj) it is greater than the threshold value T of settingi；
S8. the probability for calculating vehicle fake-license polymerize situation according to vehicle bayonet and calculates vehicle fake-license probability Z, Z=N*i*
s4, N indicates vehicle hot spot bayonet sum, and i depends on number of packet L, if L=2, i=1, if L=3, i=1/2, and otherwise i
=0；S depends on the quantity ratio of bayonet in grouping, and NAi indicates the quantity of bayonet in each group；S=max (NAi, 2)/max
(Nai, 1), max (NAi, 1) indicate quantity maximum one, and max (NAi, 2) indicates that quantity is second largest.
Further, a in the step S7, polymerization process is as follows:
Step7.1. obtaining can be with number plate of vehicle inventory；
Step7.2. a number being taken out from number plate inventory, the number plate thermal point structure region bayonet is obtained according to number
(K1, K2, K3 ... Kn) is denoted as array A0, and is arranged according to the number flashback that vehicle appears in the bayonet；
Step7.3. first bayonet K1 in A0 is added in array A1, and K1 is removed from A0；
Step7.4. judge whether the undirected probability Q (K1, Ki) in K1 and A0 between each bayonet is greater than threshold value Ti, such as
Fruit is Ki to be added to A1 and empty array A_new, and Ki is removed from array A0；
Step7.5. a bayonet Kj is taken out from array A_new, is judged undirected between each bayonet in Kj and A0
Whether probability Q (Kj, Ki) is greater than threshold value Ti, if so, Ki is added to A1 and empty array A_new, and by Ki from array A0
Middle removal removes Kj from array A_new；
Step7.6. Step7.5 is repeated, until A_new is sky；
Step7.7. increase array Ai, repeat step Step7.3-7.6；
Step7.8. Step7.7 is repeated, until A0 is sky；
Step7.9. Step7.2-7.8 is repeated, until number plate inventory is sky.
Further, in the step S2, the cleaning process of bayonet data is as follows:
2.1, cleaning repeats to record；
2.2, dirty data is cleaned, the dirty data refers to that brand number does not meet the record of naming rule；
2.3, it determines time interval, cleans the bayonet data that record is not complete in the section, these situations include bayonet damage
Bad and power-off leads to that bayonet can not work or network interruption and disk size deficiency cause data to fail to store, and cleans the card
All data of mouth.
Further, in the step S4, threshold value TiIt is set according to bayonet number plate recognition correct rate F, to bayonet KiAnd
Speech, TiValue should meet F≤∑jPij<F-Ti, wherein j meets Pij>TI,, bayonet number plate recognition correct rate F=1- (passes through card
Cross vehicle of the still number plate of mouth not in number plate storehouse records number)/(all to cross vehicles record number), TiIndicate bayonet KiThreshold value, Pij
Indicate bayonet PiTo the probability of other bayonets.
Technical concept of the invention are as follows: based on based on current Probability Detection vehicle fake-license method, utilize vehicle hot spot
Active area characteristics reduce investigation range.
In actual use, the discrimination of number plate is unable to reach since light, angle, number plate such as are stained at factors, the bayonet
100% (generally in 96%-98% or so).License plate can identify correctly in most of bayonet or period, fraction bayonet
Or the period is identified as other license plates, it is random for identifying that the bayonet point of mistake is often distributed, and number of passing through is less.Together
When, normal home vehicle can substantially travel on more fixed route, identify that correct bayonet point is often distributed in these
In route, and number is more.Therefore remove the less bayonet point of number of pass times, it can be by the bayonet of most of identification mistake
It is excluded.
Vehicle frequently by bayonet often spatially have continuity.If these bayonet points are obviously distributed presentation two
A community, and it is associated with very weak between community, explanation is likely to be two cars respectively in two regional activities, that is to say, that the bright vehicle
Deck possibility it is higher.
Beneficial effects of the present invention are mainly manifested in: the number plate of bayonet randomness can be overcome to identify to a certain extent wrong
Accidentally, it is good greatly to reduce investigation range, practicability；It is with strong applicability without relying on road network structure.
Detailed description of the invention
Fig. 1 is the flow chart of the fake-licensed car detection method of the probability polymerization based on movable vehicle hot spot region.
Fig. 2 is normal vehicle hot spot bayonet distribution, its main feature is that having stronger relevance between bayonet.
Fig. 3 is doubtful fake license plate vehicle hot spot bayonet distribution, its main feature is that being obviously scattered in the community Liang Ge.
Fig. 4 is the vehicle hot spot bayonet distribution of mistake more easy to identify, its main feature is that being scattered in multiple communities.
Specific embodiment
The invention will be further described below in conjunction with the accompanying drawings.
Referring to Fig.1~Fig. 4, a kind of fake-licensed car detection method of the probability polymerization based on movable vehicle hot spot region, including
Following steps:
S1. it the extraction of data: obtains bayonet and crosses vehicle record data, retain the dimension of needs, including bayonet number, bayonet side
To, brand number, number plate type, cross the vehicle time.(bayonet number uniquely determines bayonet crossing position, and bayonet direction determines bayonet
Shooting direction, i.e. direction of traffic, number plate type and brand number uniquely determine a motor vehicle)
The present embodiment has extracted the data of Hangzhou 1-January of January, 489 bayonets on the 27th record, and in total 129534497
Item, bayonet data format such as the following table 1:
| Field | Data type | Meaning |
| KKID | VARchar(20) | Bayonet ID |
| FXBH | VARchar(2) | Bayonet direction |
| HPHM | VARchar(10) | Brand number |
| HPLX | VARchar(2) | Number plate type |
| JGSJ | VARchar(20) | Spend the vehicle time |
Table 1
Wherein KKID+FXBH uniquely determines a bayonet, and HPHM+HPZL uniquely determines an automobile.JGSJ is accurate to the second,
(in following steps, bayonet ID contains bayonet direction, and brand number contains number plate type, repeats no more)
S2. the cleaning of data, process are as follows:
2.1, it cleans repeated data: may be generated by a plurality of data, made when capturing the vehicle of a process for bayonet
It is as shown in table 2 below at the repetition of data:
Table 2
3rd article consistent with the 4th article of the passed through bayonet ID of vehicle record excessively, and the vehicle time excessively is identical, this to belong to repetition note
Record.
The reason is that a plurality of record may be generated when capturing vehicle due to bayonet.Define same vehicle, same card
Mouthful, and be recorded as repetition of the time error no more than threshold value Δ T (the present embodiment is set as 4 seconds) records, removal repeats to record.
2.2, clean dirty data: since brand number is bayonet system according to picture recognition, there are a certain number of license plates
Character, bayonet can not identify that the brand number in record does not meet naming rule, specifically include: " can not identify ", " NULL ",
Brand number contain symbol "? ".The partial data is cleaned, part case is as shown in table 3 below:
| Serial number | Brand number | Spend the vehicle time |
| 1 | 2016-01-15 14∶52∶51 | |
| 2 | NULL | 2016-01-20 19∶32∶30 |
| 3 | Peaceful B? 711T | 2016-01-25 11∶31∶34 |
| 4 | Zhejiang A00? NT | 2016-01-25 20∶54∶04 |
| 5 | Zhejiang A025X? | 2016-01-21 14∶18∶13 |
| 6 | It can not identify | 2016-01-10 22∶49∶28 |
Table 3
3.3, cleaning records incomplete bayonet data: under practical situation, due to factors such as power-off, suspension, bayonet damages,
There is missing in the data of part bayonet.For recording incomplete bayonet data, which is cleaned.
In present case, the time interval of selection is January 1 to January 27, has 169 bayonets extremely in accumulative 489 bayonets
There is shortage of data within few 1 day, cleans all data of these bayonets.
Bayonet shortage of data is defined as follows:, should if bayonet one day vehicle record of crossing is less than average daily 1/20 for crossing vehicle and recording
This day data of bayonet missing.
S3. probability is flowed between calculating bayonet
3.1 determine wheelpath first, and extract bayonet vector
Wheelpath determination process: according to brand number, crossing vehicle time-sequencing can be obtained wheelpath.Wheelpath packet
Containing three parts (bayonet ID, brand number spend the vehicle time), the bayonet and spend the vehicle time that vehicle successively passes through are indicated.Format is such as
Under: the present embodiment shares 2123568 wheelpaths, (ellipsis part is non-display portion) as shown in table 4 below.
Table 4
It extracts bayonet vector: for two bayonets adjacent in same garage's wheel paths, successively taking out, as bayonet
Vector.For S3 case, can be taken off six vectors (31000300007402,31000300010702),
(31000300010702,31000300010904)、(31000300010904,31000300004504)、
(31000300004504,31000300004502)、(31000300004502,31000300019902)、
(31000300019902,31000300005402)。
To 2123568 wheelpaths in the present embodiment, extracted vector is distinguished, what all wheelpaths can be taken out
Vector is 85259515.
3.2. probability is flowed between calculating bayonet
Count all vector (Ki,Kj), it is available from bayonet KiOutflow, flows to bayonet KjVehicle number.Count Ki, I
It is available from KiThe vehicle fleet of outflow obtains (ellipsis part is non-display portion) as shown in table 5 below.
Table 5
Bayonet flows to probability and embodies bayonet distribution and road network structure in another dimension.If bayonet distribution or road
Significant changes occur for web frame, and probability needs recalculate (morning and evening lane, restricted driving etc. do not need to recalculate).
S4. abnormal behavior vehicle inventory is calculated
We are extracted the wheelpath of each vehicle in S3.One has the wheelpath of N number of point to correspond to N-1
Bayonet vector, each bayonet vector, a corresponding bayonet flow to probability, that is to say, that one by N number of driving rail put and formed
Mark corresponds to N-1 probability value, such as the following table 6:
| Serial number | Brand number | Bayonet ID | Spend the vehicle time | Appear in the probability |
| k1 | Zhejiang AF1*** | 31000300011101 | 2016-01-14 07∶55∶58 | |
| k2 | Zhejiang AF1*** | 31000300005401 | 2016-01-14 08∶06∶34 | 25.47% |
| k3 | Zhejiang AF1*** | 31000300004703 | 2016-01-14 08∶53∶38 | 0.00% |
| k4 | Zhejiang AF1*** | 31000300010502 | 2016-01-14 12∶12∶01 | 0.01% |
| k5 | Zhejiang AF1*** | 31000300010504 | 2016-01-14 12∶12∶34 | 13.07% |
| k6 | Zhejiang AF1*** | 31000300007301 | 2016-01-14 13∶59∶31 | 0.04% |
| k7 | Zhejiang AF1*** | 31000300004304 | 2016-01-14 16∶34∶15 | 0.28% |
| k8 | Zhejiang AF1*** | 31000300004302 | 2016-01-14 16∶34∶15 | 53.37% |
| k9 | Zhejiang AF1*** | 31000300005001 | 2016-01-14 16∶49∶02 | 1.71% |
| k10 | Zhejiang AF1*** | 31000300014819 | 2016-01-14 16∶57∶21 | 0.02% |
| k11 | Zhejiang AF1*** | 31000300010020 | 2016-01-14 17∶08∶16 | 10.55% |
| k12 | Zhejiang AF1*** | 31000300028603 | 2016-01-14 17∶17∶45 | 16.61% |
| k13 | Zhejiang AF1*** | 31000300025302 | 2016-01-14 17∶39∶38 | 2.19% |
| k14 | Zhejiang AF1*** | 31000300004702 | 2016-01-14 17∶45∶08 | 0.02% |
Table 6
In table 6, k3, k4, k6, k10 and k14 probability are below Ti.
In the present embodiment, bayonet threshold value T is calculatediCalculating process it is as follows:
4.1, all bayonets are counted first crosses vehicle sum and number plate error number (vehicle of the brand number not in number plate storehouse).
Total crosses vehicle quantity are as follows: and 87383083, number plate error number are as follows: 2524476；Bayonet number plate recognition correct rate F, which can be calculated, is
97.1%.
4.2, threshold value T is calculated to each bayoneti: for example for bayonet 31000300015401, with other each bayonets
Between probability it is as shown in table 7 below (ellipsis part is non-display portion) according to descending arrangement:
| Serial number | Ki | Kj | Probability |
| 1 | 31000300015401 | 31000300006701 | 49.594% |
| 2 | 31000300015401 | 31000300015501 | 13.035% |
| 3 | 31000300015401 | 31000300006504 | 11.874% |
| 4 | 31000300015401 | 31000300006001 | 8.771% |
| 5 | 31000300015401 | 31000300015601 | 3.733% |
| 6 | 31000300015401 | 31000300011501 | 2.199% |
| 7 | 31000300015401 | 31000300015701 | 1.821% |
| 8 | 31000300015401 | 31000300006502 | 1.490% |
| 9 | 31000300015401 | 31000300011601 | 0.957% |
| 10 | 31000300015401 | 31000300011504 | 0.854% |
| 11 | 31000300015401 | 31000300011701 | 0.832% |
| 12 | 31000300015401 | 31000300011704 | 0.730% |
| 13 | 31000300015401 | 31000300006904 | 0.728% |
| 14 | 31000300015401 | 31000300011503 | 0.309% |
| 15 | 31000300015401 | 31000300015801 | 0.268% |
| 16 | 31000300015401 | 31000300006901 | 0.256% |
| 17 | 31000300015401 | 31000300011703 | 0.255% |
| 18 | 31000300015401 | 31000300006902 | 0.135% |
| 19 | 31000300015401 | 31000300010404 | 0.133% |
| 20 | 31000300015401 | 31000300011604 | 0.128% |
| 21 | 31000300015401 | 31000300007003 | 0.096% |
| 22 | 31000300015401 | 31000300007602 | 0.087% |
| 23 | 31000300015401 | 31000300022904 | 0.082% |
| 24 | 31000300015401 | 31000300010801 | 0.073% |
| … | … | … | … |
Table 7
To bayonet KiFor, TiValue should meet F≤∑jPij<F+Ti, wherein j meets Pij>Ti, in present case, if
TiValue is 0.268%, ∑jPij=96.930% < F, if TiValue is 0.255%, F < ∑jPij=97.198% < F+Ti,
If TiValue is 0.135%, ∑jPij=97.333% > F+Ti, so for bayonet 31000300015401, threshold value Ti=
0.255%.
According to method as above, we calculate the corresponding T of each bayoneti, obtain as (ellipsis part is not show to the following table 8
Show part):
| Serial number | Bayonet ID | Ti |
| 1 | 31000300000704 | 0.262% |
| 2 | 31000300000801 | 0.222% |
| 3 | 31000300000803 | 0.283% |
| 4 | 31000300011804 | 0.273% |
| 5 | 31000300011919 | 0.236% |
| 6 | 31000300011920 | 0.191% |
| 7 | 31000300012119 | 0.147% |
| 8 | 31000300012120 | 0.249% |
| 9 | 31000300025301 | 0.146% |
| 10 | 31000300020004 | 0.255% |
| … | … | … |
Table 8
In the present embodiment, it has been found that this section includes to have 5 probability to be less than threshold value Ti in the wheelpath of 14 points.
This 14 bayonet points are shown on map, it is found that this 14 bayonet points are obviously scattered in two tracks, wherein k1,
This 8 points of k2, k4, k5, k10, k11, k12, k13 are rendered obvious by a track, and k3, k6, k7, k8, k9, k14 are obviously in one
Track.
Suspicion degree sequence: Qi be less than threshold value Ti it is primary, we are denoted as once abnormal behavior, if abnormal behavior number compared with
It is more, illustrate that the wheel paths are very high in the discontinuity in space, that is to say, that a possibility that deck is very high.Step S6 is repeated, I
Available each vehicle abnormal behavior number, according to the number descending of abnormal behavior, available abnormal behavior vehicle
Inventory, partial results such as the following table 9:
Table 9
S5. vehicle hot spot region bayonet is calculated.
5.1, statistics is grouped according to number plate, the number that vehicle occurs in each bayonet, and according to frequency of occurrence descending,
Such as the following table 10 available for vehicle " Zhejiang AN0M**** ":
Table 10
5.2, the corresponding threshold value M of vehicle is calculatedv.For vehicle " Zhejiang AN0M**** ", each bayonet point frequency of occurrence is put down
Mean value is 11.3, therefore, threshold value MvIt is set as 11.3*80%=9.1.
5.3, vehicle thermal point structure region bayonet is obtained.It is greater than this constraint of mean value 80%, sequence in table according to frequency of occurrence
It number is the movable hot spot region bayonet of vehicle vehicle " Zhejiang AN0M**** " for the bayonet of 1-15.
S6. probability current between two hot spot region bayonet points of vehicle is calculated:
Cartesian product is done to the bayonet point in vehicle thermal point structure region, obtains bayonet pair, and according to the probability in S5, give
Every a pair of bayonet chooses biggish probability in forward and reverse to progress assignment.
By taking vehicle " Zhejiang AN0M**** " as an example, add up 15 thermal point structure region bayonet points, doing after cartesian product can be with
Obtain 225 bayonets pair, then to each bayonet to assignment, such as in S5 P (31000300020402,
31000300023902)=51.4%, P (31000300023902,31000300020402)=1.4%, then Q
(31000300020402,31000300023902)=Q (31000300023902,31000300020402)=51.4%, Q
(Ki,Kj) indicate bayonet KiAnd KjBetween undirected current probability.
S7. according to probability between bayonet obtained by S6, all bayonets are polymerize.
In the present embodiment, bayonet is polymerize in such a way that probability connects.
Step7.1. obtaining can be with number plate of vehicle inventory.
Step7.2. a number being taken out from number plate inventory, the number plate thermal point structure region bayonet is obtained according to number
(K1, K2, K3 ... Kn) is denoted as array A0, and is arranged according to the number flashback that vehicle appears in the bayonet.
Step7.3. first bayonet K1 in A0 is added in array A1, and K1 is removed from A0.
Step7.4. judge whether the undirected probability Q (K1, Ki) in K1 and A0 between each bayonet is greater than threshold value Ti, such as
Fruit is Ki to be added to A1 and empty array A_new, and Ki is removed from array A0.
Step7.5. a bayonet Kj is taken out from array A_new, is judged undirected between each bayonet in Kj and A0
Whether probability Q (Kj, Ki) is greater than threshold value Ti, if so, Ki is added to A1 and empty array A_new, and by Ki from array A0
Middle removal removes Kj from array A_new.
Step7.6. Step7.5 is repeated, until A_new is sky.
Step7.7. increase array Ai, repeat step Step7.3-7.6.
Step7.8. Step7.7 is repeated, until A0 is sky.
Step7.9. Step7.2-7.8 is repeated, until number plate inventory is sky.
By the above operation, the polymerization situation of available all vehicle thermal point structures region bayonet point.
S8. according to the polymerization situation of bayonet point, vehicle fake-license probability Z is calculated.
If all bayonets all assign to A1 group, that is to say, that all thermal point structure region bayonets of vehicle can connect
Come, as shown in figure 3, so illustrating that often movable region is to be in the relatively more close place of a contiguity, the vehicle to vehicle
Deck possibility is smaller.
It is assigned in A1 combination A2 group if bayonet is relatively uniform, illustrates that vehicle thermal point structure region can significantly be divided into
The community Liang Ge, as shown in figure 4, so illustrate movable vehicle in two lower regions of contiguity, while consider the vehicle
Have repeatedly it is abnormal jump behavior, illustrate that the vehicle fake-license possibility is higher.
If bayonet is distributed in multiple groups, explanation may have multiple license plate misrecognitions at this vehicle, be grouped more, explanation
The deck possibility of the vehicle is smaller.
In the present embodiment, some numerical results such as the following table 11 (sorting according to abnormal behavior number):
Table 11
In this implementation, when suspicion degree Z is more than or equal to 1, it is believed that vehicle deck probability with higher, we
It can be found that only 6 vehicle suspicion degree Z are greater than 1, that is, 100 in most preceding 100 vehicles of behavior frequency of abnormity
The investigation range shorter of vehicle is to 6 vehicles.
Claims (4)
1. a kind of fake-licensed car detection method of the probability polymerization based on vehicle hot spot region, it is characterised in that: the deck car test
Survey method the following steps are included:
S1. it the extraction of bayonet data: obtains bayonet and crosses vehicle record data, retain the dimension of needs, including bayonet number, bayonet side
To, brand number, number plate type and cross the vehicle time；
S2. the cleaning of bayonet data；
S3. probability is flowed between calculating bayonet, process is as follows:
3.1, according to wheelpath, extract bayonet vector: by bayonet data according to brand number, bayonet is spent the vehicle time and is arranged
Sequence obtains the wheelpath of each vehicle, with the bayonet sequence (K passed througha,Kb,Kc…Ki…Kn...) indicate, extract each
Adjacent bayonet pair in wheelpath forms bayonet sequence vector (Ka,Kb)…(Ki, Kj) ..., KaIndicate a that vehicle passes through
A bayonet, vector (Ka,Kb) indicate that sequence of cars passes through two bayonet KaAnd Kb, and KaAnd KbIt is adjacent in wheelpath；
3.2, probability is flowed between calculating bayonet: to bayonet KiIt counts, obtains KiBayonet flows out vehicle summation, with count (Ki) table
Show, to vector (Ki,Kj) count, obtain bayonet KiTo bayonet KjPassing vehicle summation, with count (Ki,Kj) indicate, bayonet KiIt arrives
Bayonet KjFlow to probability be Pij=count (Ki,Kj)/count(Ki)；
S4. it calculates abnormal behavior vehicle inventory: according to the bayonet pair of each vehicle extracted in S3, calculating bayonet to general
Rate is lower than given threshold T according to probabilityiNumber, descending arrangement, N row before choosing as needed, as abnormal behavior cleaning vehicle
Single, N chooses according to actual needs；
S5. calculate vehicle hot spot region bayonet point: thermal point structure region bayonet point refers to vehicle often captured bayonet, system
The number that meter vehicle occurs in each bayonet point, if it is greater than the threshold value M of settingv, then it is assumed that the bayonet is that the hot spot of vehicle is living
Dynamic region bayonet point；
S6. vehicle probability current between 2 hot spot region bayonet points: the hot spot card calculated in S5 each vehicle is calculated
Mouth does cartesian product, and the probability calculated according to S3, carries out assignment to every a pair of of bayonet, and choose biggish in forward and reverse
Probability；If the probability P (Kj, Ki) of probability P (Ki, Kj) <bayonet Kj to bayonet Ki of bayonet Ki to bayonet Kj, then by P
(Kj, Ki) is assigned to bayonet to (Ki, Kj)；
S7. according to probability between bayonet obtained by S6, all bayonets are polymerize, it is ensured that each packets inner, for any bayonet
To (Ki, Kj), there are a paths from KiTo Kj, the probability of all adjacent bayonets pair is not less than given threshold T on pathi, respectively
Between a grouping, for any bayonet Ki, there is no bayonet K for other groupsj, meet P (Ki,Kj) it is greater than the threshold value T of settingi；
S8. the probability for calculating vehicle fake-license polymerize situation according to vehicle bayonet and calculates vehicle fake-license probability Z, Z=N*i*s4, N table
Showing vehicle hot spot bayonet sum, i depends on number of packet L, if L=2, i=1, if L=3, i=1/2, otherwise i=0；s
Depending on the quantity ratio of bayonet in grouping, NAi indicates the quantity of bayonet in each group；S=max (NAi, 2)/max (NAi, 1),
Max (NAi, 1) indicates quantity maximum one, and max (NAi, 2) indicates that quantity is second largest.
2. a kind of fake-licensed car detection method of the probability polymerization based on vehicle hot spot region as described in claim 1, feature
Be: in step S7, polymerization process is as follows:
Step7.1. number plate of vehicle inventory is obtained；
Step7.2. a number is taken out from number plate inventory, according to number, obtain the number plate thermal point structure region bayonet (K1,
K2, K3Kn), it is denoted as array A0, and arrange according to the number flashback that vehicle appears in the bayonet；
Step7.3. first bayonet K1 in A0 is added in array A1, and K1 is removed from A0；
Step7.4. judge whether the undirected probability Q (K1, Ki) in K1 and A0 between each bayonet is greater than threshold value Ti, if
It is Ki to be added to A1 and empty array A_new, and Ki is removed from array A0；
Step7.5. a bayonet Kj is taken out from array A_new, judges the undirected probability Q in Kj and A0 between each bayonet
Whether (Kj, Ki) is greater than threshold value Ti, if so, Ki is added to A1 and empty array A_new, and by Ki from array A0
It removes, Kj is removed from array A_new；
Step7.6. Step7.5 is repeated, until A_new is sky；
Step7.7. increase array Ai, repeat step Step7.3-7.6；
Step7.8. Step7.7 is repeated, until A0 is sky；
Step7.9. Step7.2-7.8 is repeated, until number plate inventory is sky.
3. a kind of fake-licensed car detection method of the probability polymerization based on vehicle hot spot region as claimed in claim 1 or 2, special
Sign is: in step S2, the cleaning process of bayonet data is as follows:
2.1, cleaning repeats to record；
2.2, dirty data is cleaned, the dirty data refers to that brand number does not meet the record of naming rule；
2.3, it determines time interval, cleans the bayonet data that record is not complete in the section, including bayonet damage and power-off cause
Bayonet can not work or network interruption and disk size deficiency cause data to fail to store, and clean all data of the bayonet.
4. a kind of fake-licensed car detection method of the probability polymerization based on vehicle hot spot region as claimed in claim 1 or 2, special
Sign is: in step S4, threshold value TiIt is set according to bayonet number plate recognition correct rate F, to bayonet KiFor, TiValue answer
Meet F≤∑jPij<F-Ti, wherein j meets Pij>Ti, bayonet number plate recognition correct rate F=1- (pass through bayonet but number plate
Vehicle of crossing not in number plate storehouse records number)/(all vehicles of crossing record number), TiIndicate bayonet KiThreshold value, PijIndicate bayonet PiIt arrives
The probability of other bayonets.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610380318.6A CN106022296B (en) | 2016-06-01 | 2016-06-01 | A kind of fake-licensed car detection method of the probability polymerization based on vehicle hot spot region |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610380318.6A CN106022296B (en) | 2016-06-01 | 2016-06-01 | A kind of fake-licensed car detection method of the probability polymerization based on vehicle hot spot region |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106022296A CN106022296A (en) | 2016-10-12 |
| CN106022296B true CN106022296B (en) | 2019-05-28 |
Family
ID=57092885
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610380318.6A Active CN106022296B (en) | 2016-06-01 | 2016-06-01 | A kind of fake-licensed car detection method of the probability polymerization based on vehicle hot spot region |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106022296B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108665699B (en) * | 2017-03-30 | 2020-04-03 | 杭州海康威视数字技术股份有限公司 | Method and device for predicting vehicle appearance place |
| CN107329977B (en) * | 2017-05-27 | 2019-08-16 | 银江股份有限公司 | A kind of false-trademark vehicle postsearch screening method based on probability distribution |
| CN107563288A (en) * | 2017-07-31 | 2018-01-09 | 东软集团股份有限公司 | A kind of recognition methods of fake-licensed car vehicle and device |
| CN108022427A (en) * | 2017-10-30 | 2018-05-11 | 深圳市赛亿科技开发有限公司 | A kind of recognition methods of fake license plate vehicle and system |
| CN107832475A (en) * | 2017-12-01 | 2018-03-23 | 浪潮软件集团有限公司 | Deployment control method, deployment control engine and deployment control system |
| CN108242153A (en) * | 2018-03-12 | 2018-07-03 | 小草数语（北京）科技有限公司 | Abnormal bayonet recognition methods and device |
| CN109741227B (en) * | 2019-01-07 | 2020-12-08 | 巩志远 | Processing method and system for predicting human-room consistency based on nearest neighbor algorithm |
| CN110310478B (en) * | 2019-05-17 | 2020-12-18 | 湖南科创信息技术股份有限公司 | Method and system for identifying fake-licensed vehicle based on big data analysis and storage medium |
| CN110164138B (en) * | 2019-05-17 | 2021-02-09 | 湖南科创信息技术股份有限公司 | Identification method and system of fake-licensed vehicle based on bayonet convection direction probability and medium |
| CN110164140A (en) * | 2019-06-05 | 2019-08-23 | 上海易点时空网络有限公司 | Deck detection system and device |
| CN112150795B (en) * | 2019-06-26 | 2022-05-06 | 杭州海康威视数字技术股份有限公司 | Method and device for detecting vehicle track abnormity |
| CN111369804B (en) * | 2019-07-05 | 2022-04-05 | 杭州海康威视系统技术有限公司 | Vehicle data processing method and device, electronic equipment and storage medium |
| CN112733047B (en) * | 2020-11-05 | 2022-10-28 | 浙江大华技术股份有限公司 | Vehicle foothold generation method, device, equipment and computer storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001034799A (en) * | 1999-07-21 | 2001-02-09 | Mitsubishi Electric Corp | Automatic toll collection system for traveling vehicle and its method |
| CN101587643A (en) * | 2009-06-08 | 2009-11-25 | 宁波大学 | Identification method of fake-licensed cars |
| CN103914986A (en) * | 2014-03-14 | 2014-07-09 | 浙江宇视科技有限公司 | Method and device for fake-license-plate analysis |
| CN104732765A (en) * | 2015-03-30 | 2015-06-24 | 杭州电子科技大学 | Real-time urban road saturability monitoring method based on checkpoint data |
-
2016
- 2016-06-01 CN CN201610380318.6A patent/CN106022296B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001034799A (en) * | 1999-07-21 | 2001-02-09 | Mitsubishi Electric Corp | Automatic toll collection system for traveling vehicle and its method |
| CN101587643A (en) * | 2009-06-08 | 2009-11-25 | 宁波大学 | Identification method of fake-licensed cars |
| CN103914986A (en) * | 2014-03-14 | 2014-07-09 | 浙江宇视科技有限公司 | Method and device for fake-license-plate analysis |
| CN104732765A (en) * | 2015-03-30 | 2015-06-24 | 杭州电子科技大学 | Real-time urban road saturability monitoring method based on checkpoint data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106022296A (en) | 2016-10-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106022296B (en) | A kind of fake-licensed car detection method of the probability polymerization based on vehicle hot spot region | |
| CN105719489B (en) | A kind of fake-licensed car detection method that probability is flowed to based on bayonet vehicle | |
| CN104200669B (en) | Fake-licensed car recognition method and system based on Hadoop | |
| CN106971552B (en) | Fake plate phenomenon detection method and system | |
| CN105225476B (en) | A kind of generation of track of vehicle, polymerization and device | |
| WO2017157119A1 (en) | Method and device for identifying abnormal behavior of vehicle | |
| WO2019153193A1 (en) | Taxi operation monitoring method, device, storage medium, and system | |
| CN109461106A (en) | A kind of multidimensional information perception processing method | |
| CN101587643B (en) | Identification method of fake-licensed cars | |
| CN112447041B (en) | Method and device for identifying operation behavior of vehicle and computing equipment | |
| CN107256394A (en) | Driver information and information of vehicles checking method, device and system | |
| CN109886204A (en) | A kind of Multidimensional Awareness system based on the application of big data police service | |
| CN107346435A (en) | A kind of suspicion fake-licensed car catching method based on vehicle characteristics storehouse | |
| CN107329977B (en) | A kind of false-trademark vehicle postsearch screening method based on probability distribution | |
| CN109326124A (en) | A kind of urban environment based on machine vision parks cars Activity recognition system | |
| CN103413440A (en) | Fake-licensed vehicle identification method based on smart city data base and identification rule base | |
| CN103514745A (en) | Fake license plate vehicle recognition method based on intelligent transportation | |
| CN106297304A (en) | A kind of based on MapReduce towards the fake-licensed car recognition methods of extensive bayonet socket data | |
| CN108074400A (en) | A kind of emphasis vehicle analysis model based on mass data analysis | |
| CN106340205A (en) | Traffic monitoring method and traffic monitoring apparatus | |
| CN112925820B (en) | Method, device and system for identifying vehicle evasion toll | |
| CN105206061A (en) | Method relying on intelligent urban transportation capable of identifying license plates through collection of RFID videos | |
| CN109726701A (en) | Vehicle identification method and system | |
| CN106710225A (en) | Identification method and monitoring platform for illegal operation related with vehicle license plate | |
| CN103700262A (en) | Automatic area acquisition method for fake-licensed vehicles |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP01 | Change in the name or title of a patent holder | ||
| CP01 | Change in the name or title of a patent holder |
Address after: 310012 1st floor, building 1, 223 Yile Road, Hangzhou City, Zhejiang Province Patentee after: Yinjiang Technology Co.,Ltd. Address before: 310012 1st floor, building 1, 223 Yile Road, Hangzhou City, Zhejiang Province Patentee before: ENJOYOR Co.,Ltd. |