CN106297304A - A kind of based on MapReduce towards the fake-licensed car recognition methods of extensive bayonet socket data - Google Patents

Abstract

The present invention relates to a kind of based on MapReduce towards the fake-licensed car recognition methods of extensive bayonet socket data, comprise the following steps: 1) obtain multiple road gate cross car information and bayonet socket self information data, and carry out pretreatment, remove abnormal and Unidentified car information excessively；2) use the big data processing platform (DPP) of Hadoop that pretreated car information and bayonet socket self information data excessively are modeled, and by MapReduce parallel programming technology, suspicious fake-licensed car is identified.Compared with prior art, the present invention has the advantages such as processing speed is fast.

Description

A kind of based on MapReduce towards the fake-licensed car recognition methods of extensive bayonet socket data

Technical field

The present invention relates to road safety field, especially relate to one based on MapReduce towards extensive bayonet socket data Fake-licensed car recognition methods.

Background technology

Along with the growth of Chinese national economy level, vehicle guaranteeding organic quantity is the most all increasing rapidly, investigates and prosecutes various traffic Illegal violation phenomenon is to ensure that the Important Action of traffic safety.In various traffic offences are violating the regulations, vehicle " deck " is to have sternly The infringement heavily endangered.Vehicle " deck " phenomenon, refers to that vehicle illegally uses showing of the license plate number identical with other legal vehicles As.According to the report of various places media, " deck " vehicle serious harm traffic transport industry and operation order, the safety to people And the interests of legal vehicle form serious threat.Lawless person applies mechanically other people vehicle license, escapes traffic accident responsibility, escapes the expenses of taxation Be engaged in criminal activity, had a strong impact on the people's lives and property safety, upset social order, compromised social safety.Control Reason " deck " vehicle, it has also become various places public security department and the vital task of vehicle supervision department.

Existing document and disclosed patent propose the recognition methods of fake-licensed car.Current main method has information of vehicles pair Place diagnostic method is travelled than method and vehicle.Information of vehicles matching type is to set up a registered vehicle information at traffic control center Storehouse, the information of vehicles comparison in the information of vehicles obtained by Internet of Things or video image analysis and data base, if be not inconsistent, then This car plate is suspicious deck.The fake-licensed car detection method based on Internet of Things proposed such as Yang Bo, uses electronic label technology, will deposit Store up the electronic tag implanter motor-car of the information such as automobile license plate and Motor Number, covered when motor vehicles sails the control point deployed troops on garrison duty into During the scope covered, the information in electronic tag is automatically read, and the information of vehicles comparison in vehicle supervision department data base, Information is not inconsistent, and is identified as fake-licensed car.Number of patent application be the method for 201310170646 be to set up vehicle feature database and vehicle Base library, according to the vehicle image identification vehicle license plate gathered, vehicle etc., and obtains with retrieving from vehicle feature database according to car plate Vehicle compare identification fake license plate vehicle.As long as the method that number of patent application is 200910099475 is to have identical car plate Number the vehicle of more than two simultaneously appear on road, the when and where according to occurring identifies whether deck.

There are some drawbacks in said method when reality is applied.Fake-licensed car recognition methods based on Internet of Things needs to motor-driven Car is implanted electronic tag and disposes wireless monitor point, relatively costly；Method based on image and video is by illumination, environmental effect relatively Greatly, accuracy rate is the highest, and vehicle travels place diagnostic method needs data volume to be processed big, it is desirable to the efficiency of processing system is sufficiently high. These methods are required for each vehicle through monitoring point is analyzed and is processed, and amount of calculation and data volume are big.Due to deck The range of activity of car is very wide, and monitoring point is the most, when the information of vehicles of collection is the most, can identify deck car plate more.Right For the city that vehicle population is huge, the vehicle data amount that every day gathers reaches TB level, uses file storage or number According to library storage mode, it is the lowest to the inquiry of data and the efficiency of analysis, calculates by single computer 100MB/SeC, Read ZTB data and need 1.5 hours, realize query analysis on this basis and be almost the most achievable task, use SQL data Storehouse is more common method, but data base needs the most powerful computer, in TB and data above process, and the number of data base Very big according to management and optimization difficulty.For realizing fast and effeciently analyzing large-scale vehicular traffic data, need a kind of new skill Art scheme meets the demand of traffic control department.

Summary of the invention

Defect that the purpose of the present invention is contemplated to overcome above-mentioned prior art to exist and provide a kind of processing speed fast Based on MapReduce towards the fake-licensed car recognition methods of extensive bayonet socket data.

The purpose of the present invention can be achieved through the following technical solutions:

A kind of based on MapReduce towards the fake-licensed car recognition methods of extensive bayonet socket data, comprise the following steps:

1) obtain multiple road gate crosses car information and bayonet socket self information data, and carries out pretreatment, removes exception Car information is crossed with Unidentified；

2) use the big data processing platform (DPP) of Hadoop that pretreated car information and bayonet socket self information data excessively are carried out Modeling, and by MapReduce parallel programming technology, suspicious fake-licensed car is identified.

Described step 2) specifically include following steps:

21) bayonet socket numbering k of any two road gate is obtained_x、k_yWith relative distance d istance, and with (k_xk_y, Distance) form uses Map interface HashMap based on Hash table to be saved in the internal memory of computer node；

22) pretreated is crossed the input as Map function of car information and bayonet socket self information data, and be converted into key Being worth the output as Map function of the form to<key, value>, wherein key value is the license plate number snapping over car, and value value is This snaps over the corresponding vehicle of car, car color, crosses car time, bayonet socket numbering, buckle longitude and buckle latitude；

23) key-value pair of Map function output is carried out shuffle sequence, merge the value value of identical key value, with < Key, list of value > form export to Reduce function；

24) use Reduce function that the data in list of value are compared two-by-two, identify suspicious fake-licensed car ?.

Described step 24) in, if the vehicle of value value is different with car color in list of value, then prove same Vehicle or color that license plate number is corresponding are different, are judged to the suspicious fake-licensed car of the first kind, and with < license plate number: car 1 occurs ground There is place, car 2 time of occurrence in point, car 1 time of occurrence, car 2 > form output result.

Described step 24) in, if the vehicle of value value is identical with color in list of value, then in office according to it Average speed between two road gate of anticipating is that index judges:

Relative distance d istance according to any two road gate and excessively car Time Calculation actual average speed per hour, and set A fixed average speed per hour threshold value, if actual average speed per hour is more than average speed per hour threshold value, is then judged to the suspicious deck of Equations of The Second Kind Car, and with the form output knot of<license plate number: place occurs in car 1: car 1 time of occurrence: place occurs in car 2: car 2 time of occurrence> Really.

Described step 2) further comprising the steps of:

25) according to the number-plate number, body color, vehicle, the time of discovery fake-licensed car and the ground of all suspicious fake license plate vehicle Dot information sets up fake-licensed car early warning information storehouse.

Described step 21) in, the calculating formula of relative distance d istance of any two road gate is:

$Dis\tan ce=2\&CenterDot;R_e\&CenterDot;\arcsin \left(\sqrt{{\sin }^2(latx-laty)+cos\left(latx\right)*cos\left(laty\right)*{\sin }^2(\ln gx-\ln gy)}\right)$

Wherein, R_eFor earth radius, latx and lngx is respectively bayonet socket k_xLongitude and latitude, laty and lngy is bayonet socket k_yLongitude and latitude.

Compared with prior art, the invention have the advantages that

Processing speed is fast: the present invention utilizes the distributed computing framework MapReduce in big data technique to realize at big number Excavate fake-licensed car according to middle parallelization and can solve fake-licensed car Mining Problems in extensive bayonet socket data, be possible not only to effectively excavate relate to The fake license plate vehicle disliked, and compared to conventional art, it is possible to it is effectively improved the recognition efficiency of fake license plate vehicle.And when Hadoop collection Node in Qun is the most, and the speed processing large-scale data is the fastest.

Accompanying drawing explanation

Fig. 1 is the flow chart of steps of the present invention.

Detailed description of the invention

The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.

Embodiment:

As it is shown in figure 1, first data are extracted, integrates, then carry out pretreatment, filter out " vehicle: car color: Spend the car time: bayonet socket is numbered: longitude: dimension " data form.Data are uploaded in HDFS, use based on Hadoop/ Fake-licensed car is excavated by the parallel programming of MapReduce, and concrete comprises the following steps:

(1) car information and the bayonet socket self excessively that gather each bayonet system extract for information about, integrate, and Initial data is carried out pretreatment, removes exception and the car data excessively not identified, use HDFS distributed file system Store.

Utilize Hadoop big data processing platform (DPP) that card makes a slip of the tongue car data and card its data is modeled and analyzes, pass through MapReduce parallel programming technology realizes suspicious fake-licensed car identification:

(2) first unit loads numbering and the latitude and longitude information of each bayonet socket, calculates according to following longitude and latitude formula (1) Go out the air line distance between each two bayonet socket.According to (k_xk_y, distance) and store every any two bayonet socket k_xAnd k_yBetween Distance distance.In the middle of the disk local presented in text dis.txt by result, after waiting parallel computation time Time is loaded in internal memory.

Formula 1:

$\begin{array}{c}Dis\tan ce\left(K1,K2\right)=2*EARTH\_RADIUS*\\ \arcsin \left(\sqrt{{\sin }^2\left(lat1-lat2\right)+\cos \left(lat1\right)*\cos \left(lat2\right)*{\sin }^2\left(\ln g1-\ln g2\right)}\right)\end{array}$

Wherein EARTH_RADIUS is the radius of the earth, lng1 and lat1 is the precision of bayonet socket 1, latitude respectively, lng2 and Lat2 is the longitude of bayonet socket 2, latitude respectively.

(3) raw data set is divided into multiple data block, the meter in MapReduce cluster by MapReduce bottom frame Calculation machine node starts multiple Mapper, and each Mapper stage processes the data block information of correspondence respectively.Start map function it Before, utilize the distance between the bayonet socket that setup function will be saved on local disk in dis.txt in step (1) to be loaded into internal memory In, use HashMap (String, Double) to be saved in internal memory.Owing to setup function is loaded with before map starts, So another node Mapper personage all of the above all can share the bayonet socket spacing in setup function.

(4) each Mapper starts to read the relevant information of car data, carries out field cutting to crossing car information, skips word What hop count did not met that standard or license plate number be " without car plate " and " 00000000 " crosses car data, will be left normal car number According to being converted into the output of<key, value>form, obtaining key value is license plate number, and value value is corresponding " vehicle: the Che Yan of this car Color: spend the car time: bayonet socket is numbered: longitude: dimension ".The value of Key and value carries out cutting one according to space and has 8 fields, The record made due to superposition due to a variety of causes in real data, less than 8 fields, in Map function, works as field This record can be skipped the when of less than 8.Key and value after reconfiguring is according to (key, value) output to Reduce End.

(5) Map output (key, value) is before entering into Reduce, can be through row's shuffle process, at the bottom of Hadoop Key value can be ranked up by layer, is then combined with the value of identical key value, can arrive with (key, list of values) Reduce end.At Reduce end, all records are all different key values, the process that reduce function can record one by one, Value in values, for each record, is compared by reduce two-by-two, if it find that vehicle in two character strings or Person's car color has different, and this just illustrates that same license plate number, vehicle or color are different, and this has fake-licensed car suspicion, we Can be with " place occurs in * * license plate number * *: car 1: car 1 time of occurrence: place occurs in car 2: car 2 time of occurrence " output result, this is defeated But going out explanation has the identical vehicle of two cars board and body color different, and deck suspicion is huge.

(6) in step (4), find, when every two vehicles recorded or body color are identical, just leaving in Bayonet socket spacing in HashMap finds out distance d of now two bayonet sockets, then utilizes two Time Calculation in record to go out one Individual time difference Δ t, defines a threshold value y=d/ Δ t, and y here can be equal to the average speed of car, but at us In physical significance be that a car can not occur in far two places of meeting, within a very short time if it is, we are just It is thought that fake-licensed car.In the algorithm, we can adjust different threshold values, such as 100km/h, 150km/h, 200km/ respectively H, 400km/h, 800km/ or more than 1000km/h, threshold value is the biggest, if also result to occur just representing deck suspicion the biggest. Finally also according to " license plate number: place occurs in car 1: car 1 time of occurrence: place occurs in car 2: car 2 time of occurrence ", result is defeated Go out.

(7) in step (4) and (5), we define two rules to judge fake-licensed car, and rule one is identical car plate Number, as long as body color or vehicle are different, we are taken as fake-licensed car.If two of rule two same license plate numbers Car occurs in two places that can not reach within a period of time, and we are considered as having deck suspicion.In step (4) and (5) result that the fake-licensed car found in the case of two kinds in finds is distinguish between, and can preferably distinguish what deck caused with let us Reason.

(8) fake-licensed car early warning information storehouse is set up, including can be with the number-plate number of fake-licensed car, body color, vehicle, discovery set The time of board car, place.

Claims (6)

1. one kind based on MapReduce towards the fake-licensed car recognition methods of extensive bayonet socket data, it is characterised in that include following Step:

1) obtain multiple road gate crosses car information and bayonet socket self information data, and carries out pretreatment, removes abnormal and not Identify crosses car information；

2) use the big data processing platform (DPP) of Hadoop that pretreated car information and bayonet socket self information data excessively are modeled, And by MapReduce parallel programming technology, suspicious fake-licensed car is identified.

The most according to claim 1 a kind of based on MapReduce towards the fake-licensed car recognition methods of extensive bayonet socket data, It is characterized in that, described step 2) specifically include following steps:

21) bayonet socket numbering k of any two road gate is obtained_x、k_yWith relative distance d istance, and with (k_xk_y, Distance) form uses Map interface HashMap based on Hash table to be saved in the internal memory of computer node；

22) pretreated is crossed the input as Map function of car information and bayonet socket self information data, and be converted into key-value pair The form of<key, value>is as the output of Map function, and wherein key value is the license plate number snapping over car, and value value is this card Snap through the corresponding vehicle of car, car color, cross car time, bayonet socket numbering, buckle longitude and buckle latitude；

23) key-value pair of Map function output is carried out shuffle sequence, merge the value value of identical key value, with < key, List of value > form export to Reduce function；

24) use Reduce function that the data in list of value are compared two-by-two, identify suspicious fake license plate vehicle.

The most according to claim 2 a kind of based on MapReduce towards the fake-licensed car recognition methods of extensive bayonet socket data, It is characterized in that, described step 24) in, if the vehicle of value value is different with car color in list of value, then prove Vehicle or color that same license plate number is corresponding are different, are judged to the suspicious fake-licensed car of the first kind, and with < license plate number: car 1 goes out There is place, car 2 time of occurrence in existing place, car 1 time of occurrence, car 2 > form output result.

The most according to claim 2 a kind of based on MapReduce towards the fake-licensed car recognition methods of extensive bayonet socket data, It is characterized in that, described step 24) in, if the vehicle of value value is identical with color, then according to it in list of value Average speed between any two road gate is that index judges:

Relative distance d istance according to any two road gate and excessively car Time Calculation actual average speed per hour, and set one Individual average speed per hour threshold value, if actual average speed per hour is more than average speed per hour threshold value, is then judged to the suspicious fake-licensed car of Equations of The Second Kind, and Form output result with<license plate number: place occurs in car 1: car 1 time of occurrence: place occurs in car 2: car 2 time of occurrence>.

5. according to a kind of described in claim 3 or 4 based on MapReduce towards the fake-licensed car identification side of extensive bayonet socket data Method, it is characterised in that described step 2) further comprising the steps of:

25) according to the number-plate number of all suspicious fake license plate vehicle, body color, vehicle, the when and where letter of discovery fake-licensed car Breath sets up fake-licensed car early warning information storehouse.

The most according to claim 2 a kind of based on MapReduce towards the fake-licensed car recognition methods of extensive bayonet socket data, It is characterized in that, described step 21) in, the calculating formula of relative distance d istance of any two road gate is:

$Dis\tan ce=2\&CenterDot;R_e\&CenterDot;arcsin\left(\sqrt{{\sin }^2(latx-laty)+cos\left(latx\right)*cos\left(laty\right)*{\sin }^2(\ln gx-\ln gy)}\right)$

Wherein, R_eFor earth radius, latx and lngx is respectively bayonet socket k_xLongitude and latitude, laty and lngy is bayonet socket k_y's Longitude and latitude.