CN104700646B - A kind of taxi exception track real-time detection method based on online gps data - Google Patents

Abstract

The invention discloses a kind of taxi exception track real-time detection method based on online gps data, including: step 1, based on the open street map data increased income, target cities is carried out road network modeling；Step 2, applies k shortest path first to obtain taxi on road network model and travels recommendation paths Candidate Set Rec；Step 3, the recommendation paths therefrom selected according to recommendation paths Candidate Set Rec and driver, real-time GPS data collection in conjunction with current taxi, and incorporate passenger's feedback to current taxi driving trace, the online taxi runed is carried out abnormal track detect in real time, and real-time testing result is reported Public Security Department；Step 4, analyzes the result of online taxi exception track detection, extracts specific taxi and travels rule, show that zonal taxi is abnormal and travels report.

Description

A kind of taxi exception track real-time detection method based on online gps data

Technical field

The present invention relates to the process of traffic big data and computer calculates field in real time, particularly a kind of based on online GPS The taxi exception track real-time detection method of data.

Background technology

Along with the fast development of sensor technology, the position locus information of magnanimity is constantly produced by sensor, particularly GPS (Global Position Systems) sensor, and by effective data analysing method, it is hidden in these trace informations Potential data value behind can be mined out.The mobile traffic data collected from the taxi loading GPS device and come For analyze hire a car drivers ' behavior and road network change provide various analysis method.And excavate specific GPS track pattern Public transport management, urban construction and the raising of public community service quality can be promoted.The potential value of GPS track is Embodied by numerous application, if J.Yuan et al. is at " T-drive:driving directions based on taxi Trajectories " in use magnanimity taxi GPS record to excavate the most intelligent travel direction to provide the user convenient fast Prompt path navigation；B.D.Ziebart et al. is at " Navigate like a cabbie:Probabilistic reasoning From observed context-aware behavior " in driving experience based on taxi driver use history GPS rail Mark provides the user the optimum drive route arrived at；F.Giannotti et al. is at " Trajectory pattern Mining " in be that Frequent Trajectory Patterns provides accurate description from the angle of room and time.

As the manager of whole taxi trade, Public Security Department is faced with numerous problem of management now, especially when The management of front taxi monitoring system.Due to current technology and the restriction of monitoring system and substantial amounts of operation taxis quantity, Traffic Administration Bureau is unable to reach efficient fine-grained taxi supervision.Compare current taxi track data increment and Processing complexity, taxi monitoring system cannot meet the supervision demand of present stage, and when passenger complains taxi department During machine, Traffic Administration Bureau have to put into the problem that sizable manpower and materials go to solve passenger.It is further noted that, the most not Effective measures and method is had to go the taxi stopping to be implemented by taxi driver to swindle the behavior of passenger.

The work having many Exception Model for GPS track to excavate at present has been achieved for some achievements, and traditional is different Often detection is to descend on data online, based on track distance, based on track density, based on track distribution and based on track deviation Etc. aspect and carry out track abnormal patterns identification and excavation, in addition, also have entered by training based on machine learning scheduling algorithm The abnormal trajectory model of row excavates.Although there is the technology of a lot of track abnormality detection present stage, but still some problem is not Effectively being solved, such as currently the majority track abnormality detection technology is primarily directed to taxi fraudulent act and occurs very Detection after Chang Shijian, so can only play the effect afterwards processed, and Traffic Administration Bureau taxi driver to be undertaken defends oneself Cost, effective management can not be provided for the supervision in taxi market.For further, current abnormal track detection Major part is all dependent on data analysis reasoning, can not process some passenger's special demands.

Summary of the invention

Goal of the invention: the technical problem to be solved is for the deficiencies in the prior art, it is provided that a kind of based on The taxi exception track real-time detection method of line gps data.

In order to solve above-mentioned technical problem, the invention discloses a kind of taxi exception track based on online gps data Real-time detection method, comprises the following steps:

Step 1: target cities is carried out road network based on open street map (OpenStreetMap) data increased income Modeling；

Step 2: for each taxi, the route received according to onboard sensor (onboard sensor in panel computer) The journey's end that starting point and passenger confirm, respectively from road network road distance and two sides of the average running time of road network road Face, applies k shortest path first (KSP) to carry out the taxi driving path recommendation of efficiently and accurately on road network model, And obtain recommendation paths Candidate Set Rec；

Step 3: the recommendation paths therefrom selected according to recommendation paths Candidate Set Rec and driver, in conjunction with current taxi Real-time GPS data collection, and incorporate passenger's feedback to current taxi driving trace, the taxi of online operation carried out Abnormal track detects in real time, and wherein testing result includes abnormal taxi license board information, taxi driver's essential information, sends out in real time The letter such as raw abnormal place (longitude and latitude), the time of generation exception, passenger loading place (longitude and latitude), taxi transaction odd numbers Breath, and this real-time testing result is uploaded and be recorded in the data base that Public Security Department specifies；

Step 4: for the result of online taxi exception track detection, according to data analysis algorithm, testing result is entered Row is analyzed, and extracts urban taxi and travel rule, including different cities, different periods, different company taxi abnormal Record statistics, and with this basis, the taxi driver of city zones of different is carried out monthly, season and year scoring, tie the most at last Fruit feeds back to the relevant departments such as management board of Department of Communications, taxi company.

In the present invention, the four kinds of Data Structures provided according to the open street map data of target cities: node Node, road Way, relationship model Relation, model label Tag, take out the elementary path road net model of target cities.Road Joint Segment is the elementary cell of every road, and it is to be had relatively short distance by what the most orderly a series of Nodes generated Section unit.And section Section to be Segment the most abstract, it is by a series of the most orderly Segments generates, and the end points of Segment is the intersection between road.So, by abstract open street map The Data Structures of data, can obtain the road network model of target cities based on Section.

In the present invention, according to the target cities road network model generated, taxi initiate carrying place O and The journey's end D determined after passenger loading, respectively from road network road distance and two sides of the average running time of road network road Face, uses k shortest path first (k-Shortest-Path, KSP), to this taxi carrying every trade subsequently for carrying out relatively The driving path answered is recommended.

Owing to the form of expression of the GPS point taxi track tr of conventional discrete cannot meet follow-up online abnormal rail Mark detects in real time, so in the present invention, discrete GPS point taxi track is converted into the Section sequence corresponding to GPS point Row, the most abstract taxi track atr, its form of expression is as follows:

Atr={sec₁→…→sec_j→…→sec_m} (1)

Wherein, sec_jThe Section of GPS point p institute projection mapping in tr, and in atr the order of Section with in tr The order of GPS point p keeps consistent.

Therefore, after O and D that given passenger determines, utilize k shortest path first (KSP) from road network road distance with And road network road average running time two aspect provides suitable k bar recommendation paths respectively, form recommendation paths candidate collection Rec, i.e. Rec={atr₁..., atr_k}。

In the present invention, according to recommendation paths candidate collection Rec and taxi driver, the final benchmark selected in Rec pushes away Recommend path atr_bench, in conjunction with the real-time GPS data collection of current taxi, respectively from current taxi travel direction and travel away from From two aspects, the taxi of online operation is carried out abnormal track to detect in real time, and the result detected incorporates the real-time of passenger Feedback so that the testing result of abnormal track is more accurate, meanwhile real-time testing result is reported Public Security Department.

From the point of view of current taxi travel direction, after given starting point O and terminal D, it is the most square that taxi travels To being determined, and when given recommendation paths candidate collection Rec and the driver final benchmark selected in Rec is recommended Path atr_benchTime, the direction that taxi travels the most more easily determines within limits.Determine according to below equation and work as Whether front taxi travel direction is abnormal:

Wherein, score_dirExpression travel direction exceptional value, temperature coefficient 0≤ξ≤1,It is by recommendation paths Candidate Set Close the set of the Section that all of atr is generated in Rec, and | atr_bench| andRepresent atr respectively_benchWithThe number of middle Section, | Sim_bench| and | Sim_cdd| represent Section and atr at current taxi place respectively_bench The number that middle Section is similar, Yi JiyuThe number that middle Section is similar.

If score_dir＞ λ_dir, then the travel direction of current taxi just can be judged as abnormal, wherein λ_dir For direction threshold value.

From the point of view of current taxi operating range, according to fixed starting point O and terminal D and current taxi institute At position p_current, distance d that taxi has been passed by_pass、p_currentFrom remaining beeline d of terminal D_minRmAnd recommend The average distance d of recommendation paths in path candidates set Rec_avgThese three distance variable can substitute into below equation and judge to work as Front taxi operating range is the most abnormal:

${\mathrm{score}}_{\mathrm{dis}}=\left\{\begin{array}{cc}\mathrm{\η}-1,& \mathrm{\η}>1\\ 0,& \mathrm{\η}\≤1\end{array}\right.---\left(3\right)$

Wherein, score_disRepresent operating range exceptional value,And|| Sec | | represent the operating range of Section.

If score_dis＞ λ_dis, then the operating range of current taxi just can be judged as abnormal, wherein λ_dis For distance threshold.

Owing to current taxi track is continually changing, so the history mobile status of this taxi track also can Affect the travel direction of current taxi and the judgement of operating range.Therefore, the condition grading Φ strengthened by introducing is to working as Front taxi state carries out judging more comprehensively and accurately.From the point of view of this taxi whole carrying route, pushing away over time Move, history score_dirAnd score_disTo current score_dirAnd score_disThere is the least impact, therefore can pass through Weight coefficient τ balances the historic state impact on current state.

Assume to concentrate the time interval Δ t of adjacent GPS point to be all equal for taxi real-time GPS data, and Initial time t₀'sWhereinIt is equal toOrSo it is the most permissible by below equation Obtain the Φ of random time,

So for the score under t_{Dir, t}And score_{Dis, t}May be replaced withWithCarry out state Judge.

If the travel direction of current taxi or operating range are judged as abnormal, then the shape that this taxi is current State then becomes suspicious state, and now system the passenger on taxi can send prompting, and request confirms current driving path Whether is off path, if passenger confirms it is abnormal driving path, then this taxi traveling record can be reported to Traffic Administration Bureau to enter Row close supervision；The abnormal driving path if passenger really admits a fault, then according to current taxi traveling-position p_currentAnd terminal D, repeats step 2 and step 3, and new recommendation paths Candidate Set can generate, and current taxi driving path can persistently be examined Survey, the D until taxi is reached home.

In the present invention, for the result of online taxi exception track detection, according to data analysis algorithm to detection knot Fruit is analyzed, and extracts specific taxi and travels rule, and analysis result feeds back to Department of Communications, taxi company etc. the most at last Relevant department.

Compared with prior art, the invention have the advantages that:

(1) taxi track detection is for online data, and carries out in real time, after detecting extremely, and energy Enough quick driving behaviors to taxi driver guide or intervene.Compared to processing track data under line, the present invention is more Have ageing.

(2) taxi track abnormality detection result adds the feedback of current passenger, effectively have recorded taxi and travel In state change, complain for subsequent analysis taxi driver's driving behavior and passenger and provide reliable foundation.

Accompanying drawing explanation

Being the present invention with detailed description of the invention below in conjunction with the accompanying drawings and further illustrate, the present invention's is above-mentioned And/or otherwise advantage will become apparent.

Fig. 1 is the present invention online abnormality detection flow chart.

Fig. 2 is the basic framework figure of the inventive method.

Detailed description of the invention

Below in conjunction with the accompanying drawings the present invention is illustrated.It is noted that described embodiment merely to explanation Purpose rather than limitation of the scope of the invention.

The invention discloses a kind of taxi exception track real-time detection method based on online gps data, this detection side Method flow chart is as it is shown in figure 1, comprise the following steps:

Step 1: target cities is carried out road network based on open street map (OpenStreetMap) data increased income Modeling.From the point of view of current technology, mainly there are two kinds of basic city decomposition techniques: gridding method and electronic chart method.Net City is specifically divided into the net region of equal size by lattice method, and represents, by the sequence of grid, the track that vehicle travels, its Advantage is operation simple, convenient, but its also some obvious problem, such as build when net region is not suitable for current city During mould, just cannot know vehicle mobile behavior within a grid；If when number of grid is the most sparse, just cannot be carried out online The analysis of road vehicle condition, and the decomposition coefficient Θ in each city is different and is difficult to determine；Specified link cannot be analyzed Road conditions and vehicle condition.So in the present invention, we have employed electronic chart method, i.e. by the ground, open street increased income Diagram data carries out road network modeling to target cities.The advantage of this method is can to carry out according to the road network in real-life Analyze, maintain the high efficiency of model.Four kinds of master data knots that open street map data according to target cities is provided Structure: node Node, road Way, relationship model Relation, model label Tag takes out the elementary path road network of target cities Model.Road joint Segment is the elementary cell of every road, and it is by having that the most orderly a series of Nodes are generated The section unit of relatively short distance.And section Section to be Segment the most abstract, it be by a series of the most in order Segments generate, and the end points of Segment is the intersection between road.For example, if giving a Nodes Sequence Node₁…Node_i…Node_j…Node_m(1≤i≤j≤m), i, j represent the mark of different Node, and m is that target cities is opened Send out total number of Node in street map data, then

sec_{I, j}={ seg_{I, i+1}..., seg_{J-1, j}}

So, by the Data Structures of abstract open street map data, can obtain with Section for the most single The road network model of the target cities of unit.

Step 2: for each taxi, according to the target cities road network model generated, the initial load of taxi The journey's end D determined after place far way from home point O and passenger loading, respectively from road network road distance and road network road average row Sail two aspects of time, use k shortest path first (KSP), to this taxi carrying every trade be subsequently carry out corresponding Driving path is recommended.

In the present invention, all of track refers both to the taxi track of carrying.Traditional GPS taxi track tr is one to be The sequence of row GPS point, i.e.

Tr={p₁→…→p_i→…→p_n}

Wherein, GPS point p_iIt is a tlv triple<lat, lng, timestamp>, and lat, lng, timestamp are respectively Refer to the latitude of GPS point, longitude and timestamp.

Owing to the form of expression of the GPS point taxi track tr of conventional discrete cannot meet follow-up online abnormal rail Mark detects in real time, so in the present invention, discrete GPS point taxi track is converted into the Section sequence corresponding to GPS point Row, the most abstract taxi track atr, its form of expression is as follows:

Atr={sec₁→…→sec_j→…→sec_m}

Wherein, sec_jThe Section of GPS point p institute projection mapping in tr, and in atr the order of Section with in tr The order of GPS point p keeps consistent.

Therefore, after O and D that given passenger determines, utilize k shortest path first (KSP) from road network road distance with And road network road average running time two aspect provides suitable k bar recommendation paths respectively, form recommendation paths candidate collection Rec, i.e. Rec={atr₁..., atr_k}.K value natural number, within generally 10.

Concrete recommendation paths is divided into three steps,

(1) by it has been determined that OD and current taxi travel direction, filter out road according to nearest Euclidean distance Intersection Node_OAnd Node_D, then with Node_OAnd Node_DK shortest path first is utilized to produce respectively as beginning and end K bar recommendation paths；

(2) in k shortest path first, according to the weighted value of each Section of road net model, path recommendation is carried out, this In weight refer to two aspects: road network road distance and the average running time of road network road, i.e.

${\mathrm{DisCost}}_{{\mathrm{atr}}_i}={\mathrm{\&Sigma;}}_{\sec \&Element;{\mathrm{atr}}_i}{w}_{\sec }\left|\right|\sec \left|\right|,$

${\mathrm{TimeCost}}_{{\mathrm{atr}}_i}={\mathrm{\&Sigma;}}_{\sec \&Element;{\mathrm{atr}}_i}\frac{\left|\right|\sec \left|\right|}{v_{\sec }},$

Wherein, | | sec | | represents at atr_iIn the length (1≤i≤k) of Section, weight coefficient w_secIt it is basis The category of roads of Section marks [http://workshop.pgrouting.org/chapters/ Advanced.html#weighted-costs], specific weight value is as shown in table 1, Section average speed coefficient v_secIt it is root The road network average speed report issued according to Public Security Department calculates.

The different weight coefficient corresponding to category of roads of table 1

Category of roads	Weight coefficient
		motorway,motorway_junction,motorway_link	0.3
trunk,trunk_link	0.4
		primary,primary_link	0.6
secondary,tertiary	0.8
		cicleway,living_street,path	1.5
pedestrian,steps,footway	2.0
		other	1

(3) basisWithTo atr_iIt is ranked up, k bar atr before selecting_iForm final pushing away Recommend path candidates set Rec.Wherein the basic step of sort algorithm is: by atr_iPress respectivelyWith Ascending sort, its ranking results is expressed as List_disAnd List_time, the most random from List_disOr List_timeMiddle selection Sort the atr of forward position_iIt is inserted in recommendation paths candidate collection Rec, if atr_iIn Rec, then repeat above-mentioned Random selection process, reaches k until element number in Rec and then stops sequence, finally give the Rec containing k element.

Step 3: the final benchmark selected in Rec is recommended according to recommendation paths candidate collection Rec and taxi driver Path atr_bench, in conjunction with the real-time GPS data collection of current taxi, respectively from current taxi travel direction and operating range Two aspects carry out abnormal track to the taxi of online operation and detect in real time, and the result detected incorporates the most anti-of passenger Feedback so that the testing result of abnormal track is more accurate, meanwhile real-time testing result is reported Public Security Department.

From the point of view of current taxi travel direction, after given starting point O and terminal D, it is the most square that taxi travels To being determined, and when given recommendation paths candidate collection Rec and the driver final benchmark selected in Rec is recommended Path atr_benchTime, the direction that taxi travels the most more easily determines within limits.Determine according to below equation and work as Whether front taxi travel direction is abnormal:

Wherein, temperature coefficient 0≤ξ≤1,Generated by atr all of in recommendation paths candidate collection Rec The set of Section, and | atr_bench| andRepresent atr respectively_benchWithThe number of middle Section, | Sim_bench| and | Sim_cdd| represent Section and atr at current taxi place respectively_benchThe number that middle Section is similar, And withThe number that middle Section is similar.In that patent, the similarity between the Section of section uses cosine similar Degree (CosineSimilarity) is measured, and if specifying sec_iAnd sec_jSimilarity cos (sec_i, sec_j) ＞ 0 (its Middle i and j represents different section Section marks respectively), then sec_iAnd sec_jSimilar, wherein

$\cos ({\sec }_i,{\sec }_j)=\frac{\stackrel{\&RightArrow;}{\mathrm{AB}}\&CenterDot;\stackrel{\&RightArrow;}{\mathrm{CD}}}{\left|\stackrel{\&RightArrow;}{\mathrm{AB}}\right|\left|\stackrel{\&RightArrow;}{\mathrm{CD}}\right|}$

Here, A, B represent sec_iEnd points, some C, D represent sec_jEnd points, and vectorRepresent sec_iGo out Hire a car travel direction, vectorRepresent sec_jTaxi travel direction,Represent the inner product of vectors of the two vector,WithRepresentation vector respectivelyWithMould.

If score_dir＞ λ_dir, then the travel direction of current taxi just can be judged as abnormal, wherein λ_dir For direction threshold value.

From the point of view of current taxi operating range, according to fixed starting point O and terminal D and current taxi institute At position p_current, distance d that taxi has been passed by_pass、p_currentFrom remaining beeline d of terminal D_minRmAnd recommend The average distance d of recommendation paths in path candidates set Rec_avgThese three distance variable can substitute into below equation and judge to work as Front taxi operating range is the most abnormal:

${\mathrm{score}}_{\mathrm{dis}}=\left\{\begin{array}{cc}\mathrm{\&eta;}-1,& \mathrm{\&eta;}>1\\ 0,& \mathrm{\&eta;}\&le;1\end{array}\right.$

Wherein, $\mathrm{\&eta;}=\frac{d_{\mathrm{pass}}+d_{\min \mathrm{Rm}}}{d_{\mathrm{avg}}},$ And $d_{\mathrm{avg}}=\frac{{\mathrm{\&Sigma;}}_{{\mathrm{atr}}_i\&Element;\mathrm{Rec}}{\mathrm{\&Sigma;}}_{\sec \&Element;{\mathrm{atr}}_i}\left|\right|\sec \left|\right|}{k},$ | | sec | | represents this Section Operating range.

If score_dis＞ λ_dis, then the operating range of current taxi just can be judged as abnormal, wherein λ_dis For distance threshold.

Owing to current taxi track is continually changing, so the history mobile status of this taxi track also can Affect the travel direction of current taxi and the judgement of operating range.Therefore, the condition grading Φ strengthened by introducing is to working as Front taxi state carries out judging more comprehensively and accurately.From the point of view of this taxi whole carrying route, pushing away over time Move, history scote_dirAnd score_disTo current score_dirAnd score_disThere is the least impact, therefore can pass through Weight coefficient τ balances the historic state impact on current state.

Assume to concentrate the time interval Δ t of adjacent GPS point to be all equal for taxi real-time GPS data, and Initial time t₀'sWhereinIt is equal toOrThen, moment t₁Condition grading ForT the most at any time_kCondition grading be,

And above-mentioned formula can expand to

So, above-mentioned computing formula can obtain the Φ of random time to be reduced to following form,

So for the score under t_{Dir, t}And score_{Dis, t}May be replaced withWithCarry out state Judge.

As shown in Figure 2, if the travel direction of current taxi or operating range are judged as abnormal, then this goes out Current state of hiring a car then becomes suspicious state, and now system the passenger on taxi can send prompting, and request confirms Whether current driving path is off path, if passenger confirms it is abnormal driving path, then this taxi traveling record can be upper The Traffic Administration Bureau that reports for work carries out close supervision；The abnormal driving path if passenger really admits a fault, then according to current taxi traveling-position p_currentAnd terminal D, repetition step 2,3, new recommendation paths Candidate Set can generate, and current taxi driving path Can persistently detect, the D until taxi is reached home.

Step 4: for the result of online taxi exception track detection, according to data analysis algorithm, testing result is entered Row is analyzed, and extracts specific taxi and travels rule, and to feed back to Department of Communications, taxi company etc. relevant for analysis result the most at last Department.

For the testing result analyzed in step 4, this invention has a using value of following several form:

(1) if finding in testing result, As time goes on and gradually the number of times of certain taxi driver's exception track subtracts Few, then Public Security Department can be inferred that this driver is a cab driving new hand, and may determine that driving of this driver Sail technical ability slowly improving over time.The most in this case, Public Security Department can pass through mail or telephone counseling Mode the guidance of some driving experience is provided pointedly, can allow this driver driving efficiency place of growing up faster.

(2) if finding in analysis result, certain taxi driver quantity of abnormal track within certain period gets more and more, and hands over Logical management board can thus infer that this taxi driver has the tendency being more biased towards obtaining dirty money in taxi swindle.This Time, Public Security Department needs to notify it or alert, and informs that this driver observes basic taxi trade and drives mark Accurate.

(3) by the examination of the most abnormal track to taxi company taxi driver, can be to every taxi The management state of car company taxi driver makes reasonably deduction such that it is able to preferably carry out for Mei Jia taxi company Rationally and effectively mark, can effectively facilitate and improve the quantity of operation of taxi company with this, promote taxi trade Benign development.

Embodiment

The present embodiment employs the data set simulation reality that all taxis in A city produced in one month Test.

Obtain the open street map data increased income on the internet first against A city, concrete network address be (http: // Www.openstreetmap.org/), and for the Node of these city map data, Way, Relation, Tag take out with Section is the road network model of ultimate unit, and erects visual use in Linux server by these data The road net model interface that family is friendly, specifically used to technology be leaflet front-end operations technology, open figure layer API (WebGISOpenlayersJavascript API)。

Table 2OD region is to essential information

	OD1	OD2	OD3	OD4	OD5
						Tracking quantity	1505	845	728	1007	895

Then, by simulating the whole process of some the true carryings of taxi, obtain these taxis in a timing The most abnormal interior track detection result.By extracting in the A city data set of month in 5 pairs of bigger OD regions of flow Taxi track data, draw the path proposed algorithm in this invention in the case of k is relatively large (test follow-up simulation time Use 5≤k≤7) there is higher path coverage, and the path meeting objective reality recommends to require and count in real time The time restriction calculated.The number of passes data chosen are as shown in table 2, the path coverage of k shortest path first (KSP) such as table 3 Shown in.

Table 3k shortest path first path coverage

	OD1	OD2	OD3	OD4	OD5
						K=3	0.4422	0.3554	0.4321	0.3429	0.3921
K=5	0.6731	0.6517	0.5883	0.6206	0.5891
						K=7	0.8231	0.7917	0.7383	0.7706	0.7152

For example, taxi Taxi_aReceive passenger at an O, and passenger specifies place of arrival D, then believe according to OD Breath, utilizes k shortest path first (KSP) to provide respectively in terms of road network road distance and the average running time of road network road two Suitably k bar recommendation paths, forms recommendation paths candidate collection Rec, i.e. Rec={atr₁..., atr_k}.Driver selects in Rec Certain paths as reference path atr_bench, then Taxi_aAccording to atr_benchThe circuit specified is driven, and is driving During, the score value Φ of current driving condition can be calculated_dirAnd Φ_disIf, Φ_dirOr Φ_disExceed the direction threshold set Value λ_dirOr distance threshold λ_dis, then can judge according to the judgement flow process shown in Fig. 2, if the traveling of current taxi Direction or operating range are judged as abnormal, then Taxi_aCurrent state then becomes suspicious state, and now system Passenger on taxi can send prompting, request confirms whether current driving path is that (passenger can pass through car to off path Carry panel computer real time inspection origin, terminal place and current location and the track run over), if passenger confirms it is different Often driving path, then this taxi traveling record can be reported to Traffic Administration Bureau to carry out close supervision；The exception if passenger really admits a fault Driving path, then according to current taxi traveling-position and terminal D, repetition step 2,3, new recommendation paths Candidate Set can be given birth to Become, and current taxi driving path can persistently detect, the D until taxi is reached home.Finally terminate at passenger getting off car After route, during this time the real-time testing result of route can bring the unitary analysis appraisal report of taxi driver into, and with What more efficient abnormal trajectory analysis was reported offers Public Security Department or taxi company, for follow-up taxi in form Drivers ' behavior assessment lays the foundation.

The invention provides a kind of taxi exception track real-time detection method based on online gps data, implement The method of this technical scheme and approach are a lot, and the above is only the preferred embodiment of the present invention, it is noted that for this skill For the those of ordinary skill in art field, under the premise without departing from the principles of the invention, it is also possible to make some improvements and modifications, These improvements and modifications also should be regarded as protection scope of the present invention.Each ingredient the clearest and the most definite in the present embodiment all can use existing Technology is realized.

Claims (8)

1. a taxi exception track real-time detection method based on online gps data, it is characterised in that: include following step Rapid:

Step 1, carries out road network modeling based on the open street map data increased income to target cities；

Step 2, the journey's end that the journey start point received according to the onboard sensor of each taxi and passenger confirm, in road Apply k shortest path first to obtain taxi on the road net model of road and travel recommendation paths Candidate Set Rec；

Step 3, the recommendation paths therefrom selected according to recommendation paths Candidate Set Rec and driver, in conjunction with the reality of current taxi Time gps data collection, and incorporate passenger's feedback to current taxi driving trace, the taxi of online operation carried out abnormal rail Mark detects in real time, and wherein testing result includes abnormal taxi license board information, taxi driver's essential information, occurs extremely in real time Place longitude and latitude, there is abnormal time, the longitude and latitude in passenger loading place, taxi transaction odd numbers information, and should Testing result is uploaded and be recorded in the data base that Public Security Department specifies in real time；

Step 4, analyzes the result of online taxi exception track detection, extracts urban taxi and travels rule, including not Same city, different periods, the taxi exception record statistics of different company.

A kind of taxi exception track real-time detection method based on online gps data the most according to claim 1, it is special Levy and be, target cities is carried out road network modeling by step 1 and includes: open street map data institute according to target cities The Data Structures provided obtains the elementary path road net model of target cities, and the elementary cell road of every road is saved Segment represents, section Section is that road joint Segment is the most abstract, and Section is the most orderly by one group Segment generates, and the end points of Segment is the intersection between road.

A kind of taxi exception track real-time detection method based on online gps data the most according to claim 2, it is special Levying and be, in step 2, the process obtaining recommendation paths Candidate Set Rec includes:

(2-1) by it has been determined that taxi place O and passenger loading after the journey's end D that determines and current taxi Car travel direction, filters out road junction Node according to nearest Euclidean distance_OAnd Node_D, use Node_OAnd Node_DTable respectively Hire a car starting point road junction and journey's end road junction are shown；

(2-2) obtaining the weighted value of each Section of road net model in employing k shortest path first, weighted value includes road network road Road distanceAnd the average running time of road network roadCalculate by equation below respectively:

${\mathrm{DisCost}}_{{\mathrm{atr}}_i}={\mathrm{\&Sigma;}}_{\sec \&Element;{\mathrm{atr}}_i}{w}_{\sec }\left|\right|\sec \left|\right|---\left(1\right)$

${\mathrm{TimeCost}}_{{\mathrm{atr}}_i}={\mathrm{\&Sigma;}}_{\sec \&Element;{\mathrm{atr}}_i}\frac{\left|\right|\sec \left|\right|}{v_{\sec }}---\left(2\right)$

Wherein, Section, the ‖ sec ‖ of sec represents that GPS records taxi tracing point p institute projection mapping represents this Section Operating range, atr represents the Section sequence corresponding to taxi tracing point that GPS records, and the form of expression is as follows:

Atr={sec₁→…→sec_j→…→sec_m} (3)

The sequence consensus of the taxi tracing point p that in atr, the order of Section and GPS record, atr_iRepresent the taxi of GPS record I bar Section sequence corresponding to wheel paths point, it is total that 1≤i≤k, k represent sequence corresponding to all tracing points of hiring a car, w_secRepresent weight coefficient, indicate according to the category of roads of Section, v_secRepresent Section average speed coefficient, according to friendship The road network average speed report that logical management board issues obtains；

(2-3) according to taxi starting point Node_OWith journey's end Node_DAnd road network road distanceWith road network road The average running time in roadTo atr_iIt is ranked up, k bar atr before selecting_iForm final recommendation paths Candidate Set Close Rec, i.e. Rec={atr₁,…,atr_k}。

A kind of taxi exception track real-time detection method based on online gps data the most according to claim 3, it is special Levying and be, in step 3, the detection in real time that the online taxi runed carries out abnormal track includes judging that current taxi travels Whether direction is abnormal, uses equation below to calculate travel direction exceptional value score_dir:

Wherein, temperature coefficient 0≤ξ≤1, atr_benchRepresent the benchmark atr that driver selects in Rec,It is by recommendation paths The set of the Section that all of atr is generated in candidate collection Rec, and | atr_bench| andRepresent respectively atr_benchWithThe number of middle Section, | Sim_bench| and | Sim_cdd| represent the Section at current taxi place respectively With atr_benchThe number that middle Section is similar, Yi JiyuThe number that middle Section is similar, λ_dirFor direction threshold value, when score_dir>λ_dir, it is determined that the travel direction of current taxi is abnormal.

A kind of taxi exception track real-time detection method based on online gps data the most according to claim 4, it is special Levying and be, in step 3, the detection in real time that the online taxi runed carries out abnormal track includes judging that current taxi travels Whether distance is abnormal, uses equation below to calculate operating range exceptional value score_dis:

${\mathrm{score}}_{dis}=\left\{\begin{array}{cc}\mathrm{\&eta;}-1,& \mathrm{\&eta;}>1\\ 0,& \mathrm{\&eta;}\&le;1\end{array}\right.---\left(5\right)$

Wherein,Andd_passRepresenting hires a car travelled away from From, d_minRmRepresent position of hiring a car from the remaining beeline of terminal, d_avgRepresent in recommendation paths candidate collection Rec and push away Recommend the average distance in path, λ_disFor distance threshold, work as score_dis>λ_dis, it is determined that the operating range of current taxi is abnormal.

A kind of taxi exception track real-time detection method based on online gps data the most according to claim 5, it is special Levying and be, step 3 includes that current taxi driving trace is sentenced by condition grading Φ the most extremely that strengthened by introducing Disconnected, employing equation below:

Wherein,Represent at moment t_kUnder enhanced situation scoring,WithIt is illustrated respectively in moment t_kUnder Travel direction exceptional value score_dirWith operating range exceptional value score_dis, it is convenient for subsequent calculations,Represent at moment t_k UnderOrAssume taxi real-time GPS data is concentrated the time interval of adjacent GPS point Δ t is equal, initial time t₀Enhanced situation scoringτ represents weight coefficient, 0 < τ < 1, so for t Time the score that inscribes_dirAnd score_disFormula (6) can be passed through and be converted into the travel direction condition grading Φ of enhancing_dirWith The driving distance condition grading Φ strengthened_disCarry out condition adjudgement.

A kind of taxi exception track real-time detection method based on online gps data the most according to claim 6, it is special Levy and be, in step 3, if the travel direction of current taxi or operating range are judged as exception, it is determined that this taxi is worked as Front state is suspicious state, and the passenger on taxi sends prompting, and request confirms whether current driving path is abnormal road Footpath, if passenger confirms it is abnormal driving path, then this taxi traveling record can be reported to Public Security Department to be monitored；If Passenger really admits a fault abnormal driving path, then according to current taxi traveling-position and journey's end, repeat step 2 and step 3, until taxi arrives journey's end.

A kind of taxi exception track real-time detection method based on online gps data the most according to claim 7, it is special Levying and be, the zonal taxi that step 4 drawn is abnormal to travel record and uploads to Public Security Department and taxi company Record in private database, and the result analyzing online taxi exception track detection with this, extract urban taxi row Sail rule, including different cities, different periods, the taxi exception record statistics of different company, and with this basis to city not Monthly, season and year scoring is carried out with the taxi driver in region.