CN104134349B - A kind of public transport road conditions disposal system based on traffic multisource data fusion and method - Google Patents

Abstract

The invention discloses a kind of public transport road conditions disposal system based on traffic multisource data fusion and method, comprise pretreatment module, source data processing module, data source evaluation module, data fusion module, pretreatment module is for the treatment of traffic multi-source data; Source data processing module is used for processing non-public transport data source road conditions calculation result data; Data source evaluation module comprises an appraisal framework for assessment of data source quality, can obtain the confidence level of multi-source data to public transport road conditions result of calculation by this appraisal framework process; Data fusion module, based on data source assessment result, is weighted multi-source data, finally obtains required public transport road condition data; Display module superposes processing the public transport road conditions figure obtained with map file, and shows full transit network traffic information.The present invention, by comprehensive multiple traffic data sources road conditions result of calculation, learns from other's strong points to offset one's weaknesses, obtains the fusion results information describing public transport road conditions, for bus service provides Data support.

Description

A kind of public transport road conditions disposal system based on traffic multisource data fusion and method

Technical field

The present invention relates to the public transport road conditions treatment technology of intelligent transportation system, particularly a kind of public transport road conditions disposal system based on traffic multisource data fusion and method.

Background technology

Data fusion technique refers to and utilizes computing machine to the some observation information obtained chronologically, in addition automatic analysis, comprehensive under certain criterion, the information processing technology of carrying out to complete required decision-making and evaluation tasks.

Because the composition in domestic and international traffic data source composition that is different, different pieces of information source is not identical yet, research point in fusion and technology also have significant difference, all there is following limitation in the data fusion method at present for field of traffic: (1) mostly is the fusion of two type data: in current main flow fusion method, mostly only relate to two type data sources, as coil/taxi, bus/taxi, and model does not possess good extensibility for newtype data source.(2) mostly be single-stage to merge: main according to varigrained data, merge in different levels, as loop data and floating car data structurally differ greatly, can only decision level fusion be carried out; And bus differs less with taxi data, it is generally the fusion carrying out data level.If multiple different grain size data occur simultaneously, then can only merge at higher level.(3) use a certain categorical data as general data: because main Demand-Oriented is general trip road conditions more, therefore in these technical methods general using taxi data as general data source, and other types data only carry out certain adjustment as auxiliary data to general data, such road conditions result of calculation information cannot meet special public transport demand.(4) need sample training and train cost high: mainly for artificial neural network algorithm, as one of now widely used algorithm, its model training process is higher to sample size requirements, there is very large problem in training effectiveness, and it is slightly poor for the processing power of unstable data, under worst case, training process needs repeated several times, even can cause neural network failure when data source increases.

Summary of the invention

The technology of the present invention is dealt with problems: overcome the deficiencies in the prior art, provides a kind of public transport road conditions disposal system based on traffic multisource data fusion and method, for promoting the accuracy rate of the calculating of public transport road conditions and result.

The technology of the present invention solution: based on the public transport road conditions disposal system of traffic multisource data fusion, comprising: pretreatment module, source data processing module, data source evaluation module, data fusion module;

Data preprocessing module: for the treatment of traffic multi-source data, so that obtain can for the data available set of subsequent module; By steps such as abnormal data cleaning, geographic coordinate conversion, sequences, obtain the traffic information that multi-data source characterizes, comprise public transport track data and hire out track data, video road section data on flows, deliver to source data processing module;

Source data processing module: the multi-source traffic data that data preprocessing module exports is of a great variety, with the angle of data source itself, traffic is described separately, source data processing module mainly processes these data, enables them to correctly reflect public transport traffic information.By steps such as curves, module exports the multi-source traffic data in order to describe public transport road conditions based on different pieces of information source, the traffic information that each data source is characterized and the traffic information of public transport data characterization are more pressed close to, thus obtain preliminary public transport traffic information, deliver to data source evaluation module;

Data source evaluation module: module is by setting up an appraisal framework for assessment of multi-data source quality, the comprehensive assessment quantized is carried out to each data source quality that the multi-source data of source data processing module output is concentrated, thus obtain the weight of each data source in data fusion process, and these weight informations are delivered to data fusion module as the basic basis merging flow process as exporting;

Data fusion module: based on the multi-source traffic data that module obtains by source data processing module, be aided with each data source weight that data source evaluation module obtains, by steps such as weight matrix division, best source calculating, singular data cleaning, data fusion, fusion calculation is carried out to multi-source data, finally obtains the public transport traffic information that can be used for showing;

Display module: public transport traffic information data fusion module process obtained is presented on map as display data, for user or research and analyse.

Described pretreatment module, for the treatment of traffic multi-source data, to obtain the traffic information that multi-data source characterizes, comprises public transport data, rental data, video data, is implemented as follows:

(1) pre-service of public transport data.Input data are public transport gps data.First, by rejecting abnormal data, geographic coordinate conversion, public transport up-downgoing correction, external sort, public transport correction gps data is obtained.Then, mated by road chain, enter the station the step such as differentiation, speed calculating, obtains public transport track data.

(2) pre-service of rental data.Input data are for hiring out gps data.First, by rejecting abnormal data, geographic coordinate conversion, external sort, obtain hiring out and revise gps data.Then, by steps such as road chain coupling, speed calculating, obtain hiring out track data.

(3) pre-service of video data.Input data are road section capture video fragment.Utilize Video processing software by vehicle identification, effectively counting etc., analyze public transportation lane data, obtain road section data on flows, and obtain vehicle travel speed data according to flow/velocity formula.

Described is for each data source by the segmentation of road chain, according to the road chain described by every bar data, is divided into several data sets, guarantees that the data of each data centralization describe same road chain road conditions feature.

Described source data processing module is implemented as:

(1) read public transport track data, determine whether to enter the station, stop, departures state.If so, then current data is rejected; If not, then retain current data under the chain of Mei Tiao road, obtain road chain identification data.Until digital independent is complete.

(2) read the road chain identification data that the first step obtains, judge whether public transport travels on public transportation lane.If so, skip data fusion module, calculate public transport road conditions by public transport data and historical data; If not, multi-source data is processed respectively, as public transport track data, hire out track data, road section data on flows etc., carry out a curve, by adjustment configuration parameter, introduce the method such as historical data, point road conditions process, calculate road travel temporal information, deliver to data source evaluation module.

Described data source evaluation module is implemented as:

(1) multi-source data characterizing public transport road conditions is input as, i.e. public transport road data hourage, taxi road travel time data, video section data on flows etc.For each data source, under corresponding evaluation index, generate evaluate parameter vector.Evaluate parameter is made up of three aspects:

● vehicle operating feature: in order to the inherent feature in data of description source, assessment result constitutive characteristic Description Matrix.Comprise confidence metric, undulatory property tolerance, temporal correlation tolerance, spatial coherence tolerance.

● historical law and event: by comparison real-time road condition information and historical law data, whether dynamic evaluation data source meets general history run rule, and binding events data judge whether that event occurs, and assessment result forms degree of variation metric matrix.Comprise degree of variation tolerance, event flag calculating.

● the mutual degree of support between data source: in order to measure the mutual degree of support between each data source, assessment result consistence of composition metric matrix.Comprise consistency metric.

(2) according to the evaluate parameter vector in step (1), each data source is assessed, obtain the confidence metric result of each data source, fluctuating range measurement results, temporal correlation measurement results, spatial coherence measurement results, degree of variation measurement results, event flag measurement results, consistency metric result, be expressed as ent, then assessment result vector $d_s=({\mathrm{\ω}}_s^{\mathrm{Conf}},{\mathrm{\ω}}_s^{\mathrm{Flu}},{\mathrm{\ω}}_s^{\mathrm{Tem}},{\mathrm{\ω}}_s^{\mathrm{Spa}},{\mathrm{\ω}}_s^{\mathrm{Var}},\mathrm{Ent},{\mathrm{\ω}}_s^{\mathrm{Cons}}),$ Deliver to data fusion module; (3) according to the result of step (2), formula P=(D is utilized ^ti ^1-Enth) ^tw obtains final data source weight vectors P, the confidence level size of each correspondence data source namely in this vector, wherein, D is feature interpretation matrix, and I is degree of variation metric matrix, H is consistency metric matrix, Ent is event flag, W is evaluate parameter weight vectors.

Data source weight vectors P, using the input as data fusion module, participates in multisource data fusion process.

Described data fusion module is implemented as:

(1) multi-source data characterizing public transport road conditions and data source assessment result vector d is input as _s, wherein use sample degree of confidence, sample fluctuation amplitude, temporal correlation constitutive characteristic Description Matrix D:

$D=\left(\begin{array}{cccc}{\mathrm{\ω}}_1^{\mathrm{Conf}}& {\mathrm{\ω}}_1^{\mathrm{Flu}}& {\mathrm{\ω}}_1^{\mathrm{Tem}}& {\mathrm{\ω}}_1^{\mathrm{Spa}}\\ {\mathrm{\ω}}_2^{\mathrm{Conf}}& {\mathrm{\ω}}_2^{\mathrm{Flu}}& {\mathrm{\ω}}_2^{\mathrm{Tem}}& {\mathrm{\ω}}_2^{\mathrm{Spa}}\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ {\mathrm{\ω}}_N^{\mathrm{Conf}}& {\mathrm{\ω}}_N^{\mathrm{Flu}}& {\mathrm{\ω}}_N^{\mathrm{Tem}}& {\mathrm{\ω}}_N^{\mathrm{Spa}}\end{array}\right)$

Use degree of variation metric matrix

$I(k,l)=\mathrm{diag}({\mathrm{\ω}}_1^{\mathrm{Var}}(k,l),{\mathrm{\ω}}_2^{\mathrm{Var}}(k,l),...,{\mathrm{\ω}}_N^{\mathrm{Var}}(k,l))$

Reflection event is to the influence degree of data;

Use case mark

Ent(k,t)＝τ

Whether reflection event occurs;

Use consistency metric matrix

$H(k,l)=\mathrm{diag}({\mathrm{\ω}}_1^{\mathrm{Cons}}(k,l),{\mathrm{\ω}}_2^{\mathrm{Cons}}(k,l),...,{\mathrm{\ω}}_N^{\mathrm{Cons}}(k,l))$

By mutual checking, obtain the order of accuarcy in Various types of data source.Wherein k represents time period sequence number, and usual 5 minutes is a time period, has 144 time periods every day; L represents road chain number.

(2) data source that the matrix acquisition confidence level after dividing according to data source assessment result vector is the highest, and be labeled as best source;

(3) fiducial interval under best source α degree of confidence is utilized to screen multi-source data result: described screening multi-source data result refers to for the data of each in multi-source data, if this data value drops on outside best source fiducial interval, then think that it is an exceptional value, should give rejecting;

(4) to the data value stayed after screening, formula is utilized calculate its weighted value, wherein for optimal data source speed average, S _sfor variance, for data cell confidence distance, p _sfor the weight of data source s, Duw _sfor data cell weighted value;

(5) average acquisition public transport road conditions result of calculation is weighted to multi-source data;

Weighted average calculation formula is wherein ω _ibe i-th data weighting, v _ifor velocity amplitude.I represents to be i-th data when pre-treatment, and n represents data total number.The public transport road conditions based on multi-source data analysis can be obtained thus, i.e. public transport road travel average velocity.

Based on the public transport road conditions disposal route of traffic multisource data fusion, performing step is as follows:

(1) data prediction step: for the treatment of traffic multi-source data, so that obtain can for follow-up data available set; By abnormal data cleaning, geographic coordinate conversion, ordered steps, obtain the traffic information that multi-data source characterizes, comprise public transport track data and hire out track data, video road section data on flows;

(2) source data treatment step: multi-source traffic data is of a great variety, with the angle of data source itself, traffic is described separately, source data treatment step processes these data, enable them to correctly reflect public transport traffic information, by a curve fitting step, export the multi-source traffic data in order to describe public transport road conditions based on different pieces of information source, the traffic information that each data source is characterized and the traffic information of public transport data characterization are more pressed close to, thus obtain preliminary public transport traffic information;

(3) data source appraisal procedure: by setting up an appraisal framework for assessment of multi-data source quality, the comprehensive assessment quantized is carried out to each data source quality that the multi-source data of source data treatment step output is concentrated, thus obtain the weight of each data source in data fusion process, and these weight informations are delivered to data fusion step as the basic basis merging flow process as exporting;

(4) data fusion step: based on the multi-source traffic data that source data processing module obtains, be aided with each data source weight that data source evaluation module obtains, calculated by weight matrix division, best source, singular data cleans, data fusion step, fusion calculation is carried out to multi-source data, finally obtains the public transport traffic information that can be used for showing;

(5) step display: public transport traffic information data fusion step process obtained is presented on map as display data, for user or research and analyse.

The present invention's advantage is compared with prior art: the present invention, by judging public transport operation state, switches, to obtain public transport traffic information more accurately between public transport list source data and traffic multi-source data.For the data fusion method that the present invention proposes, the fusion of multi-source data is mainly carried out at data level, guarantees that final fusion results granularity is enough to the concrete road chain road condition be accurate in road network.The traffic events such as traffic hazard, traffic control, occupation of land construction are considered on the impact of result in fusion process, through fusion process, realize the effect such as quick, accurate, reliable merged, data screening and average weighted computing method also ensure that final public transport road conditions result of calculation can not be subject to the impact of exceptional value in single source data.Based on the public transport road conditions that the method obtains, Data support can be provided to needing the bus service of related data, to obtain better service quality.

Accompanying drawing explanation

Fig. 1 is disposal route realization flow figure of the present invention;

Fig. 2 is public transport travelling state demarcation figure.

Embodiment

For making the object, technical solutions and advantages of the present invention clearly, below in conjunction with accompanying drawing, the present invention is described in further detail.

The incomplete traffic data that the present invention is mainly used in solving bus service field merges problem, and constrained input is wherein as follows:

● input: N, S, m, k, wherein, N is the traffic road needing to carry out data fusion; S={s _i, i=1,2 ..., N} is fused data source; M is the time period, and namely each fusion treatment uses the data of m duration; K is treatment cycle, namely carries out Single cell fusion process every k duration;

● export: R, R are the traffic informations for describing needed for bus service.

The fusion method that the present invention proposes carries out road chain division to system-wide net N, and each time period road conditions calculate in units of the chain of road, and Zai Jiangge road chain final road conditions result merges into full road network information.The flow chart of data processing of this model as shown in Figure 1.

As shown in Figure 1, the inventive method specific implementation step is as follows:

Step 100: data prediction, the traffic multi-source data collected, needs through certain series of standards process, just can become the data source that subsequent analysis uses.

The first, public transport, hire out gps data, whether, for gathering and each data stored, whether differentiation is obvious abnormal data, mainly comprises and whether surmounts zone of reasonableness, record complete etc.For rejecting abnormal data.

Second, due to the object of national policy and data security, in the public transport collected, taxi gps data, there is skew in the longitude and latitude of coordinate points, the latitude and longitude coordinates of coordinate conversion program to data therefore adopting State Bureau of Surveying and Mapping to authorize is revised on the basis of real road.

3rd, certain standard process is carried out to public transport, taxi gps data, so that subsequent analysis uses, mainly comprises temporally sheet merger, by car number sequence etc.Meanwhile, because when bus first and last station passes in and out, up-downgoing identification exists error, public transport gps data is revised up-downgoing zone bit by some data continuously.

4th, adopt dedicated video process software to carry out vehicle identification, effectively counting etc. to video data, thus obtain public transport traffic information in public transportation lane, supplementing as multisource data fusion.

Step 101: traffic source data process, after receiving the multi-source data in the process period, first will guarantee that first the traffic information calculated through all types of sensing data in order to describe public transport road conditions, therefore can carry out source data process.

definition: public transport travelling state

The regular state had when bus runs on given line, namely enter the station, stop, arrive at a station, stand between etc., see Fig. 2.Wherein, take website as Mean radius R ₁scope definition be the scope of stopping, radius R ₂with R ₁between scope be the/departures scope that enters the station, all the other sections for station between scope.

Bus is under the state of entering the station, initiatively will reduce travel speed, cause the situation occurring deviation with the road conditions of present road, in like manner, in departures situation, bus speed reduces, in the situation that stops, bus speed is zero, therefore the Bus information produced under these three states can not in order to describe public transport road condition, and the public transport road conditions under these three states should use other do not stop public bus network bus data, other data source data or historical datas to fill up and obtain.

Between station under state, by the mode of a curve, can be obtained this in order to describe the transfer equation of this linear relationship.Setting parameter with parameter by public transport data source s respectively _buswith another data source s _srcin the Speed attribute parameter that the k time period obtains, both might as well be established linear relationship can be expressed as wherein be estimate by bus rider speed value another data source velocity amplitude corresponding of obtaining, ε is unbiased esti-mator error.Consider institute generate parameter robustness and real-time, the history speed data same period and upper period velocity data be used to estimation this method linear smoothing method once returned obtains conversion parameter a _k, have wherein β is adjustment parameter, and in order to the weight relationship in equilibrium between a period velocity ratio and the history speed proportional same period, the value of β is larger, then go up a time period speed proportional at conversion parameter a _kaffect larger in computation process; Otherwise it is then less.After each cycle result has calculated, according to new result, historical law is upgraded.

Urban public transport has or partial period has special lane, public transportation lane is a kind of for ensureing the special road that bus transport passenger's efficiency designs, in specified time period, special lane travels for public transit vehicle only, ensure that and can to run along given line fast already at the period bus that blocks up.Under having public transportation lane situation to affect, public transport road conditions result of calculation can not depend on the multi-source datas such as taxi, and the full sampled data such as video data also can only use the testing result wherein belonging to public transportation lane.

On this basis, obtain the conventional method that bus rider speed value and taxi velocity amplitude transform mutually, other types data can be converted into public transport road conditions descriptor by similar source data disposal route.

As shown in Figure 1, disposal route of the present invention is specifically implemented as follows;

Step 102: data source is assessed, and usage data source of the present invention appraisal framework is assessed data source quality.

definition: data source appraisal framework

Data source appraisal framework is a parameter tuple in order to assessment data source quality and a set of appraisal procedure based on these parameter tuples.

Definition data source appraisal framework is Q=(S, A, F, Mat), wherein, and S={s ₁, s ₂..., s _nit is data source set to be assessed; A={a ₁, a ₂..., a _mit is evaluate parameter tuple; Mat is assessment result.

According to appraisal procedure $F=\{f_{a_1},f_{a_2},...,f_{a_M}\},$ Have:

Evaluate parameter tuple A in appraisal framework is formed primarily of three aspects:

● vehicle operating feature: by this component assesses result constitutive characteristic Description Matrix, in order to the inherent feature in data of description source.Comprise confidence metric, undulatory property tolerance, temporal correlation tolerance, spatial coherence tolerance;

● historical law and event: form degree of variation metric matrix and event flag by this component assesses result, by comparison real-time road data and historical law data, whether dynamic evaluation data source meets general history run rule, and binding events data judge whether that event occurs.Comprise degree of variation tolerance, event flag calculating;

● the mutual degree of support between data source: by this component assesses result consistence of composition metric matrix, in order to measure the mutual degree of support between each data source.Comprise consistency metric.

Data source assessment of the present invention mainly comprises the steps:

Step 1021: confidence metric.Theoretical according to fiducial interval, the confidence level of fiducial interval length and data source is inversely proportional to, and by the fiducial interval length of each data source under identical confidence level condition in tolerance multi-source data, obtains the degree of reliability of each source data.

Step 1022: fluctuating range is measured.In conjunction with the different qualities between multi-source data, utilize such as least square method, multi-source data undulatory property is assessed, reduce high undulatory property data source weight, effectively improve the accuracy of public transport road conditions result of calculation.

Step 1023: temporal correlation is measured.The such as method of Pearson's coefficient, measures each source data in multi-source data, reduces the more weak data source weight of space-time expending.

Step 1024: degree of variation is measured.Definition degree of variation is the degree that public transport road conditions depart from performance history same period, the data source larger to degree of variation, reduces its credible weight in fusion results.

Step 1026: event flag.Traffic flow under metrology event impact, as traffic hazard, extreme weather etc., in public transport road condition data fusion process, arranges event flag to the section being subject to events affecting.

Step 1027: data consistency is measured.Answer consistent characteristic according to the public transport traffic information that each source data calculates, the consistance between multi-source data is measured, effectively avoids singular data on the impact of final fusion results.

Step 103: the traffic data based on data source assessment merges

Traffic data based on data source assessment merges mainly through carrying out matrix trace inequality to the evaluate parameter in data source appraisal framework, and utilize the method for matrix multiple to obtain identifying the weight vectors of each data source credibility, and obtain best source with this, utilize best source to delimit interval range to clean and weight calculation multi-source data, obtain the final road conditions result data needed finally by average weighted method.

Merge based on the traffic data of data source assessment in the inventive method and mainly comprise the steps:

, for the assessment element under each data source appraisal framework, all there is a corresponding weighted value in step 1031: the matrix trace inequality under data source appraisal framework will be considered as an element, obtained the fusion matrix of a N × M by the M kind element weights information of combination N kind data source, the Mat namely in data source appraisal framework.As N=1, problems is claimed to be that single source traffic data merges problem, when N >=2 for multi-source traffic data merges problem.Be expressed as:

The inventive method is by setting up data source appraisal framework, and in dynamic measurement each time period, the degree of reliability of each data source observation data result determines that each data source participates in credibility when merging.

Wherein, use by sample degree of confidence, sample fluctuation amplitude, temporal correlation constitutive characteristic Description Matrix $D(k,l)=\left(\begin{array}{cccc}{\mathrm{\ω}}_1^{\mathrm{Conf}}(k,l)& {\mathrm{\ω}}_1^{\mathrm{Flu}}(k,l)& {\mathrm{\ω}}_1^{\mathrm{Tem}}(k,l)& {\mathrm{\ω}}_1^{\mathrm{Spa}}(k,l)\\ {\mathrm{\ω}}_2^{\mathrm{Conf}}(k,l)& {\mathrm{\ω}}_2^{\mathrm{Flu}}(k,l)& {\mathrm{\ω}}_2^{\mathrm{Tem}}(k,l)& {\mathrm{\ω}}_2^{\mathrm{Spa}}(k,l)\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ {\mathrm{\ω}}_N^{\mathrm{Conf}}(k,l)& {\mathrm{\ω}}_N^{\mathrm{Flu}}(k,l)& {\mathrm{\ω}}_N^{\mathrm{Tem}}(k,l)& {\mathrm{\ω}}_N^{\mathrm{Spa}}(k,l)\end{array}\right)$ The validity of metric data; Use degree of variation metric matrix $I(k,l)=\mathrm{diag}({\mathrm{\ω}}_1^{\mathrm{Var}}(k,l),{\mathrm{\ω}}_2^{\mathrm{Var}}(k,l),...,{\mathrm{\ω}}_N^{\mathrm{Var}}(k,l))$ Reflection event is to the influence degree of data; Use case mark Ent (k, l)=τ reflects that whether event occurs; Use consistency metric matrix $H(k,l)=\mathrm{diag}({\mathrm{\ω}}_1^{\mathrm{Cons}}(k,l),{\mathrm{\ω}}_2^{\mathrm{Cons}}(k,l),...,{\mathrm{\ω}}_N^{\mathrm{Cons}}(k,l))$ By mutual confirmation, obtain the order of accuarcy of all kinds file from the side, wherein k represents time period sequence number, and usual 5 minutes is a time period, has 144 time periods every day; L represents road chain number.

Step 1032: best source calculates, by each data source at time period k, under the multifactor associations such as the attribute description matrix D of road chain l, degree of variation moment matrix I, event flag Ent and consistency metric matrix H, carry out best source assessment:

P＝(D ^TI ^1-EntH) ^TW

Wherein, result vector P represents the degree of reliability of the data result that each data source is calculated by the data that collect in fusion process, W=(w _conf, w _flu, w _tem, w _spa) ^tfor elemental characteristic weight vectors, element w wherein _jfor data source appraisal framework corresponding element weighted value, that reflects element a _jin the significance level participating in merging.The size of these weighted values reflects the significance level of respective data sources in follow-up fusion process.

If assessment result is P=(p ₁, p ₂..., p _n) ^t, then p is marked _max(k, l)=Max{p ₁, p ₂..., p _ncorresponding data source s _max(k, l) is best source.

Step 1033: data cleansing, if best source is quantitative data, is that fiducial interval a=1-α under calculate to it in degree of confidence by its regularity of distribution, and is expressed as (L _α, R _α); If best source is typed data, we directly use its interval range as (L _α, R _α).Utilize (L _α, R _α) data of this interval range to each data source are cleaned one by one, if certain data drop on outside interval range, determine that it is a singular value, and rejected.

Step 1034: weight calculation, if s _max(k, l), for assessing the time period k obtained in previous step, the best source under the chain l condition of road, chooses the road conditions computation of mean values of this data source after data cleansing as the central value of Confidence distance assessment, suppose have N kind quantitative data to participate in assessment, at time period k, under the chain l condition of road, calculate their mean and variance respectively:

${\stackrel{\‾}{v}}_s(k,l)=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{v}_s^i(k,l)}{n},S_s^2(k,l)=\frac{1}{n-1}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(v_s^i(k,l)-{\stackrel{\‾}{v}}_s(k,l))}^2,$ Wherein n is the data volume of data source s, for certain data wherein, utilize Gauss error function (ERF function), namely and the mean and variance of N number of data source calculates the Confidence distance of wherein each data cell:

${\mathrm{Cd}}_s^i(k,l)=\mathrm{erf}\left(\frac{v_s^i-{\stackrel{\‾}{v}}_{s_{\max }}}{S_{s_{\max }}}\right)$

Wherein, for in data source s, data cell confidence distance.Suppose at time period k, under the chain l condition of road, in data source s, data cell weighted value can be calculated as follows:

${\mathrm{Duw}}_s^i(k,l)=p_s(k,l)+{\mathrm{Cd}}_s^i(k,l)$

Wherein, p _sthe weight that (k, l) is data source s.

Step 1035: Weighted Fusion, final data fusion step uses average weighted method, supposes that the data cell from all data sources is expressed as { v ₁, v ₂..., v _n, their respective weights is expressed as { ω ₁, ω ₂..., ω _n, then final at time period k, the road conditions fusion results under the chain l condition of road is:

$v_{\mathrm{output}}(k,l)=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{\ω}}_i{v}_i}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{\ω}}_i}$

After the final fusion results of output, each data source at time period k, the predicted value under the chain l condition of road, and net result will be to will be written in historical data base, for next treatment cycle fusion process.

Claims (7)

1. based on a public transport road conditions disposal system for traffic multisource data fusion, it is characterized in that comprising: pretreatment module, source data processing module, data source evaluation module, data fusion module;

Data preprocessing module: for the treatment of traffic multi-source data, so that obtain can for the data available set of subsequent module; By abnormal data cleaning, geographic coordinate conversion, ordered steps, obtain the traffic information that multi-data source characterizes, comprise public transport track data and hire out track data, video road section data on flows, deliver to source data processing module;

Source data processing module: multi-source traffic data is of a great variety, with the angle of data source itself, traffic is described separately, source data processing module processes these data, them are enable correctly to reflect public transport traffic information, by a curve fitting step, export the multi-source traffic data in order to describe public transport road conditions based on different pieces of information source, the traffic information that each data source is characterized and the traffic information of public transport data characterization are more pressed close to, thus obtain preliminary public transport traffic information, deliver to data source evaluation module;

Data source evaluation module: by setting up an appraisal framework for assessment of multi-data source quality, the comprehensive assessment quantized is carried out to each data source quality that the multi-source data of source data processing module output is concentrated, thus obtain the weight of each data source in data fusion process, and these weight informations are delivered to data fusion module as the basic basis merging flow process as exporting;

Data fusion module: based on the multi-source traffic data that source data processing module obtains, be aided with each data source weight that data source evaluation module obtains, calculated by weight matrix division, best source, abnormal data cleans, data fusion step, fusion calculation is carried out to multi-source data, finally obtains the public transport traffic information that can be used for showing;

Display module: public transport traffic information data fusion module process obtained is presented on map as display data, for user or research and analyse.

2. the public transport road conditions disposal system based on traffic multisource data fusion according to claim 1, it is characterized in that: described pretreatment module is for the treatment of traffic multi-source data, to obtain the traffic information that multi-data source characterizes, comprise public transport data, rental data, video data, be implemented as follows:

(1) pre-service of public transport data

Input data are public transport gps data

First, by rejecting abnormal data, geographic coordinate conversion, public transport up-downgoing correction, external sort, public transport correction gps data is obtained; Then, mated by road chain, enter the station differentiation, speed calculation step, obtains public transport track data;

(2) pre-service of rental data

Input data, for hiring out gps data, first, by abnormal data cleaning, geographic coordinate conversion, external sort, obtain hiring out and revise gps data; Then, by road chain coupling, speed calculation step, obtain hiring out track data;

(3) pre-service of video data

Input data are road section capture video fragment, utilize Video processing software by vehicle identification, effectively count, and analyze public transportation lane data, obtain road section data on flows, and obtain vehicle travel speed data according to flow/velocity formula.

3. the public transport road conditions disposal system based on traffic multisource data fusion according to claim 1, is characterized in that: described source data processing module is implemented as:

(1) read public transport track data, determine whether to enter the station, stop, departures state, if so, then reject current data; If not, then retain current data under the chain of Mei Tiao road, obtain road chain identification data, until digital independent is complete;

(2) read the road chain identification data that the first step obtains, judge whether public transport travels on public transportation lane; If so, skip data fusion module, calculate public transport road conditions by public transport data and historical data; If not, multi-source data is processed respectively, a curve is carried out to public transport track data, taxi track data, road section data on flows, by adjustment configuration parameter, introduce historical data, analysis road conditions disposal route, calculate road travel temporal information, deliver to data source evaluation module.

4. the public transport road conditions disposal system based on traffic multisource data fusion according to claim 1, is characterized in that: described data source evaluation module is implemented as:

(1) multi-source data characterizing public transport road conditions is input as, i.e. public transport road data hourage, taxi road travel time data, video section data on flows, for each data source, generate evaluate parameter vector under corresponding evaluation index, evaluate parameter is made up of three aspects:

● vehicle operating feature: in order to the inherent feature in data of description source, assessment result constitutive characteristic Description Matrix, comprises confidence metric, undulatory property tolerance, temporal correlation tolerance, spatial coherence tolerance;

● historical law and event: by comparison real-time road condition information and historical law data, whether dynamic evaluation data source meets general history run rule, and binding events data judge whether that event occurs, assessment result forms degree of variation metric matrix, comprises degree of variation tolerance, event flag calculating;

● the mutual degree of support between data source: in order to measure the mutual degree of support between each data source, assessment result consistence of composition metric matrix, comprises consistency metric;

(2) according to the evaluate parameter vector in step (1), each data source is assessed, obtain the confidence metric result of each data source, fluctuating range measurement results, temporal correlation measurement results, spatial coherence measurement results, degree of variation measurement results, event flag measurement results, consistency metric result, namely obtain the assessment result vector d of this data source _s; Meanwhile, by confidence metric result, sample fluctuation amplitude result, temporal correlation measurement results, spatial coherence measurement results constitutive characteristic Description Matrix, D is designated as; Form degree of variation metric matrix by degree of variation measurement results, be designated as I; Weighed by degree of variation and road chain event flag Ent is set; By consistency metric result consistence of composition metric matrix, be designated as H, by d _s, D, H, Ent, I deliver to data fusion module;

(3) according to the result of step (2), data source weight vectors P, using the input as data fusion module, participates in multisource data fusion process.

5. the public transport road conditions disposal system based on traffic multisource data fusion according to claim 1, is characterized in that: described data fusion module is implemented as:

(1) multi-source data characterizing public transport road conditions and data source assessment result vector d is input as _s, wherein use the validity of sample degree of confidence, sample fluctuation amplitude, temporal correlation constitutive characteristic Description Matrix D metric data; Event is to the influence degree of data to use degree of variation metric matrix I to reflect; Use case mark Ent reflects that whether event occurs; Use consistency metric matrix H by checking mutually, obtain the order of accuarcy in Various types of data source;

(2) data source that the matrix acquisition confidence level after dividing according to data source assessment result vector is the highest, and be labeled as best source, utilize formula P=(D ^ti ^1-Enth) ^tw obtains final data source weight vectors P, represent the degree of reliability of the data result that each data source is calculated by the data that collect in fusion process, the confidence level size of each correspondence data source namely in this vector, wherein, D is feature interpretation matrix, I is degree of variation metric matrix, H is consistency metric matrix, Ent is event flag, W is evaluate parameter weight vectors, in W, element is corresponding assessment element weights value in data source appraisal framework, reflects the significance level that this element participates in merging;

(3) fiducial interval under best source α degree of confidence is utilized to screen multi-source data result: described screening multi-source data result refers to for the data of each in multi-source data, if this data value drops on outside best source fiducial interval, then think that it is an exceptional value, should give rejecting;

(4) to the data value stayed after screening, formula is utilized calculate its weighted value, wherein v _sthe data source velocity amplitude screened, for optimal data source speed average, S _sfor variance, for data cell confidence distance, p _sfor the weight of data source s, Duw _sfor data cell weighted value;

(5) average acquisition public transport road conditions result of calculation is weighted to multi-source data;

Weighted average calculation formula is wherein ω _ibe i-th data weighting, v _ifor velocity amplitude, i represents to be i-th data when pre-treatment, and n represents data total number, can obtain the public transport road conditions based on multi-source data analysis thus, i.e. public transport road travel average velocity.

6. the public transport road conditions disposal system based on traffic multisource data fusion according to claim 1, it is characterized in that: described public transport road conditions display module is implemented as: the public transport traffic information obtained according to multisource data fusion module, with public transport before fusion, hire out traffic information and carry out contrast and show, the web page display system realized by Web and J2EE framework intuitively shows public transport road conditions, data assessment progress of real-time monitoring each time period, and chosen by target road chain, show that target road chain merges the traffic information of front and back.

7., based on the public transport road conditions disposal route of traffic multisource data fusion, it is characterized in that performing step is as follows:

(1) data prediction step: for the treatment of traffic multi-source data, so that obtain can for follow-up data available set; By abnormal data cleaning, geographic coordinate conversion, ordered steps, obtain the traffic information that multi-data source characterizes, comprise public transport track data and hire out track data, video road section data on flows;

(2) source data treatment step: multi-source traffic data is of a great variety, with the angle of data source itself, traffic is described separately, source data treatment step processes these data, them are enable correctly to reflect public transport traffic information, by a curve fitting step, export the multi-source traffic data in order to describe public transport road conditions based on different pieces of information source, the traffic information that each data source is characterized and the traffic information of public transport data characterization are more pressed close to, thus obtain preliminary public transport traffic information;

(3) data source appraisal procedure: by setting up an appraisal framework for assessment of multi-data source quality, the comprehensive assessment quantized is carried out to each data source quality that the multi-source data of source data treatment step output is concentrated, thus obtain the weight of each data source in data fusion process, and these weight informations are delivered to data fusion step as the basic basis merging flow process as exporting;

(4) data fusion step: based on the multi-source traffic data that source data processing module obtains, be aided with each data source weight that data source evaluation module obtains, calculated by weight matrix division, best source, abnormal data cleans, data fusion step, fusion calculation is carried out to multi-source data, finally obtains the public transport traffic information that can be used for showing;

(5) step display: public transport traffic information data fusion step process obtained is presented on map as display data, for user or research and analyse.