CN102881162B - Data processing and fusion method for large-scale traffic information - Google Patents

Abstract

The invention discloses a data processing and fusion method for large-scale traffic information and belongs to real-time traffic information processing technologies. The method includes: obtaining normal data of a truth value according to multisource traffic data acquired by a test vehicle and each sensor, and determining a dynamic assignment method for parameters; rejecting abnormal data in data sets acquired by the sensors, and performing historical data compensation; and fusing real-time grading information of the multisource traffic data with data compensation completed. The truth value system is set up for obtaining correct initial assignment parameters of various acquisition modes and rejecting acquired abnormal data, missing data are reasonably filled according to the historical data to guarantee accuracy and integrity of the data, and the data acquired by different acquisition modes are graded and fused gradually to guarantee reliability of the data. Moreover, quickness in the fusion process is achieved, and influences of traffic incidents, traffic control, land occupation construction and traffic accidents on data in the fusion process are considered.

Description

The data processing of large-scale traffic information and fusion method

Technical field

The present invention relates to the transport information real-time processing technique field of intelligent transportation system, particularly a kind of data processing of large-scale traffic information and fusion method.

Background technology

Information fusion is applied to military field the earliest, is that combination multi-source information completes target detection with data, associates, multi-level, many-sided process of state estimation.The object of this information fusion obtains target identification accurately, complete and situation of battlefield and threat assessment timely.Along with the development of sensor technology, computer science and infotech, the various multisensor syste towards complicated applications background emerges in multitude, acquisition of information by all kinds of means, process and fusion is made to become possibility, and at numerous areas such as Financial Management, psychological assessment and prediction, medical diagnosis, weather forecast, organization and administration decision-making, robot vision, traffic control, remote sensing remote measurements, people recognize and the informix in multiple data source are got up to improve the achievement of work.Therefore, multisource information fusion technology obtains in military field and civil area and payes attention to widely and successfully application, and its Theories and methods has become an important research direction of Intelligent Information Processing and control.

Prior art more concentrates on data level information fusion, and it directly carries out comprehensive to not pretreated sensor original observed data and analyzes.Advantage maintains object information as much as possible, substantially information dropout or omission do not occur; Shortcoming is that process data are too many, expends time in oversize, poor real.Another kind of information fusion is feature level information fusion, also known as file-level information fusion, is to after the preliminary pre-service of sensor, realizes essential characteristic extract, provide the comprehensive analysis processing that the basis of files and reports performs at sensor.Its advantage is the important information both having kept sufficient amount, again through permissible data compression, greatly dilutes data volume, can improve the real-time of processing procedure; And valuable be especially that reality, all based on feature extraction, all carries out extensive work in this respect in the application of the Modern High-Tech such as pattern-recognition, graphical analysis, computer vision.The shortcoming of feature level information fusion is, inevitably has some information loss, thus needs to propose stricter requirement to sensor pre-service.

Due to the polyphyly of traffic data, must use a kind of effective method reasonable coordination multi-source data, fully comprehensive useful information, improves the ability of correct decisions in changeable environment.Data fusion technique is a kind of good tool meeting this requirement.The polyphyly of traffic data and complicacy require that transport information process must have Data fusion technique, so just can provide the robustness of transport information.

Existing traffic data integration technology more concentrates on data level information fusion, fusion process does not carry out classification to data according to the stability of sensor, and the data do not collected sensor are carried out verification and are rejected, data merge step by step, do not consider the impact that traffic events, traffic control, occupation of land construction, traffic hazard bring to data in fusion process.

Summary of the invention

In view of this, the object of the present invention is to provide a kind of data processing and fusion method of large-scale traffic information, for realizing the efficient fusion of traffic information data.

The invention provides a kind of data processing and fusion method of large-scale traffic information, comprising:

According to the normal data of the multi-source traffic data acquisition true value system that test carriage and each sensor collect, and determine the dynamic assignment method of parameter; Reject the abnormal data in the data acquisition that described sensor collects, and carry out the compensation of historical data; To the multi-source traffic data real-time graded information fusion completing compensation data;

The step of the normal data of described acquisition true value system specifically comprises:

By the test carriage of many high sampling rates, repeatedly test at Different periods, and obtain the mean value of period t, as standard value:

$\mathrm{Et}\left(X\right)=\frac{1}{n}\underset{i,j}{\mathrm{\Σ}}\frac{\mathrm{Xij}\left(t\right)}{m}$

Wherein, Xi, j (t) represent the measured value that the jth time of i-th test carriage is tested at period t, and n car carries out m time and measure;

Describedly determine that the step of the dynamic assignment method of parameter specifically comprises:

The data that the data collect period t test carriage and each sensor collect, split out taxi data, bus data, microwave/loop data, video monitoring data and license plate identification data, as all kinds of traffic information datas of period t;

Its standard deviation is calculated at the mean value of period t for standard value respectively with taxi data, bus data, microwave/loop data, video monitoring data and license plate identification data:

$D\left(X\right)=\sqrt{\frac{1}{N}\underset{i=1,..}{\mathrm{\Σ}}{(\mathrm{Xi}-E\left(X\right))}^2}$

Wherein, E (X) tests the mean value obtaining all data for the t period, and Xi represents i-th test data of Various types of data;

Use described standard value and standard deviation, obtain the normal distribution of every class data respectively, determine credibility interval, according to normal curve, calculate the probability of data in credibility interval:

$P(a<X<b)={\&Integral;}_a^b\frac{1}{D\left(X\right)\sqrt{2\mathrm{\&Pi;}}}{e}^{\frac{1}{2}{\left(\frac{X-E\left(X\right)}{D\left(X\right)}\right)}^2}$

Wherein, a, b represent that left and right that test data departs from standard value X can accept the boundary value of data respectively, and D (X) is standard deviation, and E (X) is mean value;

With the probability of described data in credibility interval, assignment is carried out to every class data basic probability assignment function m (A), assignment procedure is m (A)=P (a<A<b), A represents every class data, and P (a<A<b) represents the probability of every class data in credibility interval.

The present invention obtains the correct initial assignment parameter of various acquisition mode by setting up true value system, abnormity removing is carried out to the data truly collected, missing data rationally fills up according to historical data the accuracy and the integrality that ensure data, to different classes of acquisition mode obtain data carry out classification step by step fusion treatment to ensure the reliability of data, the rapidity of fusion process, and in fusion process, consider the impact that traffic events, traffic control, occupation of land construction, traffic hazard bring to data.Through this fusion process, realize the effect such as quick, accurate, reliable merged, true value system solves the inaccurate problem of prior art initial parameter assignment; Rejecting abnormal data, missing data are filled up and overcome prior art acquisition terminal unstable properties, and the difficult problem that acquisition terminal damages; Adopt D-S theory to carry out level based adjustment process to the data after process, solve the polyphyly problem of data fusion; Also take into full account traffic events, traffic control, occupation of land construction, traffic hazard, overcome the impact because above factor causes data.

Accompanying drawing explanation

The data processing of the large-scale traffic information that Fig. 1 provides for the embodiment of the present invention and fusion method process flow diagram;

Fig. 2 is the method flow diagram to dynamic state of parameters assignment in the embodiment of the present invention.

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 traffic data fusion process that the embodiment of the present invention provides can realize the effect such as quick, accurate, reliable merged, and true value system solves the inaccurate problem of prior art initial parameter assignment; Rejecting abnormal data, missing data are filled up and overcome prior art acquisition terminal unstable properties, and the difficult problem that acquisition terminal damages; Adopt D-S theory to carry out level based adjustment process to the data after process, solve the polyphyly problem of data fusion; Also take into full account traffic events, traffic control, occupation of land construction, traffic hazard, overcome the impact because above factor causes data.

The data processing of the large-scale traffic information that Fig. 1 provides for the embodiment of the present invention and fusion method process flow diagram, comprise the following steps:

Step 101, acquisition true value system, and determine the dynamic assignment method of the middle parameter of D-S (Dempster-Sharer evidence theory multi-source data fusion method).

Described true value system standard data are the test carriages by many high sampling rates, repeatedly test, and obtain the mean value of period t, as standard value at Different periods:

$\mathrm{Et}\left(X\right)=\frac{1}{n}\underset{i,j}{\mathrm{\&Sigma;}}\frac{\mathrm{Xij}\left(t\right)}{m}$

X _{i, j}the measured value that t jth time that () represents i-th test carriage is tested at period t, n car carries out m time and measures.

Fig. 2 is the method flow diagram to dynamic state of parameters assignment, comprises the following steps:

The data that step 1011, the data collected by period t test carriage and each sensor collect, split out taxi data, bus data, microwave/loop data, video monitoring data, and license plate identification data, as all kinds of traffic information data of period t.Because different collecting devices all can collect the data of same type, but each collecting device acquisition precision and influenced factor difference thereof, therefore, in subsequent step, there will be substep merge.

Step 1012, calculate its standard deviation at the mean value of period t for standard value with taxi data, bus data, microwave/loop data, video monitoring data and license plate identification data respectively:

$D\left(X\right)=\sqrt{\frac{1}{N}\underset{i=1,..}{\mathrm{\&Sigma;}}{(\mathrm{Xi}-E\left(X\right))}^2}$

Wherein, E (X) tests the mean value obtaining all data, X for the t period _irepresent i-th test data of Various types of data.

The dispersion degree of standard deviation response data, the i.e. degree of stability of data.

Step 1013, use described standard deviation and standard value, obtain the normal distribution of every class data respectively, determine credibility interval (namely generating data receiver interval), according to normal curve, calculate the probability of data in credibility interval:

$P(a<X<b)={\&Integral;}_a^b\frac{1}{D\left(X\right)\sqrt{2\mathrm{\&Pi;}}}{e}^{\frac{1}{2}{\left(\frac{X-E\left(X\right)}{D\left(X\right)}\right)}^2}$

Wherein, a, b represent that left and right that test data departs from standard value X can accept the boundary value of data respectively, and D (X) is standard deviation, and E (X) is mean value.

Step 1014, with the probability of described data in credibility interval to every class data basic probability assignment function m (A) assignment.Such as taxi data basic probability assignment function m (A) assignment procedure is m (A)=P (a<A<b), A represents taxi data, P (a<A<b) represents the probability of taxi image data in credibility interval, as above calculates.

Abnormal data in the data acquisition that step 102, rejecting sensor collect.Step 101 is data that acquisition repeatedly tested under normal circumstances by test carriage, for the Various types of data initial parameter assignment in data fusion process is prepared.And because various factors can cause the exception of data in reality, therefore, first real sensor will carry out rejecting abnormal data after collecting data, it is below the process that real data is carried out.

Calculation interval t correspondence classification (comprising taxi data, bus data, microwave/loop data, video monitoring data and license plate identification data) traffic information data X (X is the data collected under actual environment, and it is the data under test carriage and all normal environment of collector that 101 steps describe X) and history information data H respectively.H is the storage of the traffic information data (comprising taxi data, bus data, microwave/loop data, video monitoring data and license plate identification data) to past 1 year.Calculate standard deviation D (X), the D (H) of mathematical expectation E (X), all kinds of traffic information data X of E (H), calculation interval t and the past history information data H 1 year same period of X, H, if:

(1) for the traffic information data that each sensor of period t and test carriage collect, mathematical expectation E (X) and mathematical expectation E (H) difference are setting within the scope of threshold values th, and D (X) is same with D (H) difference setting in threshold values th ', then think that this time data exception does not occur;

(2) for the traffic information data that each sensor of period t and test carriage collect, mathematical expectation E (X) and mathematical expectation E (H) exceed and set threshold values th, but D (X) and D (H) difference setting in threshold values th ', then wait for the traffic information data X of subsequent period " generate; If (X ") has larger difference, exceeds setting threshold values th ', then think that the transport information of period t occurs abnormal, and fallen by rejecting abnormal data for the traffic information data D (X) of subsequent period and D.

In the preferred embodiment, th threshold value setting is th=E (H)/2, th '=D (X)/2.

The compensation of step 103, historical data.Every day each sensor and the data that obtain after step 102 of the traffic information data that collects of test carriage store, as historical data, as offset data during for occurring shortage of data future.

According to historical data, calculate the road network average velocity of interior all kinds of traffic information data on the same day mutually respectively, comprise taxi data road network average velocity, bus data average velocity, microwave/loop data average velocity, generate traffic tendency curve.

Every day each sensor and after test carriage collects data, carry out in rough handling (rejecting the abnormal data in data) process to data in time, search the Q comprised the traffic data file (i.e. the traffic data with similar traffic tendency curve of preceding classifications storage) in the adjacent Z time period, calculate the road network average velocity of each time period traffic data file, and this average velocity is considered as a point, smooth connection Z dot generation transport information curve segment.This curve segment is mated with the historical data in same time section, find out P bar curve the most similar as ancillary sources, wherein Z, P and Q be according to algorithm, historical statistics, put into practice, functional value that experience, system performance draw, Z represents the time period, Q represents the file amount generated in Z time period, and P represents the most similar curve quantity.An adjacent Z time period, in units of one hour, 24 time periods.General data gathers every five minutes and generates data, and a time period can generate 12 data files, and Q can generate 288 values by 24 time period calculations, and P value is 1.

According to the time attribute of all kinds of traffic information datas that each sensor and test carriage collect, search in the ancillary sources obtained from above step with the road chain that traffic data lacks, there is identical unique number, the historical data of same time attribute fills up absence information road chain in compensatory zone.Wherein time attribute comprises cycle in week and the time window characteristic residing for this moment.Why data there will be disappearance, on the one hand because rejecting abnormal data likely causes the discontinuous of data, be on the other hand due to each data-source device due to extraneous factor or human factor short duration failure cause data not gather in time or Acquisition Error larger.And the latter is particularly important.

Step 104, to complete compensation data multi-source traffic data (multi-source data herein comprise through rejecting and compensation data after taxi data, bus data, microwave/loop data, video monitoring data, and license plate identification data) real-time graded information fusion.Three steps are the bases of real-time graded information fusion above, it is the fusion parameters that rating information merges when determining data fusion of all categories that step 101 produces true value system, and provide normal data for the rejecting of step 102 abnormal data, step 103 carries out compensation data for the road chain of shortage of data or shortage of data time period, the data produced using step 102 and step 103 as the data source of multi-source traffic information data, the data used in following steps be all through step 101,102,103 data processed.

In order to ensure data precision and coverage rate, first merging taxi data various on a large scale and bus data, then merging microwave/loop data and license plate identification data.Step is as follows:

Concept the most basic in D-S evidence theory sets up framework of identification, is denoted as U, and the element in framework of identification meets mutual exclusive condition.The assignment m (A) of proposition A to function m be set to hint obliquely at, if function m meets following condition:

m(Ф)＝0

Right $\&ForAll;A\&Element;2^U,m\left(A\right)\&GreaterEqual;0,$ And $\underset{A\&Element;2^U}{\mathrm{\&Sigma;}}m\left(A\right)=1$

Then claim the m basic probability assignment function that (A) is A (BPAF).Wherein Ф represents empty set, and m (A) represents the accurate trusting degree to proposition A, illustrates the direct support to A.

In following steps, for the initialization of taxi data, bus data, microwave/loop data, video monitoring data and license plate identification data, the same mode of step 1014 that adopts is to basic probability assignment function assignment.What step 1013 obtained is under normal circumstances by test carriage obtain to the probability of Various types of data in credibility interval, lessly consider the impact on the data value range collected such as accident, road maintenance, therefore in following fusion process, also accident will be considered, road maintenance etc.

Step 1041, first fusion taxi and bus data.Due to the high utilization rate of the equipment such as GPS, data precision and coverage rate can be ensured.

The initial value of the basic probability assignment function of the taxi data obtained according to step 1014 and bus data acquisition data, is expressed as m ₁and m (A) ₂(A);

The data that note merges are U, and taxi data and bus data are respectively U ₁and U ₂;

$U=U_1*\frac{m_1\left(A\right)}{m_1\left(A\right)+m_2\left(A\right)}+U_2*\frac{m_2\left(A\right)}{m_1\left(A\right)+m_2\left(A\right)}$

U is the income value after taxi and bus information fusion, and the accurate trusting degree of the U after fusion is:

$m\left(A\right)=\frac{\underset{A_1\&cap;A_2=A}{\mathrm{\&Sigma;}}{m}_1\left(A_1\right)m_2\left(A_2\right)}{\underset{A_1\&cap;A_2\&NotEqual;\mathrm{\&Phi;}}{\mathrm{\&Sigma;}}{m}_1\left(A_1\right)m_2\left(A_2\right)}=m_1\left(A_1\right)\&CirclePlus;m_2\left(A_2\right)$

Step 1042, incorporate microwave/loop data.Microwave/loop data ensures stability and the reliability of data further.

The initial value m of the basic probability assignment function of the microwave/loop data obtained according to step 1014 ₃(A);

The data that note merges are K, and it is K that step 1041 obtains fused data ₁(i.e. the 1014 U values obtained), microwave/loop data is K ₂;

$K=K_1*\frac{m\left(A\right)}{m\left(A\right)+m_3\left(A\right)}+K_2*\frac{m_3\left(A\right)}{m\left(A\right)+m_3\left(A\right)}$

K is the income value after taxi, bus and microwave/loop data merges, and the accurate trusting degree of the K after fusion is:

$m_k\left(K\right)=\frac{\underset{K1\&cap;K2=K2}{\mathrm{\&Sigma;}}{m}_1\left(K_1\right)m_2\left(K_2\right)}{\underset{K1\&cap;K2\&NotEqual;\mathrm{\&Phi;}}{\mathrm{\&Sigma;}}{m}_{1\left(K_1\right)m_2\left(K_2\right)}}=m_1\left(K_1\right)\&CirclePlus;m_2\left(K_2\right)$

Step 1043, incorporate license plate identification data.High-definition camera generally can be equipped with in region, crossing, incorporates the accurate reliability that high-definition camera license plate identification data can ensure the data that bayonet socket section is passed through.Method is with step 1042.

Step 1044, incorporate trunk roads, through street, secondary distributor road traffic signal timing data.High-definition camera generally can be equipped with in region, crossing, intersection is collected due to vehicle, considerable influence is had to Vehicle Speed, incorporate the accurate reliability that high-definition camera license plate identification data can ensure the data that bayonet socket section is passed through, between trunk roads, through street, secondary distributor road section, data are comparatively large by the impact of traffic signal timing data, incorporate high-definition camera data respectively and traffic signal timing data can further improve data accuracy.Due to the interference of signal lamp, the running time of vehicle on the chain of road and speed can change to some extent, and the scope of the data value collected should suitably be amplified, and suitably should relax the scope of interval [a, b] when namely step 1013 calculates, fusion method is with step 1042.

Step 1045, incorporate traffic events, traffic control, stop construction, traffic hazard data.Because traffic events, traffic control, stop construction, traffic hazard are on the impact of data, to taxi, bus, the scope of the data value that microwave/coil etc. collects should suitably be amplified, namely the scope of interval [a, b] should suitably be relaxed when step 1013 calculates, the same.

Above method has taken into full account the stability of multiple acquisition mode, accuracy and reliability.Set up true value system and obtain the correct initial assignment parameter of various acquisition mode, abnormity removing is carried out to the data truly collected, missing data rationally fills up according to historical data the accuracy and the integrality that ensure data, to different classes of acquisition mode obtain data carry out classification step by step fusion treatment to ensure the reliability of data, the rapidity of fusion process, and in fusion process, consider the impact that traffic events, traffic control, occupation of land construction, traffic hazard bring to data.

Through the fusion process of above step, realize the effect such as quick, accurate, reliable merged, true value system solves the inaccurate problem of prior art initial parameter assignment; Rejecting abnormal data, missing data are filled up and overcome prior art acquisition terminal unstable properties, and the difficult problem that acquisition terminal damages; Adopt D-S theory to carry out level based adjustment process to the data after process, solve the polyphyly problem of data fusion.Take into full account traffic events, traffic control, occupation of land construction, traffic hazard, overcome the impact because above factor causes data.

In a word, the foregoing is only preferred embodiment of the present invention, be not intended to limit protection scope of the present invention.

Claims (7)

1. the data processing of large-scale traffic information and a fusion method, is characterized in that, comprising:

According to the normal data of the multi-source traffic data acquisition true value system that test carriage and each sensor collect, and determine the dynamic assignment method of parameter; Reject the abnormal data in the data acquisition that described sensor collects, and carry out the compensation of historical data; To the multi-source traffic data real-time graded information fusion completing compensation data;

The step of the normal data of described acquisition true value system specifically comprises:

By the test carriage of many high sampling rates, repeatedly test at Different periods, and obtain the mean value of period t, as standard value:

$\mathrm{Et}\left(X\right)=\frac{1}{n}\underset{i,j}{\mathrm{\&Sigma;}}\frac{\mathrm{Xij}\left(t\right)}{m}$

Wherein, X _{i, j}t () represents the measured value that the jth time of i-th test carriage is tested at period t, n car carries out m time and measure;

Describedly determine that the step of the dynamic assignment method of parameter specifically comprises:

The data that the data collect period t test carriage and each sensor collect, split out taxi data, bus data, microwave/loop data, video monitoring data and license plate identification data, as all kinds of traffic information datas of period t;

Its standard deviation is calculated at the mean value of period t for standard value respectively with taxi data, bus data, microwave/loop data, video monitoring data and license plate identification data:

$D\left(X\right)=\sqrt{\frac{1}{N}\underset{i=1,..}{\mathrm{\&Sigma;}}{(\mathrm{Xi}-E\left(X\right))}^2}$

Wherein, E (X) tests the mean value obtaining all data, X for the t period _irepresent i-th test data of Various types of data;

Use described standard value and standard deviation, obtain the normal distribution of every class data respectively, determine credibility interval, according to normal curve, calculate the probability of data in credibility interval:

$P(a<X<b)={\&Integral;}_a^b\frac{1}{D\left(X\right)\sqrt{2\mathrm{\&Pi;}}}{e}^{\frac{1}{2}{\left(\frac{X-E\left(X\right)}{D\left(X\right)}\right)}^2}$

Wherein, a, b represent that left and right that test data departs from standard value X can accept the boundary value of data respectively, and D (X) is standard deviation, and E (X) is mean value;

With the probability of described data in credibility interval, assignment is carried out to every class data basic probability assignment function m (A), assignment procedure is m (A)=P (a<A<b), A represents every class data, and P (a<A<b) represents the probability of every class data in credibility interval.

2. the data processing of large-scale traffic information according to claim 1 and fusion method, it is characterized in that, the step of the abnormal data in the data acquisition that described rejecting sensor collects specifically comprises:

Calculation interval t correspondence classification traffic information data X and history information data H respectively, calculate mathematical expectation E (X), the E (H) of X, H, standard deviation D (X), the D (H) of all kinds of traffic information data X of calculation interval t and past history information data H 1 year same period, if:

Mathematical expectation E (X) and mathematical expectation E (H) difference are setting within the scope of threshold values th, and D (X) is same with D (H) difference setting in threshold values th ', then this time data exception does not occur;

Mathematical expectation E (X) and mathematical expectation E (H) exceed and sets threshold values th, but D (X) and D (H) difference are setting in threshold values th ', then " generation of waiting for the traffic information data X of subsequent period; If (difference of X ") exceeds and set threshold values th ', then exception occurs the transport information of period t, and is fallen by rejecting abnormal data for the traffic information data D (X) of subsequent period and D.

3. the data processing of large-scale traffic information according to claim 2 and fusion method, it is characterized in that, described threshold values th and th ' is set as: th=E (H)/2, th '=D (X)/2.

4. the data processing of the large-scale traffic information according to Claims 2 or 3 and fusion method, is characterized in that, the compensation process of described historical data specifically comprises:

Every day each sensor and after test carriage collects data, search the Q comprised a traffic data file in the adjacent Z time period, calculate the road network average velocity of each time period traffic data file, and this average velocity is considered as a point, smooth connection Z dot generation transport information curve segment; This curve segment mated with the historical data in same time section, find out P bar curve the most similar as ancillary sources, wherein, Z represents the time period, and Q represents the file amount generated in Z time period, and P represents the most similar curve quantity;

According to the time attribute of all kinds of traffic information datas that each sensor and test carriage collect, search from described ancillary sources with the road chain that traffic data lacks, there is identical unique number, the historical data of same time attribute fills up absence information road chain in compensatory zone.

5. the data processing of large-scale traffic information according to claim 4 and fusion method, is characterized in that, the step of the described multi-source traffic data real-time graded information fusion to completing compensation data specifically comprises and merges taxi and bus data:

Obtain the initial value of the basic probability assignment function of taxi data and bus data, be expressed as m ₁and m (A) ₂(A);

Set out hire a car and bus data fusion after income value be U, taxi data and bus data are respectively U1 and U2; Then

$U=U_1*\frac{m_1\left(A\right)}{m_1\left(A\right)+m_2\left(A\right)}+U_2*\frac{m_2\left(A\right)}{m_1\left(A\right)+m_2\left(A\right)}$

The accurate trusting degree of U is:

$m\left(A\right)=\frac{\underset{A_1\&cap;A_2=A}{\mathrm{\&Sigma;}}{m}_1\left(A_1\right)m_2\left(A_2\right)}{\underset{A_1\&cap;A_2\&NotEqual;\mathrm{\&Phi;}}{\mathrm{\&Sigma;}}{m}_1\left(A_1\right)m_2\left(A_2\right)}=m_1\left(A_1\right)\&CirclePlus;m_2\left(A_2\right).$

6. the data processing of large-scale traffic information according to claim 5 and fusion method, it is characterized in that, the step of the described multi-source traffic data real-time graded information fusion to completing compensation data comprises fusion microwave/loop data further:

Obtain the initial value of the basic probability assignment function of microwave/loop data, be expressed as m ₃(A);

The data arranging fusion are K, and the income value after described taxi and bus data fusion is K ₁, microwave/loop data is K ₂;

$K=K_1*\frac{m\left(A\right)}{m\left(A\right)+m_3\left(A\right)}+K_2*\frac{m_3\left(A\right)}{m\left(A\right)+m_3\left(A\right)}$

K is the income value after taxi, bus and microwave/loop data merges, and the accurate trusting degree of the K after fusion is:

$m_k\left(K\right)=\frac{\underset{K1\&cap;K2=K2}{\mathrm{\&Sigma;}}{m}_1\left(K_1\right)m_2\left(K_2\right)}{\underset{K1\&cap;K2\&NotEqual;\mathrm{\&Phi;}}{\mathrm{\&Sigma;}}{m}_{1\left(K_1\right)m_2\left(K_2\right)}}=m_1\left(K_1\right)\&CirclePlus;m_2\left(K_2\right).$

7. the data processing of large-scale traffic information according to claim 6 and fusion method, it is characterized in that, the step of the described multi-source traffic data real-time graded information fusion to completing compensation data comprises fusion license plate identification data, video monitoring data and traffic events, traffic control, stop construction, traffic hazard data further.