CN104766475B - Urban traffic bottleneck mining method - Google Patents

Abstract

An urban traffic bottleneck mining method includes the following steps that firstly, microwave data and floating car data are cleaned, wherein data interpolation is carried out to solve the data missing problem of the microwave data and the floating car data; secondly, road segment traffic states are estimated by taking set time quanta as time granularity; thirdly, the total occurrence probability of traffic congestion of each road segment in the set time quanta within a statistical cycle is calculated; fourthly, the balance degree of traffic congestion occurring at each day within the statistical cycle is calculated; fifthly, the upper bound of the balance degrees of traffic congestion within the statistical cycle is worked out; sixthly, the normalization balance degree of traffic congestion within the statistical cycle is worked out; seventhly, the severity of traffic bottlenecks is figured out. The urban traffic bottleneck mining method has good reliability.

Description

A kind of urban transportation bottleneck method for digging

Technical field

Patent of the present invention belongs to intelligent transportation field, is specifically related to a kind of based on statistical traffic bottlenecks method for digging.

Background technology

Along with the increase continuously and healthily of city vehicle, urban traffic blocking constantly aggravates, and its main cause is the increasing of vehicle Speed, much larger than the speedup of road construction, result in the situation that supply falls short of demand.Urban traffic blocking causes the Trip Costs of the public to increase Adding, and owing to trip delay causes exhaust emissions too much, be degrading living environment further, consequence is serious.In order to relax city City's traffic congestion and optimize the configuration of limited path resource, analyzes and valency that the data that utilize traffic sensor to collect contain Then value carries out effective traffic organization is important means, is the master of vehicle supervision department's rational traffic management measure of formulation Will foundation.

Traffic police road surface is on duty is one not only tradition but also effective traffic organization method with commander, is alleviate traffic congestion one Individual important means.But, which section should be paid close attention on earth, but lack the calculating means of quantification.Urban road network In often there is traffic bottlenecks section, these bottleneck roads in certain period of time current demand much larger than the handling capacity in section, Show as blocking up of the normal property sent out, find and position these sections can facilitate traffic police personnel to formulate the fixed point duties optimized pre- Case.

The method of traditional location traffic bottlenecks is by experience, with impression.This method is the most original, has following lacking Point: the discovery for traffic bottlenecks is the most comprehensive；It is not easy to find newly-increased bottleneck in time；It is not easy to find the bottle under combination condition Neck sequence, such as traffic police personnel are wished to know in advance under little rainy day, evening peak, festivals or holidays three combination conditions, which section Belong to traffic bottlenecks.Therefore, quantification is expressed the order of severity of traffic bottlenecks and is had the biggest necessity, pre-for traffic police's duties The formulation of case has important reference value.

The traffic bottlenecks of published in recent years excavate paper such as: document " WH Lee, SS Tseng, JL Shieh, HH Chen, Discovering Traffic Bottlenecks in an Urban Network by Spatiotemporal Data Mining on Location-Based Services, IEEE Transactions on Intelligent Transportation Systems, VOL.12, NO.4, DECEMBER 2011, pp.1047-1056 ", it may be assumed that WH Lee, SS Tseng, JL Shieh, HH Chen, urban transportation bottleneck space-time method for digging based on location-based service, Electrical and Electronic engineering Shi Xiehui intelligent transportation system proceedings, in December, 2,011 4 phases of volume 12, page number scope: 1047-1056, utilize location-based Taxi data etc., it is proposed that a kind of three stage space-time traffic bottlenecks mining models, are clustered and propagation of blocking up by research space-time Pattern, establishes the space-time traffic bottlenecks prediction model of a kind of complexity；Document " JC Long, ZY Gao, HL Ren, AP Lian, Urban traffic congestion propagation and bottleneck identification, Science China:Information Sciences, 2008,51 (7): 948-964 ", i.e. JC Long, ZY Gao, HLRen, AP Lian, urban traffic blocking is propagated and bottleneck identification, " Chinese science " information science version, 7 phases of volume 51 in 2008, page number model Enclose: 948-964, propose a kind of identification based on unit TRANSFER MODEL fine-grained space-time traffic bottlenecks mining model.Middle promulgated by the State Council Bright patent 201410172412.3 (publication date: 2014-10-22) proposes in a kind of road grid traffic often to send out bottleneck and accidental bottleneck Recognition methods, identifies traffic bottlenecks by the occurrence frequency that blocks up in section.

Summary of the invention

For the deficiency that the reliability overcoming existing urban transportation bottleneck excavation mode is poor, the invention provides one can By the urban transportation bottleneck method for digging that property is good.

The technical solution adopted for the present invention to solve the technical problems is:

A kind of urban transportation bottleneck method for digging, comprises the steps:

Step 1: clean microwave and floating car data: the shortage of data existed for microwave data and floating car data is asked Topic, carries out data interpolation；

Step 2: to set the time period as time granularity estimation road section traffic volume state

The traffic behavior in the section that 2.1 Floating Car cover is estimated

If section L has c automobile to pass by setting the time period, its instantaneous velocity is respectively v₁,v₂...v_c, then when setting Between section travel speed V in this section in section_LIt is expressed as:

$V_L=\underset{i=1}{\overset{c}{\Σ}}{v}_i/c$

Then work as V_LDuring less than or equal to threshold value T, it is estimated as blocking up by the traffic behavior of section L；

The traffic behavior in the section that 2.2 Floating Car and microwave radar collectively cover is estimated

If there being microwave radar sensor on the L of section, measure the section speed setting the time period returned as v_m, then when setting Between section travel speed V in this section in section_LIt is expressed as:

$V_L=(v_m+\underset{i=1}{\overset{c}{\Σ}}{v}_i/c)/2$

In like manner, V is worked as_LDuring less than or equal to threshold value T, it is estimated as blocking up by the traffic behavior of section L；

The section that 2.3 microwave radars cover

K nearest neighbor sorting technique is used to estimate the section that microwave radar covers；

Step 3: in the counting statistics cycle, every section sets the overall probability that time period traffic congestion occurs

If measurement period is N days, it is divided into the time slot setting the time period every day, makes morning peak comprise n setting time The time slot of section, if in the morning peak of N days, total m sets time period generation traffic congestion, in measurement period, N days morning peaks In time period, set the time period block up occur probability P_tIt is expressed as:

P_t=m/ (N*n)

Step 4: in the counting statistics cycle, traffic congestion is at the equilibrium degree of every day

Make C_iRepresent that setting the time period in measurement period in i-th day blocks up the number of times occurred, then at i-th day, set the time The section accounting in measurement period of blocking up is: p_i=C_i/m；In making U express measurement period, traffic congestion is at the equilibrium degree of every day, adopts Equilibrium degree is expressed by comentropy:

$U=-\underset{i}{\Σ}{p}_i\log p_i$

Step 5: the upper bound of traffic congestion equilibrium degree in the counting statistics cycle

If in measurement period N days, the number of times setting the generation of time period jam situation of every day is identical, then the accounting of every day is 1/N, in this case set time period traffic congestion equilibrium degree as maximum, maximum informational entropy U_maxIt is expressed as:

U_max=logN

Step 6: the normalization equilibrium degree of traffic congestion in the counting statistics cycle

Definition normalization equilibrium degree U_nFor actual setting time period traffic congestion information entropy divided by maximum informational entropy, it may be assumed that

$U_n=\frac{U}{U_{max}}=\frac{-\underset{i}{\Σ}{p}_i\log p_i}{\log N}$

U_nSpan (0,1] between；

Step 7: calculate the severity of traffic bottlenecks

By the severity model of traffic bottlenecks it is:

Y_j=α P_t+(1-α)U_n,

Wherein, Y_jRepresenting the bottleneck order of severity in j-th strip section, α is a parameter between (0,1).

Further, described method for digging also comprises the steps: step 8: the traffic bottlenecks in all sections of descending sort are tight Severe, distributes reference rationally for duties.

Further, in described step 1, microwave data cleaning way: first the flow collected, speed and occupation rate The threshold value that value draws according to historical data with point duty department compares, and the data beyond threshold range are defined as mistake Data, for being unsatisfactory for the data of Threshold, substitute wrong data by threshold value；Wrong data is obtained according to decision rule, right In being unsatisfactory for the data of traffic flow theory, the method using historical data average is modified.

In described step 1, floating car data cleaning way: when the Floating Car speed in a certain section a certain moment is 0, as Really three moment Floating Car speed before this moment of this section are not the most 0, and we use three moment before this moment of this section Speed average is modified；If the Floating Car speed in three moment before this moment of this section is 0, then utilize history same period Floating Car speed average be modified.

In described step 2.3, mode based on video playback marks the traffic behavior of a collection of microwave radar sample, mark After sample set be referred to as template samples collection, its form is as follows:

Flow velocity

Flow

Lane occupancy ratio

Whether block up

For a section sample S (V to be sorted_S,F_S,O_S), wherein, V_SIn representing 5 minutes, the vehicle on the S of section is put down All flow velocity, F_SRepresent its average discharge, O_SRepresent its Ratio of driveway occupancy time；First try to achieve S to all template samples Euclidean away from From, select K the section template samples that distance S is minimum, then choose the friendship of S according to the traffic state information of template samples in a vote Logical state.

In the present invention, gathering floating car data and microwave radar data, the traffic bottlenecks computation model of proposition not only considers Block up the frequency occurred, and take into account the harmony of probability distribution, contributes to filtering due to road occupying of repairing the roads, great society The sporadic traffic bottlenecks that movable and other special circumstances cause.

Beneficial effects of the present invention is mainly manifested in: reliability is good.

Accompanying drawing explanation

Fig. 1 is the flow chart of urban transportation bottleneck method for digging.

Detailed description of the invention

The invention will be further described below in conjunction with the accompanying drawings.

With reference to Fig. 1, a kind of urban transportation bottleneck method for digging, comprise the steps:

Step 1: clean microwave and floating car data

Due to mechanical breakdown or the communication issue of microwave detection equipment, data there will be error message, so in data Need before analysis microwave data is carried out；The shortage of data problem existed for microwave data and floating car data, is carried out Data interpolation.

1.1 microwave datas clean

First the threshold value flow collected, speed and occupation rate value drawn according to historical data with point duty department Compare, the data beyond threshold range are defined as wrong data, below as a example by the road traffic flow data of Hangzhou, give Go out concrete wrong data decision rule example, as shown in table 1.

Table 1

The decision rule drawn according to traffic flow theory is as shown in table 2.

Table 2

After obtaining wrong data according to decision rule, need wrong data is modified, for being unsatisfactory for Threshold Data, substitute wrong data by threshold value, for being unsatisfactory for the data of traffic flow theory, use the average method of historical data to enter Row is revised.

1.2 floating car datas clean

When Floating Car speed is 0, it is believed that Floating Car is slack, the operation of road cannot be reflected in this case Situation, therefore we need the data to Floating Car speed is 0 to be carried out.Cleaning rule is as follows:

When the Floating Car speed in a certain section a certain moment is 0, if three moment before this moment of this section float Vehicle speed is not the most 0, and three moment speed averages before we use this moment of this section are modified；If this section this time The Floating Car speed in three moment before quarter is 0, then utilize the history same period (history of the synchronization data of month) Floating Car speed average is modified.

Step 2: estimate road section traffic volume state with 5 minutes for time granularity

Traffic state estimation method designed by the present invention is carried out based on Floating Car and two kinds of data sources of microwave radar, Traffic behavior to be estimated is divided into two classes of blocking up and do not block up, and below divides three kinds of situations to be illustrated.

2.1 only Floating Car cover section traffic behavior estimate

If section L had c automobile to pass by 5 minutes, its instantaneous velocity is respectively v₁,v₂...v_c, then 5 minutes Nei Gai roads Section travel speed V of section_LIt is expressed as:

$V_L=\underset{i=1}{\overset{c}{\Σ}}{v}_i/c$

Then work as V_LDuring≤T, it is estimated as blocking up by the traffic behavior of section L.For major trunk roads, subsidiary road and branch road, T takes 15km/h；For through street, T takes 25km/h.

The traffic behavior in the section that 2.2 Floating Car and microwave radar collectively cover is estimated

If there being microwave radar sensor on the L of section, measuring the 5 minutes section speed returned is v_m, then this section in 5 minutes Section travel speed V_LIt is expressed as:

$V_L=(v_m+\underset{i=1}{\overset{c}{\Σ}}{v}_i/c)/2$

In like manner, V is worked as_LDuring≤T, it is estimated as blocking up by the traffic behavior of section L.For major trunk roads, subsidiary road and branch road, T Take 15km/h；For through street, T takes 25km/h.

2.3 sections that only microwave radar covers

The present invention uses k nearest neighbor sorting technique to estimate the section that only microwave radar covers.It is primarily based on video playback Mode mark the traffic behavior of a collection of microwave radar sample, the sample set after mark is referred to as template samples collection, and its form is such as Under:

Flow velocity

Flow

Lane occupancy ratio

Whether block up

For a section sample S (V to be sorted_S,F_S,O_S), wherein, V_SIn representing 5 minutes, the vehicle on the S of section is put down All flow velocity, F_SRepresent its average discharge, O_SRepresent its Ratio of driveway occupancy time.First try to achieve the S Euclidean distance to all template samples (all samples should be by dimension normalization), selects K the section template samples that distance S is minimum, then according to template samples Traffic state information chooses the traffic behavior of S in a vote.

Noting: microwave radar sensor belongs to section sensor, it is interval that section section speed can not represent section well Speed, therefore when differentiating traffic behavior, uses threshold value calculation method to lack reasonability.

Step 3: the overall probability that in the counting statistics cycle, 5 minutes traffic congestions in every section occur

If measurement period is N days, 288 5 minutes grooves can be divided into every day.Making morning peak is that 7:30-9:00 is totally 1 little Time 30 points, comprise 18 5 minutes grooves altogether.If in the morning peak of N days, total m there is traffic congestion in 5 minutes.Then, system In the meter cycle, in N days morning peak time periods, the probability P occurred of blocking up for 5 minutes_tCan be expressed as:

P_t=m/ (N*18)

Step 4: in the counting statistics cycle, traffic congestion is at the equilibrium degree of every day

Make C_iBlock up in representing measurement period the number of times occurred in i-th day 5 minutes, then at i-th day, within 5 minutes, block up at system Accounting in the meter cycle is: p_i=C_i/m.Traffic congestion is at the equilibrium degree of every day to make U express in measurement period, and the present invention uses letter Breath entropy expresses equilibrium degree:

$U=-\underset{i}{\Σ}{p}_i\log p_i$

Roughly, U value is the biggest, represent 5 minutes traffic congestion situations be distributed in measurement period in each sky probability more Greatly.Also may be interpreted as: bigger U value 5 minutes jam situation of explanation is not the indivedual skies being concentrated in measurement period, But 5 minutes jam situations occur in a lot of skies in measurement period.In theory, what the latter more met that traffic bottlenecks are contained is normal The basic conception that the property sent out is blocked up.

Step 5: the upper bound of traffic congestion equilibrium degree in the counting statistics cycle

If in measurement period N days, the number of times that 5 minutes jam situations of every day occur is identical, then the accounting of every day is 1/N. In this case 5 minutes traffic congestion equilibrium degrees are maximum.Can be expressed as:

U_max=logN

Step 6: the normalization equilibrium degree of traffic congestion in the counting statistics cycle

Definition normalization equilibrium degree U_nFor 5 minutes actual traffic congestion information entropys divided by maximum informational entropy, it may be assumed that

$U_n=\frac{U}{U_{max}}=\frac{-\underset{i}{\Σ}{p}_i\log p_i}{\log N}$

It is seen that, U_nSpan (0,1] between.

Step 7: calculate the severity of traffic bottlenecks

Build the assessment models of traffic congestion bottleneck, be both specifically contemplated that blocking up for 5 minutes of a section in measurement period Probability, these should be avoided to block up not again is to concentrate on a few sky, and the latter's reason such as road occupying, occasion that is likely to be due to repair the roads is led Cause, be not belonging to often issue the category that blocks up.Therefore, the severity model of traffic bottlenecks is by the present invention:

Y_j=α P_t+(1-α)U_n,

Wherein, Y_jRepresent the bottleneck order of severity in j-th strip section.α is a parameter between (0,1), its value the closer to 0, Traffic congestion balance of distribution in measurement period is more focused in the most above-mentioned expression；Its value is the closer to 1, then traffic bottlenecks is serious Degree tolerance probability of more focusing on totally blocking up for 5 minutes is the highest.From there it can be seen that why propose normalization equilibrium degree Concept, is for probability P of totally blocking up 5 minutes_tUnified with normalization equilibrium degree under identical yardstick, thus beneficially structure Build information fusion expression formula Y_j=α P_t+(1-α)U_n。

Step 8: the traffic bottlenecks severity in all sections of descending sort, distributes reference rationally for duties.

Illustrated the committed step of the present invention by two examples, embodiment 1 illustrates the computational methods of traffic bottlenecks severity, Especially illustrating the balance of distribution effectiveness when portraying traffic bottlenecks severity that blocks up, embodiment 2 illustrates how to utilize k neighbour Classification method differentiates the traffic behavior in microwave radar monitoring section.

Embodiment 1: set measurement period as 20 days, the morning peak time period is from 7:30 to 9:00.Section A is due in first 5 days Near occasion impact, cause the morning peak of the first 5 days duration that blocks up longer, its morning peak in 20 days blocks up duration respectively For:

[60,65,65,65,70,5,0,5,0,10,5,0,5,5,10,0,0,0,5,5],

Section B time of blocking up in each sky in measurement period is respectively as follows:

[20,15,15,20,30,25,15,10,15,20,20,15,15,20,20,15,25,15,20,20]。

Ask section A and section B respective bottleneck severity.

Note: owing to occasion causes the longer section A of the duration that blocks up, from concept, should not regard friendship as Bottleneck link, is not belonging to the often property sent out and blocks up, do not have generality.On the contrary, the morning peak of section B almost every day all can be blocked up 10 minutes By 30 minutes, can regard the often property sent out as and block up, in theory, its bottleneck severity should be higher than section A.Below according to the present invention The traffic bottlenecks severity computational methods proposed, to assess the bottleneck severity of section A and B, are seen and whether are met expection.

In step 3,5 minutes traffic congestion overall probability in measurement period.

A length of when always blocking up in measurement period of section A: 380 minutes, its probability that totally blocks up for 5 minutes was:

380/ (90*20)=38/180

A length of when always blocking up in measurement period of section B: 370, its probability that totally blocks up for 5 minutes is:

370/ (90*20)=37/180

In step 4, in the counting statistics cycle traffic congestion at the probability distribution equilibrium degree of every day for section A or section B, With the duration that blocks up of every day divided by the available probability distribution of blocking up of the duration that blocks up that A or B is total, it is contemplated that in follow-up Logarithmic calculation Numerical problem, i.e. the input of logarithmic function can not be 0, and the duration that always blocks up adds a small constant 0.00002, will simultaneously The duration that blocks up of every day adds 0.000001.Then probability distribution p of blocking up of section A every day in measurement period is obtained_i(1≤i ≤ 20):

[0.157 0.171 0.171 0.171 0.184 0.013 2.63e-09 0.013 2.63e-09 0.026

0.013 2.63e-09 0.013 0.013 0.026 2.63e-09 2.63e-09 2.63e-09 0.013 0.013]

And probability distribution q of blocking up of the every day that section A is in measurement period_i(1≤i≤20):

[0.054 0.040 0.040 0.054 0.081 0.067 0.040 0.027 0.040 0.054

0.054 0.040 0.040 0.054 0.054 0.040 0.067 0.040 0.054 0.054]

Using the logarithmic function with 2 as the end, the equilibrium degree that blocks up that can get section A is:

$U_A=-\underset{i}{\Σ}{p}_i\log p_i=3.029$

Using the logarithmic function with 2 as the end, the equilibrium degree that blocks up that can get section A is:

$U_B=-\underset{i}{\Σ}{q}_i\log q_i=4.2802$

Step 5: in the counting statistics cycle, the maximum of probability distributing equilibrium degree of 5 minutes traffic congestions (uses right with 2 as the end Number function):

U_max=logN=log (20)=4.3219

Step 6: in the counting statistics cycle, 5 minutes probability distribution of the normalization equilibrium degree section A of 5 minutes traffic congestions are returned One change equilibrium degree:

U_n ^A=3.029/4.3219=0.7

5 minutes probability distribution normalization equilibrium degrees of section B:

U_n ^B=4.2802/4.3219=0.9903

Step 7: the severity of calculating traffic bottlenecks:

Make parameter alpha=0.8, represent that bottleneck severity is focused on totally blocking up for 5 minutes the size of probability, but contemplate every The harmony of it probability distribution of blocking up.For section A, can obtain its bottleneck severity is:

$Y_A=0.8*P_t^A+(1-0.8)U_n^A=0.3091$

For section B, can obtain its bottleneck severity is:

$Y_B=0.8*P_t^B+(1-0.8)U_n^B=0.3625$

So far, it is seen that section B is more likely be evaluated as traffic bottlenecks than section A, concept expection is met, regular The bottleneck severity of the section B the blocked up bottleneck severity more than the sporadic section A blocked up, illustrates friendship proposed by the invention Bottleneck link mining model is effective.

Embodiment 2: traffic behavior based on microwave data and k nearest neighbor algorithm is estimated.It is provided with trunk roads as shown in the table Section template samples storehouse, wherein 5 minutes interior section speed, flow and lane occupancy ratios are recorded by microwave remote sensor, and whether section Block up and marked by video playback and obtain:

Sample sequence number	Speed (km/h)	Flow ()	Lane occupancy ratio (%)	Whether block up
					1	19	10	60	It is
2	22	23	52	It is
					3	40	87	20	No
4	64	92	12	No
					5	15	16	46	It is

Table 3

If the profile data that the upper microwave radar in certain trunk section to be sorted records: speed is 24km/h, flow is 27, lane occupancy ratio is 70%, uses k nearest neighbour method to estimate the traffic behavior in this section.Make V represent velocity vector, then it is returned One change mode is:

(x-min (V))/(max (V)-min (V)), wherein, the minimum in max function and min function return vector V respectively And maximum, x represents and treats normalized velocity amplitude.Flow is similar with the normalization mode of lane occupancy ratio.It is computed returning One speed, flow and the lane occupancy ratio changed is table 4:

Sample sequence number	Speed	Flow	Lane occupancy ratio
				1	0.08	0	0.83
2	0.14	0.16	0.69
				3	0.51	0.94	0.14
4	1	1	0
				5	0	0.07	0.59
Section to be sorted	0.18	0.21	1

Table 4

Further, section to be sorted can be tried to achieve and be respectively table 5 to the Euclidean distance of 5 samples:

Table 5

If taking k=4, then most like with section to be sorted template samples is respectively 1,2,4,5, wherein has 3 tickets to block up and 1 Ticket does not blocks up, and therefore, section to be sorted should be estimated as blocking up.

Claims (5)

1. a urban transportation bottleneck method for digging, it is characterised in that described method for digging comprises the steps:

Step 1: clean microwave and floating car data: for microwave data and the shortage of data problem of floating car data existence, enter Row data interpolation；

Step 2: to set the time period as time granularity estimation road section traffic volume state

The traffic behavior in the section that 2.1 Floating Car cover is estimated

If section L has c automobile to pass by setting the time period, its instantaneous velocity is respectively v₁,v₂...v_c, then the time period is set Section travel speed V in this section interior_LIt is expressed as:

$V_L=\underset{i=1}{\overset{c}{\mathrm{\Σ}}}{v}_i/c$

Then work as V_LDuring less than or equal to threshold value T, it is estimated as blocking up by the traffic behavior of section L；

The traffic behavior in the section that 2.2 Floating Car and microwave radar collectively cover is estimated

If there being microwave radar sensor on the L of section, measure the section speed setting the time period returned as v_m, then the time period is set Section travel speed V in this section interior_LIt is expressed as:

$V_L=(v_m+\underset{i=1}{\overset{c}{\mathrm{\Σ}}}{v}_i/c)/2$

In like manner, V is worked as_LDuring less than or equal to threshold value T, it is estimated as blocking up by the traffic behavior of section L；

The section that 2.3 microwave radars cover

K nearest neighbor sorting technique is used to estimate the section that microwave radar covers；

Step 3: in the counting statistics cycle, every section sets the overall probability that time period traffic congestion occurs

If measurement period is N days, it is divided into the time slot setting the time period every day, makes morning peak comprise n and set the time period Time slot, if in the morning peak of N days, total m sets time period generation traffic congestion, in measurement period, N days morning peak time Section in, set the time period block up occur probability P_tIt is expressed as:

P_t=m/ (N*n)

Step 4: in the counting statistics cycle, traffic congestion is at the equilibrium degree of every day

Make C_iRepresent that setting the time period in measurement period in i-th day blocks up the number of times occurred, then at i-th day, setting the time period blocks up Accounting in measurement period is: p_i=C_i/m；In making U express measurement period, traffic congestion is at the equilibrium degree of every day, uses information Entropy expresses equilibrium degree:

$U=-\underset{i}{\mathrm{\Σ}}{p}_i\log p_i$

Step 5: the upper bound of traffic congestion equilibrium degree in the counting statistics cycle

If in measurement period N days, the number of times setting the generation of time period jam situation of every day is identical, then the accounting of every day is 1/N, In this case set time period traffic congestion equilibrium degree as maximum, maximum informational entropy U_maxIt is expressed as:

U_max=log N

Step 6: the normalization equilibrium degree of traffic congestion in the counting statistics cycle

Definition normalization equilibrium degree U_nFor actual setting time period traffic congestion information entropy divided by maximum informational entropy, it may be assumed that

$U_n=\frac{U}{U_{max}}=\frac{-\underset{i}{\mathrm{\Σ}}{p}_i\log p_i}{\log N}$

U_nSpan (0,1] between；

Step 7: calculate the severity of traffic bottlenecks

By the severity model of traffic bottlenecks it is:

Y_j=α P_t+(1-α)U_n,

Wherein, Y_jRepresenting the bottleneck order of severity in j-th strip section, α is a parameter between (0,1).

2. a kind of urban transportation bottleneck method for digging as claimed in claim 1, it is characterised in that: described method for digging also includes Following steps: step 8: the traffic bottlenecks severity in all sections of descending sort, distribute reference rationally for duties.

3. a kind of urban transportation bottleneck method for digging as claimed in claim 1 or 2, it is characterised in that: in described step 1, micro- Wave datum cleaning way: first the flow collected, speed and occupation rate value are obtained according to historical data with point duty department The threshold value gone out compares, and the data beyond threshold range are defined as wrong data, for being unsatisfactory for the data of Threshold, Wrong data is substituted by threshold value；Obtaining wrong data according to decision rule, for being unsatisfactory for the data of traffic flow theory, employing is gone through The average method of history data is modified.

4. a kind of urban transportation bottleneck method for digging as claimed in claim 1 or 2, it is characterised in that: in described step 1, floating Motor-car data cleansing mode: when the Floating Car speed in a certain section a certain moment is 0, if three before this moment of this section Moment Floating Car speed is not 0, uses three moment speed averages before this moment of this section to be modified；If this section The Floating Car speed in three moment before this moment is 0, then utilize the Floating Car speed average of history same period to be modified.

5. a kind of urban transportation bottleneck method for digging as claimed in claim 1 or 2, it is characterised in that: in described step 2.3, Mode based on video playback marks the traffic behavior of a collection of microwave radar sample, and the sample set after mark is referred to as template samples Collection, its form is as follows:

Flow velocity Flow Lane occupancy ratio Whether block up

For a section sample S (V to be sorted_S,F_S,O_S), wherein, V_SVehicle mean flow on the S of section in representing 5 minutes Speed, F_SRepresent its average discharge, O_SRepresent its Ratio of driveway occupancy time；First try to achieve the S Euclidean distance to all template samples, choosing Go out K the section template samples that distance S is minimum, then choose the traffic shape of S according to the traffic state information of template samples in a vote State.