CN102800197A - Preprocessing method of road section dynamic traffic stream essential data of urban road - Google Patents

Abstract

The invention discloses a preprocessing method of road section dynamic traffic stream essential data of an urban road. The method comprises the following steps of: acquiring the traffic stream essential data dynamically for time warping; detecting the effectiveness of divided lane traffic stream essential data; collecting the time of divided lane traffic stream essential data; collecting the fracture surface space at individual vehicle driving direction of the divided lane traffic stream essential data; and estimating the dynamic traffic stream loss data at individual vehicle driving direction. According to the invention, based on the characteristics of collected data of a fixed-type vehicle detector, the essential idea of the traffic stream theory is combined by adopting a threshold method, a set of data effectiveness detecting method is provided; aiming at different data loss patterns, history data is utilized reasonably, and the continuity and integrity of data are ensured furthest; the algorithm related to data preprocessing considers the timeliness and accuracy requirements at the same time, and the data analysis processing capability is high; and the preprocessing method of road section dynamic traffic stream essential data of the urban road plays positive significance to urban intelligent transportation system construction, road traffic informationization level improvement, and road operation management level improvement.

Description

A kind of preprocess method of Urban road dynamic traffic stream basic data

Technical field

The present invention relates to the intelligent transportation application, particularly a kind of preprocess method of Urban road dynamic traffic stream basic data.

Background technology

Dynamic traffic stream basic data based on fixed wagon detector collection is important intelligent transportation system (ITS) data source.Because fixed wagon detector is influenced by uncertain factors such as self duty, Network Transmission, road traffic condition and surrounding environment, problem such as that the data of collection exist is wrong, lose, time point drift, noise are excessive.If to raw data not in addition pre-service directly use; Then can influence upper strata intelligence transport applications system to urban highway traffic situation and the estimation of performance evaluation index and prediction accuracy and reliability, cause problems such as system's operation human intervention degree is big, the application system sustainability is not strong, range of application is limited.The pretreated fundamental purpose of visual data is the quality that the control road is gathered traffic flow data; Reduce the influence of problem data to the overall data degree of accuracy; Guarantee the accurate processing and the Secure Application of road traffic flow data, also feedback information and theoretical foundation are provided for improving urban road dynamic traffic flow data acquisition system.

Two aspect contents of the pre-service most critical of dynamic traffic stream basic data are that data validation and missing data are estimated.The data validation method has experienced the evolution that combines the comprehensive test method of traffic flow theory from simple threshold test method to the threshold test method, and typical algorithm has Washington algorithm (Washington Algorithm), daily statistic algorithm (Daily Statistics Algorithm) etc.; About the estimation of missing data, early stage ITS system does not carry out estimation process to missing data, but adopts the method for ignoring, and directly removes some suspicious or misdata.Obviously ignore the sample size that missing data has not only reduced traffic flow data, also probably lose the valuable information of part.Based on this, research and engineering staff recognize theory and the realistic meaning that missing data is estimated both at home and abroad, and the disappearance algorithm for estimating has been carried out a large amount of further investigations.At present traffic flow missing data method of estimation roughly can be divided into four types of method of estimation based on experience, the method for estimation based on statistics, artificial intelligence method of estimation and combined type methods of estimation.Be the method that the traffic expert uses the traffic knowledge and experience oneself grasped to sum up out wherein based on the method for estimation of experience, like the historical method of average, the traffic missing data method of estimation that distributes based on the track etc.; Method of estimation based on statistics mainly comprises time series analysis method, EM/DA model, linear regression model (LRM), Kalman filter model etc.; The artificial intelligence method of estimation mainly comprises all kinds of neural network algorithms and genetic algorithm etc.; The combined type method of estimation is the fusion application to several different methods, like and combinational algorithm that time series analysis send out average based on history, the combinational algorithm of pairing model, genetic algorithm and neural network or regression model etc.

Sum up the research and the practice of flowing the basic data preprocess method at present both at home and abroad at dynamic traffic, also exist following several problem:

1, the algorithm that relates to of available data pre-service can't be taken into account the real-time and the accuracy requirement of ITS system applies simultaneously, lacks the support of traffic flow theory, and the data analysis processing power is weak, it is lower to excavate level;

2, data validation does not form comprehensive, the comprehensive rule system of a cover, and inspection rule itself lacks certain level;

3, missing data is estimated to lack the identification of real data disappearance pattern and the utilization of historical data, and the composition that comprises empirical is more, and adopts single algorithm application effect undesirable.

Summary of the invention

Goal of the invention

The objective of the invention is to,, adopt threshold method to combine the basic thought of traffic flow theory, form the data validation method that a cover is comprehensive, comprehensive, have level based on fixed wagon detector image data characteristics; To different data disappearance patterns, rationally utilize historical data, adopt optimum relatively method to estimate missing data, guarantee the continuity and the integrality of data to greatest extent; On this basis, take all factors into consideration the real-time and the accuracy requirement of preprocessing algorithms.

Technical scheme

The objective of the invention is to realize through following steps:

A kind of preprocess method of Urban road dynamic traffic stream basic data comprises the steps:

The divided lane traffic flow basic data of 1) gathering with the fixed wagon detector of certain hour interval acquiring, the timestamp attribute of regular data are the whole minute moment of nearest neighbor;

2) divided lane traffic flow basic data validity check, carry out validity check to the traffic flow basic data in each track successively:

2-1) data non-NULL check if data are empty, then is labeled as invalid data, changes step 2-7 over to);

2-2) date and time stamp validity check if the date and time stamp misdata then is labeled as invalid data, changes step 2-7 over to);

2-3) non-repeating data check if repeating data then is labeled as invalid data, changes step 2-7 over to);

2-4) automobile storage if traffic flow basic data attribute variable value is zero, then is labeled as invalid data in check, changes step 2-7 over to);

2-5) single-way traffic stream basic data threshold method check;

2-6) single-way traffic stream basic data traffic flow theory method check;

2-7), then change step 3) over to, otherwise carry out next track traffic flow basic data validity check if all track traffic flow basic datas have all been accomplished validity check;

3) judge whether raw data accumulative total acquisition time equals binding time at interval at interval,, then raw data accumulative total acquisition time is made zero at interval, and change step 4) over to if satisfy;

4) the divided lane traffic flow basic data time is compiled, and promptly obtains each the track binding time of effective traffic flow basic data at interval;

5) divided lane traffic flow basic data single unit vehicle travel direction sectional space compiles; If it is valid data that the result is compiled in the space; Promptly obtain the single unit vehicle section binding time of the effective traffic flow basic data at interval of going; If the result is compiled for invalid data in the space, then it is regarded as missing data, change step 6) over to;

6) single unit vehicle travel direction section traffic flow missing data is estimated, promptly obtains the single unit vehicle section binding time of the estimation traffic flow basic data at interval of going:

6-1) judge the consecutive miss time interval, if the time interval less than 15 minutes, then changes step 6-2 over to), otherwise change step 6-3 over to);

6-2) adopt the walk random method to estimate missing data;

6-3) if consecutive miss interval greater than equaling 15 minutes, smaller or equal to 30 minutes, then changes step 6-4 over to), otherwise change step 6-5 over to);

6-4) adopt the historical average real-time adjustment method estimation missing data that combines;

6-5) adopt the historical method of average to estimate missing data.

The above-mentioned consecutive miss time interval, the timing node of selecting 15 minutes, 30 minutes be because:

This is by the decision of the effect of missing data method of estimation, and when consecutive miss time interval during less than 15 minutes, the walk random method is the optimal estimation method; When consecutive miss interval greater than equaling 15 minutes smaller or equal to 30 minutes, it is historical that average to combine real-time adjustment method be the optimal estimation method; When consecutive miss during interval greater than 30 minutes, the historical method of average is the optimal estimation method.

In the step 1), certain hour is spaced apart the raw data acquisition time interval of fixed wagon detector; Dynamic traffic stream basic data is the general name of the magnitude of traffic flow, vehicle ' average velocity, three attribute variable's data of occupation rate averaging time;

Step 2-2) in, the date and time stamp misdata is defined as the data that the Data Date timestamp is not equal to effective value;

Step 2-3) in, repeating data was defined as with the traffic flow basic data attribute variable in a last raw data acquisition time interval and is worth identical data;

Said valid data and invalid data refer to effective traffic flow basic data and invalid traffic flow basic data, and traffic flow basic data attribute variable has consistent validity.

Step 2-5) in, the method for single-way traffic stream basic data threshold method check is following:

A) magnitude of traffic flow threshold test if surpass threshold range, then is labeled as invalid data, changes step 2-7 over to);

B) vehicle ' average velocity threshold test if surpass threshold range, then is labeled as invalid data, changes step 2-7 over to);

C) averaging time the occupation rate threshold test, if surpass threshold range, then be labeled as invalid data, change step 2-7 over to);

D) adjacent percentage speed variation threshold test if surpass threshold range, then is labeled as invalid data, changes step 2-7 over to), the wherein adjacent time interval (t-1, t) percentage speed variation τ _{T-1, t}Computing formula following:

${\mathrm{\τ}}_{t-1,t}=\mathrm{Min}\{\frac{X_t}{X_{t-1}},\frac{X_{t-1}}{X_t}\}$

In the formula, X _t---expression vehicle ' average velocity sequence, t=1,2,3 The object of adjacent percentage speed variation threshold test is a through street traffic flow basic data;

Above-mentioned steps a)～d) in, each threshold range need be demarcated based on the statistics of historical traffic flow basic data.

Step 2-6) in, the method for single-way traffic stream basic data traffic flow theory method check is following:

A) judge whether to satisfy that the vehicle ' average speed value equals zero and the traffic flow value is not equal to zero,, then be labeled as invalid data, change step 2-7 over to) if satisfy;

B) judge whether to satisfy that the traffic flow value equals zero and the vehicle ' average speed value is not equal to zero,, then be labeled as invalid data, change step 2-7 over to) if satisfy;

C) judge whether to satisfy traffic flow value and vehicle ' average speed value all equal zero and averaging time the occupation rate value be not equal to zero, if satisfy, then be labeled as invalid data, change step 2-7 over to);

D) judge whether to satisfy averaging time the occupation rate value be zero and the traffic flow value greater than maximum possible traffic value, if satisfy, then be labeled as invalid data, change step 2-7 over to), maximum possible volume of traffic q wherein _MaxComputing formula following:

$q_{\max }=\frac{10\×v\×{\mathrm{\Δ}}_o}{l}$

In the formula, q _Max---the maximum possible volume of traffic (/ hour) when averaging time, occupation rate was zero in the expression raw data acquisition time interval, v---expression vehicle ' average velocity (km/hour), Δ _o---expression occupation rate precision averaging time (%), l---expression driving vehicle average effective vehicle commander (rice);

E) judge whether traffic flow density calculation value in track surpasses maximum traffic flow density, if satisfy, then is labeled as invalid data, changes step 2-7 over to), wherein the computing formula of track traffic flow density k is following:

k＝q/v

In the formula, k---expression track traffic flow density (/ track/km), q---the expression track magnitude of traffic flow (/ hour), v---expression vehicle ' average velocity (km/hour); Maximum traffic flow density need be demarcated based on the statistics of historical traffic flow basic data;

F) if averaging time, occupation rate was non-vanishing, judge that whether average effective vehicle commander calculated value surpasses certain threshold range, if satisfied, then this data recording of mark is invalid, changes step 2-7 over to), wherein the computing formula of average effective vehicle commander l is following:

$l=\frac{10\×O\×v}{q}$

In the formula, l---expression average effective vehicle commander (rice), O---expression occupation rate averaging time (%), v---expression vehicle ' average velocity (km/hour), q---the expression track magnitude of traffic flow (/ hour); Average effective vehicle commander's threshold range need be demarcated based on the statistics and the traffic flow formation of historical traffic flow basic data.

In the step 3), can be defined as the integral multiple in any raw data acquisition time interval binding time at interval as required.

In the step 4); Time is compiled and refers on the basis of validity check; The divided lane traffic flow basic data in the raw data acquisition time interval is accumulated long period divided lane traffic flow basic data at interval; Concrete grammar is: successively the traffic flow basic data time of carrying out in each track is compiled, if do not have valid data at interval binding time, then this track time of mark is compiled the result and is invalid data; If there are valid data, then to compile computing formula following the single-way traffic stream basic data time:

$q=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{q}_i\×\frac{N}{n}$

$v=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(q_i\×v_i)}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{q}_i}$

$o=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{o}_i}{n}$

In the formula, q---the expression long period is the inside lane magnitude of traffic flow (/ hour) at interval, q _i---effective magnitude of traffic flow (/ hour) in expression long period at interval interior i the raw data acquisition time interval; The valid data record number in n---expression long period in the at interval interior raw data acquisition time interval; N---the expression long period is the raw data acquisition time interval data record number of interior expectation at interval; V---expression long period interval inside lane vehicle ' average velocity (km/hour), v _i---effective vehicle ' average velocity in i the raw data acquisition time interval in the expression long period interval (km/hour), o---the expression long period is inside lane occupation rate averaging time (%) at interval, o _i---occupation rate effective time (%) in expression long period at interval interior i the raw data acquisition time interval.

In the step 5), the space is compiled to refer in the time and is compiled on the basis, and divided lane traffic flow basic data is accumulated the traffic flow basic data of single unit vehicle travel direction section, and concrete grammar is following:

A) if the divided lane traffic flow basic data of carrying out compiling in the space is invalid data, then this single unit vehicle travel direction sectional space of mark compiles the result and is invalid data, if there are valid data, then changes b over to);

B) it is following to compile computing formula based on the divided lane traffic flow basic data space of valid data:

$q=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{q}_i\×\frac{N}{n}$

$v=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(q_i\×v_i)}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{q}_i}$

$o=\frac{q\×\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\left(\frac{o_i\×v_i/q_i}{n}\right)}{v}$

In the formula, q---the expression single unit vehicle travel direction section magnitude of traffic flow (/ hour), q _i---represent effective magnitude of traffic flow (/ hour) in i track; N---expression track valid data record number; N---the total number of track-lines of expression single unit vehicle travel direction section, v---expression single unit vehicle travel direction section vehicle ' average velocity (km/hour), v _i---represent effective vehicle ' average velocity (km/hour) in i track, o---expression single unit vehicle travel direction section occupation rate averaging time (%), o _i---represent occupation rate effective time (%) in i track.

Step 6-2) in, the walk random method estimates that the expression formula of missing data is:

${\hat{X}}_t=X_{t-1}$

In the formula,

---the estimated value of expression t traffic flow constantly basic data, X _T-1---the actual value of expression t-1 traffic flow constantly basic data;

Step 6-4) in, the historical average expression formula that combines real-time adjustment method to estimate missing data is:

${\hat{X}}_{t,d_i}=X_{(t-1),d_i}\×\frac{\mathrm{HIST}{\hat{X}}_{t,d_i}}{\mathrm{HIST}{\hat{X}}_{(t-1),d_i}}$

$\mathrm{HIST}{\hat{X}}_{(t-1),d_i}=\left[\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{X}_{(t-1),d_{(i-j)}}\right]/N$

$\mathrm{HIST}{\hat{X}}_{{t,d}_i}=\left[\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{\hat{X}}_{t,d_{(i-j)}}\right]/N$

In the formula,

---represent d _iThe estimated value of it t traffic flow constantly basic data,

---represent d _iThe actual value of it t-1 traffic flow constantly basic data,

With

---be expressed as d respectively _iThe history average of it t-1 and t traffic flow constantly basic data, With ---represent d respectively _(i-j)The actual value of it t-1 and t traffic flow constantly basic data, N---expression set time length of window;

Step 6-5) in, the historical method of average estimates that the expression formula of missing data is:

${\hat{X}}_{t,d_i}=\left[\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{X}_{t,d_{(i-j)}}\right]/N$

In the formula,

---represent d _iThe estimated value of it t traffic flow constantly basic data,

---represent d _(i-j)The actual value of it t traffic flow constantly basic data, N---expression set time length of window.

Technique effect:

The invention provides a kind of preprocess method of Urban road dynamic traffic stream basic data; To fixed wagon detector image data of Urban road and different types of road traffic flow operation characteristic; Confirmed that threshold value index and traffic flow with statistical significance move the restriction relation between three parameters (flow, speed, occupation rate); On this basis, provide a cover comprehensive data validation rule; From time and space two dimensions data are compiled processing, generated the traffic flow basic data of multi-level type, for the data, services of different application demand is laid a good foundation; Through data in various degree being lacked the identification in the time interval, adopt optimum relatively method to estimate missing data, guaranteed the integrality and the reliability of data to greatest extent.The inventive method is that the pre-service of Urban road traffic flow data provides the solution that a cover is comprehensive, comprehensive, have level; Can satisfy the demand of fields such as traffic operation and management, traffic programme decision-making, scientific research research, public service, also technical support is provided simultaneously for making up urban road traffic information acquisition system and traffic integration information platform for the traffic flow basic data.

Description of drawings

Fig. 1 is a kind of preprocess method process flow diagram of Urban road dynamic traffic stream basic data;

Fig. 2 is image data cross section place figure;

Fig. 3 is No. 2035 section time period flow-speed time series charts;

Fig. 4 is West 2nd Ring Road, Beijing 2046 sections flow at noon on September 4th, 2009-speed time series chart;

Fig. 5 is West 2nd Ring Road, Beijing 2041 sections in September, 2009 11-12 day speed-average effective vehicle commander's graph of a relation;

Fig. 6 is vehicle ' average velocity MAE figure;

Fig. 7 is speed MAPE figure;

Fig. 8 is speed RMSE figure.

Embodiment

Below in conjunction with accompanying drawing and specific embodiment the present invention is described further.

The present invention adopts the through street of Road Transportation in Beijing data acquisition system (DAS) microwave vehicle detecting device collection and the case verification that major trunk roads traffic flow basic data is carried out a kind of preprocess method of Urban road dynamic traffic stream basic data.The traffic flow basic data is the general name of the magnitude of traffic flow, vehicle ' average velocity, three attribute variable's data of occupation rate averaging time; The magnitude of traffic flow is the traffic value in the unit interval; Vehicle ' average velocity is the average place of vehicle instantaneous velocity in the unit interval, and averaging time, occupation rate was the time that has car to pass through to account for wagon detector acquisition time number percent at interval.The raw data acquisition time interval of Road Transportation in Beijing data acquisition system (DAS) setting is two minutes; The through street data acquisition system (DAS) to averaging time occupation rate all be recorded as 1% less than 1% data, the major trunk roads data acquisition system (DAS) to averaging time occupation rate then all be recorded as zero less than 1% data.

Shown in accompanying drawing 1, the practical implementation step is following:

Step 1. was obtained the divided lane dynamic traffic stream basic data of city expressway and major trunk roads with two minutes raw data acquisition time intervals; The line time of going forward side by side is regular; With raw data concentrate the timestamp attribute of entanglement regular be whole minute of nearest neighbor constantly, two minute datas of generation standard.

Step 2. divided lane traffic flow basic data validity check; Successively the traffic flow basic data in each track is carried out validity check; Its assay comprises two types of data; Be respectively effective traffic flow basic data and invalid traffic flow basic data, abbreviate valid data and invalid data as, and traffic flow basic data attribute variable has consistent validity.Present embodiment is distinguished through the data record is increased a data validation attribute, is defaulted as valid data, and property value is 1, if assay is an invalid data, then property value becomes 0, and concrete steps are following:

2-1) data non-NULL check if data are empty, then is labeled as invalid data, changes step 2-7 over to);

2-2) date and time stamp validity check; If the date and time stamp misdata then is labeled as invalid data, change step 2-7 over to); Wherein the date and time stamp misdata is defined as the data that the Data Date timestamp is not equal to effective value; The effective value that data time stabs promptly is worth reference time, is the predefined data upload of data acquisition system (DAS) sequence constantly;

2-3) non-repeating data check if repeating data then is labeled as invalid data, changes step 2-7 over to), wherein repeating data was defined as with the traffic flow basic data attribute variable in a last raw data acquisition time interval and is worth identical data;

2-4) automobile storage if traffic flow basic data attribute variable value is zero, then is labeled as invalid data in check, changes step 2-7 over to);

2-5) single-way traffic stream basic data threshold method check.The core concept of threshold method check is the traffic flow data based on actual acquisition, rationally formulates the very big and minimal value that fixed wagon detector is gathered traffic flow basic data attribute variable according to the mathematical statistics rule, comprises following inspection rule:

A) magnitude of traffic flow threshold test if surpass threshold range, then is labeled as invalid data, changes step 2-7 over to);

B) vehicle ' average velocity threshold test if surpass threshold range, then is labeled as invalid data, changes step 2-7 over to);

C) averaging time the occupation rate threshold test, if surpass threshold range, then be labeled as invalid data, change step 2-7 over to);

D) adjacent percentage speed variation threshold test if surpass threshold range, then is labeled as invalid data, changes step 2-7 over to), the wherein adjacent time interval (t-1, t) percentage speed variation τ _{T-1, t}Computing formula following:

${\mathrm{\τ}}_{t-1,t}=\mathrm{Min}\{\frac{X_t}{X_{t-1}},\frac{X_{t-1}}{X_t}\}$

In the formula, X _t---expression vehicle ' average velocity sequence, t=1,2,3

Present embodiment is demarcated each threshold range based on the statistics of the historical traffic flow data of Beijing's city expressway and the collection of major trunk roads microwave vehicle detecting device; The maximum of bicycle road, through street dynamic traffic stream basic data and the minimum volume of traffic (), vehicle ' average velocity (km/hour), averaging time occupation rate check (%), adjacent percentage speed variation (dimensionless) threshold values scope to be respectively [0; 80], [0; 120], [0,100], [0.4,1]; The maximum of major trunk roads bicycle road dynamic traffic stream basic data and the minimum volume of traffic, vehicle ' average velocity, averaging time, occupation rate threshold values scope was respectively [0,70], [0,120], [0,100].Because the traffic flow of city expressway belongs to the continuous traffic flow category; Have in the adjacent short period interval; Traffic flow modes changes features of smaller, therefore set adjacent percentage speed variation inspection rule, and the traffic flow of major urban arterial highway belongs to discontinuity traffic flow category; The traffic flow modes undulatory property is bigger, inapplicable this inspection rule;

2-6) single-way traffic stream basic data traffic flow theory method check.Basic thought based on the dynamic traffic flow data validity check of traffic flow theory is to utilize the comformity relation between the traffic flow operational factor to make up dynamic traffic data validation rule; The traffic flow basic data of fixed wagon detector actual acquisition is zero except some attribute variables; Other attribute variable can not be for outside zero the situation; Also exist three attribute variable's values all non-vanishing, but inconsistent situation between the attribute variable.The method of concrete test stage is following:

A) judge whether to satisfy that the vehicle ' average speed value equals zero and the traffic flow value is not equal to zero,, then be labeled as invalid data, change step 2-7 over to) if satisfy;

B) judge whether to satisfy that the traffic flow value equals zero and the vehicle ' average speed value is not equal to zero,, then be labeled as invalid data, change step 2-7 over to) if satisfy;

C) judge whether to satisfy traffic flow value and vehicle ' average speed value all equal zero and averaging time the occupation rate value be not equal to zero, if satisfy, then be labeled as invalid data, change step 2-7 over to);

D) judge whether to satisfy averaging time the occupation rate value be zero and the traffic flow value greater than maximum possible traffic value; If satisfy; Then be labeled as invalid data; Change step 2-7 over to), this rule is that occupation rate is zero situation setting averaging time greater than zero to the not enough magnitude of traffic flow that produces of fixed wagon detector occupation rate acquisition precision, as Beijing's major trunk roads data acquisition system (DAS) to averaging time occupation rate then all be recorded as zero less than 1% data.Maximum possible volume of traffic q wherein _MaxComputing formula following:

$q_{\max }=\frac{10\×v\×{\mathrm{\Δ}}_o}{l}$

In the formula, q _Max---the maximum possible volume of traffic (/ hour) when averaging time, occupation rate was zero in the expression raw data acquisition time interval, v---expression vehicle ' average velocity (km/hour), Δ _o---expression occupation rate precision averaging time (%), l---expression driving vehicle average effective vehicle commander (rice);

E) judge whether traffic flow density calculation value in track surpasses maximum traffic flow density, if satisfy, then is labeled as invalid data, changes step 2-7 over to), wherein the computing formula of track traffic flow density k is following:

k＝q/v

In the formula, k---expression track traffic flow density (/ track/km), q---the expression track magnitude of traffic flow (/ hour), v---expression vehicle ' average velocity (km/hour);

F) if averaging time occupation rate non-vanishing, judge that whether average effective vehicle commander calculated value surpasses certain threshold range, if satisfied, then be labeled as invalid data, change step 2-7 over to), wherein the computing formula of average effective vehicle commander l is following:

$l=\frac{10\×O\×v}{q}$

In the formula, l---expression average effective vehicle commander (rice), O---expression occupation rate averaging time (%), v---expression vehicle ' average velocity (km/hour), q---the expression track magnitude of traffic flow (/ hour);

Above-mentioned steps d) in, parameter l is demarcated 2.4 meters of the car average effective vehicle commanders that go for Beijing's major trunk roads, parameter Δ _oDemarcation is 1;

Above-mentioned steps e) in; The statistics of the historical traffic flow data of gathering based on Beijing's city expressway and major trunk roads microwave vehicle detecting device; It is 100/track/km that the maximum traffic flow density value in through street is demarcated, and it is 70/track/km that the maximum traffic flow density value of major trunk roads is demarcated;

Above-mentioned steps f) in; Based on Beijing's city expressway and the historical traffic flow data of major trunk roads microwave vehicle detecting device collection and the vehicle component analysis of through street, Beijing and major trunk roads; Through street and major trunk roads average effective vehicle commander threshold values scope are demarcated and are [2.4,18];

2-7), then change step 3 over to, otherwise carry out next track traffic flow basic data validity check if all track traffic flow basic datas have all been accomplished validity check.

Step 3. judges whether raw data accumulative total acquisition time equals binding time at interval at interval; If satisfy; Then raw data accumulative total acquisition time is made zero at interval, and change step 4 over to, wherein can be defined as the integral multiple in any raw data acquisition time interval binding time at interval as required.

The step 4. divided lane traffic flow basic data time is compiled; Promptly obtain each the track binding time of effective traffic flow basic data at interval, the traffic flow data time is compiled the magnitude of traffic flow, vehicle ' average velocity, the time occupancy data that refer to the short period interval (as 1 minute) that fixed wagon detector is gathered and compiles the magnitude of traffic flow, vehicle ' average velocity data and time occupancy data for long period interval (as 2 minutes or 5 minutes).The divided lane traffic flow basic data in the raw data acquisition time interval is accumulated long period divided lane traffic flow basic data at interval; Concrete grammar is: successively the traffic flow basic data time of carrying out in each track is compiled; If do not exist valid data in the interval binding time; Then this track time of mark is compiled the result for invalid data, if there are valid data, then to compile computing formula following the single-way traffic stream basic data time:

$q=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{q}_i\×\frac{N}{n}$

$v=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(q_i\×v_i)}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{q}_i}$

$o=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{o}_i}{n}$

In the formula, q---the expression long period is the inside lane magnitude of traffic flow (/ hour) at interval, q _i---effective magnitude of traffic flow (/ hour) in expression long period at interval interior i the raw data acquisition time interval; The valid data record number in n---expression long period in the at interval interior raw data acquisition time interval; N---the expression long period is the raw data acquisition time interval data record number of interior expectation at interval; V---expression long period interval inside lane vehicle ' average velocity (km/hour), v _i---effective vehicle ' average velocity in i the raw data acquisition time interval in the expression long period interval (km/hour), o---the expression long period is inside lane occupation rate averaging time (%) at interval, o _i---occupation rate effective time (%) in expression long period at interval interior i the raw data acquisition time interval.

Step 5. divided lane traffic flow basic data single unit vehicle travel direction sectional space compiles.If the result is compiled for invalid data in the space, promptly obtain the single unit vehicle section binding time of the effective traffic flow basic data at interval of going, if compiling the result, the space is invalid data, then it is regarded as missing data, change step 6 over to.Traffic flow basic data space is compiled to refer in the time and is compiled on the basis; With the magnitude of traffic flow of binding time at interval the divided lane magnitude of traffic flow, vehicle ' average velocity data, the integrated single unit vehicle travel direction of occupation rate data sink averaging time section, vehicle ' average velocity and averaging time occupation rate, concrete grammar is following:

A) if the divided lane traffic flow basic data of carrying out compiling in the space is invalid data, then this single unit vehicle travel direction sectional space of mark compiles the result and is invalid data, if there are valid data, otherwise changes b over to);

B) it is following to compile computing formula based on the divided lane traffic flow basic data space of valid data:

$q=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{q}_i\×\frac{N}{n}$

$v=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(q_i\×v_i)}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{q}_i}$

$o=\frac{q\×\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\left(\frac{o_i\×v_i/q_i}{n}\right)}{v}$

In the formula, q---the expression single unit vehicle travel direction section magnitude of traffic flow (/ hour), q _i---represent effective magnitude of traffic flow (/ hour) in i track; N---expression track valid data record number; N---the total number of track-lines of expression single unit vehicle travel direction section, v---expression single unit vehicle travel direction section vehicle ' average velocity (km/hour), v _i---represent effective vehicle ' average velocity (km/hour) in i track, o---expression single unit vehicle travel direction section occupation rate averaging time (%), o _i---represent occupation rate effective time (%) in i track.

Step 6. single unit vehicle travel direction section dynamic traffic stream missing data estimates, promptly obtains the single unit vehicle section binding time of the estimation traffic flow basic data at interval of going, and the practical implementation method is following:

6-1) judge the consecutive miss time interval, if the time interval less than 15 minutes, then changes step 6-2 over to), otherwise change step 6-3 over to);

6-2) adopt the walk random method to estimate missing data, its calculation expression is:

${\hat{X}}_t=X_{t-1}$

In the formula, ---the estimated value of expression t traffic flow constantly basic data, X _T-1---the actual value of expression t-1 traffic flow constantly basic data;

6-3) if consecutive miss interval greater than equaling 15 minutes, smaller or equal to 30 minutes, then changes step 6-4 over to), otherwise change step 6-5 over to);

6-4) adopt the historical average real-time adjustment method estimation missing data that combines, its calculation expression is:

${\hat{X}}_{t,d_i}=X_{(t-1),d_i}\×\frac{\mathrm{HIST}{\hat{X}}_{t,d_i}}{\mathrm{HIST}{\hat{X}}_{(t-1),d_i}}$

$\mathrm{HIST}{\hat{X}}_{(t-1),d_i}=\left[\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{X}_{(t-1),d_{(i-j)}}\right]/N$

$\mathrm{HIST}{\hat{X}}_{{t,d}_i}=\left[\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{\hat{X}}_{t,d_{(i-j)}}\right]/N$

In the formula,

---represent d _iThe estimated value of it t traffic flow constantly basic data, ---represent d _iThe actual value of it t-1 traffic flow constantly basic data,

With

---be expressed as d respectively _iThe history average of it t-1 and t traffic flow constantly basic data,

With

---represent d respectively _(i-j)The actual value of it t-1 and t traffic flow constantly basic data, N---expression set time length of window;

6-5) adopt the historical method of average to estimate missing data, its calculation expression is:

${\hat{X}}_{t,d_i}=\left[\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{X}_{t,d_{(i-j)}}\right]/N$

In the formula, ---represent d _iThe estimated value of it t traffic flow constantly basic data, ---represent d _(i-j)The actual value of it t traffic flow constantly basic data, N---expression set time length of window;

Set time length of window in the historical average real-time adjustment method of combination of present embodiment definition and the historical method of average is 30 days.

Based on actual acquired data, following content will further specify and verify that a kind of Urban road dynamic traffic flows the preprocess method of basic data in four aspects of missing data method of estimation recruitment evaluation under the disappearance time interval from data validation result, vehicle ' average velocity threshold test effect, adjacent percentage speed variation threshold test effect, average effective vehicle commander's threshold test effect and difference.

The data validation result

In order to check the performance of dynamic traffic stream basic data validity check rule; Present embodiment has enumerated through street, Beijing (being numbered 2010,2014,2017,2035) and major trunk roads (being numbered 22002,22005,24003,24005) are as shown in Figure 2; The data validity rule test result in 8 section on September 1st, 2009 to September 30 altogether, as shown in table 1:

Through street, table 1 Beijing and 8 sections of major trunk roads dynamic traffic in September stream basic data validity check be summary sheet as a result

Can find out that by table 1 rule 4, rule 5 and rule 6 are easier to check the data that make mistake, the amount of error that other rule tests go out is less relatively.Below will stress the effect of the average effective vehicle commander's check in the vehicle ' average velocity threshold test in the threshold method check, the check of adjacent percentage speed variation and the check of traffic flow method:

Vehicle ' average velocity threshold test effect

。

Table 2 and a mistake! Do not find Reference source.3 are respectively traffic flow basic data and the corresponding flow velocity diagram of ring direction to Beijing first lane 16:06:00 to 17:00:00 on September 8th, 2009 in No. 2035 sections in through street, Beijing.Wherein.3 that are labeled as grey in the table 2 are recorded as the wrong or suspicious data that vehicle ' average velocity threshold rule is checked out, and corresponding to 3 points in the middle of Fig. 3 circle.

Ring direction to Beijing first lane traffic flow base data table in 2035 sections of through street, table 2 Beijing

Adjacent percentage speed variation threshold test effect

It is the traffic flow basic data and the flow-speed time series graph of a relation of No. 2035 sections microwave vehicle detecting device actual acquisition at noon on the 4th September in 2009 that table 3 and Fig. 4 show.Wherein, be labeled as two mistake or suspicious datas that are recorded as through occurring behind the adjacent time percentage speed variation threshold test of grey in the table 4, and corresponding to 2 points in the middle of Fig. 4 circle.

Table 5 real time data and different traffic classification cluster centre Euclidean distance table

Average effective vehicle commander's threshold test effect

A mistake! Do not find Reference source.5 be No. 2035 sections on September 11st, 2009 and September 12 the vehicle diffusing some graph of a relation between average velocity and the average effective vehicle commander that go.Can find out by figure; The average effective vehicle commander of actual travel vehicle is seldom greater than 18 meters or less than 2.4 meters; The data that three cornet marks are represented are mistake or suspicious data through occurring behind the average effective vehicle commander threshold test, and table 6 has been enumerated the misdata that part average effective vehicle commander checks the back to occur.

2041 sections per day effective vehicle commander September 11 in 2009 in West 2nd Ring Road, table 7 Beijing checks the misdata example

Missing data method of estimation recruitment evaluation under the different disappearance time intervals

In order to assess the estimated performance of each class methods under the different disappearance time interval; Present embodiment is on the basis that guarantees selected section data integrity in September; Extract the selected section partial data in October; 11 kinds of different successive disappearances of artificial generation levels were respectively consecutive miss 10 minutes, 15 minutes, 20 minutes, 30 minutes, 45 minutes, 1 hour, 3 hours, 5 hours, 8 hours, 12 hours and 1 day, and estimated result and True Data are compared; Calculate respectively and estimate mean absolute error (MAE), average absolute percentage error (MAPE) and mean absolute error standard deviation (RMSE) value, part result of calculation is like Fig. 6,7, shown in 8.

According to the chart analysis-by-synthesis, draw the 7th consecutive miss level (promptly 3 hours) before, walk random totally is superior to the historical method of average with the real-time adjustment method of historical average combination; Walk random is superior to the historical average real-time adjustment method that combines before the 2nd consecutive miss level simultaneously; Between 7 consecutive miss levels of the 2nd consecutive miss level to the, the real-time adjustment method of historical average combination totally is superior to the walk random model; After the 7th the consecutive miss level, the historical average estimation technique is superior to the walk random estimation technique and the historical average real-time adjustment method that combines.

Claims (8)

1. the preprocess method of a Urban road dynamic traffic stream basic data is characterized in that comprising the steps:

The divided lane traffic flow basic data of 1) gathering with the fixed wagon detector of certain hour interval acquiring, the timestamp attribute of regular data are the whole minute moment of nearest neighbor;

2) divided lane traffic flow basic data validity check, carry out validity check to the traffic flow basic data in each track successively:

2-1) data non-NULL check if data are empty, then is labeled as invalid data, changes step 2-7 over to);

2-2) date and time stamp validity check if the date and time stamp misdata then is labeled as invalid data, changes step 2-7 over to);

2-3) non-repeating data check if repeating data then is labeled as invalid data, changes step 2-7 over to);

2-4) automobile storage if traffic flow basic data attribute variable value is zero, then is labeled as invalid data in check, changes step 2-7 over to);

2-5) single-way traffic stream basic data threshold method check;

2-6) single-way traffic stream basic data traffic flow theory method check;

2-7), then change step 3) over to, otherwise carry out next track traffic flow basic data validity check if all track traffic flow basic datas have all been accomplished validity check;

3) judge whether raw data accumulative total acquisition time equals binding time at interval at interval,, then raw data accumulative total acquisition time is made zero at interval, and change step 4) over to if satisfy;

4) the divided lane traffic flow basic data time is compiled, and promptly obtains each the track binding time of effective traffic flow basic data at interval;

5) divided lane traffic flow basic data single unit vehicle travel direction sectional space compiles; If it is valid data that the result is compiled in the space; Promptly obtain the single unit vehicle section binding time of the effective traffic flow basic data at interval of going; If the result is compiled for invalid data in the space, then it is regarded as missing data, change step 6) over to;

6) single unit vehicle travel direction section traffic flow missing data is estimated, promptly obtains the single unit vehicle section binding time of the estimation traffic flow basic data at interval of going:

6-1) judge the consecutive miss time interval, if the time interval less than 15 minutes, then changes step 6-2 over to), otherwise change step 6-3 over to);

6-2) adopt the walk random method to estimate missing data;

6-3) if consecutive miss interval greater than equaling 15 minutes, smaller or equal to 30 minutes, then changes step 6-4 over to), otherwise change step 6-5 over to);

6-4) adopt the historical average real-time adjustment method estimation missing data that combines;

6-5) adopt the historical method of average to estimate missing data.

2. the preprocess method of a kind of Urban road dynamic traffic stream basic data according to claim 1 is characterized by:

In the step 1), certain hour is spaced apart the raw data acquisition time interval of fixed wagon detector; Dynamic traffic stream basic data is the general name of the magnitude of traffic flow, vehicle ' average velocity, three attribute variable's data of occupation rate averaging time;

Step 2-2) in, the date and time stamp misdata is defined as the data that the Data Date timestamp is not equal to effective value;

Step 2-3) in, repeating data was defined as with the traffic flow basic data attribute variable in a last raw data acquisition time interval and is worth identical data;

Said valid data and invalid data refer to effective traffic flow basic data and invalid traffic flow basic data, and traffic flow basic data attribute variable has consistent validity.

3. the preprocess method of a kind of Urban road dynamic traffic stream basic data according to claim 1 is characterized by step 2-5) in, the method for single-way traffic stream basic data threshold method check is following:

A) magnitude of traffic flow threshold test if surpass threshold range, then is labeled as invalid data, changes step 2-7 over to);

B) vehicle ' average velocity threshold test if surpass threshold range, then is labeled as invalid data, changes step 2-7 over to);

C) averaging time the occupation rate threshold test, if surpass threshold range, then be labeled as invalid data, change step 2-7 over to);

D) adjacent percentage speed variation threshold test if surpass threshold range, then is labeled as invalid data, changes step 2-7 over to), the wherein adjacent time interval (t-1, t) percentage speed variation τ _T-1, the computing formula of t is following:

In the formula, X _t---expression vehicle ' average velocity sequence, t=1,2,3 The object of adjacent percentage speed variation threshold test is a through street traffic flow basic data;

Above-mentioned steps a)～d) in, each threshold range need be demarcated based on the statistics of historical traffic flow basic data.

4. the preprocess method of a kind of Urban road dynamic traffic stream basic data according to claim 1 is characterized by step 2-6) in, the method for single-way traffic stream basic data traffic flow theory method check is following:

A) judge whether to satisfy that the vehicle ' average speed value equals zero and the traffic flow value is not equal to zero,, then be labeled as invalid data, change step 2-7 over to) if satisfy;

B) judge whether to satisfy that the traffic flow value equals zero and the vehicle ' average speed value is not equal to zero,, then be labeled as invalid data, change step 2-7 over to) if satisfy;

C) judge whether to satisfy traffic flow value and vehicle ' average speed value all equal zero and averaging time the occupation rate value be not equal to zero, if satisfy, then be labeled as invalid data, change step 2-7 over to);

D) judge whether to satisfy averaging time the occupation rate value be zero and the traffic flow value greater than maximum possible traffic value, if satisfy, then be labeled as invalid data, change step 2-7 over to), maximum possible volume of traffic q wherein _MaxComputing formula following:

In the formula, q _Max---the maximum possible volume of traffic (/ hour) when averaging time, occupation rate was zero in the expression raw data acquisition time interval, v---expression vehicle ' average velocity (km/hour), Δ _o---expression occupation rate precision averaging time (%), l---expression driving vehicle average effective vehicle commander (rice);

E) judge whether traffic flow density calculation value in track surpasses maximum traffic flow density, if satisfy, then is labeled as invalid data, changes step 2-7 over to), wherein the computing formula of track traffic flow density k is following:

k＝q/v

In the formula, k---expression track traffic flow density (/ track/km), q---the expression track magnitude of traffic flow (/ hour), v---expression vehicle ' average velocity (km/hour); Maximum traffic flow density need be demarcated based on the statistics of historical traffic flow basic data;

F) if averaging time, occupation rate was non-vanishing, judge that whether average effective vehicle commander calculated value surpasses certain threshold range, if satisfied, then this data recording of mark is invalid, changes step 2-7 over to), wherein the computing formula of average effective vehicle commander l is following:

In the formula, l---expression average effective vehicle commander (rice), O---expression occupation rate averaging time (%), v---expression vehicle ' average velocity (km/hour), q---the expression track magnitude of traffic flow (/ hour); Average effective vehicle commander's threshold range need be demarcated based on the statistics and the traffic flow formation of historical traffic flow basic data.

5. the preprocess method of a kind of Urban road dynamic traffic stream basic data according to claim 1 is characterized by in the step 3), can be defined as the integral multiple in any raw data acquisition time interval binding time at interval as required.

6. the preprocess method of a kind of Urban road dynamic traffic stream basic data according to claim 1; It is characterized by in the said step 4); Time is compiled and refers on the basis of validity check, and the divided lane traffic flow basic data in the raw data acquisition time interval is accumulated long period divided lane traffic flow basic data at interval, and concrete grammar is: successively the traffic flow basic data time of carrying out in each track is compiled; If do not exist valid data in the interval binding time; Then this track time of mark is compiled the result for invalid data, if there are valid data, then to compile computing formula following the single-way traffic stream basic data time:

In the formula, q---the expression long period is the inside lane magnitude of traffic flow (/ hour) at interval, q _i---effective magnitude of traffic flow (/ hour) in expression long period at interval interior i the raw data acquisition time interval; The valid data record number in n---expression long period in the at interval interior raw data acquisition time interval; N---the expression long period is the raw data acquisition time interval data record number of interior expectation at interval; V---expression long period interval inside lane vehicle ' average velocity (km/hour), v _i---effective vehicle ' average velocity in i the raw data acquisition time interval in the expression long period interval (km/hour), o---the expression long period is inside lane occupation rate averaging time (%) at interval, o _i---occupation rate effective time (%) in expression long period at interval interior i the raw data acquisition time interval.

7. the preprocess method of a kind of Urban road dynamic traffic stream basic data according to claim 1; It is characterized by in the said step 5); The space is compiled to refer in the time and is compiled on the basis; Divided lane traffic flow basic data is accumulated the traffic flow basic data of single unit vehicle travel direction section, and concrete grammar is following:

A) if the divided lane traffic flow basic data of carrying out compiling in the space is invalid data, then this single unit vehicle travel direction sectional space of mark compiles the result and is invalid data, if there are valid data, then changes b over to);

B) it is following to compile computing formula based on the divided lane traffic flow basic data space of valid data:

In the formula, q---the expression single unit vehicle travel direction section magnitude of traffic flow (/ hour), q _i---represent effective magnitude of traffic flow (/ hour) in i track; N---expression track valid data record number; N---the total number of track-lines of expression single unit vehicle travel direction section, v---expression single unit vehicle travel direction section vehicle ' average velocity (km/hour), v ⁱ---represent effective vehicle ' average velocity (km/hour) in i track, o---expression single unit vehicle travel direction section occupation rate averaging time (%), o _i---represent occupation rate effective time (%) in i track.

8. the preprocess method of a kind of Urban road traffic flow basic data according to claim 1 is characterized by said step 6-2) in, the walk random method estimates that the expression formula of missing data is:

In the formula, ---the estimated value of expression t traffic flow constantly basic data, X _T-1---the actual value of expression t-1 traffic flow constantly basic data;

Said step 6-4) in, the historical average expression formula that combines real-time adjustment method to estimate missing data is:

In the formula,

---represent d _iThe estimated value of it t traffic flow constantly basic data, ---represent d _iThe actual value of it t-1 traffic flow constantly basic data, With ---be expressed as d respectively _iThe history average of it t-1 and t traffic flow constantly basic data,

With

---represent d respectively _(i-j)The actual value of it t-1 and t traffic flow constantly basic data, N---expression set time length of window;

Said step 6-5) in, the historical method of average estimates that the expression formula of missing data is:

In the formula,

---represent d _iThe estimated value of it t traffic flow constantly basic data,

---represent d _(i-j)The actual value of it t traffic flow constantly basic data, N---expression set time length of window.

Images