CN101833858A - Method for judging road traffic state based on annular coil of signal lamp system - Google Patents

Abstract

The invention relates to a method for judging the road traffic state based on an annular coil of a signal lamp system. In the method, traffic historical data is acquired by using an annular coil detector and subjected to cluster analysis to obtain clustering centers corresponding to different traffic states; and a Euclidean distance between the real-time traffic data and each clustering center is directly calculated when the real-time traffic state is judged so as to judge that the current traffic state is the traffic state corresponding to the clustering center of which the Euclidean distance is the shortest.

Description

A kind of method of differentiating road traffic state based on annular coil of signal lamp system

Technical field

The present invention relates to intelligent transportation monitoring field, be specifically related to a kind of method of differentiating road traffic state based on annular coil of signal lamp system.

Background technology

What the lamp system in China city still adopted at present mostly is to be the intelligence control system in information acquisition source with the toroid winding detecting device, toroidal installation site be generally the intersection apart from 20 meters of stop line.For the toroid winding of lamp system service is laid as shown in Figure 1.

The traffic characteristic parameter of coil collection comprises flow, speed and occupation rate.Because the change of traffic behavior can cause the change of traffic flow essential characteristic parameter, as speed under the congestion state obviously reduce, occupation rate increase etc.Changing traffic flow basic parameter is obtained by the toroid winding collection, after then the certain algorithm of these parameter utilizations being handled, just can realize the differentiation to road traffic state.

Coil data processing algorithm commonly used at present has California algorithm, McMaster algorithm, exponential smoothing and standard deviation method.The California algorithm utilizes upstream detection cross section occupation rate to increase, and detected downstream cross section occupation rate descends and differentiates crowdedly, but this algorithm can't be applied to urban road and rate of false alarm is higher.The McMaster algorithm adds the humorous fluid uncommon 13. that accounts for of new suitable bitter edible plant whetstone coltfoal with the flow Xun

Condyle Zuo occupation rate X-Y scheme is divided into four zones, a kind of traffic behavior of each Regional Representative, and this algorithm dictates is in three continuous sampling periods, the speed of a motor vehicle is reduced to below the threshold value, or occupation rate surpasses threshold value, or flow and occupation rate are then judged and crowded all outside non-congested area; In continuous two sampling periods, any two threshold values that surpass separately of the speed of a motor vehicle, flow and occupation rate judge then and crowd that the shortcoming of this algorithm is the definite very difficult of threshold value, in case definite improper, will cause bigger rate of false alarm.Exponential smoothing is earlier original traffic data to be carried out smoothly, removes the traffic of short-term and disturbs, and then data processed and pre-set threshold is compared, judge crowded, this algorithm is easy to use, but the noise reduction of exponential smoothing is limited, still has higher rate of false alarm.The arithmetic mean of the traffic parameter data in n sampling period is as the predicted value of traffic parameter at moment t before the standard deviation method utilization moment t, measure the change degree of traffic parameter with the standard normal deviation again with respect to its former mean value, when deviation surpasses pre-set threshold, then think sporadic traffic congestion has taken place, this algorithm uses also easier, but do not see the variation tendency of traffic variable, error is very big.

The algorithm major part that above-mentioned parameter based on the coil collection is carried out traffic state judging belongs to space-time scenography category.In these traffic state judging algorithms, used the standard of comparison of different traffic behavior classification results respectively as algorithm.The common defective of these algorithms just is that the traffic behavior criteria for classification determines, can't change according to the variation of the magnanimity traffic flow essential characteristic parameter that collects.

The traffic characteristics of urban road is constantly to change in time and change, quantitative change is not only all taking place in the essential characteristic of traffic flow all the time, and in a quite long period, the qualitative change that the quantitative change meeting of continuous accumulation brings an integral body of whole traffic system traffic flow character.Therefore want to improve the algorithm accuracy rate, need carry out regular revision, thereby guarantee coincideing of this state classification standard and actual conditions, for the judgement of traffic behavior provides reliable comparison foundation to the traffic behavior criteria for classification.

And urban road is influenced by signal lamp, and coil detects data the obvious periodic variation, and the judgement of traffic behavior also must be considered the influence of lamp system signal period, still lacks this respect research in the existing algorithm.

Summary of the invention

The present invention proposes a kind of automatic traffic behavior decision method of taking into account accuracy and real-time, and it is specially a kind of method based on annular coil of signal lamp system differentiation road traffic state, comprising:

A, by occupation rate and track average velocity in flow, track cycle in the toroid winding detector acquisition track cycle;

B, with the track in the cycle flow, track in the cycle occupation rate and track average velocity be scaled section average velocity, section average occupancy and bicycle average occupancy respectively, through after misdata reparation and the data smoothing processing, deposit section average velocity, section average occupancy and bicycle average occupancy in database as historical data again;

C, after the historical data in the accumulation predetermined period, at first to the data normalization of predetermined period, with three kinds of different raw data be converted into no unit and the identical data of the order of magnitude, to find the solution cluster centre based on the data after the standardization again, so that the Weighted distance quadratic sum of each sample to three cluster centre reaches minimum, three cluster centres are corresponding unimpeded respectively, crowded, three class traffic behaviors block up, according to the variable standardization formula cluster centre is carried out reverse computing, obtain three kinds of traffic behaviors by section average velocity, the section average occupancy, the cluster centre vector that the bicycle average occupancy is formed, and judge respectively corresponding which kind of traffic behavior of each vector based on vector;

D, current real time traffic data is carried out after the pre-service, calculate Euclidean distance respectively with described three cluster centres, obtain the cluster centre with current real time traffic data Euclidean distance minimum, judge the traffic behavior of current traffic behavior for the cluster centre correspondence of this Euclidean distance minimum;

E, the current traffic behavior that will judge are reflected on the electronic chart with colour code intuitively.

Wherein, finding the solution cluster centre among the step C comprises the steps:

C11, initialization are according to the given exponential factor m of result that parameter is determined, given iteration cut-off error ε＞0, algorithm maximum iteration time T_MaxAnd initial cluster center V0={v_1,0, v_2,0, v_3,0;

C12, calculating U_t=[u_{Ik, t}], wherein, u_IkBe the degree of membership of k sample to i cluster centre; Make U_t=[u_{Ik, t}], wherein, d_IkBe k sample and i distances of clustering centers; If d_Ik=0, u then_{Ik, t}=1, and to j ≠ i, u_{Ik, t}=0,1≤i≤3,1≤k≤n; If d_{Ik, t}＞0, then

$u_{\mathrm{ik},t}={\left(\underset{r=1}{\overset{3}{\mathrm{\Σ}}}\frac{{\left(d_{\mathrm{ik},t}\right)}^{1/(m-1)}}{{\left(d_{\mathrm{jk},t}\right)}^{1/(m-1)}}\right)}^{-1},$1≤i≤3，1≤k≤n；

C13, calculating V_T+1=[v_{1, t+1}, v_{2, t+1}, v_{3, t+1}],

$v_{i,t+1}=\frac{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\left(u_{\mathrm{ik},t}\right)}^m{x}_k}{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\left(u_{\mathrm{ik},t}\right)}^m},$1≤i≤3；

C14, judgement, if$\left|\right|V_{t+1}-V_t\left|\right|=\underset{i=1}{\overset{3}{\mathrm{\&Sigma;}}}\left|\right|v_{i,t+1}-v_{i,t}\left|\right|<\mathrm{\&epsiv;},$Or t reaches maximum iteration time T_Max, termination of iterations then; Otherwise, make t=t+1, get back to C11;

C15, the V when iteration stops are the cluster centre of asking.

Further, determine among the step C11 that the m value comprises the steps:

C111, the historical data sample X that collects according to detecting device calculate λ_Max[C_x];

C112: if λ_Max[C_x]＜0.5, then

If

λ_Max[C_x] 〉=0.5, then m ∈ [1.5,2.5];

C113: with step-length is to carry out progressive optimizing between the m location among 0.1 couple of step2, calculates the distance between the corresponding all kinds of cluster centres of different m values, seeks the maximum m value of distance between all kinds of cluster centres that make classification.

Wherein, the computing formula of real-time traffic characteristic parameter and cluster centre Euclidean distance is among the step D:

$\left|\right|x-v_i\left|\right|=\sqrt{\underset{j=1}{\overset{3}{\mathrm{\&Sigma;}}}{(x_j-v_{\mathrm{ij}})}^2}$

Wherein: x represents through pretreated real-time traffic parameter vector; v_iRepresent i cluster centre vector; x_j, v_IjRepresent j parameter in real-time traffic parameter vector and the cluster centre vector respectively, j=1,2,3 represents section average velocity, section average occupancy and bicycle average occupancy respectively.

Simultaneously, step D also comprises: have only the traffic behavior in continuous two cycles all inconsistent with original traffic behavior, and these two cycle traffic behavior unanimities, the just traffic behavior of change output.

The present invention also proposes a kind of device based on annular coil of signal lamp system differentiation road traffic state, it is characterized in that comprising:

Data acquisition module is used for by toroid winding detector acquisition track flow, track occupation rate and track average velocity in the cycle in the cycle;

Data preprocessing module, be used for the track in the cycle flow, track in the cycle occupation rate and track average velocity be scaled section average velocity, section average occupancy and bicycle average occupancy respectively, through after misdata reparation and the data smoothing processing, deposit section average velocity, section average occupancy and bicycle average occupancy in database as historical data again;

Historical data cluster analysis module, be used for after the historical data in the accumulation predetermined period, at first to the data normalization of predetermined period, with three kinds of different raw data be converted into no unit and the identical data of the order of magnitude, to find the solution cluster centre based on the data after the standardization again, so that the Weighted distance quadratic sum of each sample to three cluster centre reaches minimum, three cluster centres are corresponding unimpeded respectively, crowded, three class traffic behaviors block up, according to the variable standardization formula cluster centre is carried out reverse computing, obtain three kinds of traffic behaviors by section average velocity, the section average occupancy, the cluster centre vector that the bicycle average occupancy is formed, and judge respectively corresponding which kind of traffic behavior of each vector based on vector;

The road traffic state discrimination module, be used for current real time traffic data is carried out after the pre-service, calculate Euclidean distance respectively with described three cluster centres, obtain the cluster centre with current real time traffic data Euclidean distance minimum, judge the traffic behavior of current traffic behavior for the cluster centre correspondence of this Euclidean distance minimum;

The road traffic state display module is used for the current traffic behavior that will judge and is reflected to electronic chart with colour code intuitively.

The present invention is from existing annular coil of signal lamp system data, selection can reflect the characteristic parameter of traffic behavior, consider signal lamp cycle, after raw data is carried out pre-service, auto-correlation pattern in green light cycle inner analysis coil image data, realize the automatic revision of traffic state judging standard, thereby determine road traffic state comparatively accurately.

Above-mentioned and other purpose, feature and advantage of the present invention will become clear and definite for those skilled in the art after the detailed description below having read in conjunction with the accompanying drawing that shows and describe specific embodiments of the invention.

Description of drawings

Fig. 1 is that the toroid winding of lamp system service is laid synoptic diagram;

Fig. 2 is automatic traffic state judging system chart;

Fig. 3 is a historical data cluster analysis process flow diagram;

Fig. 4 is that road traffic state is differentiated process flow diagram.

Embodiment

The enforcement of the embodiment of the invention needs the historical data accumulation of certain hour, is that basic cluster goes out rational discrimination standard with historical data.

With time is horizontal ordinate (time shaft need comprise the crowded period), historical data with each traffic characteristic parameter of annular coil of signal lamp system collection is the ordinate mapping, can compare, crowded generation, lasting reaching in the evanishment, speed that the annular coil of signal lamp system detecting device collects and occupation rate parameter are all stronger to crowded time sensitivity and degree susceptibility, the susceptibility of flow then obviously a little less than.But occupation rate is divided by flow, and promptly the bicycle average occupancy then changes comparatively obvious.Therefore speed, occupation rate, bicycle average occupancy are the most suitable traffic characteristic parameters.Therefore, the present invention chooses the foundation that these parameters are used as judging traffic.

As shown in Figure 2, system administers and maintains module, Dynamic Data Acquiring module, dynamic data pretreatment module, data normalization module, historical data update module, historical data cluster analysis module, road traffic state discrimination module, road junction traffic behavior display module, data memory module by static data and constitutes.

Wherein static data is meant that road network geometric configuration, network topology structure etc. are long-term and changes little data, and relatively stable attribute datas such as highway section number of track-lines, grade.These data need be spent the bigger time initialized the time, need to carry out traffic study in case of necessity.The data structure of the highway section data that need safeguard is as follows:

Highway section between adjacent two crossings is except that the one-way road is single highway section, and the highway section of different trends numbering is different, and the start-stop node is opposite.

System's parameter conversion, misdata reparation, smoothing processing through the dynamic data pretreatment module after collecting the real-time green light cycle parameter data of annular coil of signal lamp system detecting device stores database afterwards into.

The pretreated idiographic flow of data is as follows.

A, valid data time set

Consider the laying place of annular coil of signal lamp system and the periodic characteristic of urban highway traffic, think valid data of data in the green time, because lamp system and toroid winding link, therefore can directly in system, set, only allow toroid winding return the interior detection data of green time.

B, input parameter convert

What the annular coil of signal lamp system detecting device provided is occupation rate and track average velocity in flow in the cycle of track, track cycle, so need further be scaled section average velocity, section average occupancy and bicycle average occupancy.

The section flow is:

$q_s=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{q}_i$

Section average velocity is:

$v_s=\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{v}_i{q}_i}{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{q}_i}$

The section average occupancy is:

$o_s=\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{o}_i}{n}$

The bicycle average occupancy is:

$o_e=\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{o}_i}{q_s}$

In the formula: q_i, o_i, v_iRepresent the magnitude of traffic flow, time occupancy and average velocity in each lane detection cycle respectively; N represents number of track-lines.

C, misdata reparation

Random disturbance is more in the urban road, and misdata might appear in the raw data that directly collects by the toroid winding detecting device, if these data are directly used in the differentiation of traffic behavior, can influence the result of traffic behavior greatly.

Therefore can determine the last lower threshold value of section flow, section average occupancy and bicycle average occupancy according to the actual conditions of every road, thereby determine a valid interval,, then think misdata in case exceed this interval.Need utilize the weighted estimation of historical data and measured data to repair to misdata, its computing formula is:

$x_{\left(t\right)}^{\mathrm{new}}={\mathrm{\&alpha;x}}_{(t-1)}+(1-\mathrm{\&alpha;})x_{\left(t\right)}^{\mathrm{old}}$

In the formula: x_(t)^NewBe the data after t answers constantly; x_(t-1)It is t-1 measured data constantly; x_(t)^OldBe t historical data constantly, i.e. t the previous day data constantly; α is a weighting coefficient, has reacted t-1 measured data and t historical data role in data repair constantly constantly, and α is big more, and measured data is big more to the data influence after repairing, and vice versa.α can determine desired value according to actual measurement.

D, data smoothing are handled

Smoothing processing adopts the method for moving average, its objective is for the traffic of removing short-term to disturb, and concrete formula is as follows:

$x_{\left(t\right)}^{\&prime;}=\frac{1}{m}(x_{\left(t\right)}+x_{(t-1)}+\&CenterDot;\&CenterDot;\&CenterDot;+x_{(t-m+1)})$

In the formula: x '_(t)It is the moving average of t period.M is the data number that moving average calculation is got; The size of m is very big to the smooth effect influence, and m obtains little, and smooth curve is highly sensitive, but the poor performance of anti-random disturbance; M obtains very big, and it is good that anti-random disturbance gets performance, but sensitivity is low, and is insensitive to new variation tendency.So the selection of m is the key of the method for moving average.At concrete problem, when selecting m, should consider the length that how much reaches data collection cycle of object time series data.In general, m gets 3.

When data accumulation to certain fate, but just starting algorithm.The whole algorithm flow process of the cluster analysis module of historical data as shown in Figure 3.

At first historical data being carried out data normalization handles, because the unit and the order of magnitude of speed, occupation rate, three parameters self of bicycle average occupancy are all inequality, so need carry out the standardization computing to three kinds of original different pieces of informations, with its all be converted into no unit and the identical data of the order of magnitude import as variable, like this could better utilization the data digging method of cluster analysis study.

The data normalization utilization be the mean difference principle.For one group of data (x₁, x₂..., x_n), its standardization formula is as follows:

$\stackrel{\&OverBar;}{E}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}|\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i}{n}-x_i|$

$E_i=|\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{xi}}{n}-x_i|$

$y_i=\frac{E_i}{\stackrel{\&OverBar;}{E}}$

In the formula: E is a mean difference; E_iBe x_iDifference with respect to arithmetic mean; y_iBe x_iMetric after the standardization.

After standardization is finished, carry out the data clusters analysis by historical data cluster analysis module.Fuzzy cluster analysis

Historical data in N before adopting (N can according to the actual conditions value, general desirable 15) day is as the basis of cluster analysis, and these data all are through depositing historical data base in after the data pre-service certainly.Only need from historical data base, to transfer to get final product.

The target of algorithm is that historical data is divided into three groups automatically, so that the Weighted distance quadratic sum of each sample to three cluster centre reaches minimum.

A, model are constructed

With X={x₁, x₂..., x_nThe set of preceding N days historical data of expression annular coil of signal lamp system, n represents the number of samples in the historical data set.Now these historical datas to be divided into three groups automatically, three cluster centres are promptly arranged, corresponding respectively unimpeded, three class traffic behaviors crowd, block up.If these three cluster centre vectors are V={v₁, v₂, v₃.u_IkBe the degree of membership of k sample to i cluster centre, U=[u_Ik] be one 3 * n matrix.Then the objective function of fuzzy clustering algorithm is:

$\min J_m(U,V;X)=\stackrel{3}{\mathrm{\&Sigma;}}\stackrel{n}{\mathrm{\&Sigma;}}{\left({\mathrm{\&mu;}}_{\mathrm{ik}}\right)}^m{\left|\right|x_k-v_i\left|\right|}^2$

(1)

Constraint condition is:

$s.t.\left\{\begin{array}{c}{u}_{\mathrm{ik}}\&Element;\left[\mathrm{0,1}\right]\\ \underset{i=1}{\overset{3}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ik}}=1\\ 1\&le;i\&le;3\\ 1\&le;k\&le;n\end{array}\right.---\left(2\right)$

Jm is the weighted sum of squares objective function of error in the class, and m is the fuzzy index that can control cluster result, and (1, ∞), fuzzy index is big more, and the fog-level of cluster is just big more for m ∈.

|| x_k-v_i|| the sample x that expression is gathered_kWith the cluster centre vector v_iEuclidean distance.

$\left|\right|x_k-v_i\left|\right|=\sqrt{\underset{j=1}{\overset{3}{\mathrm{\&Sigma;}}}{(x_{\mathrm{kj}}-v_{\mathrm{ij}})}^2}$

Wherein j=1,2,3 represents 3 indexs in sample and the cluster centre vector respectively: section flow, section average occupancy and section average velocity.

The target of this algorithm is the minimal value of asking formula (1) under condition (2) constraint, and when trying to achieve minimum value, an optimum that just can obtain X is fuzzy to be divided

B, parameter are determined

The value of fuzzy exponent m has considerable influence to cluster result, and its selection depends on the statistic λ of the historical data sample X of annular coil of signal lamp system detecting device_Max[C_x].

$C_X=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\frac{[x_i-\stackrel{\&OverBar;}{x}]{[x_i-\stackrel{\&OverBar;}{x}]}^T}{n{\left|\right|x_i-\stackrel{\&OverBar;}{x}\left|\right|}^2},$||x_i-x||≠0

λ wherein_iBe Matrix C_xCharacteristic root, work as λ_Max[C_x]＜0.5 o'clock, then

$1\&le;m\&le;\frac{1}{1-2{\mathrm{\&lambda;}}_{\max }\left[C_X\right]}$

Value not like this, then the cluster result of fuzzy clustering algorithm is not ideal enough.Work as λ_Max[C_x] 〉=0.5 o'clock, then m＞1.According to a large amount of analysis of experimental data, the optimal selection interval [1.5,2.5] of m value.

This shows, determine that the step of m value is as follows:

Step1: the historical data sample X that collects according to detecting device calculates λ_Max[C_x];

Step2: if λ_Max[C_x]＜0.5, then

$1\&le;m\&le;\frac{1}{1-2{\mathrm{\&lambda;}}_{\max }\left[C_X\right]}$

If λ_Max[C_x] 〉=0.5 a m ∈ [1.5,2.5];

Step3: adopting heuristic, is to carry out progressive optimizing between the m location among 0.1 couple of step2 with step-length, calculates the distance between the corresponding all kinds of cluster centres of different m values, seeks the maximum m value of distance between all kinds of cluster centres that make classification.

C, find the solution cluster centre

Make J_mTo v_iAnd u_IkPartial derivative be 0, then can release the computing formula of cluster centre and degree of membership according to the objective function of fuzzy clustering.

The cluster centre computing formula is:

$v_i=\frac{\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}{{\mathrm{\&mu;}}_{\mathrm{ik}}}^m{x}_k}{\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}{{\mathrm{\&mu;}}_{\mathrm{ik}}}^m}---\left(3\right)$

The degree of membership computing formula is:

${\mathrm{\&mu;}}_{\mathrm{ik}}={\left(\underset{r=1}{\overset{3}{\mathrm{\&Sigma;}}}\frac{{\left|\right|y_k-v_i\left|\right|}^{2/(m-1)}}{{\left|\right|y_k-v_r\left|\right|}^{2/(m-1)}}\right)}^{-1}---\left(4\right)$

Then the solution procedure of cluster centre is as follows:

Step1: initialization, according to the given exponential factor m of result that parameter is determined, given iteration cut-off error ε＞0, algorithm maximum iteration time T_MaxAnd initial cluster center V0={v_1,0, v_2,0, v_3,0, v wherein_1,0, v_2,0, v_3,0Corresponding to cluster centre initial value unimpeded, the three class traffic behaviors that crowd, block up.

Step2: calculate U_t=[u_{Ik, t}], make d_{Ik, t}=|| x_k-v_i||²If, d_Ik=0, u then_{Ik, t}=1, and to j ≠ i, u_{Ik, t}=0,1≤i≤3,1≤k≤n; If d_{Ik, t}＞0, then

$u_{\mathrm{ik},t}={\left(\underset{r=1}{\overset{3}{\mathrm{\&Sigma;}}}\frac{{\left(d_{\mathrm{ik},t}\right)}^{1/(m-1)}}{{\left(d_{\mathrm{jk},t}\right)}^{1/(m-1)}}\right)}^{-1}$

1≤i≤3，1≤k≤n。

Step3: calculate V_T+1=[v_{1, t+1}, v_{2, t+1}, v_{3, t+1}],

$v_{i,t+1}=\frac{\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}{\left(u_{\mathrm{ik},t}\right)}^m{x}_k}{\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}{\left(u_{\mathrm{ik},t}\right)}^m},$1≤i≤3。

Step4: judge, if

$\left|\right|V_{t+1}-V_t\left|\right|=\underset{i=1}{\overset{3}{\mathrm{\&Sigma;}}}\left|\right|v_{i,t+1}-v_{i,t}\left|\right|<\mathrm{\&epsiv;},$Or t reaches maximum iteration time T_Max, termination of iterations then; Otherwise, make t=t+1, get back to the first step.

V when iteration stops is the cluster centre of asking.Because the input data are through the data behind the data normalization, so the cluster centre after the iteration end also is the dimensionless number certificate.Can carry out reverse computing to cluster centre according to the variable standardization formula, obtain the cluster centre vector of forming by section average velocity, section average occupancy, bicycle average occupancy of three kinds of traffic behaviors.

N days in the past historical data automatic Iterative goes out after the cluster centre, can this cluster centre be that standard is differentiated the road traffic state when the day before yesterday just.Road traffic state is differentiated flow process as shown in Figure 4.Only need cluster once every day, only real time data need be carried out after the pre-service during differentiation, calculates the Euclidean distance with three cluster centres respectively.The computing formula of real-time traffic characteristic parameter and cluster centre Euclidean distance is:

$\left|\right|x-v_i\left|\right|=\sqrt{\underset{j=1}{\overset{3}{\mathrm{\&Sigma;}}}{(x_j-v_{\mathrm{ij}})}^2}$

Wherein: x represents through pretreated real-time traffic parameter vector; v_iRepresent i cluster centre vector; x_j, v_IjRepresent j parameter in real-time traffic parameter vector and the cluster centre vector respectively, j=1,2,3 represents section average velocity, section average occupancy and bicycle average occupancy respectively.Apart from the minimum current traffic behavior that is.This method of discrimination calculates very easy, and operation time is extremely short, has guaranteed the real-time of algorithm.But unlikely too frequent in order to guarantee that traffic behavior changes, think that the traffic behavior that has only continuous two cycles is all inconsistent with original traffic behavior, and these two cycle traffic behavior unanimities, the traffic behavior of change output just now.

The road traffic state display module is responsible for the traffic behavior of the final output of road traffic state discrimination module is reflected on the electronic chart with colour code intuitively, and red expression is blocked up, and yellow expression is crowded, and green expression is unimpeded.

The traffic state judging algorithm that the embodiment of the invention proposes is applicable to urban road more, can realize the automatic revision of traffic state judging standard.

The above is preferred embodiment of the present invention only, is not limited to the present invention, all any modifications of being made within the present invention spirit and principle, is equal to replacement and improvement etc., all is contained within protection scope of the present invention.

Claims (6)

1. differentiate the method for road traffic state based on annular coil of signal lamp system for one kind, it is characterized in that comprising:

A, by occupation rate and track average velocity in flow, track cycle in the toroid winding detector acquisition track cycle;

B, with the track in the cycle flow, track in the cycle occupation rate and track average velocity be scaled section average velocity, section average occupancy and bicycle average occupancy respectively, through after misdata reparation and the data smoothing processing, deposit section average velocity, section average occupancy and bicycle average occupancy in database as historical data again;

C, after the historical data in the accumulation predetermined period, at first to the data normalization of predetermined period, with three kinds of different raw data be converted into no unit and the identical data of the order of magnitude, to find the solution cluster centre based on the data after the standardization again, so that the Weighted distance quadratic sum of each sample to three cluster centre reaches minimum, three cluster centres are corresponding unimpeded respectively, crowded, three class traffic behaviors block up, according to the variable standardization formula cluster centre is carried out reverse computing, obtain three kinds of traffic behaviors by section average velocity, the section average occupancy, the cluster centre vector that the bicycle average occupancy is formed;

D, current real time traffic data is carried out after the pre-service, calculate Euclidean distance respectively with described three cluster centres, obtain the cluster centre with current real time traffic data Euclidean distance minimum, judge the traffic behavior of current traffic behavior for the cluster centre correspondence of this Euclidean distance minimum;

E, the current traffic behavior that will judge are reflected on the electronic chart with colour code intuitively.

2. the method for claim 1 is characterized in that: find the solution cluster centre among the step C and comprise the steps:

C11, initialization are according to the given exponential factor m of result that parameter is determined, given iteration cut-off error ε＞0, algorithm maximum iteration time Tmax and initial cluster center V0={v1,0, v2,0, v3,0};

C12, calculating U_t=[u_{Ik, t}], wherein, uik is the degree of membership of k sample to i cluster centre; Make d_{Ik, t}=|| x_k-v_i||², wherein, dik is k sample and i distances of clustering centers; If dik=0, uik then, t=1, and to j ≠ i, uik, t=0,1≤i≤3,1≤k≤n; If dik, t＞0, then

1≤i≤3，1≤k≤n；

C13, calculating Vt+1=[v1, t+1, v2, t+1, v3, t+1],

1≤i≤3;

C14, judgement, if

Or t reaches maximum iteration time Tmax, then termination of iterations; Otherwise, make t=t+1, get back to C11;

C15, the V when iteration stops are the cluster centre of asking.

3. method as claimed in claim 2 is characterized in that: determine among the step C11 that the m value comprises the steps:

C111, the historical data sample X that collects according to detecting device calculate λ max[C];

C112: if λ is max[Cx]＜0.5, then

If λ is max[C] 〉=0.5, m ∈ [1.5,2.5] then;

C113: with step-length is to carry out progressive optimizing between the m location among 0.1 couple of step2, calculates the distance between the corresponding all kinds of cluster centres of different m values, seeks the maximum m value of distance between all kinds of cluster centres that make classification.

4. the method for claim 1, it is characterized in that: the computing formula of real-time traffic characteristic parameter and cluster centre Euclidean distance is:

$\left|\right|x-v_i\left|\right|=\sqrt{\underset{j=1}{\overset{3}{\mathrm{\&Sigma;}}}{(x_j-v_{\mathrm{ij}})}^2}$

Wherein: x represents through pretreated real-time traffic parameter vector; Vi represents i cluster centre vector; Xj, vij represent j parameter in real-time traffic parameter vector and the cluster centre vector respectively, and j=1,2,3 represents section average velocity, section average occupancy and bicycle average occupancy respectively.

5. the method for claim 1 is characterized in that, step D also comprises: have only the traffic behavior in continuous two cycles all inconsistent with original traffic behavior, and these two cycle traffic behavior unanimities, the just traffic behavior of change output.

6. differentiate the device of road traffic state based on annular coil of signal lamp system for one kind, it is characterized in that comprising:

Data acquisition module is used for by toroid winding detector acquisition track flow, track occupation rate and track average velocity in the cycle in the cycle;

Data preprocessing module, be used for the track in the cycle flow, track in the cycle occupation rate and track average velocity be scaled section average velocity, section average occupancy and bicycle average occupancy respectively, through after misdata reparation and the data smoothing processing, deposit section average velocity, section average occupancy and bicycle average occupancy in database as historical data again;

Historical data cluster analysis module, be used for after the historical data in the accumulation predetermined period, at first to the data normalization of predetermined period, with three kinds of different raw data be converted into no unit and the identical data of the order of magnitude, to find the solution cluster centre based on the data after the standardization again, so that the Weighted distance quadratic sum of each sample to three cluster centre reaches minimum, three cluster centres are corresponding unimpeded respectively, crowded, three class traffic behaviors block up, according to the variable standardization formula cluster centre is carried out reverse computing, obtain three kinds of traffic behaviors by section average velocity, the section average occupancy, the cluster centre vector that the bicycle average occupancy is formed, and judge respectively corresponding which kind of traffic behavior of each vector based on vector;

The road traffic state discrimination module, be used for current real time traffic data is carried out after the pre-service, calculate Euclidean distance respectively with described three cluster centres, obtain the cluster centre with current real time traffic data Euclidean distance minimum, judge the traffic behavior of current traffic behavior for the cluster centre correspondence of this Euclidean distance minimum;

The road traffic state display module is used for the current traffic behavior that will judge and is reflected to electronic chart with colour code intuitively.

Images