CN101188064A - 3D integrated freeway traffic event automatic detection method - Google Patents

Abstract

The invention discloses an automatic detection method of three-dimensional integrated-highway traffic incident, and has a fast speed and low rate of missing report. The invention has the technical proposal that whether alarming signals are sent out or not is determined by the analysis on the detected data form an expressway detector; and the detail steps are that firstly, the system can be initialized to read the driveway flow- the value of the speed valve, the driveway share- the value of the flow valve, driveway speed- the value of the share valve; secondly, the energy utilization sign can be started; thirdly, the real-time data of the traffic flow can be read, and data of all detectors can be read respectively; fourthly, a horizontal dimension, namely, the driveway flow-speed calculation module, a time dimension, namely, the driveway share-flow calculation module, a vertical dimensional detection module, namely, the driveway flow-speed calculation module, and a integrated judgment module can be used for judging based on the data of all detectors; finally, whether an alarming device is proposed to start to give an alarming signal is determined according to the judgment results.

Description

3 D integrated freeway traffic event automatic detection method

Technical field

The invention belongs to freeway traffic event detection method field, especially a kind of 3 D integrated freeway traffic time automatic testing method.Specifically through street that fixed traffic flow detecting device has been installed or highway being carried out traffic events detects automatically.

Background technology

Traffic events detects the important component part that (AID) is through street (highway) traffic administration and control system automatically, how to detect time, place and the order of severity that traffic events takes place accurately and rapidly, so that take active and effective rescue measure, traffic administration and control system had crucial meaning.

The event detecting method of using various technology has been arranged at present, substantially can be divided into method based on mode-matching technique, based on the method for statistical study, based on the method for traffic flow model with based on the method for artificial intelligence. these methods all are to be based upon on the data basis that the traffic flow detecting device gathers mostly, and employed basic input parameters is flow, speed and the occupation rate sequence etc. in time or space.

Typical mode identification method mainly contains: California algorithm and McMaster algorithm; Representational statistical prediction methods has: standard normal deviate method, bayes method etc.; Method based on traffic flow model has: Dynamic algorithm etc.; In recent years, some new methods have appearred, wherein methods such as neural network method and wavelet transformation again.The emphasis of traditional AID research only is the generation of identification incident, and the information relevant with event feature is not provided.California algorithm that has reasonable AID performance utilizes direct comparison method to obtain testing result.Yet, the common problem of relative method be used for event recognition initial value determine that these can limit the practical application of relative method.Utilizing the artificial neural network training data to carry out event detection has certain advantage, yet a large amount of event datas of neural network needs are trained and just can be made it have good transplantability, and this is difficult to realize in actual applications.

Detection sectional plane according to detection method is divided, and event detecting method can be divided into single method of section and double sections method.When the volume of traffic was smaller, vehicle was less to roadway occupancy, and the traffic parameter of adjacent detector variation difference was littler when incident took place, and was not suitable for adopting the double sections method, should adopt single method of section, and under the bigger situation of the volume of traffic, should adopt the double sections method.

Sum up existing event detecting method, the main problem that exists: (1) adopts single method of section, perhaps adopts the multi-section method, and both of these case all might take place in fact, should consider comprehensively; (2) each detection method can only be applicable to a kind of situation.Only adopt a kind of system robustness of detection method not high, can adopt the integrated of a plurality of detection methods, to improve the precision and the robustness of system; (3) traffic flow model and the threshold value in the existing method generally all is preset parameter, and in fact these parameters change along with time period difference, link characteristics and vehicle feature, should make its dynamic change, change to reflect nearest traffic flow; (4) existing method is as based on only relying on traffic flow parameter of occupation rate in the method for California algorithm, and the McMaster algorithm has been considered occupation rate and flow.But the occupation rate height may be congested in traffic performance, also might be the situation that cart passes through; Speed is low might to be congested in traffic, also may be the very careful situation of driver; The traffic flow flow is low may to be that traffic flow is unimpeded, also might be the very crowded situation of traffic.So, must could be to the traffic flow comprehensive test in conjunction with its dependent variable; (5) variation of existing method traffic flow parameter between same section adjacent lane when nearly all not consideration incident takes place, and in fact this variation is tangible.

Summary of the invention

The purpose of this invention is to provide a kind of 3 D integrated freeway traffic event automatic detection method quick, that rate of failing to report is low that detects.

Technical scheme of the present invention is: 3 D integrated freeway traffic event automatic detection method, it is characterized in that judging whether to send alerting signal according to the analysis that the highway detecting device is detected data, and specifically comprise the following steps:

----system initialization: read track flow-threshold speed, lane occupancy ratio-flow threshold, track speed-occupation rate threshold value;

----starts enabler flags: the automatic measuring ability of start-up routine, and the manual or undated parameter automatically when needs, " detection of end " button zero clearing sign by the interface finishes automatic measuring ability, treats that parameter update finishes, and can restart;

----reads the traffic flow real time data, reads all detector data respectively;

-----call horizontal dimension is that track flow-rate algorithm module, time dimension are lane occupancy ratio-flow algoritic module, vertically to tie up detection module be track speed-occupation rate algoritic module and integrated judge module, to whole monitor data is that the basis is judged, makes according to judged result whether starting warning device and sending alerting signal.

Described horizontal dimension is that track flow-rate algorithm module specifically comprises the following steps: to add up with one month historical data, obtains average discharge ratio and the velocity rate of each period, and then the threshold value of the throughput ratio in each track and velocity ratio is made as:

Volume_Limit _i＝max(0，mean(VOLUMN_RITO _i)-2*std(VOLUMN_RITO _i))；(i＝1，2，3，4)

Speed_Limit _i＝max(0，mean(SPEED_RITO _i)-2*std(SPEED_RITO _i))；(i＝1，2，3，4)

Mean represents to average, and std represents to ask variance; When certain track flow and speed drop to certain threshold value, show that event occurs, by calculating the velocity rate and the flow rate ratio in each track of detecting device section, and compare with lower velocity limit that configures and flux lower limit parameter, come the detection incident; The computing formula of flow and velocity rate is as follows:

$\mathrm{VOLUME}\_{\mathrm{RITO}}_i=\frac{{\mathrm{VOLUME}}_i}{\mathrm{\Σ}{\mathrm{VOLUME}}_i}(i=\mathrm{1,2,3,4});$

$\mathrm{SPEED}\_{\mathrm{RITO}}_i=\frac{{\mathrm{SPEED}}_i}{{\mathrm{\ΣSPEED}}_i}(i=\mathrm{1,2,3,4});$

VOLUME_RITO wherein _iBe the flow rate ratio in certain track, VOLUME _iBe the flow in certain track, module settings is for when certain track flow and speed drop to certain threshold value twice continuously, and the affair alarm request is sent in this track.

Described time dimension is that lane occupancy ratio-flow algoritic module comprises the following steps: that specifically traffic flow model is:

$f=k*v_f{(1-{\left(\frac{o}{o_j}\right)}^r)}^m*o;$

In the formula, o is an occupation rate, and f is a flow, v _fThe expression free stream velocity, o _jOccupation rate is blocked in expression, and r, m are constants, v _f, o _j, r, m are parameters undetermined, according to the real road situation, the variation range of these four parameters is carried out certain limitation, be translated into a constrained optimization problem, the objective function of model optimization is the absolute error sum, and this value has reflected the size of error comprehensively, is shown below

$\mathrm{SAE}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}|f\left(i\right)-f^1\left(i\right)|$

Wherein, f (i) is the flow value that obtains according to traffic flow model, f ¹(i) be the flow value of actual measurement; In constrained optimization problem, employing is found the solution based on K-T (knhn-Tucker) equation], occupation rate according to continuous four cycles draws corresponding grade of living in and LevChange (t, t-1, i) (the grade that expression t jumps with respect to t-1 moment occupation rate constantly, i represents the track), judge whether to satisfy one of following three conditions then:

If after satisfying, then the affair alarm request is sent in this track.

Described vertical dimension detection module is that track speed-occupation rate algoritic module comprises the following steps: that specifically the concrete steps that method is implemented are: use the central value that fuzzy clustering method provides the cluster of a large amount of traffic flow speed-occupation rate historical data; Obtain the Euclidean distance at traffic flow speed, occupation rate data and four centers of real-time upstream and downstream respectively, with which center recently (Euclidean distance minimum) be exactly under grade accordingly; Judge the rank difference of upstream and downstream, when superfine continuous two cycles of upstream and downstream traffic behavior differ 3 ranks, show that event occurs, then send the affair alarm request.

The synthetic integrated rule of described integrated judge module is: does not just report to the police in the t cycle so if certain detecting device was reported to the police in the t-1 cycle (1), thinks same incident; (2) sometime, reported to the police as if certain detecting device, just do not report to the police in its downstream so, thinks same incident; (3) if certain detecting device was reported to the police in the t-1 cycle, just do not report to the police in its downstream of t cycle so, think same incident; (4) if the crossing is arranged in the middle of the detecting device in highway section, through street, then do not report to the police or the threshold range increasing in the highway section so; (5) if in the middle of the detecting device in highway section, through street import and export are arranged, kerb lane does not report to the police so.Traffic flow (comprising flow, speed, occupation rate (density)) is an aleatory variable that changes along with two-dimensional space (adjacent lane and upstream and downstream track) and one dimension time, and is interrelated between three variablees of traffic flow, determines the feature of traffic flow jointly.When incident took place in highway a certain highway section, the case point traffic capacity descended immediately.If drop to when being lower than transport need, traffic flow is affected, and produces sporadic traffic congestion, the traffic flow abnormality will occur in the certain limit in downstream, the site of an accident: the upstream vehicle is obstructed because of traffic and slows down, and the downstream vehicle is rare and quicken; The magnitude of traffic flow of meeting accident on the track reduces, and adjacent lane increases because of traffic flow merging vehicle flowrate; The time occupancy of upstream road increases, and the time occupancy in downstream reduces or the like (as Fig. 1).In a word, when traffic abnormity takes place, traffic flow parameter within the specific limits will be bigger than normal or less than normal than normal value.

Also comprise the following steps: alarm event information according to historical data, draw the performance index that comprise verification and measurement ratio and rate of false alarm, with manual or upgrade the key parameter of three detection modules automatically, make event detecting method can closer reflect the variation of traffic flow, the undated parameter rule is as follows: (1) utilizes up-to-date historical data and statistical method to upgrade automatically and laterally ties up is the parameter of track flow-rate algorithm module, in certain cycle of being separated by, the data and the true traffic events of having reported to the police are added up, draw verification and measurement ratio and rate of false alarm, adjust univers parameter and be in reasonable levels to make rate of false alarm and alarm rate as far as possible; (2) upgrade traffic flow model automatically according to one month up-to-date historical data, time dimension is that three threshold values of lane occupancy ratio-flow algoritic module obtain upgrading automatically, historical data input traffic flow model optimizer with up-to-date one month directly obtains three up-to-date threshold values; (3) carrying out the vertical dimension of the automatic renewal of fuzzy clustering detection module again according to one month up-to-date historical data is track speed-4 cluster centres of occupation rate algoritic module, be exactly with one month up-to-date historical speed and occupation rate data input fuzzy clustering program, directly obtain 4 up-to-date fuzzy clustering centers.

Effect of the present invention is: existing variety of event automatic testing method is mostly considered a kind of variable, perhaps considers the variation of traffic flow on a dimension, will inevitably cause rate of failing to report height and detection time longer like this.The three-dimensional integration events automatic testing method of the present invention, the technical matters that is solved is for take all factors into consideration the contact between the traffic flow three elements matter comprehensively, respectively from flow-speed, flow-occupation rate, three aspects of speed-occupation rate detect, the last integrated event automatic detection method of incident can reduce rate of failing to report and accelerate detection time.

The present invention is described further below in conjunction with drawings and Examples.

Description of drawings

Fig. 1 is a hardware block diagram of the present invention;

Fig. 2 is a program flow chart of the present invention;

Fig. 3 is program principle figure of the present invention;

Fig. 4 is that time dimension detects occupation rate-discharge model grade classification figure;

Fig. 5 is integrated judge module process flow diagram;

Fig. 6 is that module parameter upgrades FB(flow block);

Fig. 7 laterally dimension is track flow-rate algorithm module parameter flow chart.

Embodiment

As shown in Figure 1, hardware components of the present invention comprises monitor, computing machine and warning bump, computing machine to the detector data analysis that is provided with on the highway after, determine to start warning device and do not send alerting signal.

As shown in Figure 2,3 D integrated freeway traffic event automatic detection method, it is characterized in that according to the analysis that the highway detecting device is detected data, judge whether to send alerting signal, specifically comprise the following steps: system initialization: read track flow-threshold speed, lane occupancy ratio-flow threshold, track speed-occupation rate threshold value; Start enabler flags; Read the traffic flow real time data, read all detector data respectively; To call horizontal dimension and be track flow-rate algorithm module, time dimension be lane occupancy ratio-flow algoritic module, vertically tie up detection module is track speed-occupation rate algoritic module and integrated judge module, to whole monitor data is that the basis is judged, makes according to judged result whether starting warning device and sending alerting signal.

According to shown in Figure 3, introduce each module and principle of work of the present invention in detail.

1, submodule is described

(1) laterally ties up (adjacent lane) detection module

The traffic flow difference of same section adjacent lane is little under the normal condition, and when certain track traffic events took place, the magnitude of traffic flow in this track descended, speed descends.So, when certain track flow and speed drop to certain threshold value, show that event occurs, applied statistical method is provided the initial value of this threshold value; By calculating the velocity rate and the flow rate ratio in each track of detecting device section, and compare, come the detection incident with lower velocity limit that configures and flux lower limit parameter.The computing formula of flow and velocity rate is as follows:

$\mathrm{VOLUME}\_{\mathrm{RITO}}_i=\frac{{\mathrm{VOLUME}}_i}{{\mathrm{\ΣVOLUME}}_i}(i=\mathrm{1,2,3,4})---\left(1\right)$

$\mathrm{SPEED}\_{\mathrm{RITO}}_i=\frac{{\mathrm{SPEED}}_i}{{\mathrm{\ΣSPEED}}_i}(i=\mathrm{1,2,3,4})---\left(2\right)$

VOLUME_RITO wherein _iBe the flow rate ratio in certain track, VOLUME _iFlow for certain track.

After tested, and for avoiding detector data instantaneous unusual, this module settings shows the incident generation for when certain track flow and speed drop to certain threshold value twice continuously.

Laterally the ultimate principle of dimension (adjacent lane) detection module is that nearest one month historical data is added up, and obtains average discharge ratio and the velocity rate of each period, and then the throughput ratio in each track and speed are made as than the initial value of lower limit

Volume_Limit _i＝max(0，mean(VOLUMN_RITO _i)-2*std(VOLUMN_RITO _i))；(i＝1，2，3，4)(3)

Speed_Limit _i＝max(0，mean(SPEED_RITO _i)-2*std(SPEED_RITO _i))；(i＝1，2，3，4)(4)

Mean represents to average, and std represents to ask variance, with 0 to ask maximum be for fear of the situation that negative occurs.

Throughput ratio and velocity ratio when certain track, smaller or equal to the throughput ratio lower limit and the velocity ratio lower limit in this track, affair alarm is sent in this track simultaneously.

(2) time dimension (time variation) detection module

Under the normal condition, it is continuous that the traffic flow in the same track of same section changes.Under the situation that incident takes place, this variation is jumped.Occupation rate-discharge model is the important models of reflection traffic flow variation characteristic.The utilization of this module is divided into different 4 grades with traffic flow with different threshold values based on occupation rate-discharge model of optimizing, as shown in Figure 4, will come the generation of decision event with the variation size of grade under traffic flow modes is on time dimension.

The effect of module two needed three threshold value O1, O2, O3 is a grade of judging occupation rate with this, be traffic flow model, historical data training according to occupation rate and flow and the theoretical gained of constrained optimization, wherein, O2 is the occupation rate when accounting for the flow maximum, and O1 and O3 represent that respectively flow is peak flow values one a half occupation rate.

According to traffic flow model and optimization method, obtain o1, o2 and o3, the traffic flow model that the present invention adopts is:

$f=k*v_f{(1-{\left(\frac{o}{o_j}\right)}^r)}^m*o---\left(5\right)$

(5) in the formula, o is an occupation rate, and f is a flow, v _fThe expression free stream velocity, o _jOccupation rate is blocked in expression, and r, m are constants.v _f, o _j, r, m are parameters undetermined.According to the real road situation, the variation range of these four parameters is carried out certain limitation, be translated into a constrained optimization problem.

The objective function of model optimization is absolute error sum (sum of absolute error), and this value has reflected the size of error comprehensively, as the formula (6).

$\mathrm{SAE}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}|f\left(i\right)-f^1\left(i\right)|---\left(6\right)$

F (i) is the flow value that obtains according to traffic flow model, f ¹(i) be the flow value of actual measurement.

In constrained optimization problem, adopt and find the solution based on K-T (knhn-Tucker) equation.The K-T equation separate the basis that has formed many Nonlinear Programming Algorithm.These algorithms directly calculate Lagrange multiplier.Use the quasi-Newton method renewal process, give K-T equation accumulation second order information, can guarantee the superlinear convergence of constraint quasi-Newton method.These methods are called Sequential Quadratic Programming method (SQP).For given planning problem, the main thought of seqential quadratic programming (SQP) is the quadratic programming subproblem that forms based on the Lagrangian function second approximation.The realization of SQP method generally was divided into for 3 steps, and promptly the renewal of Lagrangian function Hess matrix, quadratic programming problem are found the solution the calculating with linear search and objective function.

According to the occupation rate in continuous four cycles draw corresponding grade of living in and LevChange (t, t-1 i) (with respect to the t-1 occupation rate grade of jumping constantly, i represents the track to expression t constantly), judge whether to satisfy one of following three conditions then:

If after satisfying, then report to the police.

(3) vertically tie up (upstream and downstream track) detection module

Because under the normal condition, the traffic flow difference of identical track upstream and downstream is little, when certain track traffic events takes place, the traffic flow modes of its upstream detector worsens, and occupation rate increases, and speed reduces, the traffic behavior of downstream detector improves, and occupation rate reduces, and speed improves.Avoid California algorithm only to utilize the drawback of occupation rate and in conjunction with the fuzzy behaviour of traffic flow modes, this module will utilize fuzzy clustering that traffic flow speed-occupation rate model is divided into four grades, will come the generation of decision event according to rank difference size under the upstream and downstream traffic flow modes.

Use the central value that fuzzy clustering method provides the cluster of a large amount of traffic flow speed-occupation rate historical data, be followed successively by the central value of 1 grade, 2 grades, 3 grades and 4 grades according to occupation rate order from small to large.

The notion of fuzzy clustering is proposed by Ruspini the earliest.Fuzzy c average (Fuzzy C-means abbreviates FCM as) cluster is a kind of more typical fuzzy clustering algorithm, is proposed in 1981 by Bezdek, is used for the data point of multidimensional data space distribution is divided into the class of given number.In fuzzy clustering, each data point is to belong to a certain class to a certain degree, and it represents that with degree of membership each data point belongs to the degree of certain cluster.FCM is n vector x _i(i=1,2 ... ..n) be divided into c ambiguity group, and ask every group cluster centre, make the objective function of non-similarity index reach minimum.Specific algorithm is described below:

If objective function is as the formula (8):

$J(U,{\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="S2007101799774E00041.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,...c_c)=\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}{J}_i=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{c}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}^m{d}_{\mathrm{ij}}^2---\left(8\right)$

$u_{\mathrm{ij}}=\frac{1}{\underset{k=1}{\overset{c}{\mathrm{\&Sigma;}}}{\left(\frac{d_{\mathrm{ij}}}{d_{\mathrm{kj}}}\right)}^{2/(m-1)}}---\left(9\right)$

$\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}=1,\&ForAll;j=1,....,n---\left(10\right)$

U wherein _Ij∈ [0,1] represents that j data point belongs to the degree of membership of i cluster centre; c _iBe the cluster centre of ambiguity group i, d _Ij=‖ c _i-c _j‖ represents the Euclidean distance between i cluster centre and j data point; M ∈ [1, be a weighted index ∞), the present invention is made as 2.

The structure Lagrange multiplier, set up new objective function as the formula (11):

$J(U,{\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="S2007101799774E00041.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,...c_c,{\mathrm{\&lambda;}}_1,...,{\mathrm{\&lambda;}}_n)=J(U,{\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="S2007101799774E00041.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,...,c_c)+\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_i(\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}-1)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{c}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}^m{d}_{\mathrm{ij}}^2+\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_i(\underset{i=1}{\overset{c}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}-1)---\left(11\right)$

To all input parameter differentiates, the necessary condition that makes former objective function reach minimum is:

$c_i=\frac{\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}^m{x}_j}{\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{u}_{\mathrm{ij}}^m}---\left(12\right)$

The iterative process of FCM algorithm is as follows:

Step 1: random initializtion c data cluster centre.

Step 2: calculate the U battle array with formula (9).

Step 3: calculate c new cluster centre c with formula (12) _i, i=1 ..., c.

Step 4: according to formula (11) calculating target function, if less than certain threshold value of determining, or objective function change last time amount is less than certain threshold value relatively, and then algorithm stops, otherwise, return step 2.

The concrete steps that method is implemented are: use the central value that fuzzy clustering method provides the cluster of a large amount of traffic flow speed-occupation rate historical data; Obtain the Euclidean distance at traffic flow speed, occupation rate data and four centers of real-time upstream and downstream respectively, with which center recently (Euclidean distance minimum) be exactly under grade accordingly; Judge the rank difference of upstream and downstream, when superfine continuous two cycles of upstream and downstream traffic behavior differ 3 ranks, show that event occurs, then report to the police.

Integrated judge module

This method is reducing omission and might cause repetition of alarms in detection time because consider comprehensively comprehensively.So must research integrated between the three-dimensional method.When taking place according to traffic events to the influence of three-dimensional traffic state, the geometric properties in the highway section of studying, and the experience of traffic administration, following several the synthetic integrated rules of this method regulation:

(1) if certain detecting device was reported to the police in the t-1 cycle, just do not report to the police in the t cycle so, think same incident;

(2) sometime, reported to the police as if certain detecting device, just do not report to the police in its downstream so, thinks same incident;

(3) if certain detecting device was reported to the police in the t-1 cycle, just do not report to the police in its downstream of t cycle so, think same incident, as shown in the table: (having √ to represent to report to the police)

Table 1 space-time instruction card

(4) if the crossing is arranged in the middle of the detecting device in highway section, through street, then do not report to the police or the threshold range increasing in the highway section so;

(5) if in the middle of the detecting device in highway section, through street import and export are arranged, the warning confidence level of kerb lane reduces so;

Its software implementing course is as follows:

Earlier all detecting devices on a certain highway section are deposited in the chained list by the upstream and downstream relation with the data type of structure, used structure SAlarm_Attribute represents the event attribute of each detecting device, has to give a definition:

typedef?struct {

// interior ring

Int_pre_inc_id; The Case Number of the generation incident in // this last cycle of detecting device, 0 is no incident

CString alg_id; The algorithm sign of // this incident, 0 no incident 1,2,3 is represented algorithm one, two, three

Int alarm_count; The alarm times of // this detecting device

Int IncidentID; The Case Number of the generation incident of // this detecting device, 0 is no incident

// outer shroud

Int pre_inc_id2; The Case Number of the generation incident in // this last cycle of detecting device, 0 is no incident

CString alg_id2; The algorithm sign of // this incident, 0 no incident 1,2,3 is represented algorithm one, two, three

Int alarm_count2; The alarm times of // this detecting device

Int IncidentID2; The Case Number of the generation incident of // this detecting device, 0 is no incident

CString posid; // this detecting device numbering

CString UpPOSID; // upstream POSID

CString DownPOSID; // downstream POSID

CString roadname; Highway section name under // this detecting device

Int curve; // straight curve sign, 1 represents straight way, and 0 represents bend

DOUBLE MIDLONGITUDE; // with table in the longitude in next detecting device centre position

DOUBLE MIDLATITUDE; // with table in the latitude in next detecting device centre position

DOUBLE LONGITUDE; // this POSID longitude

DOUBLE LATITUDE; // this POSID latitude

SAlarm_Attribute; // detecting device event attribute structure

At first, whether the alarm times alarm_count that judges this detecting device is 1:

If not, whether judge it again greater than 1, if greater than 1, illustrate that then this detecting device reports to the police, just can be merged into same incident according to its Case Number IncidentID, if less than 1, promptly equal 0, no incident then is described, continue other detecting devices of judgement;

If, illustrate that then this detecting device reports to the police for the first time, then judge earlier to have or not the upstream: if no upstream, then this incident is a new events, warning; If the upstream is arranged, can judge earlier whether this incident was reported, if reported, then be merged into same incident according to its Case Number IncidentID, if do not report, judge the alarm condition of upstream again in the last cycle.The process flow diagram of integration module as shown in Figure 5.

The present invention tentatively adopts the method for statistics to provide the threshold value of module 1 (laterally dimension), provides the threshold value of module 2 (time dimension) according to traffic flow model and optimization method, adopts the threshold value of fuzzy clustering method module 3 (vertically dimension).Will that a situation arises be different with the period according to the actual traffic incident: to the threshold value of method with key value is finely tuned and upgrade automatically.

The parameter learning step is introduced:

Alarm event information according to historical data, draw the method performance index, verification and measurement ratio and rate of false alarm etc. are arranged, thus can be by hand or upgrade three key parameters that detect the submethod modules automatically, make event detecting method can closer reflect the variation of traffic flow, can satisfy customer requirements with the method performance index of making, system flowchart as shown in Figure 6.

And the undated parameter rule is as follows:

(1) utilizes the automatic parameter of upgrading module 1 of up-to-date historical data and statistical method, in certain cycle of being separated by as three months, the data and the true traffic events of having reported to the police are added up, draw verification and measurement ratio and rate of false alarm, adjust univers parameter and be in reasonable levels to make rate of false alarm and alarm rate as far as possible.Concerning algorithm one, software flow pattern as shown in Figure 7.

(2) upgrade traffic flow model automatically according to one month up-to-date historical data, three threshold value O1, O2 of module 2 and O3 obtain upgrading automatically, briefly, be exactly that up-to-date one month historical data is imported the traffic flow model optimizer, directly obtain three up-to-date threshold values.

(3) carry out 4 cluster centres of the automatic update module of fuzzy clustering again according to one month up-to-date historical data.Briefly, be exactly with one month up-to-date historical speed and occupation rate data input fuzzy clustering program, directly obtain 4 up-to-date fuzzy clustering centers.

Claims (6)

1. 3 D integrated freeway traffic event automatic detection method is characterized in that judging whether to send alerting signal according to the analysis that detecting device behind highway or the through street is detected data, specifically comprises the following steps:

----system initialization: read track flow-threshold speed, lane occupancy ratio-flow threshold, track speed-occupation rate threshold value;

----starts enabler flags: the automatic measuring ability of start-up routine, and the manual or undated parameter automatically when needs, " detection of end " button zero clearing sign by the interface finishes automatic measuring ability, treats that parameter update finishes, and can restart;

----reads the traffic flow real time data, reads all detector data respectively;

-----call horizontal dimension is that track flow-rate algorithm module, time dimension are lane occupancy ratio-flow algoritic module, vertically to tie up detection module be track speed-occupation rate algoritic module and integrated judge module, to whole monitor data is that the basis is judged, makes according to judged result whether starting warning device and sending alerting signal.

2. 3 D integrated freeway traffic event automatic detection method according to claim 1, it is characterized in that described horizontal dimension is that track flow-rate algorithm module specifically comprises the following steps: to add up with one month historical data, obtain average discharge ratio and the velocity rate of each period, then the threshold value of the throughput ratio in each track and velocity ratio is made as:

Volume_Limit _i＝max(0，mean(VOLUMN_RITO _i)-2*std(VOLUMN_RITO _i))；(i＝1，2，3，4)

Speed_Limit _i＝max(0，mean(SPEED_RITO _i)-2*std(SPEED_RITO _i))；(i＝1，2，3，4)

Mean represents to average, and std represents to ask variance;

When certain track flow and speed drop to certain threshold value, show that event occurs, by calculating the velocity rate and the flow rate ratio in each track of detecting device section, and compare with lower velocity limit that configures and flux lower limit parameter, come the detection incident; The computing formula of flow and velocity rate is as follows:

$\mathrm{VOLUME}\_{\mathrm{RITO}}_i=\frac{{\mathrm{VOLUME}}_i}{{\mathrm{\&Sigma;VOLUME}}_i}(i=\mathrm{1,2,3,4});$

$;\mathrm{SPEED}\_{\mathrm{RITO}}_i=\frac{{\mathrm{SPEED}}_i}{{\mathrm{\&Sigma;SPEED}}_i}(i=\mathrm{1,2,3,4});$

VOLUME_RITO wherein _iBe the flow rate ratio in certain track, VOLUME _iBe the flow in certain track, module settings is for when certain track flow and speed drop to certain threshold value twice continuously, and the affair alarm request is sent in this track.

3. 3 D integrated freeway traffic event automatic detection method according to claim 1 is characterized in that described time dimension is that lane occupancy ratio-flow algoritic module comprises the following steps: that specifically traffic flow model is:

$f=k*v_f{(1-{\left(\frac{o}{o_j}\right)}^r)}^m*o$

In the formula, o is an occupation rate, and f is a flow, v _fThe expression free stream velocity, o _jOccupation rate is blocked in expression, and r, m are constants, v _f, o _j, r, m are parameters undetermined, according to the real road situation, the variation range of these four parameters is carried out certain limitation, be translated into a constrained optimization problem, the objective function of model optimization is the absolute error sum, and this value has reflected the size of error comprehensively, is shown below

$\mathrm{SAE}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}|f\left(i\right)-f^1\left(i\right)|$

Wherein, f (i) is the flow value that obtains according to traffic flow model, f ¹(i) be the flow value of actual measurement; In constrained optimization problem, employing is found the solution based on K-T (knhn-Tucker) equation, occupation rate according to continuous four cycles draws corresponding grade of living in and LevChange (t, t-1, i) (the grade that expression t jumps with respect to t-1 moment occupation rate constantly, i represents the track), judge whether to satisfy one of following three conditions then:

If after satisfying, then the affair alarm request is sent in this track.

4. 3 D integrated freeway traffic event automatic detection method according to claim 1 is characterized in that described vertical dimension detection module is that track speed-occupation rate algoritic module comprises the following steps: that specifically the concrete steps that method is implemented are: use the central value that fuzzy clustering method provides the cluster of a large amount of traffic flow speed-occupation rate historical data; Obtain the Euclidean distance at traffic flow speed, occupation rate data and four centers of real-time upstream and downstream respectively, with which center be recently the Euclidean distance minimum be exactly under corresponding grade; Judge the rank difference of upstream and downstream, when superfine continuous two cycles of upstream and downstream traffic behavior differ 3 ranks, show that event occurs, then send the affair alarm request.

5. 3 D integrated freeway traffic event automatic detection method according to claim 1, it is characterized in that the synthetic integrated rule of described integrated judge module is: report to the police in the t-1 cycle as if certain detecting device (1), just do not report to the police in the t cycle so, think same incident; (2) sometime, reported to the police as if certain detecting device, just do not report to the police in its downstream so, thinks same incident; (3) if certain detecting device was reported to the police in the t-1 cycle, just do not report to the police in its downstream of t cycle so, think same incident; (4) if the crossing is arranged in the middle of the detecting device in highway section, through street, then do not report to the police or the threshold range increasing in the highway section so; (5) if in the middle of the detecting device in highway section, through street import and export are arranged, kerb lane does not report to the police so.

6. 3 D integrated freeway traffic event automatic detection method according to claim 1, it is characterized in that also comprising the following steps: alarm event information according to historical data, draw the performance index that comprise verification and measurement ratio and rate of false alarm, with manual or upgrade the key parameter of three detection modules automatically, make event detecting method can closer reflect the variation of traffic flow, the undated parameter rule is as follows: (1) utilizes up-to-date historical data and statistical method to upgrade automatically and laterally ties up is the parameter of track flow-rate algorithm module, in certain cycle of being separated by, the data and the true traffic events of having reported to the police are added up, draw verification and measurement ratio and rate of false alarm, adjust univers parameter and be in reasonable levels to make rate of false alarm and alarm rate as far as possible; (2) upgrade traffic flow model automatically according to one month up-to-date historical data, time dimension is that three threshold values of lane occupancy ratio-flow algoritic module obtain upgrading automatically, historical data input traffic flow model optimizer with up-to-date one month directly obtains three up-to-date threshold values; (3) carrying out the vertical dimension of the automatic renewal of fuzzy clustering detection module again according to one month up-to-date historical data is track speed-4 cluster centres of occupation rate algoritic module, be exactly with one month up-to-date historical speed and occupation rate data input fuzzy clustering program, directly obtain 4 up-to-date fuzzy clustering centers.

Images