CN103514743B - A kind of abnormal traffic state characteristic recognition method of real-time index-matched memory range - Google Patents

A kind of abnormal traffic state characteristic recognition method of real-time index-matched memory range Download PDF

Info

Publication number
CN103514743B
CN103514743B CN201310451579.9A CN201310451579A CN103514743B CN 103514743 B CN103514743 B CN 103514743B CN 201310451579 A CN201310451579 A CN 201310451579A CN 103514743 B CN103514743 B CN 103514743B
Authority
CN
China
Prior art keywords
traffic
index
real
abnormal
time
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201310451579.9A
Other languages
Chinese (zh)
Other versions
CN103514743A (en
Inventor
吴超腾
沈峰
肖永来
张莉
矫晓丽
林瑜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xinjiang City Branch Intelligent Polytron Technologies Inc
Original Assignee
上海电科智能系统股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 上海电科智能系统股份有限公司 filed Critical 上海电科智能系统股份有限公司
Priority to CN201310451579.9A priority Critical patent/CN103514743B/en
Publication of CN103514743A publication Critical patent/CN103514743A/en
Application granted granted Critical
Publication of CN103514743B publication Critical patent/CN103514743B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

The present invention relates to a kind of abnormal traffic state characteristic recognition method of real-time index-matched memory range, it is characterized in that, step is: all historical traffic exponent datas in step 1, certain section obtained in some time periods, creates the index memory space corresponding with this moment and calculate the intermediate value of each index memory space according to all historical traffic indexes under synchronization; Step 2, each index memory space to be at least divided between positive buffer zone and between negative buffer, and the region being less than minimal index value is defined as negative overflow area, the region being greater than greatest exponential value is defined as positive overflow area; Step 3, obtain the real-time traffic index Index (Tn) of this section under current time Tn, judge traffic behavior according to subregion.The invention provides a kind of new traffic index application process, fully excavate the quick identification of potentiality realization to urban road real-time traffic exception of traffic index.

Description

A kind of abnormal traffic state characteristic recognition method of real-time index-matched memory range
Technical field
The present invention relates to a kind of real-time identification method of urban road abnormal traffic state feature, the historical law data that the method accumulates with urban road real-time traffic index monitoring road conditions are empirically remembered, and set up " memory space " according to the statistical property of data memory, real-time traffic index-matched current time section " memory space ", and the target road object monitored with this discrimination index is current, and whether to be in traffic behavior abnormal, belongs to intelligent transport system field.
Background technology
Traffic index is a kind of class relativity index can expressing urban road traffic state or traffic congestion with serial number, according to practical application request, choose special traffic parameter and build according to certain functional rule, along with the development of transport information technology, domestic each main cities is monitored road traffic running status in real time as Beijing, Shanghai, Hangzhou, Shenzhen etc. all construct the traffic index model meeting city characteristic.Traffic index tool three significant mathematical features of real-time monitoring dynamic road operation characteristic, the first, index is present in fixing numerical intervals, as [0,100], [0,5] etc., choosing of data interval is determined jointly by exponential model and issue requirement; The second, index results is continuous print numerical value, can cover all real number values in data interval in theory; 3rd, index has monotonicity, and namely exponential quantity dullness reacts the trend that road conditions improve or degenerate, and there is not ambiguity, the present invention carries out so that the larger road conditions of index are worse.Just based on these three mathematical features, index can with the real-time traffic states in the form record object road moment of numerical point, and then the curve of a reflection whole day traffic behavior Change and Development trend can be drawn, can not only differentiate that target road or road network region blocked up moment on peak and degree in one day by this curve, can also compare different road object or road network colony, assessment normality heavy congestion road and unimpeded road.
To urban highway traffic in general, by population distribution, road layout, traffic trip rule etc., factor is metastable affects, traffic behavior often has certain time space distribution, namely the road occurred frequently that blocks up in the ordinary course of things is relatively-stationary, and the time occurred frequently of jam road is also relatively-stationary.Traffic control department can deploy to ensure effective monitoring and control of illegal activities according to the police strength of this kind of characteristic formulation normality, and optimum management resource, smooth work is protected in commander's unimpeded.But the road network system that city level is numerous, the scale of construction is huge has the feature of traffic congestion occurred frequently, traffic rule sudden change at random, especially in the face of inclement weather, shortly before festivals or holidays, the factor such as lager-scale social event time, the road that normality is not blocked up can produce the heavy congestion being difficult to predict, or normality congested link can present significantly unimpeded etc., this situation often to be deployed to ensure effective monitoring and control of illegal activities section and being ignored in early days because the road network that blocks up does not belong to emphasis, causes the continuous worsening serious consequence even causing long-time, large area and block up of road conditions.Traffic index can realize the digital expression to road conditions as a kind of efficient information tool, if can the change of the further personalized traffic behavior feature of the every bar road of discovery of intelligence on the basis of real-time road monitoring, the abnormal state degree of quantitative analysis assessment road, brings important intelligence value and economic worth by for confirming that cause of problem and site traffic manage fast.By the retrieval to prior art and system, do not find to obtain the known correlation technique that can meet the abnormal ONLINE RECOGNITION of real-time traffic states and method.
Summary of the invention
The technical problem to be solved in the present invention is to provide a kind of new traffic index application process, fully excavates the quick identification of potentiality realization to urban road real-time traffic exception of traffic index.
In order to solve the problems of the technologies described above, technical scheme of the present invention there is provided a kind of abnormal traffic state characteristic recognition method of real-time index-matched memory range, and it is characterized in that, step is:
Step 1, the traffic flow parameter of a real time data in the interval dynamic fluctuation of fixed data is chosen arbitrarily as traffic status identification parameter in all traffic flow parameters, obtain all historical traffic state recognition supplemental characteristics of certain section in some time periods or Regional Road Network, create the traffic status identification parameters memorizing space corresponding with this moment according to all traffic status identification parameters under synchronization and calculate the intermediate value in each traffic status identification parameters memorizing space, wherein, the traffic status identification parameters memorizing space that moment T is corresponding is IMS (T), the negative edge that minimal index value MIN (T) in all historical traffic state recognition supplemental characteristics of moment T and greatest exponential value MAX (T) is traffic status identification parameters memorizing space IMS (T) and positive boundary, the average that the intermediate value MEDIAN (T) of traffic status identification parameters memorizing space IMS (T) is traffic status identification supplemental characteristics all under moment T,
Step 2, each traffic status identification parameters memorizing space to be at least divided between positive buffer zone and between negative buffer, and the region being less than minimal index value is defined as negative overflow area, the region being greater than greatest exponential value is defined as positive overflow area; Be PB (T) between the positive buffer zone of traffic status identification parameters memorizing space IMS (T), span be (MEDIAN (T), MAX (T)]; Be NB (T) between negative buffer, span be [MIN (T), MEDIAN (T)); Negative overflow area is NO (T), span be [MIN, MIN (T)); Positive overflow area is PO (T), span be (MAX (T), MAX], wherein, MIN is that traffic status identification parameter is minimum may value, and MAX is the maximum possible value of traffic status identification parameter;
Step 3, obtain the real-time traffic states identification parameter Index (Tn) of this section under current time Tn, to judge between the positive buffer zone whether real-time traffic states identification parameter Index (Tn) is positioned at traffic status identification parameters memorizing space IMS (Tn) NB (Tn) between PB (Tn) or negative buffer, if, then the traffic behavior of this section under current time Tn is normality, if not, then the traffic behavior of this section under current time Tn is abnormal, this exception refers to extremely unimpeded or abnormal blocking up, if to be traffic status identification parameter more high more blocks up for the traffic status identification parameter model that this section adopts, if then real-time traffic states identification parameter Index (Tn) is positioned at positive overflow area PO (Tn), the traffic behavior of this section under current time Tn blocks up for abnormal, if real-time traffic states identification parameter Index (Tn) is positioned at negative overflow area NO (Tn), the traffic behavior of this section under current time Tn is abnormal unimpeded, if it is more high more unimpeded that the traffic status identification parameter model that this section adopts is traffic status identification parameter, if then real-time traffic identification parameter Index (Tn) is positioned at positive overflow area PO (Tn), the traffic behavior of this section under current time Tn is abnormal unimpeded, if real-time traffic identification parameter Index (Tn) is positioned at negative overflow area NO (Tn), the traffic behavior of this section under current time Tn blocks up for abnormal.
Preferably, in described step 1, influence factor is utilized to mark MARK, all historical traffic identification parameter data in certain section in some time periods are marked, influence factor mark MARK is the one group of cartesian product configuration item be made up of the non-traffic factor having extensively impact to traffic behavior feature in influence factor config set, all historical traffic identification parameters under synchronization are according to influence factor mark MARK classification, create traffic status identification parameters memorizing space respectively for each class and calculate the intermediate value in each traffic status identification parameters memorizing space,
In described step 3, after obtaining the real-time traffic identification parameter Index (Tn) under current time Tn, first utilize influence factor to mark MARK to it and mark, obtain the traffic identification parameter Index (Tn) with mark result mARK, utilize traffic identification parameter Index (Tn) mARKtraffic status identification parameters memorizing space corresponding with its classification under finding moment Tn, then judge that whether the traffic behavior of this section under current time Tn be abnormal according to the relation of between itself and Zhong Zheng buffer zone, traffic status identification parameters memorizing space and between negative buffer and negative overflow area and positive overflow area.
Preferably, described Weather information Weather, calendar information Calendar and action message Event are at least comprised on the non-traffic factor that traffic behavior feature has an extensively impact, wherein, calendar information Calendar is main factor, each moment has different influence factor mark MARK, and the influence factor of moment T is labeled as MARK t(1,2,3)=MARK t(Calendar, Weather, Event).
Preferably, in described step 2, each traffic status identification parameters memorizing space is at least divided between positive buffer zone, normal state is interval, bear between normality interval and negative buffer; Between the positive buffer zone of then traffic status identification parameters memorizing space IMS (T), the span of PB (T) is (MPSD (T), MAX (T)], MPSD (T)=MEDIAN (T)+stdev (T), stdev (T) is the standard deviation of historical traffic identification parameters all under moment T; Normal state interval is PN (T), and span is [MEDIAN (T), MPSD (T)]; Negative normality interval is NN (T), span be [MNSD (T), MEDIAN (T)), MNSD (T)=MEDIAN (T)-stdev (T); Between negative buffer the span of NB (T) be [MIN (T), MNSD (T));
In described step 3, obtain the real-time traffic identification parameter Index (Tn) of this section under current time Tn, judge whether real-time traffic identification parameter Index (Tn) is positioned at the interval PN (Tn) of normal state or the negative normality interval (Tn) of traffic status identification parameters memorizing space IMS (Tn), if, then the traffic behavior of this section under current time Tn is normal, if not, then the traffic behavior of this section under current time Tn is abnormal, this exception refers to extremely unimpeded or abnormal blocking up, more block up if the traffic identification parameter model that this section adopts is that traffic identification parameter is more high, if then real-time traffic identification parameter Index (Tn) be positioned at traffic status identification parameters memorizing space IMS (Tn) positive buffer zone between PB (Tn) or positive overflow area PO (Tn), the traffic behavior of this section under current time Tn blocks up for abnormal, if NB (Tn) or negative overflow area NO (Tn) between the negative buffer that real-time traffic identification parameter Index (Tn) is positioned at traffic status identification parameters memorizing space IMS (Tn), the traffic behavior of this section under current time Tn is abnormal unimpeded, if it is more high more unimpeded that the traffic identification parameter model that this section adopts is traffic identification parameter, if then real-time traffic identification parameter Index (Tn) be positioned at traffic status identification parameters memorizing space IMS (Tn) positive buffer zone between PB (Tn) or positive overflow area PO (Tn), the traffic behavior of this section under current time Tn is abnormal unimpeded, if NB (Tn) or negative overflow area NO (Tn) between the negative buffer that real-time traffic identification parameter Index (Tn) is positioned at traffic status identification parameters memorizing space IMS (Tn), the traffic behavior of this section under current time Tn blocks up for abnormal.
Preferably, in described step 3, if NB (Tn) between PB (Tn) or negative buffer between the positive buffer zone that real-time traffic identification parameter Index (Tn) is positioned at traffic status identification parameters memorizing space IMS (Tn), then any one adopting in following two methods judges the traffic behavior of current road segment under current time Tn:
First method: if real-time traffic identification parameter Index (Tn) is positioned at PB (Tn) between positive buffer zone, then with threshold value [MAX (Tn)+MPSD (Tn)]/[2 × MEDIAN (Tn)] for boundary line, be greater than this value and sentence "abnormal", be less than this value and be greater than MPSD (Tn)/MEDIAN (Tn) and then judge " doubtful exception "; If real-time traffic identification parameter Index (Tn) is positioned at NB between negative buffer (Tn), then with threshold value [MIN (Tn)+MNSD (Tn)]/[2 × MEDIAN (Tn)] for boundary line, be less than this value and sentence "abnormal", be greater than this value and be less than MNSD (Tn)/MEDIAN (Tn) and then judge " doubtful exception ";
Second method: judge whether to there is reference point Ref t, reference point Ref tenter between positive buffer zone by real-time traffic identification parameter Index (Tn) first by the interval PN (Tn) of normal state or the interval NN (Tn) of negative normality, between negative buffer, when positive overflow area or negative overflow area, create with MPSD (Tn) or MNSD (Tn) crossing interpolation, the interpolation moment is rounded to the previous moment in crossing moment, if not, then sentence " normality ", if, then judge that real-time traffic identification parameter Index (Tn) to be positioned between positive buffer zone NB (Tn) between PB (Tn) or negative buffer, if be positioned at PB (Tn) between positive buffer zone, then calculate [Index (Tn)-Ref t] × T 0/ (Tn-t)-[MPSD (Tn)-Ref t] × T 0/ (Tn-t), wherein, T 0for the update cycle of traffic identification parameter, t is the interpolation moment, if its result is greater than A, then sentences "abnormal", between (0, A] between then sentence " doubtful exception ", A is empirical value, if be positioned at NB between negative buffer (Tn), then calculate [Index (Tn)-Ref t] × T 0/ (Tn-t)-[MNSD (Tn)-Ref t] × T 0/ (Tn-t), if its result is not more than-A, then sentences "abnormal", between (-A, 0] between then sentence " doubtful exception ".
Preferably, in described step 3, utilize three-dimensional extremely to combine the traffic behavior differentiating and judge certain section, then described step 3 comprises:
Step 3.1, obtain this section or the real-time traffic identification parameter Index (Tn) of Regional Road Network under current time Tn;
Step 3.2, judge whether real-time traffic identification parameter Index (Tn) is positioned between positive buffer zone NB (Tn) between PB (Tn) or negative buffer, if not, then judges whether to there is reference point Ref t, reference point Ref tto be entered between positive buffer zone, between negative buffer by real-time traffic identification parameter Index (Tn), positive overflow area or negative overflow area time, create with MPSD (Tn) or MNSD (Tn) crossing interpolation, if exist, directly enter next step, if do not exist, then create reference point Ref tafter enter next step; If real-time traffic identification parameter Index (Tn) to be positioned between positive buffer zone NB (Tn) between PB (Tn) or negative buffer, then judge whether to there is reference point Ref tif exist, then sentence " normality ", completing steps 3, if do not exist, then created reference point Ref tafter enter step 3.4;
Step 3.3, the first dimension abnormality juding:
If real-time traffic identification parameter Index (Tn) is positioned at positive overflow area PO (Tn), then calculate Index (Tn) MAX (Tn), if this difference is less than A, then sentence " doubtful exception ", enter next step, otherwise, sentence "abnormal", enter next step; If real-time traffic identification parameter Index (Tn) is positioned at negative overflow area NO (Tn), then calculate MIN (Tn)-Index (Tn), if this difference is less than A, then sentence " doubtful exception ", enter next step, otherwise, sentence "abnormal", enter next step, A is empirical value;
Step 3.4, the second dimension abnormality juding:
If real-time traffic identification parameter Index (Tn) is positioned at PB (Tn) or positive overflow area PO (Tn) between positive buffer zone, then with threshold value [MAX (Tn)+MPSD (Tn)]/[2 × MEDIAN (Tn)] for boundary line, Index (Tn)/MEDIAN (Tn) is greater than this value and sentences "abnormal", enter next step, Index (Tn)/MEDIAN (Tn) is less than this value and is greater than MPSD (Tn)/MEDIAN (Tn) and then judges " doubtful exception ", enters next step; If real-time traffic identification parameter Index (Tn) is positioned at NB between negative buffer (Tn) or negative overflow area NO (Tn), then with threshold value [MIN (Tn)+MNSD (Tn)]/[2 × MEDIAN (Tn)] for boundary line, Index (Tn)/MEDIAN (Tn) is less than this value and sentences "abnormal", enter next step, Index (Tn)/MEDIAN (Tn) is greater than this value and is less than MNSD (Tn)/MEDIAN (Tn) and then judges " doubtful exception ", enters next step;
Step 3.5, third dimension abnormality juding:
If be positioned at PB (Tn) or positive overflow area PO (Tn) between positive buffer zone, then calculate [Index (Tn)-Ref t] × T 0/ (Tn-t)-[MPSD (Tn)-Ref t] × T 0/ (Tn-t), wherein, T 0for the update cycle of traffic identification parameter, t is the interpolation moment, if its result is greater than A, then sentences "abnormal", enters next step, between (0, A] between then sentence " doubtful exception ", enter next step; If be positioned at NB between negative buffer (Tn) or negative overflow area NO (Tn), then calculate [Index (Tn)-Ref t] × T 0/ (Tn-t)-[MNSD (Tn)-Ref t] × T 0/ (Tn-t), if its result is not more than-A, then sentences "abnormal", enters next step, between (-A, 0] between then sentence " doubtful exception ", enter next step;
Step 3.6, adopt " abnormal a voting adopted fixed " or " the minority is subordinate to the majority " to carry out exception to combine differentiation, wherein, " an abnormal voting adopted is fixed " refers to: at traffic behavior from " normality " to "abnormal" cognitive phase, the first dimension abnormality juding result described in step 3.3 has a ticket power to make decision; And returning to " normality " stage from "abnormal", be then on " normality " basis in the result of the first dimension abnormality juding, jointly confirm abnormal restoring by the second dimension abnormality juding described in step 3.4 and the third dimension abnormality juding described in step 3.5;
" the minority is subordinate to the majority " refers to: the first dimension abnormality juding result described in step 3.3, the second dimension abnormality juding result described in 3.4 and the third dimension abnormality juding result described in step 3.5 have identical weight, and the majority according to exporting result of determination judges final recognition result; If the result of determination of three dimensionality exports all not identical, then or according to " criterion of pessimism " towards "abnormal" direction discernment, or according to " criterion of optimism " towards " normality " direction discernment.
The abnormal traffic state feature real-time identification method that the present invention proposes match index memory space a kind ofly utilizes real-time index to combine the intelligent method identifying traffic behavior exception with history index, difference is, history index is on the basis of classifying through the influence factor such as date, weather, for each time slice creating out " memory space " in data cover region, and carry out off-note identification as total environment and algorithm triggers condition." memory space " marks off six domain logics by minimal value, negative bias difference, intermediate value, positively biased difference, maximum value five controlling values: just overflow, just cushion, normal state, negative normality, negative buffering, negatively to overflow, " six territories " that form division index span wherein just overflows, negatively to overflow for " memory space " exterior domain.Real-time index-matched " memory space " also differentiates the need of the analysis of trigger condition startup multidimensional abnormity diagnosis according to place domain logic, multidimensional exception diagnosis algorithm comprises the easily extensible diagnostic methods such as differential analysis, proportion grading, trend analysis, and each dimension independence Output rusults is all that fuzzy logic judges.Gather multidimensional abnormity diagnosis analysis result, input associating distinguished number carries out aggregative weighted differentiation, and the traffic behavior feature that its final Output rusults is time slice exports.(whether note: all use the moment above, unify, and the difference slightly in time slice and moment is used in the moment on real-time index, and time slice was used on all statistics dates in memory space corresponding moment.This my sensation still keeps present literary style, because be easier to understand)
The abnormal traffic state feature real-time identification method beneficial effect of the match index memory range that the present invention proposes can be embodied in the following aspects:
The first, realize the comprehensive abnormal monitoring to urban road the whole network, section, path, section from traffic information system aspect, eliminate road conditions abnormal monitoring space-time blind area;
The second, for the priority scheduling of resource of city relevant departments provides information instrument, auxiliary relevant departments, to the genetic analysis of off-note and decision-making, especially have important value to the early warning of some pernicious traffic congestion;
Three, greatly can improve the recognition efficiency of the overall traffic abnormity of urban traffic network, improve traffic administration and to note abnormalities the rapidity of traffic problems and accuracy, combine disposal save time for multidisciplinary;
Four, anomalous identification result exports after confirming, for statistics road traffic year, the frequency of monthly generation abnormality feature and degree provide quantification analysis index helpful.
Accompanying drawing explanation
Fig. 1 real-time traffic abnormality recognition method logical architecture;
Fig. 2 real-time traffic abnormality recognition method main-process stream;
Fig. 3 index memory space builds sub-process;
Fig. 4 real-time index-matched memory space sub-process;
Fig. 5 is that multidimensional abnormity diagnosis analyzes sub-process;
Fig. 6 abnormal state connection is appraised and is analysed sub-process;
North and south, Fig. 7 Shanghai City elevated bridge section normal case of June 19 (Wednesday) status flag in 2013;
Middle Ring Line, Fig. 8 Shanghai City section abnormal case of April 19 (Thursday) morning peak status flag in 2012.
Embodiment
For making the present invention become apparent, hereby with preferred embodiment, and accompanying drawing is coordinated to be described in detail below.
Method provided by the invention all has effect, such as traffic index or the road travel speed of a motor vehicle etc. for having the traffic flow parameter of real time data in the interval dynamic fluctuation of fixed data.In the present embodiment, be described further inventing the method provided for traffic index, traffic index model adopts traffic index larger, more blocks up, if traffic index is less of more to block up, its principle is identical with the present embodiment, as long as arranged on the contrary relevant position.Basic terms below with regard to using in the present invention provide definition:
Time slice (TimeSlice): namely with each moment value of whole day that real-time traffic index update cycle T O obtains for step-length, Tn=n × TO (n=1,2 ...).In existing traffic index model, update cycle TO is the longest is half an hour, for the present invention, and update cycle TO more short better (suggestion is 2min, 5min).By time slice alignment index historical data base create memory space, if current time is T, measurement period is 100 days, then the index history value in corresponding 100 days on each moment T forms a time slice.
Influence factor mark MARK. influence factor mark MARK is the one group of cartesian product configuration item be made up of the non-traffic factor that traffic behavior feature has extensively impact the Weather information Weather in influence factor config set, calendar information Calendar, action message Event etc.Influence factor mark MARK in, calendar information Calendar column number is 1, comprise Monday, Tu. ..., Sun., lunar calendar red-letter day etc.; Weather information Weather column number is 2, comprises fine day, light rain, moderate rain, heavy rain, heavy rain, slight snow, heavy snow, dense fog, hail etc.; Action message Event column number is 3, containing normality, World Expo, the Olympic Games, car exhibition, college entrance examination, winter and summer vacation etc.; MARK is labeled as the T moment t(1,2,3)=MARK t(Calendar, Weather, Event)=MARKT (Monday, fine, normality), wherein, calendar information Calendar is main factor, also bears the basic function to historical data classification.
Index memory space IMS (IndexMemorySpace): in the complete or collected works of the index historical data measurement period (as 2 years) after MARK mark, the maximum value that a time slice in office occurred and minimal value the index extreme difference interval closed.If index history value is 2 years, wherein have 112 days Monday, then the memory space in 8:00 moment on Monday is exactly the data interval that this 112 8:00 moment indexes cover.
The minimal index value that minimal index value MIN. time slice occurred in index memory space IMS, lower limit 0.
Greatest exponential value MAX: the minimal index value that time slice occurred in index memory space IMS, higher limit 100.
Intermediate value MEDIAN: the index memory space IMS on time slice, according to ascending sequence, concentrate the Mean value of index added up and obtain as intermediate value in general classification.
Negative bias difference MNSD (MedianNegativeStandardDeviation): standard deviation function is used to the index memory space IMS on time slice, obtain stdev () value, MNSD=MEDIAN-stdev (), if MNSD<=MIN, then get MIN value.
Positively biased difference MPSD (MedianPositiveStandardDeviation): standard deviation function is used to the index memory space IMS on time slice, obtain stdev () value, MPSD=MEDIAN+stdev (), if MPSD>=MAX, then get MAX value.
Positive overflow area PO (PositiveOverflow): index codomain time slice being greater than more than maximum value MAX is interval, and span (MAX, 100], if MAX=100 under extreme case, then just overflowing and equaling maximum value MAX.Positive overflow area PO is the exterior domain of index memory space IMS, belong to the just half side of overflow area, real-time index drops on this interval, then monitoring is described, and road is current has occurred the poorest traffic behavior, belong to suspected outlier, need to start multidimensional abnormity diagnosis and analyze and associating distinguished number.
PB (PositiveBuffer) between positive buffer zone: time slice is greater than more than positively biased difference MPSD and to be less than the index codomain of maximum value MAX interval, span (MPSD, MAX].Between positive buffer zone, PB is index memory space IMS inner region, belongs to the just half side of buffer zone, though to drop on this interval abnormal for real-time index, is also directly do not export as abnormal, need export monitor in conjunction with the analysis of multidimensional abnormity diagnosis and associating distinguished number.
The interval PN (PositiveNormal) of normal state: time slice is less than or equal to positively biased difference MPSD and to be more than or equal to the index codomain of intermediate value MEDIAN interval, span [MEDIAN, MPSD].The interval PN of normal state is index memory space IMS main areas, belongs to the just half side of normality district, and real-time index drops on this interval for normality, only need maintain monitoring without the need to launching other measures.
The interval NN (NegativeNormal) of negative normality: time slice is more than or equal to negative bias difference MNSD and to be less than the index codomain of intermediate value MEDIAN interval, span [MNSD, MEDIAN).Negative normality is index memory space IMS main areas, belongs to the negative half side of normality district, and real-time index drops on this interval for normality, only need maintain monitoring without the need to launching other measures.
NB (NegativeBuffer) between negative buffer: time slice is more than or equal to more than minimal value MIN and to be less than the index codomain of negative bias difference MNSD interval, span [MIN, MNSD).Between negative buffer, NB is index memory space IMS inner region, belongs to the negative half side of buffer zone, though to drop on this interval abnormal for real-time index, is also directly do not export as abnormal, need start the analysis of multidimensional abnormity diagnosis and associating distinguished number and export and monitor.
Negative overflow area NO (NegativeOverflow): index codomain time slice being less than below minimal value MIN is interval, and span [0, MIN), if MIN=0 under extreme case, then negative spilling equals MIN.Negative spilling is index memory space IMS exterior domain, belong to the negative half side of overflow area, real-time index drops on this interval, then monitoring is described, and road is current has occurred best traffic behavior, belong to suspected outlier, need to start multidimensional abnormity diagnosis and analyze and associating distinguished number.
Reference point Ref t(Reference): the temporary fiducial point being used for carrying out Trend judgement, position is on positively biased difference MPSD or negative bias difference MNSD, when real-time traffic index Index (Tn) enters buffer zone or overflow area by normality region, create with positively biased difference MPSD or the crossing interpolation of negative bias difference MNSD, the interpolation moment, t was rounded to the previous moment in crossing moment.Reference point Ref thave after generation and only have one, its numerical value and moment t all constant, when only having the traffic index Index (Tn) on certain time slice on the same day again to enter normality district by buffer zone or overflow area, reference point is automatically deleted and is emptied.
As shown in Figures 1 and 2, the abnormal traffic state characteristic recognition method of a kind of real-time index-matched memory range provided by the invention, the steps include:
All historical traffic indexes in step 1, certain section obtained in certain time period, the influence factor exported according to traffic impact set of factors mark MARK t(1, 2, 3) (not containing the up-to-date moment) is divided into groups to the historical traffic index of moment Tn, Main classification influence factor is calendar information Calendar, the factors such as Weather information Weather and action message Event are as auxiliary packet factor, with sorted historical traffic index for Exponential Sample overall building index memory space IMS (Tn), composition graphs 3, comprise intermediate value MEDIAN (Tn), positive extreme value MAX (Tn), negative pole value MIN (Tn), positively biased difference MPSD (Tn)=MEDIAN (Tn)+stdev (Tn), if MPSD (Tn) >=MAX (Tn), then get MAX (Tn) value, negative bias difference MNSD (Tn)=MEDIAN (Tn)-stdev (Tn) is if MNSD (Tn) <=MIN (Tn), then get MIN (Tn) value.After " five values " line creates, Nature creating " six territories ", i.e. NB (Tn), negative overflow area NO (Tn) between PB (Tn), the interval PN (Tn) of normal state, the interval NN (Tn) of negative normality, negative buffer between positive overflow area PO (Tn), positive buffer zone.If MPSD (Tn)=MAX (Tn), then do not have PB (Tn) between positive buffer zone.In like manner, if MNSD (Tn)=MIN (Tn), then NB between negative buffer (Tn) is not had.Create index memory space IMS (Tn) totally to carry out based on sub-index historical data, therefore do not need to carry out in real time, unify for each road, each cycle carry out index memory space IMS (Tn) and create in system idles phase morning every day, result data is that second day real-time index-matched prepares.
Step 2, mark real-time index results: the influence factor mark MARK exported by traffic impact set of factors t(1,2,3), mark the real-time traffic index Index (Tn) of moment Tn, generate the data Index (T) with mark result mARKand stored in historical data base, carry out the index memory space IMS (Tn) mating moment Tn simultaneously.
Step 3, in real time index-matched index memory space IMS (Tn), composition graphs 4: with MARK tthe real-time traffic index Index (Tn) that (1,2,3) mark mARKmatch index memory space IMS (Tn), according to positive overflow area PO (Tn), just between buffer zone, between PB (Tn), the interval PN (Tn) of normal state, the interval NN (Tn) of negative normality, negative buffer, NB (Tn), negative overflow area NO (Tn) differentiate Index (T) mARKregion.
If real-time traffic index Index (Tn) mARKbe positioned at the interval PN (Tn) of normal state or the interval NN (Tn) of negative normality, then directly export after " normality " differentiates result and empty reference point Ref t, and jump out whole step and enter next cycle.If real-time traffic index Index (Tn) mARKto be positioned between positive buffer zone NB (Tn) between PB (Tn) or negative buffer, then to differentiate whether there is reference point Ref t, if there is no, then generate reference point Ref t, and directly export and jump out whole step after " normality " differentiates result and enter next cycle.If there is reference point Ref t, then the second dimension in the multidimensional abnormity diagnosis analysis of setting up procedure 4 and the third dimension and extremely combine differentiation, obtain after status flag differentiates result, jump out whole step and enter next cycle.If real-time traffic index Index (Tn) mARKbe positioned at positive overflow area PO (Tn) or negative overflow area NO (Tn), then differentiate whether there is reference point Ref t, if there is no, then generate reference point Ref t, and the first dimension, the second dimension and the third dimension in the multidimensional abnormity diagnosis analysis of setting up procedure 4 and abnormal combine differentiation, obtain after status flag differentiates result, jump out whole step and enter next cycle.If reference point Ref texist, then the first dimension in the multidimensional abnormity diagnosis analysis of setting up procedure 4, the second dimension and the third dimension and extremely combine differentiation, export " doubtful exception " or "abnormal" result, do not empty reference point Ref t, jump out whole step and enter next cycle.
Step 4, composition graphs 5, multidimensional abnormity diagnosis is analyzed: as real-time traffic index Index (Tn) mARKnot when the interval PN (Tn) of normal state or negative normality interval NN (Tn), start the analysis of multidimensional abnormity diagnosis.It is all " normality " that multidimensional abnormity diagnosis analyzes each dimension default value, different according to the trigger condition of real-time matching IMS process, starts the abnormity diagnosis analytical algorithm of different dimensions respectively, finally enters all dimension results all conduct inputs when extremely combining differentiation.
First dimension is that difference differentiates: calculate real-time traffic index Index (Tn) mARKwith the deviation of greatest exponential value MAX (Tn) or minimal index value MIN (Tn), if Index (Tn) mARK>MAX (Tn), deviation is positivity bias, then real-time traffic index Index (Tn) mARKbe positioned at positive overflow area PO (Tn), calculate Index (Tn) mARK-MAX (Tn), and by difference compared with empirical value (in this implementation column, empirical value is taken as 5).If this difference <5, then export as " doubtful exception of blocking up ", if this difference >=5, then export as " exception of blocking up ".If Index (Tn) mARk<MIN (Tn), deviation is negative sense deviation, then real-time traffic index Index (Tn) mARKbe positioned at negative overflow area NO (Tn), calculate MIN (Tn)-Index (Tn) mARK, and by difference compared with empirical value (in this implementation column, empirical value is taken as 5).If this difference <5, then export as " doubtful exception is unimpeded ", if this difference >=5, then export as " abnormal unimpeded ".Empirical value can be demarcated according to concrete road object.
Second dimension ratio differentiates: calculate real-time traffic index Index (Tn) mARKwith the scale-up factor of intermediate value MEDIAN (Tn), i.e. Index (Tn) mARK/ MEDIAN (Tn), if this scale-up factor is greater than MPSD (Tn)/MEDIAN (Tn), then real-time traffic index Index (Tn) mARKbe positioned at PB (Tn) between positive overflow area PO (Tn) or positive buffer zone.Be boundary line according to threshold value [MAX (Tn)+MPSD (Tn)]/[2 × MEDIAN (Tn)], being greater than this value exports as " exception of blocking up ", be less than this value and be greater than MPSD (Tn)/MEDIAN (Tn), then exporting as " doubtful exception of blocking up ".If scale-up factor is less than MNSD (Tn)/MEDIAN (Tn), then real-time traffic index Index (Tn) mARKbe positioned at NB (Tn) between negative overflow area NO (Tn) or negative buffer, be boundary line according to threshold value [MIN (Tn)+MNSD (Tn)]/[2 × MEDIAN (Tn)], be less than this value and then export " abnormal unimpeded ", be greater than this value and be less than MNSD (Tn)/MEDIAN (Tn) and then export " doubtful exception is unimpeded ".
Third dimension trend discrimination: calculate real-time traffic index Index (Tn) mARKwith MPSD (Tn) or MNSD (Tn) relative to reference point Ref tslope differences, if Index (Tn) mARKrelative to reference point Ref tslope >MPSD (Tn) relative to reference point Ref tslope, deviation is positivity bias, then real-time traffic index Index (Tn) mARKbe positioned at PB (Tn) between positive overflow area PO (Tn) or positive buffer zone, then according to current time index and positively biased mathematic interpolation [Index (Tn)-Ref t] × T 0/ (Tn-t)-[MPSD (Tn)-Ref t] × T 0/ (Tn-t), wherein, T 0for the update cycle of traffic index, t is the interpolation moment, if its result >5, then export " exception of blocking up ", if its result between (0,5] between, then export " doubtful exception of blocking up ".If Index (Tn) mARKrelative to reference point Ref tslope <MNSD (Tn) relative to reference point Ref tslope, deviation is negative sense deviation, then real-time traffic index Index (Tn) mARKbe positioned at NB (Tn) between negative overflow area NO (Tn) or negative buffer, then according to current time index and negative bias mathematic interpolation [Index (Tn)-Ref t] × T 0/ (Tn-t)-[MNSD (Tn)-Ref t] × T 0/ (Tn-t), if its result <=-5, then exports " abnormal unimpeded ", if its result between (-5,0] between, then export " doubtful exception of blocking up ".Threshold value can be demarcated according to concrete road object.
Step 6, exception combine differentiation: according to the judgement demand of blocking up abnormal, the abnormality diagnostic decision rule of customizable expansion multidimensional.If Index (Tn) mARK>=MEDIAN (Tn), then the decision rule of all dimensions all exports: { " normality ", " doubtful exception of blocking up ", " exception of blocking up " }, if Index (Tn) mARK<MEDIAN (Tn), then export: { " normality ", " doubtful exception is unimpeded ", " abnormal unimpeded " }.Being input as of associating differentiation: { the first dimension exports, second dimension exports, and third dimension exports ..., according to the difference to IMS degree of dependence, two kinds of logic identification abnormal traffic state features can be divided into: " an abnormal voting adopted is fixed " and " the minority is subordinate to the majority ".
" abnormal a voting adopted fixed ": from " normality " to "abnormal" cognitive phase, the first dimension result of determination has a ticket power to make decision, as long as namely think Index (Tn) mARKjust be not identified as doubtful abnormal or abnormal in index memory space IMS (Tn); And returning to " normality " stage from "abnormal", be then on " normality " basis in the first dimension result of determination, jointly confirm abnormal restoring by other dimensions.
" the minority is subordinate to the majority ": all dimension diagnostic results have identical weight, judges final recognition result according to exporting the majority judged.For three dimensions, two " normalities " are namely identified as normality, if three results export all not identical, then can respectively according to " criterion of pessimism " towards "abnormal" direction discernment, or according to " criterion of optimism " towards " normality " direction discernment.
In actual applications, the target road network scale identified as required is different, Different Logic can be used respectively, as to through street whole network, surface road the whole network or administrative section anomalous identification because overall index fluctuation range is less, recommendation " an abnormal voting adopted is fixed " identifies, as to certain section or path, because index fluctuation range is larger, recommendation " the minority is subordinate to the majority " identifies.
7) history index upgrades: real-time traffic index Index (Tn) mARKafter completing index memory space IMS (Tn) coupling, be stored in original historical data base, matching stage uses index memory space IMS (Tn) not comprise current time latest index value.When index memory space IMS (Tn) whole updating, the measurement period that time slice covers is according to removing one day the earliest, add and slide for up-to-date one day, namely according to forgetting one day farthest, remembeing up-to-date one day, ensure that control by kinds measurement period is constant value.
In addition, the exponential quantity that the inventive method uses is all through data check screening and repairing, and quality of data work for the treatment of completed before each link involved in the present invention.The unitary construction of small part date or time shortage of data not Intrusion Index memory space IMS (Tn).
The present invention is illustrated below with a specific embodiment.
Step 1) mark real-time index results: the influence factor mark MARK that (1) is exported by traffic impact set of factors t(1,2,3), mark T moment up-to-date real-time index Index (T), and (2) generate the data Index (T) with mark result mARKand stored in historical data base (step 2), (3) carry out coupling IMS simultaneously t(step 4).
Step 2) create index classification history library: (1) in the original historical data base of index, according to MARK t(1,2,3) mark result and (not containing the up-to-date moment) is divided into groups to historical data in the same time, Main classification influence factor is calendar, weather and the factor such as movable as auxiliary packet factor, (2) with sorted historical data for Exponential Sample overall building IMS (step 3).
Step 3) build IMS:1) " memory space " establishment is carried out to each sorted T moment history index sample population (step 2), (1) intermediate value line MEDIAN (T) is comprised, positive extreme value MAX (T), negative pole value MIN (T), positively biased difference MPSD (T)=MEDIAN (T)+stdev (T) is if MPSD (T) >=MAX (T), then get MAX (T) value, negative bias difference MNSD (T)=MEDIAN (T)-stdev (T) is if MNSD (T) <=MIN (T), then get MIN (T) value, (2) after " five values " line creates, Nature creating " six territories " PO (T), PB (T), PN (T), NN (T), NB (T), NO (T), if MPSD (T)=MAX (T), then do not have PB (T) interval, if in like manner MNSD (T)=MIX (T), then do not have NB (T) interval, (3) create IMS memory space totally to carry out based on sub-index historical data, therefore do not need to carry out in real time, unify for each road, each cycle carry out IMS establishment (step 7) in system idles phase morning every day, result data is that second day real-time index-matched prepares.
Step 4) real-time index-matched IMS: with MARK tthe real-time index Index (T) that (1,2,3) mark mARK(step 1) coupling IMS t(step 3), differentiates Index (T) according to " six territories " mARKbetween location; (1) if [Index (T) mARK∈ PN (T)] ∩ [Index (T) mARK∈ NN (T)], directly export " normality " and differentiate result, empty reference point Ref tand jump out differentiate wait enter the T+I cycle; (2) if [Index (T) mARK∈ PB (T)] ∩ [Index (T) mARK∈ NB (T)], then differentiate whether there is reference point Ref t, if there is no, then generate reference point Ref t, and export " normality " result wait for enter the T+1 cycle, if there is Ref t, then start multidimensional abnormity diagnosis analysis (the second dimension, the third dimension) (step 5) and extremely combine differentiation (step 6), obtaining status flag and differentiate result, wait for and enter the T+I cycle; (3) if [Index (T) mARX∈ PO (T)] ∩ [Index (T) mARK∈ NO (T)], then differentiate whether there is reference point Ref t, if there is no, then generate reference point Ref t, and start multidimensional abnormity diagnosis analysis (all dimensions) (step 5) and extremely combine differentiation (step 6), obtain status flag and differentiate result, and wait enters the T+I cycle, if reference point Ref texist, then start multidimensional abnormity diagnosis and analyze (all dimensions) (step 5) and extremely combine differentiation (step 6), export " doubtful exception " or "abnormal" result, do not empty reference point Ref twait enters the T+1 cycle.
Step 5) analysis of multidimensional abnormity diagnosis: as Index (T) mARKnot when PN (T) or NN (T) (step 4), start the analysis of multidimensional abnormity diagnosis.It is all " normality " that multidimensional abnormity diagnosis analyzes each dimension default value, different according to the trigger condition of real-time matching IMS process, start the abnormity diagnosis analytical algorithm of different dimensions respectively, finally enter all dimension results all conduct inputs when extremely combining differentiation (step 6).
(1) first dimension is that difference differentiates: calculate Index (T) in real time mARKwith the positive error of MAX (T) or MIN (T), if Index (T) mARK∈ PO (T), then Index (T) mARK-MAX (T) also judges to be in " doubtful exception of blocking up " (<5) or " exception of blocking up " (>=5), if Index (T) according to threshold value (as poor in 5 vertex degrees) mARK∈ NO (T), then MN (T)-Index (T) mARKand be in " doubtful exception is unimpeded " (<5) or " abnormal unimpeded " (>=5) according to threshold decision; Threshold value can be demarcated according to concrete road object.
(2) second dimension ratios differentiate: calculate Index (T) in real time mARKwith the scale-up factor of MEDIAN (T), i.e. Index (T) mARK/ MEDIAN (T), if [Index (T) mARK∈ PO (T)] ∩ [Index (T) mARK∈ PB (T)], be boundary line according to threshold value [MAX (T)+MPSD (T)]/(2MEDIAN (T)), be greater than and sentence this value and sentence " exception of blocking up ", be less than this value and be greater than MPSD (T)/MEDIAN (T) and then judge " doubtful exception of blocking up ", if [Index (T) mARK∈ NO (T)] ∩ [Index (T) mARK∈ NB (T)], be boundary line according to threshold value [MN (T)+MNSD (T)]/(2MEDIAN (T)), be less than and sentence this value and sentence " abnormal unimpeded ", be greater than this value and be less than MNSD (T)/MEDIAN (T) and then judge " doubtful exception is unimpeded ".
(3) third dimension trend discrimination: calculate Index (T) in real time mARKwith MPSD (T) or MNSD (T) relative to reference point Ref tslope differences, if [Indcx (T) mARK∈ pO(T)] ∩ [INdex (T) mARK∈ PB (T)], then according to current time index and positively biased mathematic interpolation [Index (T) mARK-REF t] T 0/ (T-t)-[MPSD (T)-REF t] T 0/ (T-t), according to threshold value 5 vertex degree judge, if >5, sentence " exception of blocking up ", between (0,5] between sentence " doubtful exception of blocking up ", if [Index (T) mARK∈ NO (T)] ∩ [Index (T) mARK∈ NB (T)], then according to current time index and negative bias mathematic interpolation [INdex (T) mARK-REF t] T 0/ (T-t-[MNSD (T)-REF t] T 0/ (T-t), judges according to threshold value-5 vertex degree, if <=-5, sentences " abnormal unimpeded ", between (-5,0] between sentence " doubtful exception of blocking up "; Threshold value can be demarcated according to concrete road object.
Step 6) extremely combine differentiation: according to the judgement demand of blocking up abnormal, the decision rule of customizable expansion multidimensional abnormity diagnosis (step 5).If Index (T) mARK>=MEDIAN (T), then the decision rule of all dimensions all exports: { " normality ", " doubtful exception of blocking up ", " exception of blocking up " }, if Index (T) mARK<MEDIAN (T), then export: { " normality ", " doubtful exception is unimpeded ", " abnormal unimpeded " }.Being input as of associating differentiation: the first dimension exports, and the second dimension exports, and third dimension exports ..., according to the difference to IMS degree of dependence, two kinds of logic identification abnormal traffic state features can be divided into: " a ticket power to make decision " and " the minority is subordinate to the majority ".
In actual applications, the target road network scale identified as required is different, Different Logic can be used respectively, as to through street whole network, surface road the whole network or administrative section anomalous identification because overall index fluctuation range is less, " an abnormal voting adopted is fixed " is used to identify, as to certain section or path, because index fluctuation range is larger, " the minority is subordinate to the majority " is used to identify.
Step 7) history index renewal: Index (T) in real time mARKafter completing IMS coupling, be stored in original historical data base, matching stage uses IMS memory space not comprise current time latest index value (step 4).The measurement period that time slice T covers, according to removing one day the earliest, added and slided for up-to-date one day when IMS memory space unitary construction every day, ensured that control by kinds measurement period is constant value.

Claims (6)

1. an abnormal traffic state characteristic recognition method for real-time index-matched memory range, is characterized in that, step is:
Step 1, the traffic flow parameter of a real time data in the interval dynamic fluctuation of fixed data is chosen arbitrarily as traffic status identification parameter in all traffic flow parameters, obtain all historical traffic state recognition supplemental characteristics of certain section in some time periods or Regional Road Network, create the traffic status identification parameters memorizing space corresponding with this moment according to all traffic status identification parameters under synchronization and calculate the intermediate value in each traffic status identification parameters memorizing space, wherein, the traffic status identification parameters memorizing space that moment T is corresponding is IMS (T), minimal index value MIN (T) in all historical traffic state recognition supplemental characteristics of moment T and greatest exponential value MAX (T) is respectively negative edge and the positive boundary of traffic status identification parameters memorizing space IMS (T), the average that the intermediate value MEDIAN (T) of traffic status identification parameters memorizing space IMS (T) is traffic status identification supplemental characteristics all under moment T,
Step 2, each traffic status identification parameters memorizing space to be at least divided between positive buffer zone and between negative buffer, and the region being less than minimal index value is defined as negative overflow area, the region being greater than greatest exponential value is defined as positive overflow area; Be PB (T) between the positive buffer zone of traffic status identification parameters memorizing space IMS (T), span be (MEDIAN (T), MAX (T)]; Be NB (T) between negative buffer, span be [MIN (T), MEDIAN (T)); Negative overflow area is NO (T), span be [MIN, MIN (T)); Positive overflow area is PO (T), span be (MAX (T), MAX], wherein, MIN is that traffic status identification parameter is minimum may value, and MAX is the maximum possible value of traffic status identification parameter;
Step 3, obtain the real-time traffic states identification parameter Index (Tn) of this section under current time Tn, to judge between the positive buffer zone whether real-time traffic states identification parameter Index (Tn) is positioned at traffic status identification parameters memorizing space IMS (Tn) NB (Tn) between PB (Tn) or negative buffer, if, then the traffic behavior of this section under current time Tn is normality, if not, then the traffic behavior of this section under current time Tn is abnormal, this exception refers to extremely unimpeded or abnormal blocking up, if to be traffic status identification parameter more high more blocks up for the traffic status identification parameter model that this section adopts, if then real-time traffic states identification parameter Index (Tn) is positioned at positive overflow area PO (Tn), the traffic behavior of this section under current time Tn blocks up for abnormal, if real-time traffic states identification parameter Index (Tn) is positioned at negative overflow area NO (Tn), the traffic behavior of this section under current time Tn is abnormal unimpeded, if it is more high more unimpeded that the traffic status identification parameter model that this section adopts is traffic status identification parameter, if then real-time traffic states identification parameter Index (Tn) is positioned at positive overflow area PO (Tn), the traffic behavior of this section under current time Tn is abnormal unimpeded, if real-time traffic states identification parameter Index (Tn) is positioned at negative overflow area NO (Tn), the traffic behavior of this section under current time Tn blocks up for abnormal.
2. the abnormal traffic state characteristic recognition method of a kind of real-time index-matched memory range as claimed in claim 1, it is characterized in that, in described step 1, influence factor is utilized to mark MARK, all historical traffic state recognition supplemental characteristics in certain section in some time periods are marked, influence factor mark MARK is the one group of cartesian product configuration item be made up of the non-traffic factor having extensively impact to traffic behavior feature in influence factor config set, all historical traffic state recognition parameters under synchronization are according to influence factor mark MARK classification, create traffic status identification parameters memorizing space respectively for each class and calculate the intermediate value in each traffic status identification parameters memorizing space,
In described step 3, after obtaining the real-time traffic states identification parameter Index (Tn) under current time Tn, first utilize influence factor to mark MARK to it and mark, obtain the traffic status identification parameter Index (Tn) with mark result mARK, utilize traffic status identification parameter Index (Tn) mARKtraffic status identification parameters memorizing space corresponding with its classification under finding moment Tn, then judge that whether the traffic behavior of this section under current time Tn be abnormal according to the relation of between itself and Zhong Zheng buffer zone, traffic status identification parameters memorizing space and between negative buffer and negative overflow area and positive overflow area.
3. the abnormal traffic state characteristic recognition method of a kind of real-time index-matched memory range as claimed in claim 2, it is characterized in that, described Weather information Weather, calendar information Calendar and action message Event are at least comprised on the non-traffic factor that traffic behavior feature has an extensively impact, wherein, calendar information Calendar is main factor, each moment has different influence factor mark MARK, and the influence factor of moment T is labeled as MARK t(1,2,3)=MARK t(Calendar, Weather, Event).
4. the abnormal traffic state characteristic recognition method of a kind of real-time index-matched memory range as claimed in claim 1, it is characterized in that, in described step 2, each traffic status identification parameters memorizing space is at least divided between positive buffer zone, normal state is interval, bear between normality interval and negative buffer; Between the positive buffer zone of then traffic status identification parameters memorizing space IMS (T), the span of PB (T) is (MPSD (T), MAX (T)], MPSD (T)=MEDIAN (T)+stdev (T), stdev (T) is the standard deviation of historical traffic state recognition parameters all under moment T; Normal state interval is PN (T), and span is [MEDIAN (T), MPSD (T)]; Negative normality interval is NN (T), span be [MNSD (T), MEDIAN (T)), MNSD (T)=MEDIAN (T)-stdev (T); Between negative buffer the span of NB (T) be [MIN (T), MNSD (T));
In described step 3, obtain the real-time traffic states identification parameter Index (Tn) of this section under current time Tn, judge whether real-time traffic states identification parameter Index (Tn) is positioned at the interval PN (Tn) of normal state or the interval NN (Tn) of negative normality of traffic status identification parameters memorizing space IMS (Tn), if, then the traffic behavior of this section under current time Tn is normal, if not, then the traffic behavior of this section under current time Tn is abnormal, this exception refers to extremely unimpeded or abnormal blocking up, if to be traffic status identification parameter more high more blocks up for the traffic status identification parameter model that this section adopts, if then real-time traffic states identification parameter Index (Tn) be positioned at traffic status identification parameters memorizing space IMS (Tn) positive buffer zone between PB (Tn) or positive overflow area PO (Tn), the traffic behavior of this section under current time Tn blocks up for abnormal, if NB (Tn) or negative overflow area NO (Tn) between the negative buffer that real-time traffic states identification parameter Index (Tn) is positioned at traffic status identification parameters memorizing space IMS (Tn), the traffic behavior of this section under current time Tn is abnormal unimpeded, if it is more high more unimpeded that the traffic status identification parameter model that this section adopts is traffic status identification parameter, if then real-time traffic states identification parameter Index (Tn) be positioned at traffic status identification parameters memorizing space IMS (Tn) positive buffer zone between PB (Tn) or positive overflow area PO (Tn), the traffic behavior of this section under current time Tn is abnormal unimpeded, if NB (Tn) or negative overflow area NO (Tn) between the negative buffer that real-time traffic states identification parameter Index (Tn) is positioned at traffic status identification parameters memorizing space IMS (Tn), the traffic behavior of this section under current time Tn blocks up for abnormal.
5. the abnormal traffic state characteristic recognition method of a kind of real-time index-matched memory range as claimed in claim 4, it is characterized in that, in described step 3, if NB (Tn) between PB (Tn) or negative buffer between the positive buffer zone that real-time traffic states identification parameter Index (Tn) is positioned at traffic status identification parameters memorizing space IMS (Tn), then any one adopting in following two methods judges the traffic behavior of current road segment under current time Tn:
First method: if real-time traffic states identification parameter Index (Tn) is positioned at PB (Tn) between positive buffer zone, then with threshold value [MAX (Tn)+MPSD (Tn)]/[2 × MEDIAN (Tn)] for boundary line, be greater than this value and sentence "abnormal", be less than this value and be greater than MPSD (Tn)/MEDIAN (Tn) and then judge " doubtful exception "; If real-time traffic states identification parameter Index (Tn) is positioned at NB between negative buffer (Tn), then with threshold value [MIN (Tn)+MNSD (Tn)]/[2 × MEDIAN (Tn)] for boundary line, be less than this value and sentence "abnormal", be greater than this value and be less than MNSD (Tn)/MEDIAN (Tn) and then judge " doubtful exception ";
Second method: judge whether to there is reference point Ref t, reference point Ref tby real-time traffic states identification parameter Index (Tn) by the interval PN(Tn of normal state) or the interval NN(Tn of negative normality) enter between positive buffer zone first, between negative buffer, when positive overflow area or negative overflow area, create with MPSD (Tn) or MNSD (Tn) crossing interpolation, the interpolation moment is rounded to the previous moment in crossing moment, if not, then sentence " normality ", if, then judge that real-time traffic states identification parameter Index (Tn) to be positioned between positive buffer zone NB (Tn) between PB (Tn) or negative buffer, if be positioned at PB (Tn) between positive buffer zone, then calculate [Index (Tn)-Ref t] × T 0/ (Tn-t)-[MPSD (Tn)-Ref t] × T 0/ (Tn-t), wherein, T 0for the update cycle of traffic status identification parameter, t is the interpolation moment, if its result is greater than A, then sentences "abnormal", between (0, A] between then sentence " doubtful exception ", A is empirical value, if be positioned at NB between negative buffer (Tn), then calculate [Index (Tn)-Ref t] × T 0/ (Tn-t)-[MNSD (Tn)-Ref t] × T 0/ (Tn-t), if its result is not more than-A, then sentences "abnormal", between (-A, 0] between then sentence " doubtful exception ".
6. the abnormal traffic state characteristic recognition method of a kind of real-time index-matched memory range as claimed in claim 4, is characterized in that, in described step 3, utilize three-dimensional extremely to combine the traffic behavior differentiating and judge certain section, then described step 3 comprises:
Step 3.1, obtain this section or the real-time traffic states identification parameter Index (Tn) of Regional Road Network under current time Tn;
Step 3.2, judge whether real-time traffic states identification parameter Index (Tn) is positioned between positive buffer zone NB (Tn) between PB (Tn) or negative buffer, if not, then judges whether to there is reference point Ref t, reference point Ref tto be entered between positive buffer zone, between negative buffer by real-time traffic states identification parameter Index (Tn), positive overflow area or negative overflow area time, create with MPSD (Tn) or MNSD (Tn) crossing interpolation, if exist, directly enter next step, if do not exist, then create reference point Ref tafter enter next step; If real-time traffic states identification parameter Index (Tn) to be positioned between positive buffer zone NB (Tn) between PB (Tn) or negative buffer, then judge whether to there is reference point Ref tif exist, then sentence " normality ", completing steps 3, if do not exist, then created reference point Ref tafter enter step 3.4;
Step 3.3, the first dimension abnormality juding:
If real-time traffic states identification parameter Index (Tn) is positioned at positive overflow area PO (Tn), then calculate Index (Tn)-MAX (Tn), if this difference is less than A, then sentence " doubtful exception ", enter next step, otherwise, sentence "abnormal", enter next step; If real-time traffic states identification parameter Index (Tn) is positioned at negative overflow area NO (Tn), then calculate MIN (Tn)-Index (Tn), if this difference is less than A, then sentence " doubtful exception ", enter next step, otherwise, sentence "abnormal", enter next step, A is empirical value;
Step 3.4, the second dimension abnormality juding:
If real-time traffic states identification parameter Index (Tn) is positioned at PB (Tn) or positive overflow area PO (Tn) between positive buffer zone, then with threshold value [MAX (Tn)+MPSD (Tn)]/[2 × MEDIAN (Tn)] for boundary line, Index (Tn)/MEDIAN (Tn) is greater than this value and sentences "abnormal", enter next step, Index (Tn)/MEDIAN (Tn) is less than this value and is greater than MPSD (Tn)/MEDIAN (Tn) and then judges " doubtful exception ", enters next step; If real-time traffic states identification parameter Index (Tn) is positioned at NB between negative buffer (Tn) or negative overflow area NO (Tn), then with threshold value [MIN (Tn)+MNSD (Tn)]/[2 × MEDIAN (Tn)] for boundary line, Index (Tn)/MEDIAN (Tn) is less than this value and sentences "abnormal", enter next step, Index (Tn)/MEDIAN (Tn) is greater than this value and is less than MNSD (Tn)/MEDIAN (Tn) and then judges " doubtful exception ", enters next step;
Step 3.5, third dimension abnormality juding:
If be positioned at PB (Tn) or positive overflow area PO (Tn) between positive buffer zone, then calculate [Index (Tn)-Ref t] × T 0/ (Tn-t)-[MPSD (Tn)-Ref t] × T 0/ (Tn-t), wherein, T 0for the update cycle of traffic status identification parameter, t is the interpolation moment, if its result is greater than A, then sentences "abnormal", enters next step, between (0, A] between then sentence " doubtful exception ", enter next step; If be positioned at NB between negative buffer (Tn) or negative overflow area NO (Tn), then calculate [Index (Tn)-Ref t] × T 0/ (Tn-t)-[MNSD (Tn)-Ref t] × T 0/ (Tn-t), if its result is not more than-A, then sentences "abnormal", enters next step, between (-A, 0] between then sentence " doubtful exception ", enter next step;
Step 3.6, adopt " abnormal a voting adopted fixed " or " the minority is subordinate to the majority " to carry out exception to combine differentiation, wherein, " an abnormal voting adopted is fixed " refers to: at traffic behavior from " normality " to "abnormal" cognitive phase, the first dimension abnormality juding result described in step 3.3 has a ticket power to make decision; And returning to " normality " stage from "abnormal", be then on " normality " basis in the result of the first dimension abnormality juding, jointly confirm abnormal restoring by the second dimension abnormality juding described in step 3.4 and the third dimension abnormality juding described in step 3.5;
" the minority is subordinate to the majority " refers to: the first dimension abnormality juding result described in step 3.3, the second dimension abnormality juding result described in 3.4 and the third dimension abnormality juding result described in step 3.5 have identical weight, and the majority according to exporting result of determination judges final recognition result; If the result of determination of three dimensionality exports all not identical, then or according to " criterion of pessimism " towards "abnormal" direction discernment, or according to " criterion of optimism " towards " normality " direction discernment.
CN201310451579.9A 2013-09-28 2013-09-28 A kind of abnormal traffic state characteristic recognition method of real-time index-matched memory range Active CN103514743B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310451579.9A CN103514743B (en) 2013-09-28 2013-09-28 A kind of abnormal traffic state characteristic recognition method of real-time index-matched memory range

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310451579.9A CN103514743B (en) 2013-09-28 2013-09-28 A kind of abnormal traffic state characteristic recognition method of real-time index-matched memory range

Publications (2)

Publication Number Publication Date
CN103514743A CN103514743A (en) 2014-01-15
CN103514743B true CN103514743B (en) 2016-01-06

Family

ID=49897408

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310451579.9A Active CN103514743B (en) 2013-09-28 2013-09-28 A kind of abnormal traffic state characteristic recognition method of real-time index-matched memory range

Country Status (1)

Country Link
CN (1) CN103514743B (en)

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105387865B (en) * 2015-10-16 2020-07-03 上海博泰悦臻网络技术服务有限公司 Route planning method and system based on traffic road data
CN108122408B (en) * 2016-11-29 2021-02-05 中国电信股份有限公司 Road condition monitoring method and device and system for monitoring road conditions
CN109923595B (en) * 2016-12-30 2021-07-13 同济大学 Urban road traffic abnormity detection method based on floating car data
CN107123264A (en) * 2017-05-31 2017-09-01 温州市鹿城区中津先进科技研究院 A kind of method that abnormal congestion is judged based on traffic big data
CN108109375B (en) * 2017-12-21 2020-10-09 北京航空航天大学 Urban traffic network reliability prediction method based on seepage analysis
CN108961761B (en) * 2018-08-14 2020-04-24 百度在线网络技术（北京）有限公司 Method and apparatus for generating information
CN110349410B (en) * 2019-08-16 2020-11-13 甘肃万华金慧科技股份有限公司 Traffic congestion index measuring method
CN110567472A (en) * 2019-08-16 2019-12-13 武汉东湖大数据交易中心股份有限公司 Vehicle-mounted meteorological data management method and device, terminal equipment and storage medium
CN113808384B (en) * 2020-06-16 2023-02-10 英业达科技有限公司 Traffic condition detection method
CN114724361A (en) * 2022-03-16 2022-07-08 阿里巴巴（中国）有限公司 Road condition determination method and device and computer readable storage medium
CN115830873B (en) * 2023-01-10 2023-05-02 西南交通大学 Urban road traffic event classification method, device, equipment and readable storage medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101540099A (en) * 2008-03-17 2009-09-23 上海宝康电子控制工程有限公司 Method and system for judging road traffic states
CN101739815A (en) * 2009-11-06 2010-06-16 吉林大学 On-line recognition method of road traffic congestion state
CN101751777A (en) * 2008-12-02 2010-06-23 同济大学 Dynamic urban road network traffic zone partitioning method based on space cluster analysis
CN202075864U (en) * 2011-04-28 2011-12-14 北京市劳动保护科学研究所 Abnormal traffic state automatic detection system

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2010191614A (en) * 2009-02-17 2010-09-02 Sumitomo Electric Ind Ltd Traffic information estimation system, estimation method, provision system and computer program
JP5491104B2 (en) * 2009-08-28 2014-05-14 株式会社アイ・トランスポート・ラボ Traffic condition analysis device, traffic condition analysis program, and traffic condition analysis method
WO2013011796A1 (en) * 2011-07-20 2013-01-24 住友電気工業株式会社 Traffic evaluation device, computer program and traffic evaluation method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101540099A (en) * 2008-03-17 2009-09-23 上海宝康电子控制工程有限公司 Method and system for judging road traffic states
CN101751777A (en) * 2008-12-02 2010-06-23 同济大学 Dynamic urban road network traffic zone partitioning method based on space cluster analysis
CN101739815A (en) * 2009-11-06 2010-06-16 吉林大学 On-line recognition method of road traffic congestion state
CN202075864U (en) * 2011-04-28 2011-12-14 北京市劳动保护科学研究所 Abnormal traffic state automatic detection system

Also Published As

Publication number Publication date
CN103514743A (en) 2014-01-15

Similar Documents

Publication Publication Date Title
CN103514743B (en) A kind of abnormal traffic state characteristic recognition method of real-time index-matched memory range
Ozbayoglu et al. A real-time autonomous highway accident detection model based on big data processing and computational intelligence
Almatar Traffic congestion patterns in the urban road network:(Dammam metropolitan area)
Weijermars Analysis of urban traffic patterns using clustering
CN103413443A (en) Short-term traffic flow forecasting method based on hidden Markov model
CN112927513B (en) Real-time online traffic simulation method and system
CN113033840B (en) Method and device for judging highway maintenance
Olayode et al. PREDICTION OF VEHICULAR TRAFFIC FLOW USING LEVENBERG-MARQUARDT ARTIFICIAL NEURAL NETWORK MODEL: ITALY ROAD TRANSPORTATION SYSTEM.
Xie et al. A spatiotemporal apriori approach to capture dynamic associations of regional traffic congestion
Zhang et al. Hybrid traffic speed modeling and prediction using real-world data
CN113868492A (en) Visual OD (origin-destination) analysis method based on electric police and checkpoint data and application
Saracoglu et al. Estimation of traffic incident duration: a comparative study of decision tree models
Kodi Evaluating the mobility and safety benefits of adaptive signal control technology (ASCT)
Shen Freeway travel time estimation and prediction using dynamic neural networks
Dias et al. From data to action: Exploring AI and IoT-driven solutions for smarter cities
Zhang et al. Finding Paths With Least Expected Time in Stochastic Time-Varying Networks Considering Uncertainty of Prediction Information
Jiang et al. Artificial intelligence-enabled smart city construction
Liu et al. Highway traffic congestion detection and evaluation based on deep learning techniques
Wang et al. Time-space relationship analysis model on the bus driving characteristics of different drivers based on the traffic performance index system
Bavar et al. Evaluating the spatial effects of environmental influencing factors on the frequency of urban crashes using the spatial Bayes method based on Euclidean distance and contiguity
Silva et al. Interpreting traffic congestion using fundamental diagrams and probabilistic graphical modeling
Lei et al. Research on multi-objective bus route planning model based on taxi GPS data
Moreira-Matias et al. An online learning framework for predicting the taxi stand's profitability
Golze et al. Impact analysis of accidents on the traffic flow based on massive floating car data
Neumann A cost-effective method for the detection of queue lengths at traffic lights

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
TR01 Transfer of patent right

Effective date of registration: 20170411

Address after: 830000 the Xinjiang Uygur Autonomous Region Urumqi City, Kanas economic and Technological Development Zone No. 455 Hubei road Xinjiang Powerise Software Park building block A room 6007

Patentee after: Xinjiang City Branch intelligent Polytron Technologies Inc

Address before: 200063 Wuning Road, Putuo District, No. 505, building 53, building 7, building

Patentee before: Shanghai SEARI Intelligent System Co., Ltd.

TR01 Transfer of patent right