CN101587644A - Automatic detection method for traffic accident on urban expressway based on non-continuous sliding sequence - Google Patents

Abstract

The invention provides an automatic detection method for traffic accident on urban expressway based on non-continuous sliding sequence, which collects real time traffic flow information through a detection system and an earth coil; whether a traffic accident happens is judged according to the obtained average time occupancy data with process using a non-continuous sliding one-minute occupancy sequence. If an accident happens, the alarm sends out for announcing the traffic management department to take corresponding measures for the accident, otherwise the data is continued to be collected for detection. The method provided by the invention has the advantages of high detection rate for detection, low false alarm rate, short testing time, selected index with clear meaning, and is suitable for realizing engineering.

Description

City expressway automatic detection method for traffic accident based on non-continuous sliding sequence

Technical field

The present invention relates to a kind of city expressway automatic detection method for traffic accident, particularly be based on the city expressway automatic detection method for traffic accident of 1 minute occupation rate sequence of discontinuous slip, can be used for whether existing traffic hazard to carry out real time automatic detection, relate to traffic intelligent management and control technology city expressway.

Background technology

Traffic hazard is meant: continue for some time, cause road passage capability to reduce, and background traffic flow situation is produced the accident of appreciable impact.City expressway is being brought into play enormous function as the skeleton of urban road traffic network.The burst traffic hazard not only is embodied in safety of person vehicle is formed threat the influence of city expressway, and more important is to produce comparatively serious traffic congestion bottleneck, and causes the Urban Expressway System operational efficiency to reduce.Reduce the negative effect of burst accident to Urban Expressway System, one of key link is the accident of finding fast and accurately.

Traffic hazard detects the element that (Automatic Incident Detection) is advanced person's traffic control and management system automatically, its objective is by finding traffic hazard as early as possible, carries out the scene cleaning as early as possible, and the minimizing accident is to the influence of traffic flow.Along with the appearance of (often claiming vehicle checker) of traffic data real-time automatic collecting equipment, various automatic accident detection method ultimate principles based on vehicle checker are identical: by the variation of identification accident generation back road traffic condition, reach the purpose of the accident of detecting.Automatic accident detection method, owing to do not need people's participation, and the variation to traffic has automatic identification capability, discovery accident is quick and precisely gone forward side by side and is acted so report to the police, therefore take the emergency processing measure as early as possible after more helping accident to take place, reduce duration of fault, the loss that the reduction accident is caused becomes the focus that various countries are studied.

Automatically the development of accident detection algorithm mainly comprises state recognition, traffic and theoretical model, statistics, artificial intelligence and image recognition etc. so far at present.But have its limitation separately: the state recognition method requires rule of thumb calibration index threshold value, but is using the later stage to be difficult to revise; Catastrophe theory is difficult enforcement on the city expressway of traffic stream characteristics complexity; Mathematical statistics and artificial intelligence all need a large amount of detailed accident samples to learn; Image recognition algorithm is subjected to weather effect bigger.In the above algorithm, have only the state recognition algorithm to obtain the application verification on the engineering, other algorithms only stay in theory or simulation study stage mostly.

Summary of the invention

The objective of the invention is to overcome the deficiency that existing automatic accident detection algorithm exists, propose a kind of city expressway automatic detection method for traffic accident based on 1 minute occupation rate sequence of discontinuous slip, its verification and measurement ratio, detection time and rate of false alarm all are better than existing algorithm.

For reaching above purpose, solution of the present invention is:

A kind of city expressway automatic detection method for traffic accident based on non-continuous sliding sequence, it may further comprise the steps:

1) detection system is gathered the telecommunication flow information of a fixed step size in real time by earth coil;

2) judge according to the real-time traffic stream information of gathering whether have traffic hazard take place, if judgement has an accident then give the alarm, notice vehicle supervision department takes corresponding measure if detecting the highway section; Otherwise continue to read real time data, judge next time;

The collection of described real-time traffic stream information generates 1 minute occupation rate data sequence of discontinuous slip as the basic data of differentiating algorithm.

Structure serves as the 1 minute occupation rate sequence of discontinuous slip at interval of sliding with minimum period 20s, and its construction method is:

$\mathrm{if}\left\{\begin{array}{ccc}{o}_i^{20s}>1.5*o_{i-1}^{\mathrm{mov}5m}& \mathrm{or}& o_i^{20s}<0.5*o_{i-1}^{\mathrm{mov}5m}\\ o_{i+1}^{20s}>1.5*o_{i-1}^{\mathrm{mov}5m}& \mathrm{or}& o_{i+1}^{20s}<0.5*o_{i-1}^{\mathrm{mov}5m}\\ o_{i+2}^{20s}>1.5*o_{i-1}^{\mathrm{mov}5m}& \mathrm{or}& o_{i+2}^{20s}<0.5*o_{i-1}^{\mathrm{mov}5m}\end{array}\right.,O_i=o_{i+2}^{1m}$

else， $O_i=o_i^{\mathrm{mov}1m}$

Wherein: O _i---i value of 1 minute occupation rate sequence of discontinuous slip;

o _i ^1m---i value of average 1 minute occupation rate sequence;

o _i ^Mov1m---with 20s is at interval, i value of the 1 minute occupation rate sequence of sliding;

o _i ^20s---i value of 20s occupation rate sequence;

o _i ^Mov5m---with 20s is at interval, i value of the 5 minutes occupation rate sequences of sliding.

Judge the method whether highway section has an accident that detects:

1) handles the detection road section traffic volume stream information of gathering in real time, generate 1 minute occupation rate data sequence of discontinuous slip;

2) determine the threshold value of judge index;

Whether 3) the correlated judgment index of calculating 1 minute occupation rate data sequence of discontinuous slip is compared with decision threshold, judge to have an accident.

The correlated judgment index of 1 minute occupation rate data sequence of described discontinuous slip is respectively occupation rate O _{T, i}, 1 minute occupation rate of i section t discontinuous slip constantly; Occupation rate time difference d _{T, i} ^t, i section t constantly with the difference of 1 minute occupation rate of t-1 discontinuous slip constantly; Occupation rate space difference d _{T, i} ^s, the difference of t moment upstream i section and 1 minute occupation rate of the discontinuous slip of adjacent downstream i+1 section; Occupation rate space time difference d _{T, i} ^s→ t, t be occupation rate space difference and t-1 moment occupation rate space difference poor constantly.

Threshold value determination method is: based on the off-line calibration of historical casualty data or based on predetermined threshold value and in one day accident and non-casualty data online review one's lessons by oneself just definite.

Described occupation rate O _{T, i}, occupation rate time difference d _{T, i} ^t, occupation rate space difference d _{T, i} ^sWith occupation rate space time difference d _{T, i} ^{S → t}When all being not less than threshold value, the judgement accident takes place.

According to the needs that accident detects automatically, the basic data that is used for the judgement accident should satisfy following two basic demands: under accident situation, have good mutability, be convenient to carry out accident identification; Under non-accident situation, have good stationarity, reduce the false alarm under the normal traffic situation.Adopt 1 minute occupation rate sequence of discontinuous slip to carry out data processing, can realize this two targets effectively.

Owing to adopted such scheme, the present invention to have following characteristics: accident verification and measurement ratio height of the present invention, false alarm rate is low, and detection time is short, and index for selection has clear and definite meaning, is fit to Project Realization.

Description of drawings

Fig. 1 is 1 minute occupation rate sequence of discontinuous slip (O _i) algorithm flow that makes up;

Fig. 2 is the signal of many indexs serial accident determination methods;

Fig. 3 is automatic accident trace routine flow process;

Fig. 4 is the off-line calibration method based on historical casualty data;

Fig. 5 is the online from modification method of threshold value;

Fig. 6 is that threshold value is from the correction algorithm flow process.

Embodiment

The present invention is further illustrated below in conjunction with the accompanying drawing illustrated embodiment.

Studies show that, for 20 seconds be the time average occupation rate data of step-length, its undulatory property is bigger, is difficult to directly apply to automatic accident detection algorithm.If employing was the time average occupation rate data sequence of step-length with 1 minute, have mutability under the best accident, but sense cycle was extended for 1 minute from 20 seconds; And 20 seconds serving as that the 1 minute occupation rate sequence of slip at interval of sliding has stationarity under the best non-accident, but the sudden change under the accident is obvious inadequately.The minimum period 20s that 1 minute occupation rate sequence of discontinuous slip is obtained with data slides at interval, combines the advantage of above two kinds of data sequences, and its construction method is as follows:

$\mathrm{if}\left\{\begin{array}{ccc}{o}_i^{20s}>1.5*o_{i-1}^{\mathrm{mov}5m}& \mathrm{or}& o_i^{20s}<0.5*o_{i-1}^{\mathrm{mov}5m}\\ o_{i+1}^{20s}>1.5*o_{i-1}^{\mathrm{mov}5m}& \mathrm{or}& o_{i+1}^{20s}<0.5*o_{i-1}^{\mathrm{mov}5m}\\ o_{i+2}^{20s}>1.5*o_{i-1}^{\mathrm{mov}5m}& \mathrm{or}& o_{i+2}^{20s}<0.5*o_{i-1}^{\mathrm{mov}5m}\end{array}\right.,O_i=o_{i+2}^{1m}$

else， $O_i=o_i^{\mathrm{mov}1m}$

In the formula: O _i---i value of 1 minute occupation rate sequence of discontinuous slip;

o _i ^1m---i value of average 1 minute occupation rate sequence;

o _i ^Mov1m---with 20s is at interval, i value of the 1 minute occupation rate sequence of sliding;

o _i ^20s---i value of 20s occupation rate sequence.

o _i ^Mov5m---with 20s is at interval, i value of the 5 minutes occupation rate sequences of sliding;

Before algorithm begins to make up 1 minute occupation rate sequence of continuous slip, should at least have 5 minutes 20s at interval raw data exist.Be exemplified below:

20s original occupation rate data sequence at interval:

18 19 17 18 19 20 21 19 19 20 21 19 21 17 22 24 21 28 35 42 43 45 48 41 36 20 17 19 18 20

1 minute occupation rate sequence (since the 5th minute end) of the corresponding discontinuous slip that makes up:

20.0 20.0 20.0 24.3 24.3 40.0 40.0 40.0 44.7 44.7 44.7 44.7 44.7 24.3 24.3

With the algorithm that 1 minute occupation rate data sequence of discontinuous slip is a fundamental construction, it is consistent with the shortest interval 20s of raw data that sense cycle can reach, and guaranteed the higher detection frequency, helps the timely discovery of accident.

As the basic data that the judgement accident takes place, determine four indexs of the automatic accident detection that this algorithm adopts with 1 minute occupation rate data sequence of discontinuous slip, adopt " serial judgement " to carry out accident and judge as many indexs accident associating determination methods:

(1) occupation rate (O _{T, i}), 1 minute occupation rate of i section t discontinuous slip constantly.

(2) occupation rate time difference (d _{T, i} ^t), i section t constantly with the difference of 1 minute occupation rate of (t-1) discontinuous slip constantly. $d_{t,i}^t=O_{t,i}-O_{t-1,i}$

(3) occupation rate space difference (d _{T, i} ^s), the difference of t moment upstream i section and 1 minute occupation rate of the discontinuous slip of adjacent downstream i+1 section. $d_{t,i}^s=O_{t,i}-O_{t,i+1}$

(4) occupation rate space time difference (d _{T, i} ^{S → t}), t is occupation rate space difference and t-1 moment occupation rate space difference poor constantly. $d_{t,i}^{s\&RightArrow;t}=d_{t,i}^s-d_{t-1,i}^s$

Order is judged above-mentioned 4 indexs, is judged as accident condition when satisfying condition simultaneously, and wherein any one index is ineligible then jumps out judgement, thinks that accident free takes place.

Implementation method of the present invention is as follows, with reference to shown in Figure 3:

1. arrange checkout equipment (earth coil) in city expressway surveyed area upstream and downstream, gather occupation rate parameter averaging time with certain sample unit (as 20 seconds), as the basic data of algorithm.

2. determine the threshold value of judge index.To some sections, the elementary section in the identical track in same detection highway section, adopt the same threshold group, promptly same threshold vector.To determining of metrics-thresholds, the present invention proposes the off-line calibration method of two kinds of method: A. based on historical casualty data, and B. is based on predetermined threshold value and accident and non-casualty data online from modification method in a day.

A. based on the off-line calibration method of historical casualty data

At first, delimit highway section and track to be calibrated,, transfer the data in respective stretch track in the historical data base according to the accident record of confirming with artificial supervision, telephone call or other modes;

Subsequently, calculate the maximal value that obtains each key parameter.Adopt limited exhaustive thought, avoid doing too much unnecessary computing.Promptly earlier determine the scope that it is exhaustive, needn't carry out computing surpassing its peaked scope according to each key parameter maximal value; According to its situation of change each key parameter is determined corresponding step-size in search respectively again, thereby obtain the threshold vector of some various combinations.

At last, these threshold vectors are moved detection algorithm one by one, obtain the detection effect of each threshold value grouping, choose and detect one group of best threshold value of effect as recommending threshold value to historical accident.

Be exemplified below table:

Table 1 is based on historical data off-line calibration key parameter

Parameter	Occupation rate (%)	Occupation rate time difference (%)	Occupation rate space difference (%)	Occupation rate space time difference (%)
					Maximal value	71	44	44	34
Exhaustive scope	0-75	0-45	0-45	0-35
					Step-size in search	5	5	5	5
The final threshold value of determining	30	20	25	5

B. based on predetermined threshold value and accident and non-casualty data online in a day, judge relatively by three grades mainly to constitute that the algorithm detailed process is with reference to figure 6 from modification method.

At first, according to historical data analysis,, obtain predetermined threshold value vector T=[T for the default one group of threshold value in each track, highway section ₁, T ₂, T ₃, T ₄], T wherein _iIt is the predetermined threshold value of i index; Carry out one-day operation according to T,, acquire the set of bare maximum separately of following four indexs of normal traffic situation in a day 24 hours by confirming to detecting actual the having an accident in highway section $t^{\max }=[{\mathrm{<figure-callout\; id="1"\; label="t"\; filenames="A2009100523520011H1.png"\; state="\{\{state\}\}">t</figure-callout>}}_1^{\max },{\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="A2009100523520011H1.png"\; state="\{\{state\}\}">t</figure-callout>}}_2^{\max },t_3^{\max },{\mathrm{<figure-callout\; id="4"\; label="t"\; filenames="A2009100523520014H1.png"\; state="\{\{state\}\}">t</figure-callout>}}_4^{\max }],$ And the set of the relative minimum of following four indexs of accident situation, promptly can distinguish the minimum value set of accident condition and normal condition $t^{\min }=[{\mathrm{<figure-callout\; id="1"\; label="t"\; filenames="A2009100523520011H1.png"\; state="\{\{state\}\}">t</figure-callout>}}_1^{\min },{\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="A2009100523520011H1.png"\; state="\{\{state\}\}">t</figure-callout>}}_2^{\min },t_3^{\min },{\mathrm{<figure-callout\; id="4"\; label="t"\; filenames="A2009100523520014H1.png"\; state="\{\{state\}\}">t</figure-callout>}}_4^{\min }].$ Wherein $t_i^{\min }=\min \left({\max }_j\left(t_i\right)\right),$ Max _j(t _i) represent that j plays accident i key parameter t takes place in back 3 minutes _iObtained maximal value.

Subsequently, with T and t ^Max, t ^MinCarry out vector respectively relatively.Wherein i, j, m, n, k, l ∈ 1,2,3,4}:

(1) when the predetermined threshold value of all indexs $T_i\&le;t_i^{\max }$ The time, show that then preset threshold value might be judged as accident with normal condition:

1. if t _i ^MinBe sky, then accident free takes place in 24 hours same day. $T_i^{\&prime;}=t_i^{\max }+\mathrm{\&epsiv;},$ ε＝1；

2. if $t_i^{\min }<T_i,$ T then _iBe provided with still higherly, can not detect whole accidents, need turn down to T '=t ^Min

3. if $t_i^{\min }>T_i,$ T then _iCan detect whole accidents, at this moment:

If $t_i^{\max }<t_i^{\min },$ Show that then the i index occurs all having under each accident greater than the situation of mxm. under the normal condition, can heighten T _iExtremely $T_i^{\&prime;}=t_i^{\max }+\mathrm{\&epsiv;},$ ε＝1；

If $t_i^{\max }=t_i^{\min },$ Show that then the i index just equals maximal value under the normal condition at the relative minimum under each accident, get this moment $T_i^{\&prime;}=t_i^{\max };$

If $t_i^{\max }>t_i^{\min },$ Show that then relative minimum under this emergency conditions less than the maximal value under the normal condition, will cause the appearance of false alarm.Get this moment $T_i^{\&prime;}=t_i^{\min },$ To guarantee that whole accidents can detect;

If t ^MaxIn the k index $t_k^{\max }<T_k^{\min },$ The l index $t_l^{\max }\&GreaterEqual;T_l^{\min },$ Then get respectively $T_k^{\&prime;}=t_k^{\min },$ $T_l^{\&prime;}=t_l^{\min },$ Promptly get T '=t ^Min

4. if t ^MinIn the m index $t_m^{\min }\&GreaterEqual;T_m,$ Refer to the n index $t_n^{\min }<T_n,$ Then order respectively $T_m^{\&prime;}=t_m^{\min },$ $T_n^{\&prime;}=t_n^{\min },$ Promptly get T '=t ^Min

(2) when the predetermined threshold value of all indexs $T_i>t_i^{\max }$ The time, show that predetermined threshold value can not be judged as accident with normal condition:

1. if t _i ^MinBe sky, then accident free takes place in 24 hours same day.Get T ' _i=T _i, i.e. T '=T;

2. if $t_i^{\min }\&GreaterEqual;T_i,$ Then all accidents can both be detected, and get T ' _i=T _i, i.e. T '=T;

3. if t ^MinIn the m index $t_m^{\min }\&GreaterEqual;T_m,$ And the n index $t_n^{\min }<T_n,$ Then order respectively $T_m^{\&prime;}=t_m^{\min },$ $T_n^{\&prime;}=t_n^{\min },$ Promptly get T '=t ^Min

4. if $t_i^{\min }<T_i,$ Then there is accident to be failed to report.At this moment:

If $t_i^{\max }<t_i^{\min },$ Show that then the i index occurs all having under each accident greater than the situation of mxm. under the normal condition, can heighten T _iExtremely $T_i^{\&prime;}=t_i^{\max }+\mathrm{\&epsiv;},$ ε＝1；

If $t_i^{\max }=t_i^{\min },$ Show that then the i index just equals maximal value under the normal condition at the relative minimum under each accident, get this moment $T_i^{\&prime;}=t_i^{\max };$

If $t_i^{\max }>t_i^{\min },$ Show that then relative minimum under this emergency conditions less than the maximal value under the normal condition, will cause the appearance of false alarm.Reduce T this moment _iValue, order $T_i^{\&prime;}=t_i^{\min },$ To guarantee that whole accidents can detect;

If t ^MaxIn the k index $t_k^{\max }<T_k^{\min },$ The l index $t_l^{\max }\&GreaterEqual;T_l^{\min },$ Then get respectively $T_k^{\&prime;}=t_k^{\min },$ $T_l^{\&prime;}=t_l^{\min },$ Promptly get T '=t ^Min

(3) when $t_i^{\max }\&GreaterEqual;T_i,$ $t_j^{\max }<T_j,$ Then show part predetermined threshold value (T _i) normal condition might be judged as accident:

1. if t ^MinBe sky, then accident free takes place in 24 hours same day.Get T ' _k=T _k, i.e. T '=T;

2. if $t_k^{\min }\&GreaterEqual;T_k,$ Then all accidents can both be detected, and get T ' this moment _k=T _k, i.e. T '=T;

3. if $t_k^{\min }<T_k,$ Then there is accident to be failed to report.This season $T_k^{\&prime;}=t_k^{\min },$ To guarantee that whole accidents can detect;

4. if t ^MinIn the m index $t_m^{\min }\&GreaterEqual;T_m,$ And the n index $t_n^{\min }<T_n,$ Then the m index can detect all accidents, and the n index has accident and fails to report.At this moment:

If $t_k^{\max }<t_k^{\min },$ Then can improve $T_k^{\&prime;}=t_k^{\max }+\mathrm{\&epsiv;},$ ε＝1；

If $t_k^{\max }=t_k^{\min },$ Then $T_k^{\&prime;}=t_k^{\max },$ Be T '=t ^Max

If $t_k^{\max }>t_k^{\min },$ Then have normal condition by reported by mistake for accident may, be detected for guaranteeing whole accidents this moment, gets $T_k^{\&prime;}=t_k^{\min },$ Be T '=t ^Min

If t ^MaxIn the k index $t_k^{\max }<T_k^{\min },$ The l index $t_l^{\max }\&GreaterEqual;T_l^{\min },$ Then get respectively $T_k^{\&prime;}=t_k^{\min },$ $T_l^{\&prime;}=t_l^{\min },$ Promptly get T '=t ^Min

At last, with adjusted metrics-thresholds vector T '=[T ' ₁, T ' ₂, T ' ₃, T ' ₄] as new metrics-thresholds T, the accident that is used for next day is differentiated.

3. calculate four index O that the serial of 1 minute occupation rate data sequence of discontinuous slip is judged _{T, i}, d _{T, i} ^t, d _{T, i} ^sAnd d _{T, i} ^{S → t}

4. order is judged successively, relatively O _{T, i}With T ₁, d _{T, i} ^tWith T ₂, d _{T, i} ^sWith T ₃And d _{T, i} ^{S → t}With T ₄If all indexs have all been passed through serial arithmetic, satisfy simultaneously: O _{T, i}〉=T ₁, $d_{t,i}^t\&GreaterEqual;{\mathrm{<figure-callout\; id="2"\; label="T"\; filenames="A2009100523520011H1.png"\; state="\{\{state\}\}">T</figure-callout>}}_2,d_{t,i}^s\&GreaterEqual;T_3$ With $d_{t,i}^{s\&RightArrow;t}\&GreaterEqual;{\mathrm{<figure-callout\; id="4"\; label="T"\; filenames="A2009100523520014H1.png"\; state="\{\{state\}\}">T</figure-callout>}}_4,$ Then thinking has accident to take place, and sends warning.

Sample calculation analysis

The observation accident of Shanghai City city expressway is chosen in this experiment, and many indexs accident detection method that this research is set up and existing basic California algorithm and California 7# algorithm carry out parallel testing.Data are the Shanghai Urban through street that derives from through street, Shanghai supervisory system, are the coil section occupation rate averaging time data of step-length with 20 seconds.1 minute occupation rate data sequence of discontinuous slip algorithm adopts preceding 1 heavenly calendar historical event in advance so data are carried out the threshold value demarcation.

Test result is as shown in table 2.

Discontinuous slip 1 minute occupation rate data sequence algorithm of table 2 and the contrast of California test of heuristics

(detection zone 1 day data comprises 3 accidents)

Testing result shows:

(1) do not report by mistake in 13477 times are detected, rate of false alarm is 0; 2 wrong reports have taken place in basic California algorithm in 7200 times are detected, rate of false alarm is 0.03%; 9 wrong reports have taken place in California 7# algorithm in 7200 times are detected, rate of false alarm is 0.12%.The many indexs detection method that shows this method is better than basic California algorithm and California 7# algorithm on the performance of avoiding reporting by mistake;

(2) this algorithm all can effectively detect 3 test accidents with basic California algorithm, and California 7# algorithm can only detect 2 test accidents;

(3) this algorithm detection time and two kinds of California algorithms are basic identical, all find constantly early than artificial accident.

The above-mentioned description to embodiment is can understand and apply the invention for ease of those skilled in the art.The person skilled in the art obviously can easily make various modifications to these embodiment, and needn't pass through performing creative labour being applied in the General Principle of this explanation among other embodiment.Therefore, the invention is not restricted to the embodiment here, those skilled in the art should be within protection scope of the present invention for improvement and modification that the present invention makes according to announcement of the present invention.

Claims (6)

1, a kind of city expressway automatic detection method for traffic accident based on non-continuous sliding sequence, it is characterized in that: it may further comprise the steps:

1) detection system is gathered the telecommunication flow information of a fixed step size in real time by earth coil;

2) judge according to the real-time traffic stream information of gathering whether have traffic hazard take place, if judgement has an accident then give the alarm, notice vehicle supervision department takes corresponding measure if detecting the highway section; Otherwise continue to read real time data, judge next time;

The collection of described real-time traffic stream information generates 1 minute occupation rate data sequence of discontinuous slip as the basic data of differentiating algorithm.

2, the city expressway automatic detection method for traffic accident based on non-continuous sliding sequence as claimed in claim 1 is characterized in that: making up with minimum period 20s is the 1 minute occupation rate sequence of discontinuous slip at interval of sliding, and its construction method is:

$\mathrm{if}=\left\{\begin{array}{ccc}{o}_i^{20s}>1.5*o_{i-1}^{\mathrm{mov}5m}& \mathrm{or}& o_i^{20s}<0.5*o_{i-1}^{\mathrm{mov}5m}\\ o_{i+1}^{20s}>1.5*o_{i-1}^{\mathrm{mov}5m}& \mathrm{or}& o_{i+1}^{20s}<0.5*o_{i-1}^{\mathrm{mov}5m}\\ o_{i+2}^{20s}>1.5*o_{i-1}^{\mathrm{mov}5m}& \mathrm{or}& o_{i+2}^{20s}<0.5*o_{i-1}^{\mathrm{mov}5m}\end{array}\right.,$ $O_i=o_{i+2}^{1m}$

else， $O_i=o_i^{\mathrm{mov}1m}$

Wherein: O _i---i value of 1 minute occupation rate sequence of discontinuous slip;

o _i ^1m---i value of average 1 minute occupation rate sequence;

o _i ^Mov1m---with 20s is at interval, i value of the 1 minute occupation rate sequence of sliding;

o _i ^20s---i value of 20s occupation rate sequence;

o _i ^Mov5m---with 20s is at interval, i value of the 5 minutes occupation rate sequences of sliding.

3, the city expressway automatic detection method for traffic accident based on non-continuous sliding sequence as claimed in claim 1 is characterized in that: judge the method whether highway section has an accident that detects:

1) handles the detection road section traffic volume stream information of gathering in real time, generate 1 minute occupation rate data sequence of discontinuous slip;

2) determine the threshold value of judge index;

Whether 3) the correlated judgment index of calculating 1 minute occupation rate data sequence of discontinuous slip is compared with decision threshold, judge to have an accident.

4, the city expressway automatic detection method for traffic accident based on non-continuous sliding sequence as claimed in claim 3 is characterized in that: the correlated judgment index of 1 minute occupation rate data sequence of described discontinuous slip is respectively occupation rate O _{T, i}, 1 minute occupation rate of i section t discontinuous slip constantly; Occupation rate time difference d _{T, i} ^t, i section t constantly with the difference of 1 minute occupation rate of t-1 discontinuous slip constantly; Occupation rate space difference d _{T, i} ^s, the difference of t moment upstream i section and 1 minute occupation rate of the discontinuous slip of adjacent downstream i+1 section; Occupation rate space time difference d _{T, i} ^{S → t}, t is occupation rate space difference and t-1 moment occupation rate space difference poor constantly.

5, the city expressway automatic detection method for traffic accident based on non-continuous sliding sequence as claimed in claim 3, it is characterized in that: threshold value determination method is: based on the off-line calibration of historical casualty data or based on predetermined threshold value and in one day accident and non-casualty data online review one's lessons by oneself just definite.

6, the city expressway automatic detection method for traffic accident based on non-continuous sliding sequence as claimed in claim 4 is characterized in that: described occupation rate O _{T, i}, occupation rate time difference d _{T, i} ^t, occupation rate space difference d _{T, i} ^sWith occupation rate space time difference d _{T, i} ^sWhen → t all was not less than threshold value, the judgement accident took place.

Images