JP2002342872A - Device and method for detecting abnormality of traffic flow - Google Patents

Abstract

PROBLEM TO BE SOLVED: To comprehensively decide a change of a traffic situation appearing when a sudden event occurs on a road by perceiving the change of the traffic situation from various aspects. SOLUTION: Sudden abnormality of a traffic flow is detected by two or more different systems on the basis of traffic measuring data 11, weighted mean operation to an abnormality congestion evaluation value of each of the systems is conducted 12, and comprehensive decision is conducted on the basis of the weighted mean value 15-17. Since a defect of each of the systems can be compensated by combining congestion results of each of the systems, the occurrence of the sudden event on the road can be detected more accurately.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention collects traffic measurement data by installing a vehicle sensor or the like on a road, and uses the traffic measurement data to detect a traffic flow abnormality caused by an unexpected event. And a method and apparatus for detecting an abnormality.

[0002]

2. Description of the Related Art When an unexpected event such as a traffic accident or disaster occurs on a road, an abnormality in traffic flow based on the incident is detected promptly to notify a following vehicle or guide a following vehicle. There is a need to. 2. Description of the Related Art Conventionally, a camera is installed on a road and image processing is performed to detect an abnormality in a traffic flow (see JP-A-7-21488, JP-A-10-40490, etc.). Installing cameras over a wide area is costly and difficult to detect at night or in bad weather.

[0003] In view of this, it has been practiced to measure the traffic volume of a road, the speed of a vehicle, and the like using a vehicle sensor installed at various places on the road, and to monitor the abnormality of the traffic flow based on these measured values. . According to this monitoring device, it is determined that an accident has occurred when the speed drops sharply despite a small traffic volume and the state continues for a certain period of time.

[0004]

However, in the above-described monitoring device, the determination is made based on the traveling speed of the vehicle.
There is a problem that it is difficult to distinguish when an unexpected event occurs during a natural traffic jam, and the detection accuracy is reduced. Therefore,
In order to solve the above-mentioned problem, the present inventor has proposed a traffic flow abnormality detection algorithm using time variation of the number of vehicles existing in a road section sandwiched between vehicle sensors (Japanese Patent Application No. 2000-199416).
No.), a traffic flow abnormality detection algorithm that focuses on the deviation of the usage rate of a lane with a road (Japanese Patent Application No. 2000-278352), and determines the degree of matching (matching) of the same vehicle group that has passed up and down two points Algorithm for detecting traffic flow anomalies (Japanese Patent Application No. 2000-289296), and an algorithm for detecting traffic flow anomalies using the appearance of compression waves (disordered periodicity, spectrum changes) in traffic volume and average vehicle speed (Japanese Patent Application No. 2000-289296). Request 2000-3
No. 14139) and filed a patent application.

[0005] If any of the four traffic flow abnormality detection algorithms and the conventional traffic flow abnormality detection algorithm utilizing changes in traffic volume and vehicle speed, etc., are used alone, it is possible to use the algorithm on roads. Can be detected with a certain degree of certainty. However, none of these algorithms can be said to be all-purpose, and the detection accuracy may be reduced depending on conditions. With only one algorithm, the probability of erroneously detecting that an unexpected event has occurred while it has not occurred is high, or the probability of detecting omission that it is determined that it has not occurred although an unexpected event has occurred is high. . Some algorithms have a short detection delay time and some have a long detection delay time.

For example, in the traffic flow abnormality detection algorithm using the time variation of the number of vehicles existing in a road section, it is considered that if the change in the inflow and outflow of vehicles after occurrence of traffic congestion is reduced, the detection accuracy is reduced. In the traffic flow abnormality detection algorithm that pays attention to the bias of the usage rate, if the sudden event occurrence point is too far from the vehicle detector, the detection accuracy deteriorates. The traffic flow abnormality detection algorithm that determines the matching of the same vehicle group and the traffic flow abnormality detection algorithm that uses the spectrum change are effective during congestion, but are more effective during free running when traffic is not congested. It is considered that the detection accuracy may be deteriorated.

Therefore, the inventor grasps changes in traffic conditions that appear when an unexpected event occurs on a road from various aspects, and comprehensively judges them. Given that the occurrence can be detected, the present invention is proposed.

[0008]

SUMMARY OF THE INVENTION A traffic flow abnormality detection device according to the present invention comprises: abnormality detection means for detecting traffic flow abnormality in two or more different systems based on traffic measurement data; It is provided with a comprehensive judgment means for judging the occurrence of an abnormality in the traffic flow by combining the detection results of the methods (claim 1). According to the above configuration, by combining the detection results of the respective methods, it is possible to compensate for the disadvantages of the respective methods and to make a more accurate determination.

[0009] The comprehensive determination means may perform a comprehensive determination based on the number of systems in which an abnormality is detected. Since the occurrence of traffic flow abnormality is determined based on how many of the multiple methods have detected an abnormality,
The reliability of the determination is improved. The comprehensive determination means may determine the occurrence of a traffic flow abnormality if the number of detected abnormalities exceeds a threshold value. If the number of detected abnormalities is larger than the threshold value, the occurrence of an abnormality in the traffic flow is determined, so that the reliability of the determination is further improved. If the threshold value is selected by a majority of a plurality of methods, it can be determined by a so-called majority decision.

[0010] The comprehensive determination means may determine the occurrence of an abnormality in the traffic flow based on the sum of the evaluation values of the respective methods. If each method is used independently, even if the evaluation value is slightly small, it may be possible to determine that a traffic flow abnormality has occurred if the sum of the evaluation values of each method is large. This is an effective determination method. The comprehensive determination means of the present invention may perform a weighted average calculation on the evaluation value of each method, and may perform a comprehensive determination based on the weighted average value.
Since the detection accuracy of each method differs according to the occurrence situation of the sudden event, a weighted average calculation is performed on the evaluation value of each method, and a comprehensive judgment is performed based on the weighted average value. It can be further improved.

[0011] The comprehensive judgment means may judge the occurrence of a traffic flow abnormality if the weighted average value exceeds a threshold value. If the weighted average value is larger than the threshold value, the occurrence of a traffic flow abnormality is determined, so that the reliability of the determination is further improved. The weighting factor is
The function may be a function of any one of (a) to (g) or a combination thereof, and may be automatically determined (claim 7). (a) Traffic measurement data: The detection accuracy of each method may differ depending on traffic measurement data such as traffic volume Q, speed V, and occupancy O. There are an advantageous detection method and an unfavorable detection method when the traffic volume Q is large.

(B) Road alignment: The detection accuracy of each method may differ depending on the degree of road bend, road width, and the like. For example, on a road with many sharp curves, the speed is slow and traffic congestion is apt to occur. In such a case, there are advantageous detection methods and disadvantageous detection methods. (c) Day of the week: Since there are roads that are crowded or vacant depending on the day of the week, it is possible to determine which detection method is important. (d) Time zone: Since there are roads that are crowded or vacant depending on the time zone, it is possible to determine what kind of detection method is important.

(D) Degree of traffic congestion: There are an advantageous detection method and a disadvantageous detection method depending on the degree of traffic congestion, and it is possible to determine which detection method is important. (e) Detection accuracy of each method: Since there are a high-precision method and a low-accuracy method based on the technical evaluation of each method, past results, and the like, it is possible to determine which detection method is important. (f) Weather: A more accurate method depending on the weather conditions such as rain and snow,
Since there is a method with low accuracy, it is possible to determine which detection method is to be emphasized.

Further, the comprehensive determination means of the present invention may make a comprehensive determination based on the order of time when each system detects an abnormality (claim 8). Depending on each system, there is a system that can detect at an early stage and a system that detects at a late stage.
Therefore, if the probability of detection is determined based on the order of time detected by each method, the reliability of determination can be improved. Further, according to the present invention, it is preferable that the detection result of each method is obtained based on the traffic measurement data before and after the actual occurrence of the sudden event, and the result is stored as actual data (claim 9). This makes it possible to make an evaluation based on the results of each method.

[0015] The performance data may include one or more of a correct detection rate, a detection omission rate, a false detection rate, and a detection delay time. These values are
This is a useful parameter for evaluating each method. It is preferable that the traffic flow abnormality detection device of the present invention further includes an information providing unit that outputs a result of the determination by the comprehensive determination unit as abnormality information (claim 11). This is to prevent the accident from spreading by notifying the driver and the like.

The comprehensive determination means makes a stepwise determination according to the magnitude (likelihood) of the value on which the determination is based, and the information providing means makes a stepwise determination by the comprehensive determination means. It is preferable to change the content of the abnormality information according to the result (claim 12). By changing the content of information provision such as “accident / stop ahead” or “caution ahead” according to the likelihood (certainty) of the occurrence of the sudden event, more appropriate information can be given to the driver or the like.

When the traffic flow abnormality is detected in a plurality of road sections, the comprehensive determination means may specify the abnormality occurrence section in accordance with the magnitude of the value on which the determination is based. ). By setting the section having the highest likelihood (certainty) of the occurrence of the sudden event as the abnormality occurrence section, it is possible to notify a subsequent driver or the like. Further, the traffic flow abnormality detection method of the present invention performs traffic flow abnormality detection by two or more different methods based on traffic measurement data, and determines the occurrence of a traffic flow abnormality by combining the detection results of each method. (Claim 14). This method is a method according to the same invention as the traffic flow abnormality detection device according to the first aspect.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The embodiments of the present invention will be described below in detail with reference to the accompanying drawings, taking monitoring of a traffic flow on a highway as an example. 1. FIG. 1 is a schematic diagram showing a traffic flow monitoring system for detecting a traffic flow abnormality.

On the highway 1 having two lanes, a vehicle detector 5 of a two-loop embedding type is installed for each lane at intervals. An ultrasonic vehicle sensor 3 for measuring the vehicle height from above the vehicle is also provided for each lane. The sections of the expressway where these vehicle sensors 3 and 5 are installed are displayed as sections 1, 2, ‥‥, i, ‥‥ (i is an integer of 2 or more). Let the distance of each section be Li. These vehicle sensors 3, 5 and the camera are connected to the primary processing unit 4,
The primary processing device 4 performs counting of the number of vehicles passing, detection of vehicle speed, and the like.

The highway 1 is provided with a variable display panel 6 for notifying the vehicle of accident information and road surface information. In addition, a roadside beacon 7 that performs bidirectional communication with a vehicle.
Is provided. Further, on the general road 2 connected to the highway 1, accident information and road surface information of the highway 1 are provided.
A variable display board 9 is provided to notify a vehicle entering the highway 1.

The computer 11 in the traffic management center 10 communicates with the primary processing unit 4, the roadside beacon 7, the variable display panel 6, etc. installed in each section and the wired communication network 12 (which may be a wireless communication network). It is connected. Also,
The computer 11 is connected to related organizations 13 such as the Ministry of Construction, the National Police Agency, and the Fire Service through a communication line, and is also connected to a broadcasting station 14 through a communication line. In the system example described above, a highway is assumed, but a general road may be used. Although a road having two lanes is assumed, the number of lanes is not limited to two, and may be one lane or three or more lanes.

In addition, a vehicle sensor of a Doppler type installed on the side of a road may be used instead of the vehicle sensor 5 of a plurality of embedded types. Alternatively, a television camera may be installed on the road to detect the number of vehicles passing, vehicle height, vehicle length, passing speed, and the like by image processing. All or some of the functions of the computer 11 described below are realized by the computer 11 executing a program recorded on a recording medium such as a program ROM.

2. Traffic Management Center FIG. 2 shows a computer 11 in the traffic management center 10.
3 is a functional block diagram of FIG. In the input processing unit 21 of the computer 11, the detection signal of the vehicle detector 5 is
Is entered via The input processing unit 21 includes the vehicle detector 5
The traffic volume (number of vehicles passing per unit time) Q, the average speed V, the occupancy O (the vehicle crossed the vehicle sensor within a certain time T) Sum of time tk Σtk divided by time T: Σtk
/ T), the characteristic amount of the vehicle, and the like. The “average” speed is used to average the speed of each vehicle that has passed within a certain period of time. Hereinafter, the “average speed” is simply referred to as “speed”.

Focusing on section i, the distance of section i is represented by Li,
The inflow traffic volume of the first lane to the section i at the time t is Q1,
i (t), the inflow traffic volume in the second lane is Q2, i (t), and the inflow traffic volume to the section i at time t, which is the sum of both the first lane and the second lane, is Qi (t). I do. Similarly, the outflow traffic volume in the first lane from section i is Q1, i + 1 (t), and the outflow traffic volume in the second lane is Q2, i + 1 (t). And the outflow traffic volume from section i at time t is Qi + 1
(t).

Qi (t) = Q1, i (t) + Q2, i (t) Qi + 1 (t) = Q1, i + 1 (t) + Q2, i + 1 (t) The inflow speed of the first lane into the vehicle is V1, i (t), and the inflow speed of the second lane is V2, i (t). Similarly, let the outflow speed of the first lane from the section i be V1, i + 1 (t) and the outflow speed of the second lane be V2, i + 1 (t). 1st lane and 2nd
The inflow velocity into the section i at the time t at which the lane is weighted and averaged by each traffic volume is Vi (t), and the outflow velocity from the section i is Vi + 1 (t).

The occupancy of the vehicle in the first lane is represented by O1 (t),
The occupancy of the vehicle in the second lane is O2 (t). The occupancy of both lanes is O (t). Further, the input processing unit 21 calculates a feature amount of the vehicle. That is, the maximum vehicle height of each vehicle is calculated based on the output of the vehicle sensor 3, and the speed of the vehicle is measured based on the output time difference between the two loops of the vehicle sensor 5. Is calculated based on the vehicle length. Since the vehicle height and vehicle length are calculated each time the vehicle passes, one or a plurality of vehicle height and vehicle length data strings are obtained for each lane.

The input processing unit 21 of the computer 11
In the above, the traffic volume Q and the average speed V are calculated, but these calculation processes may be performed in the primary processing device 4. The detected data of the traffic volume, speed, occupancy, vehicle height, vehicle length, and the like are referred to as “traffic measurement data”. 3. Individual determination method The computer 11 is provided with a determination unit 22.
The determination unit 22 includes a plurality of determination algorithms (hereinafter, referred to as a plurality of determination algorithms).
According to methods A to E), determination units A to E for determining traffic flow abnormalities are provided.

3.1 Detection of Traffic Flow Abnormality Using Time Variation of the Number of Vehicles in Road Section (Method A) As described below, the determination unit A uses different methods A1 and A2 to determine vehicle traffic. The time variation of the number is calculated. <Method A1> The determination unit A calculates a difference ΔQi (t) between the inflow traffic volume Qi (t) and the outflow traffic volume Qi + 1 (t).

.DELTA.Qi (t) = Qi (t) -Qi + 1 (t) Then, the determination unit A obtains the time variation of the number of vehicles Ei (t). ΔQi (t) itself represents the time variation of the number of vehicles Ei (t). The determination unit A records ΔQi (tk) at regular time intervals (for example, one minute) at times t1, t2, t3, ‥‥, tk, ‥‥ (use a suffix k when representing), and The variance (varia) over a period T (eg, 10 minutes)
nce). This variance is written as σ1 (tk).

<Method A2> The determination unit A divides the section i into the first half and the second half, and calculates the number of vehicles present in each. In the first half of the section, the number of existing vehicles Ei1 (t) in the first half of the section is obtained using the inflow traffic Qi (t) and the inflow velocity Vi (t). Ei1 (t) = Qi (t) / Vi (t) · Li / 2 In the latter half, the outflow traffic Qi + 1 (t) and the outflow velocity Vi + 1 (t)
Then, the existence number Ei2 (t) in the latter half of the section is obtained.

Ei2 (t) = Qi + 1 (t) / Vi + 1 (t) · Li / 2 Then, the sum of both the existing numbers Ei1 (t) and Ei2 (t) is calculated, and the vehicle in section i is calculated. It is assumed that the number of vehicles is Ei (t). Ei (t) = Ei1 (t) + Ei2 (t) The determination unit A records Ei (tk) at each time tk in order to find the time variation of the number of vehicles Ei (t), Calculate the variance over T. This variance is written as σ2 (tk) (end of method A2).

The determination unit A determines the occurrence of a sudden event based on the variances σ1 (tk) and σ2 (tk) obtained from the methods A1 and A2. FIG. 3 is a flowchart illustrating a process performed by the determination unit A to monitor the occurrence of a sudden event. The determination unit A repeats this process at each time tk. FIG.
According to this, the determination unit A determines that the variances σ1 (tk) and σ2 (t
k) is obtained (step S1), and it is determined whether σ1 (tk) and σ2 (tk) are each equal to or larger than a threshold (step S1).
2, 4). If both are greater than or equal to the threshold value, a constant α is added to the evaluation value (step S3), and if only one is greater than or equal to the threshold value, a constant β (β <α) is added to the evaluation value (step S5). ). α and β are weighting coefficients of the method A1 and the method A2. The reason for adding to the evaluation value is to prevent instantaneous erroneous detection.

If both σ1 (tk) and σ2 (tk) are less than the threshold value, the evaluation value is reset to 0 (step S6).
It is determined whether or not the evaluation value is equal to or greater than a certain value (step S7). If the evaluation value is equal to or more than the certain value, the determination unit A determines that a sudden event has occurred (step S8). If not, the determination is carried over to the next time tk + 1. The calculated evaluation value and the result of the determination are sent to the overall determination unit 23.

3.2 Abnormality Detection of Traffic Flow Focusing on Unevenness of Lane Use Rate (Method B) The determination unit B determines the times t1, t2,
Traffic volume Q1 (t), Q2 (t), speed V1 for each of t3, ‥‥, tk, ‥‥ (use “t” without subscript when representing)
(t), V2 (t), occupancy rates O1 (t), O2 (t) are recorded, and lane utilization rates r1 (t), r2 (t) of the first lane and the second lane are calculated. Here, the lane utilization rates r1 (t) and r2 (t) may be defined as the ratio of the traffic volume of the lane to the traffic volume of all lanes, and the occupancy of the lane with respect to the sum of the occupancy rates of the lanes. It may be defined by a ratio of rates. In the former case, r1 (t) = Q1 (t) / Q (t) and r2 (t) = Q2 (t) / Q (t). In the latter case, r1 (t) = O1 (t) / O (t) and r2 (t) = O2 (t) / O (t).

The determination unit B determines whether the past period T (for example, 10
Calculate the average value of the lane utilization rate over minutes). The average value is written as r1, r2. The determination unit B determines the occurrence of a sudden event according to the procedure shown in FIG. FIG. 4 is a flowchart illustrating a process performed by the determination unit B to monitor the occurrence of a sudden event. The following description focuses on the installation point of a certain vehicle sensor 5 and describes the processing performed at that point. The same applies to each of the installation points of the vehicle sensor 5 upstream and downstream. It is refused in advance that the processing is performed.

The determination section B repeats this process at each time tk. Explaining with reference to FIG.
It is determined whether or not the traffic volume Q (t) is equal to or more than a certain value (may be determined based on the occupancy O (t)) (step T1). If the traffic Q (t) is not equal to or more than the predetermined value, the determination is stopped. Traffic Q
If (t) is too low, the accuracy of the calculated lane utilization rate also decreases, so that reliable determination of the occurrence of the sudden event cannot be made. If the traffic volume Q (t) is above a certain level and the accuracy of the lane utilization can be expected, the average lane utilization r1 or r2
Is calculated, and it is determined whether or not the lane utilization ratio r1 or r2 is biased (step T2). As a criterion, r1
Or, whether r2 is close to 0.5 or whether the ratio of r1 to r2 is close to 1. For example, if r1 <0.4 or r1> 0.6, it is determined that there is a bias. If 0.4 ≦ r1 ≦ 0.6, it is determined that there is no bias. 0.6 and 0.4 are threshold values. When the threshold value is too close to 0.5, the chances of determining that the lane utilization rate is biased increase, and the determination rate of the occurrence of the sudden event increases. As a result, it leads to an excessive determination. If the threshold value is too far from 0.5, the chance that the lane utilization rate is determined to be uneven decreases, and the determination rate of the occurrence of the sudden event decreases. As a result, it is easy to overlook occurrence of a sudden event.

In view of this, the threshold value is used to record lane utilization data before and after the actual occurrence of a sudden event on the road, and the occurrence of the sudden event on the road is most accurately performed by implementing the present invention. The value should be selected so that it can be detected well. In addition, data on lane occupancy under various conditions in the past, such as time of day, day of the week, and days with no events, is accumulated, and there is no optimal data according to time of day, day of the week, season, weather, presence of events, etc. A threshold may be set.

When it is determined that the vehicle is unbalanced, a correlation with the lane utilization detected by the vehicle sensor 5 installed upstream is calculated (step T3). This correlation is given, for example, by the reciprocal of the difference between the lane utilization ratio r1 and the upstream lane utilization ratio r1 '(ABS represents an absolute value). 1 / ABS (r1-r1 ') If the lane utilization ratio r1 and the upstream lane utilization ratio r1' take a close value, the correlation becomes large, and the lane utilization ratio r1 and the upstream lane utilization ratio r1 'become larger. If the values are far apart, the correlation becomes small.

Next, this correlation is compared with a fixed reference value.
(Step T4). In this manner, the correlation is compared with the reference value if the lane utilization ratio r1 and the upstream lane utilization ratio r1 ′ take a value that is apart from each other (that is, the correlation is small). It is based on the fact that a change occurs in the utilization rate, and it can be determined that an unexpected event has occurred during that time. As a “reference value”, a value that can be detected with high accuracy in the occurrence of a sudden event in a section is selected from experience. Specifically, it is automatically selected from stored data or the like.

If the correlation with the upstream point is below the reference value
(NO in step T4), the vehicle detector 5 installed downstream
(Step T5). If the correlation with the downstream point is also equal to or less than the reference value (NO in step T6), it is determined that an unexpected event has occurred near the point (step T7). This result is brought to the comprehensive judgment described later. If the correlation with the upstream point is equal to or more than the reference value (when it is determined that there is a correlation) (step T
4: YES), there is no change in the lane utilization rate from the upstream point to the point, and it is unlikely that an unexpected event has occurred during that time. Therefore, the process of determining the occurrence of an unexpected event at the point is stopped (in this case, the occurrence of the unexpected event may be detected in the event monitoring process performed at the upstream point).

If the correlation with the downstream point is equal to or greater than a predetermined reference value (when it is determined that there is a correlation) (YES in step T6), the lane utilization rate does not change from the point to the downstream point. It is unlikely that a sudden event occurred. Therefore, also in this case, the process of judging the occurrence of the sudden event at the point is stopped (in this case, the occurrence of the sudden event may have been detected in the sudden event occurrence monitoring process performed at the downstream point). .

It should be noted that a traffic flow abnormality detection method which focuses on the deviation of the lane utilization rate can be applied to a three-lane road. In this case, the lane utilization rates r1 (t), r2 (t) and r3 (t) of each lane may be defined by the ratio of the traffic volume of the lane to the traffic volume of all lanes. May be defined as the ratio of the occupancy of the lane to the vehicle. In the former case, r1 (t) = Q1 (t) / Q (t), r2 (t) = Q2 (t) / Q (t), r3 (t) = Q3 (t) / Q (t), Q (t) = Q1 (t) + Q2 (t) + Q3 (t). In the latter case, r1 (t) = O1 (t) / O (t), r2 (t) = O2 (t) / O (t), r3 (t) = O3 (t) / O (t), O (t) = O1 (t) + O2 (t) + O3 (t). Even on a road having four or more lanes, the lane utilization rate of each lane can be defined.

3.3 Traffic Flow Abnormality Detection by Determining Vehicle Group Matching (Method C) A vehicle group matching processing method performed by the determination unit C will be described with reference to a flowchart (FIG. 5). Various methods are known as the vehicle group matching processing method in addition to the method described below. In the following, as an example, a method of associating vehicles with a p-dimensional alignment problem is performed. (Kobayashi et al., "Global Traffic Flow Analysis Algorithm from Two-Point Vehicle Observation Information" (published by Information Processing Society of Japan 72nd Algorithm Study Group, March 21, 2000)).

Hereinafter, attention will be focused on only one lane. A vehicle that has entered the lane from another lane or a vehicle that has exited from the lane to another lane is treated in the same manner as a vehicle that has entered the lane from an intersection or exited from the lane to an intersection. The vehicle height detected by the vehicle detectors 3, 5 upstream of the traveling direction,
The vehicle length data sequence is input (step U1), and the vehicle height and vehicle length data sequences detected by the vehicle sensors 3, 5 downstream in the traveling direction are input (step U2).

The number of vehicles to be processed detected by the upstream vehicle detector is defined as n, and the vehicle height data sequence is defined as hA1, hA2, hA3,.
The vehicle length data sequence is represented by lA1, lA2, lA3,..., lAn (representative is represented by "lA"). The number of vehicles to be processed detected by the downstream vehicle sensor is m, and the vehicle height data sequence is hB1,
hB2, hB3,..., hBm (represented by “hB” when represented), and the vehicle length data sequence is represented by lB1, lB2, lB3,.
‥, IBm (representatively represented by IB).

The probability that a vehicle having a vehicle height h detected by the upstream vehicle sensor as a vehicle height hA is observed as a vehicle height hB by a downstream vehicle sensor is based on the observation error of the vehicle sensor following a Gaussian distribution. If it is assumed, it is expressed by the following equation (1).

[0047]

(Equation 1)

Here, σ ² h, A is the variance of the vehicle height observed by the upstream vehicle sensor, and σ ² h, B is the variance of the vehicle height observed by the upstream vehicle sensor. Similarly, the probability that the vehicle having the vehicle length l detected by the upstream vehicle detector as the vehicle length lA is observed as the vehicle length IB by the downstream vehicle detector is expressed by the following equation (2).

[0049]

(Equation 2)

Here, σ ² l, A is the variance of the vehicle length observed by the upstream vehicle sensor, and σ ² l, B is the variance of the vehicle length observed by the upstream vehicle sensor. (a) When the difference in vehicle height or vehicle length is used as a cost evaluation criterion The posterior probability density Ph in which a vehicle detected as a vehicle height hA by an upstream vehicle sensor is detected as a vehicle height hB by a downstream vehicle sensor
(ha, hb) is represented by the following equation (3).

[0051]

(Equation 3)

Similarly, the posterior probability density Pl (la, lb) that the vehicle detected as the vehicle length IA by the upstream vehicle sensor is detected as the vehicle length IB by the downstream vehicle sensor is represented by the following (4). It is expressed by an equation.

[0053]

(Equation 4)

Σ ² h, A + σ ² h, B = σ ² h, σ ² l, A + σ
² l, B = Writing and sigma ² l, wherein (3) (4), respectively, (5)
It can be rewritten as in equation (6).

[0055]

(Equation 5)

[0056]

(Equation 6)

(B) When the ratio of vehicle height or vehicle length is used as a cost evaluation criterion The following expressions (7) and (8) are used instead of the expressions (5) and (6).

[0058]

(Equation 7)

[0059]

(Equation 8)

The natural logarithm of the probability density listed above is called "match cost". When only the vehicle height is used, the match cost d (h) between the vehicle detected as the vehicle height hA by the upstream vehicle sensor and the vehicle detected as the vehicle height hB by the downstream vehicle sensor.
a, hb) becomes d (ha, hb) = ln [Ph (ha, hb)] (9) When only the vehicle length is used, the match cost d (la, lb) between the vehicle detected as the vehicle length IA by the upstream vehicle detector and the vehicle detected as the vehicle length IB by the downstream vehicle detector is: d (la, lb) = ln [Pl (la, lb)] (10)

When both the vehicle height and the vehicle length are used, the vehicle height hA and the vehicle length IA are detected by the upstream vehicle sensor, and the vehicle height hB and the vehicle length IB are detected by the downstream vehicle sensor. The cost d (ha, la; hb, lb) of the match with the vehicle is d (ha, la; hb, lb) = ln [Ph (ha, hb)] + ln [Pl (la, lb)] (11) Becomes The determination unit C calculates the match cost for all combinations of the vehicle sensed by the upstream vehicle sensor and the vehicle sensed by the downstream vehicle sensor, according to any of the formulas (9), (10), and (11). (Step U3).

Next, the judgment section C acquires the “gap cost”. This gap cost is stored as a constant corresponding to the characteristics of the vehicle sensor. The match cost and the gap cost are collectively referred to as “score”.
The gap cost includes the internal gap cost (I
GC) and external gap cost (EXGC). The IGC is set when overtaking occurs between the two points and the order is changed, so that no correspondence can be obtained. This is based on the assumption that either the vehicle height or the vehicle length will not be more than 3σ, and if it is more than 3σ, it will be regarded as another vehicle.

When only the vehicle height is used, the specific value of IGC is expressed by the following equation (12). d (ha,-) = d (-, hb) = (1/2) ln [Ph (0,3.sigma.h)] =-(1/2) ln ((2.pi.) ^{1 /} 2.sigma.h) -9/4 ( 12) The reason why 1/2 is used in the equation is that the cost is equivalent to one branch because two branches are used when the gap cost is used (the same applies hereinafter). When only the vehicle length is used, the specific value of the IGC is expressed by Expression (13).

D (la, −) = d (−, lb) = (1/2) ln [Pl (0.3σl)] = − (1/2) ln ((2π) ^{1 /} 2σl) -9 / 4 (13) When both the vehicle height and the vehicle length are used, the specific value of the IGC is expressed by the equation (14). d (ha, −; −, −) = d (−, la; −, −) = d (−, −; hb, −) = d (−, −; −, lb) = (1/2) ｛ ln [Ph (0,3σh)] + ln [Pl (0,0)]｝ = (1/2) ｛ln [Ph (0,0)] + ln [Pl (0,3σl)]｝ = − (1 / 2) ln (2πσhσl) -9/4 (14) EXGC is set when a vehicle that has passed the upstream point has not yet passed the downstream point and cannot be associated with a vehicle.
This value of EXGC is calculated by calculating the maximum score of the alignment when it is confirmed that the optimal matching is actually obtained by visual inspection or the like at the time of initial setting after installation of the apparatus. Is determined based on

Further, it is conceivable that the correspondence score between vehicles that cannot be handled in reality is set to −∞ (for example, when compared with the travel time estimated from experience, which is negative for travel time, Too short or long, etc.).
When the above costs are obtained, the n vehicles detected by the upstream vehicle detector and the m vehicles detected by the downstream vehicle detector are set so that the sum of the match costs or gap costs is maximized.
It associates with one vehicle. For this reason, it is formulated as a two-dimensional character string alignment problem.

The vehicles passing the upstream point are referred to as a1, a2, a
The vehicle passing through the downstream point is denoted by b1,
Expressed as b2, b3,..., bn. Matrix (aibj) (1 <i <n, 1 <
j <m) is called alignment. The two-dimensional character string alignment problem can be reduced to the longest path problem on a two-dimensional grid-like directed graph (the general term is "shortest path problem,
Here we are looking for the longest path of the sum of the costs,
"The longest road problem"). FIG. 6 is a diagram illustrating a two-dimensional grid-like directed graph. In FIG. 6, diagonal branches that correspond to the vehicle ai and the vehicle bj are represented by solid lines. This branch length corresponds to the above-mentioned match cost d (any one of equations (9) to (11)). The vertical and horizontal broken lines inside the frame indicate the vehicle ai
And the vehicle bj do not correspond, and the branch length corresponds to the above-mentioned IGC. The dashed-line branch outside the frame indicates a case where there is no corresponding vehicle, and the branch length corresponds to the above-described EXGC.

The longest path of the sum of the costs from the upper left point to the lower right point in FIG. 6 is the solution indicating the plausible association of the vehicle. The longest path problem is based on dynamic programming (DP; D
dynamic programming) (step U4) (see [1] and [2] below). [1] D. Gusfield. "Aigorithms on Strings, Trees,
and Sequences. "Cambridge University Prass, 1997. [2] SBNeedleman and CDWunsch." A general meth
od applicable to thesearch for similarities in the
amino acid sequences of two proteins. "Journal of
Molecular Biology, 48, pp. 443-453, 1970. For example, it is assumed that n = 3 and m = 4, and the longest path problem is solved to obtain a path (thick solid line) as shown in FIG. From FIG. 7, vehicle a1 corresponds to b1, vehicle a2 does not correspond to vehicle b2 (internal gap), and vehicle a3
Corresponds to b3 and there is no corresponding to vehicle b4 (external gap). It is considered that this is because vehicles a2 and a3 have been switched after passing the upstream point, and vehicle b4 has entered from an intersection on the way or has entered due to movement between lanes.

The above analysis result is output (step U
5). Then, based on the output result, the travel time can be estimated by focusing on the matching vehicles.
Here, a method of determining the value of EXGC will be described. EX
Let the value of GC be g. It is assumed that n vehicles are observed upstream and m vehicles are observed downstream. Note that n <m. Vehicles a1-observed at the upstream point by visual inspection
An is assumed to correspond to the vehicles b1 to bn observed at the downstream point. However, vehicles an + 1 to am observed at a relatively late time at the upstream point
Has not reached downstream yet at the time of observation of the downstream point, and there is no corresponding one.

In this case, the score of the matching portion in the alignment is written as S '(A), and the score of the non-matching portion is written as g · x (A). S '(A) is a value calculated from the measured value of the vehicle sensor. g is the value of EXGC to be obtained, x (A) indicates the number of non-matching units, and x (A) =
nm. The value g of the EXGC is changed to “the score S ′ of the matching portion when the optimal matching is obtained.
(A) divided by "n + mx (A)".

G = S ′ (A) / [n + mx (A)] (15) For example, if a two-dimensional lattice-like directed graph where n = 4 and m = 3 is drawn, the result is as shown in FIG. In FIG. 8, the score S '(A) of the matching part and the value g · x of the EXGC
(A) is shown. Where x (A) = nm = 4-
Since 3 = 1, g.x (A) in FIG. 8 indicates g itself. In the example of FIG.
The value g of GC is g = S '(A) / [4 + 3-1] = S' (A) / 6 based on (15).

Actually, a process of observing a group of vehicles for which matching has been confirmed upstream and downstream to obtain an EXGC value g is performed a plurality of times, and an average of the obtained plurality of g is obtained. May be finally determined as the EXGC value g. In the embodiment using the IGC and the EXGC, the vehicle association portion (data with the largest m) on the latest time point (latest observation time) is EXGC
Greatly depends on the value of. The value of EXGC fluctuates because it is a value obtained statistically as described above.

A method of extending the algorithm so that EXGC at the latest time point in the time series is not required will be described. As described above, the value g of EXGC is represented by g = S '(A) / [n + mx (A)] (15). On the other hand, the total score S (A, g) is represented by S (A, g) = S '(A) + g.x (A) (17) By substituting equation (15) into equation (17), S (A, g) = S '(A) + S' (A) x (A) / [n + mx (A)] = (n + m) S '( A) / [n + mx (A)] (18) Since n + m is constant, S '(A) / [n + m-
x (A)] is the optimal solution.

The above is illustrated as follows. FIG. 9 is a diagram illustrating a two-dimensional grid-like directed graph excluding the external gap at the latest time point in the time series. In the graph of FIG. 9, the upstream passing vehicles a1,
The subscripts of ‥‥ and an are i (1 ≦ i ≦ n), and the corresponding point on the latest time point is vi. The maximum score of the alignment at each point vi is obtained, and the point at which the value obtained by dividing the score by (i + m) -x (A) becomes the maximum is defined as the latest corresponding point vi ₀ .

Although the embodiments of the present invention have been described above, the embodiments of the present invention are not limited to the above embodiments. For example, in the present invention, in addition to vehicle length and vehicle height data, camera image data can be acquired by a camera and used for alignment. In this case, it is necessary to determine a score between images, but the “distance between images” devised in fields such as image search, specifically, match distance (see [3] below), EMD (Earth Mover's Dista
nce) (see [4] below).

[3] M. Werman, S. Peieg, and A. Rosenfe
ld. "A distance metric for multi-dimensional histgr
ams. "Computer, Vision, Graphics, and Image Proces
sing, 32, pp.328-336, 1985. [4] Y. Rubner, C Tomasi, and LJ Guibas. "The Eart
h Mover's Distance as a Metric for Image Retrieva
l. "Technical Report STAN-CS-TN-98-86, Department
of Computer Science, Stanford University, Septembe
r 1998. The match distance is a distance using color information of a pixel in an image, and is given as an L1 distance between cumulative histograms of two image histograms. Specifically, a luminance histogram and a hue histogram are created from images captured from behind each vehicle. In the luminance histogram, all pixels are divided into 32 levels from the luminance, and the frequency is the number of pixels at each level. The hue histogram divides the hue into 30 levels, and the frequency is the sum of the saturations of the pixels included in each level. The image distance was a weighted sum of the distances based on the histogram of luminance and hue.

The EMD is a generalized distance of the match distance, and uses not only the color information of the pixel but also the position information. The pixels of each image are clustered by color and proximity (see [5] below) and given as the least cost stream between them. This time, each image was clustered into 16 clusters. [5] M. Inaba, H. Imai, and N. Katoh. "Application of
weighted Voronoi diagrams and randomization to va
riance-based k-clastering. "In Proceedingsof the 1
0th ACM Symposium on Computational Geometry, pp.33
2-339, 1994 When the distance between the images of the vehicle ai and the vehicle bj is Dij,
As the alignment score Dtotal, the above (9) (10) (11)
A linear sum of any of the equations and the distance Dij between the images is adopted.

Dtotal = d (ha, la; hb, lb) + λDij (18) Here, the constant λ is a weighting coefficient. Further, assuming that the gap cost of the vehicle length / height is G and the maximum value of the distance between the corresponding vehicles is M, the total gap cost Gtotal is given as Gtotal = G−λM / 2 (19).

In the above-described embodiment, the number of vehicle observation points is two. However, the number of vehicle observation points can be extended to any number of points. If the vehicle observation point is a point p (p is an integer of 2 or more), the association of the vehicle can be reduced to a p-dimensional alignment problem. As described above, the association between the n vehicles passing the upstream point and the m vehicles passing the downstream point was performed.

Next, the traffic flow abnormality detection processing performed by the determination unit C will be described with reference to a flowchart (FIG. 10). This abnormality detection processing is performed in synchronization with associating a vehicle passing through the upstream point with a vehicle passing through the downstream point. The determination unit C inputs the results of association at the upstream and downstream points (step V1). Then, of the vehicles that have passed the downstream point, it is checked what percentage is associated with the vehicle that has passed the upstream point. This ratio is called a matching ratio. For example, of the m vehicles that have passed the downstream point, p vehicles are:
If it corresponds to the vehicle that has passed the upstream point, the matching rate will be p / m.

Here, the travel time is measured, and when the travel time is relatively short, that is, when it is close to free flow, the threshold value is made relatively small, and when the travel time is relatively long, that is, when there is congestion, Make the threshold relatively large (Step V2
V4). The determination unit C compares the matching rate with the threshold (step V2), and if it is lower than the threshold, determines that an unexpected event has occurred between the upstream and downstream points (step V2).
4). If it is higher than the threshold value, the past (for example, 1
It is checked whether or not the matching rate obtained by associating the vehicles passing through the upstream and downstream points before the round is higher than the threshold value, and if it is higher, it is determined that there is no abnormality. If it is low, it is determined that the sudden event has continued.

As described above, when the matching ratio is higher than the threshold value twice or more consecutively, it is determined that there is no abnormality in order to avoid erroneous determination in case of accidental matching even if a sudden event occurs. It is. The threshold is
What is necessary is just to check the matching rate when an unexpected event occurs, and determine an appropriate value. As described above, the matching rate and the determination result of the occurrence of the sudden event are carried to the comprehensive determination described later.

As a method of calculating the matching ratio, another known method may be employed. For example, a method of taking an image of a vehicle and recognizing a feature of the vehicle, for example, a color or a plate number, using image processing may be adopted. 3.4 Abnormal detection of traffic flow using compressional waves (spectrum change) appearing in the average speed of the vehicle (method D) The determination unit D determines the times t1, t2,
Traffic volume Q (tn), speed V (tn), time occupancy O (t) for each of t3, ‥‥, tn, ‥‥ (subscript n is used for representative)
Record n) and calculate their frequency spectrum. In this calculation, as is known, the average value of the traffic volume Q (tn), the speed V (tn) or the time occupancy O (tn) is obtained, the sample self-covariance shifted by time k is obtained, and the sample spectrum distribution pj If you ask for

Specifically, one of the traffic volume Q (tn), the speed V (tn) and the time occupancy O (tn) is selected, and is selected as yn.
Notation. Let μ be the average value E (yn) of yn. μ = E (yn) = (1 / N) Σyn (sum is n = 1 to N
Take up to. (N is the number of samples) The sample self-covariance function Ck is as follows: Ck = (1 / N) Σ (yn-μ) (yn-k-μ) However, the sum Σ is from n = k + 1 to N. The sample spectrum pj is as follows: pj = ΣCk exp (−2πikfj). The sum Σ is taken from k = -N + 1 to N-1. Expressed as a cosine function, pj = C0 + 2ΣCk cos (2πkfj). The sum Σ is taken from k = 1 to N-1. fj is a frequency, and fj = j / Nj = 0, 1, ‥‥, [N / 2] ([] is a Gaussian symbol).

The determining section D determines the occurrence of a sudden event based on the sample spectrum distribution. FIG. 11 is a graph showing the variation of the traffic Q (tn) calculated every day on the day when the sudden event occurs. The measurement result of 500 m downstream of the sudden event occurrence point is shown. Before 8:00, the time when the sudden event occurred, Q (tn) was stable at about 50 cars / minute. However, immediately after the time 8:00 when the sudden event occurred, Q
(tn) is significantly reduced. Time 8 when the traffic obstacle was removed
After 20 minutes, Q (tn) rises to about 60 units / minute, and after a while, returns to the original steady value of around 50 units / minute.

FIG. 12 is a graph showing a three-dimensional distribution of a sample spectrum (power spectrum) pj in a 30-minute sample (N = 30). The horizontal axis represents frequency and time. The vertical axis is the power spectrum.
FIG. 13 is a graph in which the locus of the peak value of the sample spectrum pj is plotted with time. Before 8:00, the time when the accident occurred, the peak value hardly appeared. After 8:00, the peak value rises and keeps a value near the time after 8:20 when the traffic obstacle is removed.

FIG. 14 is a graph plotting the peak frequency (unit: Hz) of the sample spectrum. The peak frequency takes a relatively high value before 8:00, at which the sudden event occurs, but drops sharply to 0.05 Hz or less after 8:00, at which the sudden event occurs. It remains low even after traffic obstructions have been removed. FIG. 15 is a flowchart illustrating a process performed by the determination unit D to determine occurrence of a sudden event. FIG. 15 is described on the assumption that a sample spectrum distribution of the traffic Q (tk) is obtained.

The determination section D repeats this process at each time tn. The description will be made with reference to FIG.
Calculates the sample spectrum distribution pj (j is equivalent to frequency) (step W0). The determination unit D initializes an evaluation value (referred to as a score) stored in a memory attached to the determination unit D to 0 (step W1), and checks a previous incident event occurrence determination flag or a traffic condition attention determination flag (step W1). W
2). If the previous determination flag is ON, the average value of the previous power spectrum is used as a reference value (step W3),
The constant value α is added to the score (step W4).

Next, it is determined whether the peak value of the spectrum (see FIG. 13) is greater than any of the past 30 minutes (step W5). If it is larger, a constant value β is added to the score (step W6). Next, it is determined whether or not the amount of change (for example, the differential value) of the peak value of the spectrum is larger than the average value of the past 30 minutes (step W7). If it is larger, a constant value γ is added to the score (step W8).

Next, it is determined whether or not the peak frequency of the spectrum is smaller than a threshold value (for example, 0.1 Hz).
(Step W9). If smaller, a constant value ε is added to the score (step W10). The addition to the score is for preventing instantaneous erroneous detection. The constant values α, β,
How to take γ and ε can be automatically determined after statistically examining the operating highway, how accurately the sudden event can be actually detected.

Then, the occurrence of a sudden event is determined based on the score (step W11), and the determination result is output (step W12). The score and the determination result are used for comprehensive determination. In the above-described embodiment, the sample spectrum distribution of the traffic Q (tk) is obtained. However, the same processing is performed by obtaining the sample spectrum distribution of the speed V (tk) and the time occupancy O (tk). , A sudden event occurrence determination can be made. Further, besides the above, even when the sample spectrum distribution of the headway time interval, the number of vehicles existing in the road section, the space occupancy rate of the vehicle, the spatial average speed, and the headway distance is obtained, the sudden event occurrence determination is performed by the same processing. I can do it.

In the sample spectrum distribution, a similar effect can be obtained by using the value of the spectral density and the phase characteristic in addition to the value of the power spectrum itself and the value of the peak frequency. 3.5 Conventional Traffic Flow Abnormality Detection Based on Measured Values of Road Traffic Volume, Vehicle Speed, Image Data, etc. (Method E) The difference between the measured value of traffic volume or vehicle speed and the measured value on the downstream side is calculated, and when the difference is greater than or equal to this difference threshold value, the occurrence of an unexpected event in the section is determined, and the results are integrated. Send to judgment.

The determination unit E measures the image of the road with a camera installed on the road, detects a stopped / low-speed vehicle, tracks a lane-changed vehicle, etc. by image processing, and based on the result, determines an accident or the like. May be detected, and the result may be sent to the comprehensive judgment (see [6] below). [6] Yamada, Miyao, Sakai, Nishiyama, Ishishita, Negishi "Development of an abnormal traveling detection system in a tunnel" Sumitomo Electric No. 145,
pp.124-129, September 1994 In addition to the above, a method using a neural network can be adopted. These known methods are collectively referred to as “method E”.

4. Comprehensive Judgment The computer 11 is provided with an overall judgment section 23. The comprehensive determination unit 23 includes a plurality of determination algorithms A to E
Based on the determination result, the likelihood (probability) of occurrence of the sudden event is calculated and determined. 4.1 Normalization of Evaluation Value Traffic Flow Abnormality Detection Methods A to described in 3.1 to 3.5 above
In E, although terms are different, evaluation values (FIG. 3), correlations (FIG. 4), matching rates (FIG. 10), and scores (FIG. 1)
5) Calculate values called differences and set threshold values in advance, and when these values exceed those threshold values (in the case of correlation, lower than the threshold values), Abnormality was detected.

In the case of “correlation”, it is necessary to take the reciprocal of the correlation in order to match with another method. Further, if the ranges that these values can take are different, comprehensive judgment cannot be made, so the possible ranges need to be standardized. Thereafter, these values are normalized so as to take a range from 0 to 1. Then, the value after the normalization is referred to as an “evaluation value” with a unified name. And method A
The evaluation value calculated by the method P _A , the evaluation value calculated by the method B is P _B , the evaluation value calculated by the method C is P _C , the evaluation value calculated by the method D is P _D , and the evaluation value is calculated by the method E the evaluation value will be referred to as P _E.

As examples of the comprehensive judgment method using these evaluation values P _{A to} P _E , the following three judgment methods will be described. In the following, the well-known method E is excluded for convenience of explanation, and the four methods A to D are described as a basis. However, the embodiment of the present invention is not limited to this embodiment, and any one of the methods A to D is performed. May be replaced by method E or an unknown method, or method E or an unknown method may be added to methods A to D to implement the present invention.

4.2 Comprehensive Judgment 1 (Majority Decision Method) This method is a method of making an overall judgment based on the number of abnormalities detected by methods A to D. FIG. 16 is a flowchart for explaining the majority decision method. If all the processes of the methods A to D have been completed (step (1)), it is determined whether an abnormality has been detected by three or more methods (step (2)). If the abnormality is detected by three or more methods, the likelihood of occurrence of the sudden event is sufficiently high and it is determined that "the occurrence of the sudden event" (step (6)).

If the abnormality is detected only by two or less methods, the evaluation values calculated by each method are added (step
(3)). If the added value P _A + P _B + P _C + P _D exceeds the threshold value, the likelihood of occurrence of the sudden event is moderately high, and it is determined that “attention state in which the possibility of occurrence of the sudden event is high” ( Step (7)). If the threshold is too high, the number of detection omissions increases, and if the threshold is too low, erroneous detections increase.
The threshold must be set to an appropriate value. As this threshold value, for example, a value of １ ／ times the total number of methods (4 in this case) may be set.

If the added value does not exceed the threshold value, the likelihood of occurrence of a sudden event is low, and it is determined that "the occurrence of a sudden event has not occurred" (step (5)). In the above processing, in step (2), it was determined whether abnormality was detected by three or more of the four methods.
The number is not limited to three, but may be two or four. Further, in step (4), a large number of thresholds may be provided, and a plurality of likelihoods of the occurrence of the sudden event may be obtained stepwise.

4.3 Comprehensive Judgment 2 (Weighting Method) In this method, the evaluation values calculated by methods A to D are weighted to obtain an average. The weighting factors are α and
beta, gamma, and [delta], and displays the weighted average value P _{tota l.} P _total = (αP _A + βP _B + γP _C + δP _D ) / (α + β
+ Γ + δ) FIG. 17 is a flowchart for explaining the weighting method. If all the processes of the methods A to D have been completed (step (11)), P _total is calculated according to the above equation.
Then, this P _total is compared with the detection threshold value (step
(2)), if the detection threshold is exceeded, the likelihood of occurrence of the sudden event is sufficiently high and it is determined that the sudden event has occurred (step (1)
Five)). If this detection threshold is too high, detection omissions will increase, and if the detection threshold is too low, false detections will increase. This detection threshold value may be automatically determined based on the results of the detection omission rate and the erroneous detection rate described later with reference to FIG.

If the detection threshold is not exceeded, this P
_{The total} is compared with the attention threshold (step (14)). Attention threshold <detection threshold. If the caution threshold is exceeded, the likelihood of a sudden event is moderately high,
Judge as "Attention state that is likely to have an unexpected event"
(Step (16)). If the caution threshold value is not exceeded, the likelihood of occurrence of a sudden event is low, and it is determined that "the occurrence of a sudden event has not occurred" (step (17)).

Here, how to determine the weighting factors α, β, γ, δ will be described. As a premise for making this determination, traffic measurement data is actually collected, and based on the traffic measurement data at the time of the occurrence of the sudden event, an abnormality is detected by each of the methods A to D,
It is necessary to check the actual results such as whether it was correctly detected. FIG. 18 is a flowchart for explaining a method of recording the detection rate and the like. First, traffic measurement data is constantly accumulated (step (21)). If it is known that the sudden event has actually occurred (YES in step (22)), the traffic measurement data before and after the occurrence time is referred to (step (23)), and the processing of method A to method D is performed (step (22)). 24)), it is determined whether the evaluation value exceeds the threshold value and the abnormality of the traffic flow is detected by each method. The above processing is performed every time a sudden event occurs.

As a result, the probability of correctly detecting the total number of incidents is defined as the “correct detection rate”, the probability of not being detected for the total number of incidents is defined as the “missing detection rate”, and the total number of detections. On the other hand, the probability of erroneous detection is referred to as “false detection rate”, and the time from the occurrence of a sudden event until detection is referred to as “detection delay time” (step (25)). Comprehensive judgment unit 23
Classifies these values for each of the methods A to D and classifies and records them according to traffic conditions, days of the week, seasons, weather, and time zones. Table 1 below shows an example of evaluating the recorded contents.

[0103]

[Table 1]

The overall judgment section 23 determines the weight coefficient at the current time. The weighting factors are traffic volume Q, speed V, occupancy O,
It is a function such as road alignment (curve, zigzag, etc.), day of the week, time zone, degree of congestion, past detection results (Table 1), and the like. FIG. 19 is a flowchart for explaining the weight coefficient determination processing. This process is a process performed in real time. Correction is made to the initial values of α to δ (for example, all the values are the same).

First, the weight of the road section to be detected is added according to the road alignment (step (31)). For example, if the road alignment is likely to be a bottleneck, the weight coefficient α of method A (the number of vehicles) and the weight coefficient β of method B (lane use rate) are increased. Next, the weight based on the day of the week is added (step (32)). For example, if the current day is a weekday, the weight coefficient α of the method A (the number of vehicles), the weight coefficient β of the method B (lane use rate), and the weight coefficient γ of the method C (matching) are increased. If it is a holiday, the weight coefficient δ of the method D (spectrum) is increased.

Next, weights based on time periods are added.
(Step (33)). For example, in the morning, the weight coefficient α of the method A (the number of vehicles) is increased. If it is midnight, the weight coefficient β of the method B (lane use rate) is reduced. Next, weights based on past results are added (step (34)). For example, a weight coefficient of a method having a high detection rate in the section is increased. Then, check the current traffic condition (degree of congestion) (Step (3
Five)). If there is no congestion (NO in step (36)), the weight coefficient β for method B (lane use rate) and the weight coefficient δ for method D (spectrum) are increased (step (37)).

If there is a traffic jam, it is determined whether the traffic jam is early.
(Step (38)) If the traffic jam is early, method A (the number of vehicles)
(Step (40)). If it is during the traffic jam or at the end, the weight coefficient δ of the method D (spectrum) is increased (step (39)). As described above, the weight coefficient is automatically determined. 4.4 Comprehensive judgment 3 (detection order method) This method is a comprehensive judgment method that focuses on the detection delay time of method A to method D.

For simplicity, two methods (methods A to D)
However, here, methods A and D are used. ) Will be described as an example. FIG. 20 is a graph showing, based on the past results, a ratio at which a sudden event was detected by the methods A and D at the same time, a ratio at which the method A was detected earlier than the method D, and a ratio at which the method A was detected later than the method D. is there. The rate at which sudden events were detected simultaneously by Method A and Method D was 43%, the rate at which Method A detected earlier than Method D was 34%, and the rate at which Method A detected later than Method D was 23%. The reason is that the detection rate tends to be high in the early stage of the traffic jam after the occurrence of the sudden event, while the detection ratio tends to be high in the middle to the end period of the traffic jam after the occurrence of the sudden event in the method A. It is considered based.

FIG. 21 (a) is a graph showing the weighted average detection success rate and the failure rate when the methods A and D are simultaneously detected. The detection success rate and the failure rate are, for example, 89
% And 11%. FIG. 21 (b) is a graph showing the detection success rate and the failure rate by weighted averaging when the method A detects earlier than the method D. The detection success rate and the failure rate are, for example, 100% and 0%. FIG. 21 (c)
9 is a graph showing a weighted average detection success rate and a failure rate when method D detects data earlier than method A. The detection success rate and the failure rate are, for example, 80% and 20%.

As described above, the detection success rate when the method A is detected earlier than the method D is as high as 100%. Therefore, according to the present detection order method, when the method A is actually detected earlier than the method D, , And it is unconditionally determined that a sudden event has occurred. That is, the likelihood of occurrence of an abnormality is set to the maximum value. For the first time in other cases, the judgment is made by the method of the comprehensive judgment 1 or the comprehensive judgment 2. FIG. 22 is a flowchart for explaining the detection order method determination method. All the processes of the methods A to D are completed (step (51)), and at least the methods A and D
Has detected the occurrence of an unexpected event (YE in step (52))
S), referring to the data of the detection time of each of the methods A and D (step (53)).

It is checked which method has been detected earlier (step (54)). At this time, if the detection times of the methods are so far apart that the same sudden event is not detected, it is meaningless to perform the comprehensive determination. Therefore, it is preferable that the comprehensive judgment 3 is made only when the detection time difference of each method is within a certain time or within a certain processing cycle, and it is preferable not to make this judgment otherwise. If the method A has detected earlier than the method B, the likelihood is the highest and the occurrence of a sudden event is determined (step (55)).

If the simultaneous detection is performed or the method B is detecting earlier than the method A, the occurrence of the sudden event is determined by using the above-described method of the comprehensive judgment 1 or the comprehensive judgment 2 (step (56)). ). When it is determined that an unexpected event has occurred (YES in step (57)), it is determined that the likelihood is medium and there is a possibility that an unexpected event has occurred.
8)). When the occurrence of the sudden event is not determined, the likelihood is low and it is determined that the state is normal (step (59)).

As described above, in the comprehensive judgment 3, the method A is the method B
If it is detected earlier, the occurrence of a sudden event is determined without giving the above-mentioned overall judgment 1 or overall judgment 2 with emphasis on the past tendency. 4.5 Identification of Sudden Event Occurrence Section When the occurrence of the above-described sudden event is determined in a plurality of sections, the likelihood of the determination in each section is compared, and the section with the highest likelihood is identified as the sudden event occurrence section. Can be specified as

When the occurrence of a sudden event and the section in which the sudden event occurs are determined as described above, the traffic management center 10 displays the occurrence of the sudden event such as "accident / stop ahead" on the variable display boards 6 and 9. Then, the occurrence of the sudden event is notified to the vehicle through the roadside beacon 7. In addition, it communicates with related organizations 13 and broadcasting stations 14 through communication lines. In the case of traffic condition attention, the traffic management center 10 sets the variable display board 6
9, a message such as “attention to the front” that calls attention of the driver is displayed, and the vehicle is notified through the roadside beacon 7 that the vehicle is in the traveling caution section.

If road construction or the like is already planned and an unexpected event is expected to occur, the traffic management center 10 determines the occurrence of the unexpected event at that time, The occurrence of a sudden event is not displayed on the display boards 6 and 9, and no notification is made to the related organization 13 or the broadcasting station 14.

[0116]

As described above, according to the traffic flow abnormality detecting apparatus or method of the present invention, the faults of each system can be compensated by combining the detection results of each system.
The occurrence of a sudden event on the road can be detected with higher accuracy.

[Brief description of the drawings]

FIG. 1 is a schematic diagram showing a traffic flow monitoring system for detecting a traffic flow abnormality.

FIG. 2 is a functional block diagram of a computer 11 in the traffic management center 10.

FIG. 3 is a flowchart illustrating a process performed by a determination unit for monitoring occurrence of a sudden event.

FIG. 4 is a flowchart illustrating a process performed by a determination unit B to monitor the occurrence of a sudden event.

FIG. 5 is a flowchart illustrating a vehicle group matching processing method performed by a matching determination unit C;

FIG. 6 is a diagram illustrating a two-dimensional grid-like directed graph.

FIG. 7 is a diagram of a two-dimensional grid-like directed graph for showing the result of solving the longest path problem with n = 3 and m = 4.

FIG. 8 is a two-dimensional grid-like directed graph when n = 4 and m = 3 and only vehicle b4 is not matched.

FIG. 9 is a diagram illustrating a two-dimensional lattice-like directed graph for explaining an algorithm that does not require an external gap at the latest time point in the time series.

FIG. 10 is a flowchart illustrating a process performed by a determination unit C to monitor the occurrence of a sudden event.

FIG. 11 is a graph showing a variation in traffic volume Q (tn) calculated every day on the day when an accident actually occurred.

FIG. 12 is a graph showing the time change of a sample spectrum (power spectrum) pj.

FIG. 13 is a graph in which a sample spectrum pj is obtained and peak values of the spectrum are plotted.

FIG. 14 is a graph in which peak frequencies of a sample spectrum are plotted.

FIG. 15 is a flowchart illustrating a process performed by the determination unit D to monitor the occurrence of a sudden event.

FIG. 16 is a flowchart for explaining a majority decision method which is one of the comprehensive judgment methods.

FIG. 17 is a flowchart illustrating a weighting method that is one of the comprehensive determination methods.

FIG. 18 is a flowchart illustrating a method of recording a detection rate or the like based on past results.

FIG. 19 is a flowchart illustrating a weight coefficient determination process.

FIG. 20 shows a method using the method A,
6 is a graph showing a ratio of simultaneous detection by method D, a ratio of method A detected earlier than method D, and a ratio of method A detected later than method D.

FIG. 21A is a graph showing a weighted average detection success rate and a failure rate when methods A and D are simultaneously detected. (b) is a graph showing the detection success rate and the failure rate by the weighted average when the method D detects earlier than the method A. (c) is a graph showing a weighted average detection success rate and a failure rate when the method D detects earlier than the method A.

FIG. 22 is a flowchart illustrating a determination method of a detection order method.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Expressway 2 General road 3 Vehicle detector 4 Primary processing unit 5 Vehicle detector 6 Variable display board 7 Roadside beacon 9 Variable display board 10 Traffic management center 11 Computer 13 Related organization 14 Broadcasting station 21 Input processing unit 22, A to E Judgment unit 23 Total judgment unit 25 Output processing unit

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Kenji Amame 1-3-1 Shimaya, Konohana-ku, Osaka-shi F-term in Sumitomo Electric Industries, Ltd. Osaka Works 5H180 AA01 BB15 CC04 CC18 DD04 EE03 EE12 EE15

Claims (14)

[Claims] An abnormality detecting means for detecting an abnormality of a traffic flow in two or more different methods based on traffic measurement data, and a detection result of each method by the abnormality detecting means are combined.
An apparatus for detecting an abnormality in a traffic flow, comprising: a general determination means for determining occurrence of an abnormality in a traffic flow. 2. The system according to claim 1, wherein said comprehensive determination means makes a comprehensive determination based on the number of methods in which an abnormality is detected.
Abnormality detection device for traffic flow described. 3. The traffic flow abnormality detection device according to claim 2, wherein said comprehensive determination means determines the occurrence of a traffic flow abnormality if the number of detected abnormalities exceeds a threshold value. 4. The traffic flow abnormality according to claim 1, wherein said comprehensive determination means determines the occurrence of an abnormality in the traffic flow based on an added value obtained by adding the evaluation values of the respective methods. Detection device. 5. The traffic flow according to claim 1, wherein said comprehensive judgment means performs a weighted average operation on the evaluation value of each method, and makes a comprehensive judgment based on the weighted average value. Anomaly detection device. 6. The traffic flow abnormality detecting device according to claim 5, wherein said comprehensive determination means determines the occurrence of a traffic flow abnormality if the weighted average value exceeds a threshold value. 7. The method according to claim 5, wherein the weighting coefficient is a function of one of the following (a) to (g) or a combination thereof, and is automatically determined. Abnormality detector for traffic flow. (a) Traffic measurement data, (b) Road alignment, (c) Day of week, (d) Time zone, (d) Degree of congestion, (e) Detection accuracy of each method (f) Weather 8. The traffic flow abnormality detection device according to claim 1, wherein said comprehensive determination means makes a comprehensive determination based on the order of time when each system detects an abnormality. 9. The traffic flow abnormality detection according to claim 1, wherein the detection result of each method is obtained based on traffic measurement data before and after the actual occurrence of the sudden event, and the result is accumulated as actual data. apparatus. 10. The traffic flow abnormality according to claim 9, wherein the actual data includes one or more of a correct detection rate, a detection omission rate, a false detection rate, and a detection delay time. Detection device. 11. The traffic flow abnormality detection device according to claim 1, further comprising information providing means for outputting a result of the determination by said comprehensive determination means as abnormality information. 12. The comprehensive judging means makes a stepwise judgment in accordance with the magnitude of the value on which the judgment is based, and the information providing means makes a decision based on the result of the stepwise judgment by the comprehensive judging means. The traffic flow abnormality detection device according to claim 11, wherein the content of the abnormality information is changed. 13. The system according to claim 1, wherein said general determination means specifies an abnormality occurrence section in accordance with a magnitude of a value serving as a basis of the determination when a traffic flow abnormality is detected in a plurality of road sections. The traffic flow abnormality detection device according to claim 1. 14. A traffic flow, comprising detecting traffic flow abnormalities in two or more different modes based on traffic measurement data, and determining the occurrence of traffic flow abnormalities by combining the detection results of the respective modes. Abnormality detection method.