JP2018180601A - Detector, detection method, and detection program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a detector, a detection method, and a detection program that can detect a traffic abnormality such as the congestion induced by some vehicles.SOLUTION: A detector 1 comprises: a prediction unit 12 that predicts an exit route to an approach route of a vehicle at each of intersections included in a road network of an observation area on the basis of past statistical information; a calculation unit 13 that obtains the real exit route of the vehicle and calculates a prediction accuracy for each of the intersections; and a determination unit 14 that determines a traffic abnormality on the basis of the comparison results between the current value of the prediction accuracy and the past record of it.SELECTED DRAWING: Figure 2

Description

  The present invention relates to an apparatus, method and program for detecting a traffic abnormality.

Traditionally, traffic jams are caused, for example, by time zones or by bottlenecks present in the road network. Such patterns of congestion are predictable and usually occur over a wide range.
Moreover, in order to detect the traffic congestion which generate | occur | produced, the various methods using GPS data are proposed (for example, refer nonpatent literature 1-3).

On the other hand, there are traffic congestions caused by unexpected situations such as accidents, disasters and incidents. In addition, when vehicles on the road network are configured as autonomous vehicles in the future, it is conceivable that a traffic jam or the like may be caused by a cyber attack using the vehicles as a botnet.
Such traffic congestion usually occurs only in a narrow area of the road network and is difficult to predict, but it is desirable to be able to detect an event early and cope with the situation.

Sanjay Chawla, Yu Zheng, and Jiafeng Hu. 2012. Inferring the Root Cause in Road Traffic Anomalies. In Proceedings of the 2012 IEEE 12th International Conference on Data Mining (ICDM '12). IEEE Computer Society, Washington, DC, USA, 141-150. Linsey Xiaolin Pang, Sanjay Chawla, Wei Liu, and Yu Zheng. 2013. On detection of emerging anomalous traffic patterns using GPS data. Data Know. Eng. 87 (September 2013), 357-373. Bei Pan, Yu Zheng, David Wilkie, and Cyrus Shahabi. 2013. Crowd sensing of traffic anomalies based on human mobility and social media. In Proceedings of the 21st ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (SIGSPATIAL '13). ACM, New York, NY, USA, 344-353.

  However, since conventional methods for detecting traffic anomalies can only detect when an event (congestion) appears completely in the traffic pattern, it is timely to detect botnet attacks that cause congestion. could not.

  An object of the present invention is to provide a detection device, a detection method, and a detection program capable of early detecting a traffic abnormality such as a traffic jam induced by a part of vehicles.

  The detection device according to the present invention includes a prediction unit that predicts an exit route with respect to an approach route of a vehicle based on past statistical information, at each of the intersections included in the road network of the observation area; A calculation unit that acquires an exit route and calculates a prediction accuracy for each intersection, and a determination unit that determines a traffic abnormality based on a result of comparing the current value of the prediction accuracy with a history.

  The determination unit may determine that the traffic is abnormal if the difference between the current value of the prediction accuracy and the central value of the history is equal to or more than a reference value.

  The determination unit may determine that the traffic is abnormal when the number of intersections at which the difference is equal to or more than the reference value is equal to or more than a predetermined number.

  The determination unit may set, as the reference value, a value based on a comparison result with the history of the prediction accuracy at another intersection.

  The determination unit may perform determination periodically, and output a warning when the number of times of determination of traffic abnormality within a predetermined observation period is equal to or more than a predetermined value.

  The detection method according to the present invention comprises the steps of: predicting, at each intersection included in the road network in the observation area, an exit route for the approach route of the vehicle based on past statistical information; The computer executes a calculating step of acquiring a leaving route and calculating a prediction accuracy for each intersection, and a determination step of determining a traffic abnormality based on a result of comparing a current value of the prediction accuracy with a history.

  The detection program according to the present invention comprises, at each of the intersections included in the road network in the observation area, predicting the exit route with respect to the approach route of the vehicle based on past statistical information; For causing a computer to execute a calculation step of acquiring an exit route and calculating prediction accuracy for each intersection, and determining a traffic abnormality based on a result of comparing a current value of the prediction accuracy with a history It is a thing.

  According to the present invention, it is possible to early detect a traffic abnormality such as a traffic jam induced by a part of vehicles.

It is a figure which shows the observation object of the traffic condition by the detection apparatus which concerns on embodiment. It is a figure showing the functional composition of the detecting device concerning an embodiment. It is a figure which illustrates the notification algorithm of the GPS location from the vehicle which concerns on embodiment. It is a figure which illustrates the algorithm of the abnormality detection which concerns on embodiment.

Hereinafter, an example of the embodiment of the present invention will be described.
FIG. 1 is a diagram showing an observation target of a traffic situation by the detection device 1 according to the present embodiment.
As shown in the figure, the detection device 1 detects a traffic abnormality by observing a route of entry and exit for each vehicle at an intersection in a road network.

  Here, a vehicle entering an intersection is in the process of transitioning from one approach route to another exit route. The detection device 1 uses the route through which the past vehicle that entered the same intersection has passed and the route through which the past vehicle that has followed the same route as the vehicle has passed, and use the exit route to which the vehicle follows next Predict. The detection device 1 calculates the prediction accuracy after the set time, and determines the traffic abnormality at the intersection using the difference compared with the reference value based on the history.

The following terms are defined for the road network in the observation area.
Segment: The segment tεE is a directed edge defined by the endpoints s, eεV when traffic flows from s to e.
Road traffic graph: The road traffic graph G = (V, E) is a directed graph. V is a set of graph vertices that represent the end points of the segment. E is a set of edges representing a segment.
Path: Path p is an ordered list of connected segments as follows.
p = <t ₁ , t ₂ , t ₃ ,..., t _n >, t _k . e = t _{k + 1} . s, 1 ≦ k <n

GPS location: GPS location gl is a set of data consisting of <Long (longitude), Lat (latitude), TS (time stamp)>.
Trajectory: Trajectory tr is a series of GPS locations. Trajectory tr is transformed to path p using a known map matching algorithm.
Block: A block is a rectangular area defined, for example, as B = <gl ₁ , gl ₂ , gl ₃ , gl ₄ > using GPS location gl _i .
Grid: A grid is a collection of non-overlapping blocks and covers the entire area of the road traffic graph G.

Intersection: Intersection I is a set of end points of each segment when a plurality of segments are connected.
I = [t1. s, t1. e, t2. s, ..., tn. e]

FIG. 2 is a diagram showing a functional configuration of the detection device 1 according to the present embodiment.
The detection device 1 is an information processing device (computer) such as a server device or a personal computer. The detection device 1 includes an input / output device for various data, a communication device, and the like in addition to the control unit 10 and the storage unit 20.
The control unit 10 is a part that controls the entire detection device 1, and realizes various functions in the present embodiment by appropriately reading and executing various programs stored in the storage unit 20. The control unit 10 may be a CPU.

  The storage unit 20 is a storage area for various programs for causing the hardware group to function as the detection device 1 and various data, and may be a ROM, a RAM, a flash memory, a hard disk (HDD), or the like. Specifically, the storage unit 20 is a detection program that causes the control unit 10 to execute each function of the present embodiment, as well as a history database 21 and a prediction result database 22 etc. obtained by the processing described later executed by the program. Remember.

  The control unit 10 includes a preprocessing unit 11, a prediction unit 12, a calculation unit 13, a determination unit 14, and an analysis unit 15.

The preprocessing unit 11 collects trace data of GPS location for each vehicle passing through the road traffic graph, and stores path history data in the history database 21. The collected data is associated with the vehicle identifier v _id and the block. Every vehicle in the block has a unique identifier v _id in the block and is assigned only while staying in the block.

Here, GPS tracing of a vehicle is usually performed at a constant time period. In this embodiment, since it is necessary to observe a path when the vehicle passes through the intersection, the GPS location at the time of entering the intersection and after passing is further acquired.
For example, a vehicle is equipped with a sensor that detects that the vehicle is at an intersection, and the vehicle can refer to this sensor information to determine the notification timing of the GPS location in addition to the notification every constant time.

FIG. 3 is a diagram illustrating an algorithm for notifying a GPS location from a vehicle according to the present embodiment.
First, a periodic GPS _time for notifying the GPS location of the vehicle and a time Tran _time at which it can be expected that the vehicle is moving from the intersection (end point of the approach route) to the next segment to be the exit route are input (line number : 1), the following steps are repeatedly performed.

When the vehicle detects that it is at an intersection (line number: 4), it notifies the GPS location (line number: 5), and after a predetermined time Tran _time (line number: 6) notifies the GPS location again (Line number: 7).
If the vehicle is not at an intersection, the vehicle notifies the GPS location (line number: 10) each _time the predetermined period GPS _time elapses (line number: 8).

  The prediction unit 12 determines, based on past statistical information obtained from the history database 21, the next segment to be the exit route for the path which is the entry route of the vehicle at each of the intersections included in the block which is the observation area. , And stored in the prediction result database 22.

  When acquiring the actual exit route of the vehicle, the calculation unit 13 calculates the prediction accuracy for each intersection by comparison with the prediction result, and stores the prediction accuracy in the prediction result database 22.

  The determination unit 14 determines traffic abnormality based on the result of comparing the current value of the prediction accuracy with the history stored in the prediction result database 22.

  When the traffic abnormality is determined, the analysis unit 15 performs extraction of a subgraph that is a set of intersections recognized as an abnormality, cause analysis of the traffic abnormality, and visualization of a traffic state.

Next, processing steps performed in the traffic abnormality detection method by the detection device 1 will be described in detail.
[Preprocessing]
The preprocessing unit 11 performs map matching to specify the position of the vehicle on the map. In the process of map matching, the acquired GPS location is adjusted and mapped on the road of the digital map.

In addition, the specific method of the pre-processing task for specifying the path | pass containing map matching can be selected from many known methods including the thing shown by following document A, for example.
The output of this step is the path of the vehicle and is the input to the path prediction step. Further, the output path of the vehicle is also stored in the history database 21.

  (Document A) Yu Zheng. 2015. Trajectory Data Mining: An Overview. ACM Trans. Intell. Syst. Technol. 6, 3, Article 29 (May 2015), 41 pages.

Path prediction
The block B and the identifier v _{id of the} vehicle at the end point d of the segment at time t _n are given. In this step, the prediction unit 12 predicts the next segment to be an exit route among all the segments starting from the same intersection as the vehicle entering the intersection which is the end point of the segment. This step is executed for all v _id ∈V, where V is a set of vehicles located at the end point of the segment at time t _n .

The prediction unit 12 predicts the next segment to be an exit route for the entry route of each vehicle included in the set V, that is, the path p up to the current time t _n . At this time, the prediction unit 12 uses history information of past paths obtained from a vehicle that has passed through similar paths during a time zone similar to t _n . Here, the path of the vehicle is obtained by segmentation of the traffic in pre-processing. The path length is controlled, for example, by the maximum time, the maximum length and the minimum length.

  In addition, the specific method for path | pass prediction can be selected from the known methods including the thing shown by the above-mentioned literature A and the following literature B and C. For example, data-driven techniques using decision trees, Markov models or neural networks are applicable to path prediction.

(Reference B) Philip Pecher, Michael Hunter, and Richard Fujimoto. 2016. Data-Driven Vehicle Trajectory Prediction. In Proceedings of the 2016 annual ACM Conference on SIGSIM Principles of Advanced Discrete Simulation (SIGSIM-PADS '16).
(Document C) Anna Monreale, Fabio Pinelli, Roberto Trasarti, and Fosca Giannotti. 2009. WhereNext: a location predictor on trajectory pattern mining. In Proceedings of the 15th ACM SIGKDD international conference on Knowledge discovery and data mining (KDD '09). ACM, New York, NY, USA, 637-646.

[Error detection]
When the block B to be observed is given, the determination unit 14 detects a traffic abnormal event such as ongoing traffic congestion occurring in the block B.

FIG. 4 is a diagram illustrating an algorithm of abnormality detection according to the present embodiment.
When the block B is input (line number: 1), the control unit 10 sets the reference time (currTime) to the current time (systime) (line number: 2), and the number of observations (Obs _cnt ) and the number of abnormality determinations After initializing (Anomaly _cnt ) to zero (line numbers 3 and 4), the following processing is repeatedly executed.

When the observation period (Obs _time ) has elapsed from the reference time (line number: 6), the control unit 10 updates the reference time (line number: 7) and initializes the number of alarming intersections (Alarm _cnt ) (line Number: 8). Furthermore, after the control unit 10 specifies a set of vehicles (V _tran ) that has shifted between segments during the observation cycle and a set of intersections (E _tran ) that are between the shifted segments (line numbers: 9, 10) The path prediction and the evaluation of the prediction result are performed for each of the identified intersections (line number: 11).

The PredictAndEvaluate function evaluates the prediction result relative to the actual and evaluates the prediction accuracy Acc _Val at each intersection (line number: 12). As the evaluation criteria, for example, false positive rate, accuracy rate, recall rate, F-scale, sensitivity, etc. may be used.
After that, the determination unit 14 calculates the difference of the current value from the median (or average value) Acc _{μ of} the past prediction accuracy using an appropriate distance function.

If the difference is equal to or greater than the reference value (line number: 13), the determination unit 14 counts up the number of warning intersections (line number: 14).
In addition, when the number of warning intersections exceeds the predetermined threshold value Obs _thres (line number: 17), the determination unit 14 determines that the traffic is abnormal, and counts up the number of abnormality determinations (line number: 18).

The above-mentioned reference value which compares a difference may be set up a value based on a comparison result with a history of prediction accuracy in other intersections. For example, the reference value is set by raising the sensitivity by the square root and multiplying the average value of differences at a plurality (N) of intersections by multiplying the predetermined coefficient A nom _sf .

The determination unit 14 periodically determines the traffic abnormality, and detects it as a traffic abnormality event when the number of abnormality determinations within a predetermined observation period is equal to or more than a predetermined threshold value Anom _thes (line numbers: 23, 23), Output a warning (line number: 24).
The output data is, for example, a set of end points (intersections) of segments determined to have abnormal traffic.

[Abnormal subgraph extraction]
Traffic anomalies are unlikely to be isolated at one intersection, and there is a high probability that multiple intersections in the area will exhibit abnormal traffic patterns. The analysis unit 15 extracts a subgraph to which the detected abnormal intersection is connected.

[Cause analysis]
The analysis unit 15 uses social media and the video of the surveillance camera to identify the cause of the traffic abnormality.
In social media, individuals involved in traffic anomalies may be posting their experiences. The analysis unit 15 acquires the information posted by the user in the area determined to be abnormal, and extracts the term used by the user. These terms are used as descriptions of detected traffic anomalies.

  Here, the analysis unit 15 may use known text mining technology used to extract related terms that describe a set of documents. The analysis unit 15 treats past posts as a series of documents and extracts related terms. The latest posts during traffic anomalies may be extracted as a set of documents and related terms. The analysis unit 15 can extract the term related to the traffic abnormality, which frequently appears in recent posts without past posts, by comparing these sets with the past posts.

[Visualization]
The analysis unit 15 visualizes the information of the detected traffic abnormality and presents it to the persons concerned. The data to be visualized may include, for example, a mark indicating the severity of the traffic abnormality based on the magnitude of the difference in prediction accuracy and the like.
In addition, the analysis unit 15 may acquire and present an image of a surveillance camera in an area where a traffic abnormality is detected. This image can be used for cause analysis of traffic abnormalities.

According to the present embodiment, at each of the intersections included in the road network of the observation area, the detection device 1 predicts a segment to be an exit route with respect to the approach route of the vehicle based on past statistical information. And calculate the prediction accuracy for each intersection as compared with the actual exit route. Then, the detection device 1 determines traffic abnormality based on the result of comparing the current value of the prediction accuracy with the history.
In the conventional method of detecting traffic congestion by observing the flow of traffic between two points in a block, it was necessary to wait for the vehicle to move from one point to another. On the other hand, since the detection device 1 performs prediction and detection of abnormality based on fine-grained information of the combination of the approach route and the exit route for each vehicle at the intersection on the traffic graph, it is possible to It is possible to more quickly detect a traffic abnormality such as a traffic jam induced by an abnormal behavior of a part of vehicles caused.

  At this time, the detection device 1 can reduce the number of vehicles that need to be observed for abnormality detection as the prediction accuracy of transition between segments is enhanced. If the prediction accuracy is close to 100%, and if several consecutive vehicles make a transition between segments different from the prediction, external factors such as cyber attacks are considered, and the detection device 1 detects such traffic anomalies. The onset can be detected quickly before the impact spreads.

  In addition, since the detection apparatus 1 uses the identifier of a valid vehicle only in a block, it can not trace the specific user on a road network over a wide range, but can ensure a user's privacy.

  The detecting device 1 determines that the traffic is abnormal if the difference between the current value of the prediction accuracy and the central value of the history is equal to or greater than the reference value, so that the prediction accuracy is different from the history, that is, there is an unpredictable movement of the vehicle. The case can be easily detected as a traffic anomaly.

  The detection device 1 determines that there is a traffic abnormality when the number of intersections at which the difference is equal to or greater than the reference value is a predetermined number or more, and thus can appropriately detect a traffic abnormality caused by a cyber attack or the like affecting a plurality of intersections.

  The detection device 1 sets a value based on the comparison result with the history of prediction accuracy at another intersection as a reference value for comparing the difference, so that the unique traffic can be determined by comparison with another intersection. Therefore, it is possible to properly detect traffic anomalies caused by unusual traffic in a narrow area, as compared with the widely spread normal traffic.

  The detection device 1 periodically determines the traffic abnormality, and outputs a warning when the number of times the traffic abnormality is determined within a predetermined observation period is equal to or more than a predetermined value. Therefore, the detection device 1 can reduce false detection by outputting a warning when traffic abnormality is determined continuously.

  As mentioned above, although embodiment of this invention was described, this invention is not limited to embodiment mentioned above. Further, the effects described in the present embodiment only list the most preferable effects arising from the present invention, and the effects according to the present invention are not limited to those described in the present embodiment.

  In the present embodiment, the various reference values and the values of the thresholds and the like that are initially set are not limited. For example, the observation period, period, block size and the like can be appropriately set according to the normal traffic conditions and the like. Furthermore, depending on the determination result, the parameters may be dynamically adjusted, for example, for identification of an abnormal vehicle or identification of an abnormal point.

Further, the reference value may be a fixed value, or may be a variable value based on surrounding conditions as described above, or may be calculated by an appropriate function.
Furthermore, in order to improve the prediction accuracy, the length of the path used as the history may be set appropriately.
By adjusting these parameters, an appropriate detection system with reduced false positives can be constructed according to the road conditions for each area.

  The detection method by the detection device 1 is realized by software. When implemented by software, a program that configures this software is installed in an information processing apparatus (computer). Also, these programs may be recorded on removable media such as a CD-ROM and distributed to the user, or may be distributed by being downloaded to the user's computer via a network. Furthermore, these programs may be provided to the user's computer as a web service via a network without being downloaded.

Reference Signs List 1 detection device 10 control unit 11 preprocessing unit 12 prediction unit 13 calculation unit 14 determination unit 15 analysis unit 20 storage unit 21 history database 22 prediction result database

Claims (7)

A prediction unit that predicts an exit route with respect to the approach route of the vehicle based on past statistical information at each of the intersections included in the road network of the observation area;
A calculation unit that acquires an actual exit route of the vehicle and calculates a prediction accuracy for each intersection;
A detection unit that determines a traffic abnormality based on a result of comparing a current value of the prediction accuracy with a history.   The detection device according to claim 1, wherein the determination unit determines that the traffic is abnormal when the difference between the current value of the prediction accuracy and the central value of the history is equal to or more than a reference value.   The detection device according to claim 2, wherein the determination unit determines that the traffic is abnormal when the number of the intersections at which the difference is equal to or more than the reference value is a predetermined number or more.   The detection apparatus according to claim 2, wherein the determination unit sets a value based on a comparison result with a history of the prediction accuracy at another intersection as the reference value.   The detection device according to any one of claims 1 to 4, wherein the determination unit periodically performs determination, and outputs a warning when the number of times of determination of traffic abnormality within a predetermined observation period is a predetermined number or more. . A prediction step of predicting an exit route with respect to an approach route of the vehicle based on past statistical information at each of the intersections included in the road network of the observation area;
Calculating the actual exit route of the vehicle and calculating the prediction accuracy for each intersection;
A determination step of determining a traffic abnormality based on a result of comparing a current value of the prediction accuracy with a history, and a computer executing the method. A prediction step of predicting an exit route with respect to an approach route of the vehicle based on past statistical information at each of the intersections included in the road network of the observation area;
Calculating the actual exit route of the vehicle and calculating the prediction accuracy for each intersection;
A detection program for causing a computer to execute a determination step of determining a traffic abnormality based on a result of comparing a current value of the prediction accuracy with a history.