JP6812337B2 - Programs, devices and methods for estimating the tendency of a predetermined event to occur in a moving object - Google Patents

Description

The present invention relates to a technique of collecting position information of a large amount of moving objects as big data and estimating the occurrence tendency of a predetermined event such as a traffic accident.

Conventionally, a technology for predicting a traffic accident by learning a traffic accident occurrence pattern from control information (traffic volume, traffic accident, road construction) and traffic accident information for each time zone using a sensor installed on the road has been used. (See, for example, Patent Document 1). According to this technique, there are a plurality of groups consisting of an accident occurrence pattern learning unit and an accident occurrence forecasting unit according to the tendency of traffic accidents, and these groups are switched according to the time zone.

There is also a technology of an accident prevention system using an accident prevention device mounted on an automobile and a monitoring device for vehicle photography installed on a road (see, for example, Patent Document 2). According to this technique, a monitoring device collects driving characteristics of each vehicle from captured images. Then, a collision accident between vehicles is predicted from the relationship between the driving locus information of the own vehicle and the driving characteristic information of another vehicle close to the own vehicle. The predicted collision possibility information is notified to each driver as an alarm.

Furthermore, the distance is measured from the round-trip time of the wireless signal between the traffic light and the communication device using the wireless tag installed on the traffic light and the communication device and projector of the automobile (others, pedestrians, bicycles, bikes). However, there is also a technique for projecting position information on a projector (see, for example, Patent Document 3).

Japanese Unexamined Patent Publication No. 2016-194845 JP-A-2015-225366 Japanese Unexamined Patent Publication No. 2016-162151

According to the above-mentioned conventional technology, unless a sensor or a wireless tag is installed on a road or a traffic light, the movement of a moving object cannot be detected, and of course, an accident between moving objects cannot be predicted. That is, there is a problem that infrastructure equipment on the road is required.

On the other hand, in the mobile phone communication system that constantly collects the position information of a large number of moving objects, the inventors of the present application tend to generate a predetermined event such as a traffic accident even if there is no real-time property. I wondered if it could be estimated.

Therefore, the present invention estimates the tendency of a predetermined event to occur in a geographical range by using location information that can be collected only by an existing mobile phone communication system without requiring infrastructure equipment around the moving object. It is intended to provide programs, devices and methods that can be used.

According to the present invention, it is a program for operating a computer mounted on a device for estimating the occurrence tendency of an accident event in a plurality of moving objects.
A position group generation means for generating a position group from the position information of each of a plurality of moving objects,
For each position group output from the position group generation means based on the position information of the moving object in which an accident event has occurred in the past, element group data of a plurality of moving objects in the direction of travel is input to the learning engine as teacher data. And a teacher data input means to learn the tendency of accident events,
For the position group output from the position group generation means based on the position information of the moving object to be estimated, the element group data of the traveling direction of a plurality of moving objects is input to the learning engine, and the output is in the position group. It is characterized in that a computer functions as an event occurrence tendency estimation means for acquiring the occurrence tendency of an accident event.

According to other embodiments in the program of the present invention
The occurrence of an accident event is based on a plurality of first moving objects and a plurality of second moving objects of different types.
The position group generation means generates a position group from the position information of the first moving object and the second moving object.
The teacher data input means has a plurality of first moving objects for each position group output from the position group generating means based on the position information of the first moving object and the second moving object in which an accident event has occurred in the past. The element group data of the traveling direction of the moving object and the traveling direction of the plurality of second moving objects are input to the learning engine as teacher data.
The event occurrence tendency estimating means has a plurality of first moving objects for the position group output from the position group generating means based on the position information of the first moving object and the second moving object to be estimated. It is also preferable to make the computer function to input the element group data of the traveling direction and the traveling direction of the plurality of second moving objects into the learning engine.
According to other embodiments in the program of the present invention
The first moving object is a vehicle,
The second object to move is a person
It is also preferable.

According to other embodiments in the program of the present invention
It is also preferable to make the computer function so that the element group data further includes the velocity of the moving object.

According to other embodiments in the program of the present invention
It is also preferable that the position group causes the computer to function as a cluster in which element group data is clustered by DBSCAN (Density-Based Spatial Clustering of Applications with Noise) based on the position information.

According to other embodiments in the program of the present invention
It is also preferable to make the computer function so that the position group is a geographical range divided based on latitude and longitude, or a road range divided based on a road map.

According to other embodiments in the program of the present invention
The teacher data input means inputs the total number of the total element group data counted for each position group and the representative value of each element into the learning engine.
It is also preferable that the event occurrence tendency estimation means causes the computer to function so as to input the total number of the total element group data counted for each position group and the representative value of each element into the learning engine.

According to other embodiments in the program of the present invention
It is also preferable to make the computer function so that the representative value is the average value, the median value, the mode value, the minimum value or the maximum value of each element of the element group data included in the position group.

According to other embodiments in the program of the present invention
The position group generating means further divides the position group for each predetermined time zone and divides the position group.
The event occurrence tendency estimation means inputs the element group data in the position group based on the predetermined time zone to be estimated into the learning engine, and acquires the occurrence tendency of the accident event in the position group in the predetermined time zone as the output. It is also preferable to make the computer function.

According to other embodiments in the program of the present invention
The position group generation means further divides the position groups according to the weather conditions.
The event occurrence tendency estimation means inputs the element group data in the position group based on the weather condition to be estimated into the learning engine, and obtains the occurrence tendency of the accident event in the position group of the weather condition as the output of the computer. It is also preferable to make the function work.

According to other embodiments in the program of the present invention
A position information collecting means that collects position information for each time from each of a plurality of first moving objects and a plurality of second moving objects, and
For each of the first moving object and the second moving object, the movement / stay is estimated from the displacement of the position information, and only for the first moving object and the second moving object estimated to be "moving". it is also preferable to further function a computer as estimation data selecting means you output to the position group generation unit as an element group data.

According to the present invention, it is a device for estimating the occurrence tendency of an accident event in a plurality of moving objects.
A position group generation means for generating a position group from the position information of each of a plurality of moving objects,
For each position group output from the position group generation means based on the position information of the moving object in which an accident event has occurred in the past, element group data of a plurality of moving objects in the direction of travel is input to the learning engine as teacher data. And a teacher data input means to learn the tendency of accident events,
For the position group output from the position group generation means based on the position information of the moving object to be estimated, the element group data of the traveling direction of a plurality of moving objects is input to the learning engine, and the output is in the position group. It is characterized by having an event occurrence tendency estimation means for acquiring the occurrence tendency of an accident event.

According to the present invention, it is a method for estimating an event occurrence tendency of a device in a plurality of moving objects.
The device is
The first step of generating a position group from the position information of each of a plurality of moving objects,
For each position group output by the first step based on the position information of the moving object in which the accident event occurred in the past, the element group data of a plurality of moving objects in the traveling direction is input to the learning engine as teacher data. And the second step to learn the tendency of accident events,
For the position group output by the first step based on the position information of the moving object to be estimated, the element group data of the traveling direction of a plurality of moving objects is input to the learning engine, and the output is in the position group. It is characterized by executing the third step of acquiring the occurrence tendency of the accident event.

According to the program, apparatus and method of the present invention, the location information that can be collected only by the existing mobile phone communication system without the need for infrastructure equipment around the moving object is specified in the geographical range. It is possible to estimate the tendency of events to occur.

It is a system block diagram in this invention. It is a functional block diagram of the estimation apparatus in this invention. It is a 1st explanatory diagram which shows the process of a learning stage in this invention. It is a map showing each predetermined range. It is a 2nd explanatory diagram which shows the process of the learning stage in this invention. It is 1st explanatory drawing which shows the processing of the estimation stage in this invention. It is a 2nd explanatory diagram which shows the processing of the estimation stage in this invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a system configuration diagram in the present invention.

The estimation device 1 as a server collects the position information of a large number of mobile terminals. According to FIG. 1, as a plurality of first mobile terminals and a plurality of second mobile terminals of different types, for example, the following are shown.
First mobile terminal: For example, mounted on a vehicle (first moving object) Second mobile terminal: For example, possessed by a person (second moving object) The estimation device 1 is a position group in which a large number of mobile terminals coexist. Estimate the tendency of a predetermined event to occur for (geographical range). Here, as a predetermined event, for example, a traffic accident between a vehicle (first moving object) and a person (second moving object) is assumed. In addition, the probability of occurrence is assumed as the tendency of occurrence.
That is, according to the present invention, the estimation device 1 acquires only the position information of the mobile terminal 2 and estimates the tendency (probability of occurrence) of a traffic accident as a predetermined event.

In the case of a mobile phone communication system, the position information of the mobile terminal 2 can be constantly collected by base station positioning or access point positioning. In that case, the estimation device 1 can be incorporated as a part of the function of the mobile phone communication system.

The mobile terminal 2 may have a positioning unit and may acquire and estimate the current position information by receiving a positioning radio wave such as GPS (Global Positioning System). It may also be based on base station positioning. Further, the position information may be acquired via CAN (Controller Area Network). The mobile terminal 2 always transmits the position information to the estimation device 1.
Further, as another embodiment, the mobile terminal 2 uses the location information together with, for example, VICS (registered trademark, Vehicle Information and Communication System) information and traffic condition information (traffic jam, traffic regulation, etc.) such as ITS Connect. It may be transmitted to 1.

FIG. 2 is a functional configuration diagram of the estimation device according to the present invention.

The estimation device 1 has each functional unit divided into a <learning stage> and a <estimation stage>. Here, the "position group generation unit 101" that classifies a plurality of moving objects into position groups is commonly used at each stage.
The estimation device 1 has an event occurrence data storage unit 120, a teacher data input unit 121, and a learning engine 100 as a <learning stage>.
Further, the estimation device 1 has an event occurrence tendency estimation unit 131, a position information collection unit 132, and an estimation data selection unit 133 in addition to the learning engine 100 as the <estimation stage>.
These functional components are realized by executing a program that makes the computer mounted on the device function. In addition, the processing flow of these functional components can be understood as an event occurrence tendency estimation method.

<Learning stage>
FIG. 3 is a first explanatory diagram showing the processing of the learning stage in the present invention.

[Event occurrence data storage unit 120]
The event occurrence data storage unit 120 has a "position (latitude / longitude)", "traveling direction of the first moving object", and "traveling direction of the second moving object" in which a predetermined event (for example, a traffic accident) has occurred in the past. This is a collection of a plurality of "element group data" having the above as an element. Of course, it is not always necessary to use different types as in the first mobile terminal and the second mobile terminal, and only "position information" and "movement direction of the moving object" may be used.

Further, it is also preferable that the event occurrence data storage unit 120 further includes the "velocity of the first moving object" (and the speed of the second moving object) as the element group data. The velocity is calculated from the distance difference between the positions of the moving objects with respect to the time difference between the two time points.
Then, the element group data of the event occurrence data storage unit 120 is referred to by the position group generation unit 101 and the teacher data input unit 121.

[Position group generation unit 101]
The position group generation unit 101 generates a position group from the position information of the moving object (the first moving object and the second moving object).
(Learning stage) A position group is generated based on the position information of the first moving object and the second moving object in which a predetermined event has occurred in the past.
(Estimation stage) A position group is generated based on the position information of the first moving object and the second moving object to be estimated.
The position group in the learning stage and the position group in the estimation stage have no geographical commonality and are generated independently in each stage. Therefore, the learning stage may learn based on the position group created in the geographical range "Tokyo", and the estimation stage may estimate the event occurrence tendency based on the position group created in the geographical range "Osaka". it can.

In addition, the position group generation unit 101 has the following two methods for generating a position group.
(Generation Method 1) Position-Based Clustering The position group may be a cluster in which element group data is clustered by DBSCAN (Density-Based Spatial Clustering of Applications with Noise) based on position information.

DBSCAN is a density-based clustering method that uses the following two parameters.
Distance threshold: ε (Eps-neighborhood of a point)
Target number threshold: minPts (a minimum number of points)
Since it is not necessary to set the number of clusters k in advance, it is possible to classify a data group whose number of clusters k is unknown.
The points of data are classified into the following three types.
Core points: Points with at least minPts of adjacent points within a radius of ε Reachable points: There are not as many adjacent points as there are minPts within a radius of ε,
Outliers with Core points within radius ε: Points with no adjacent points within radius ε Create a cluster from a collection of core points and assign reachable points to each cluster.
That is, in DBSCAN, when a set of points is given to a certain space, the points in the high density region are grouped together, and the points in the low density region are set as outliers.

(Generation Method 2) Classification into Predetermined Ranges Most simply, the position group may be a geographical range divided based on latitude and longitude, or a road range divided by a road map.

FIG. 4 is a map showing each predetermined range.
According to FIG. 4A, the predetermined range is defined as a geographical range (latitude / longitude) obtained by dividing the map into tiles.
According to FIG. 4B, the predetermined range is defined as the road range in which the road map is divided.
A position group may be formed in units of such a geographical range or a road range.

FIG. 5 is a second explanatory diagram showing the processing of the learning stage in the present invention.

[Teacher data input unit 121]
The teacher data input unit 121 generates element group data of a plurality of moving objects in the direction of progress for each position group output from the position group generating unit 101 based on the position information of the moving object in which a predetermined event has occurred in the past. , Input to the learning engine as teacher data.
Here, for each of the different types of moving objects, the element group data of the traveling direction of the first moving object and the traveling direction of the second moving object are input to the learning engine as teacher data. May be good.
As a result, the teacher data input unit 121 causes the learning engine 100 to learn the element group data in units of position groups, and builds a learning model based on the tendency of occurrence of a predetermined event.
Of course, when the element group data includes "speed", the speed is also input to the learning engine 100 along with the direction of travel.

Further, the teacher data input unit 121 may count the "total number" of the entire element group data and the "representative value" of each element for each position group. The representative value may be the average value, median value, mode value, minimum value or maximum value of each element of the element group data included in the position group. For example, it may be an average value for traveling or a median value for speed.
Then, the teacher data input unit 121 inputs the total number of the entire element group data and the representative value of each element to the learning engine 100 for each position group to construct a learning model.

[Learning engine 100]
The learning engine 100 is used from each of the learning stage and the estimation stage.
(Learning stage) The learning engine 100 obtains element group data (for example, traveling direction and speed, or the total number and representative values thereof) in units of position groups of moving objects in which a predetermined event (for example, a traffic accident) has occurred in the past. Enter and build your own learning model.
The learning engine 100 has a machine learning algorithm for statistical classification. Specifically, general machine learning algorithms used for identification problems, such as Random Forest, Support Vector Machine, Naive Bayes, and K-nearest neighbor, may be used.

(Estimation stage) The learning engine 100 inputs element group data in units of position groups of moving objects to be estimated, and estimates the occurrence tendency of a predetermined event.
The occurrence tendency of a predetermined event is the likelihood (probability of occurrence or score) output from the learning engine 100. In the learning stage, only the element group data in which the predetermined event has occurred is input, so that the estimation stage knows the probability of occurrence of the predetermined event from the input element group data of the position group. be able to.

<Estimation stage>
FIG. 6 is a first explanatory diagram showing the processing at the estimation stage in the present invention.

[Location information collection unit 132]
The position information collecting unit 132 collects position information for each time from each of the plurality of moving objects. This position information may be based on base station positioning or may be based on GPS positioning.
The collected position information is output to the estimation data selection unit 133.

[Estimated data selection unit 133]
The estimation data selection unit 133 estimates the movement / stay of each moving object from the displacement of the position information, and outputs only the moving object estimated to be “moving” to the position group generation unit 101 as element group data. ..
Specifically, based on the time-series position information of the moving object, if the displacement of the position information is equal to or more than a predetermined distance within a predetermined time (for example, 30 minutes), it is determined as "movement". Further, when the speed of the moving object is equal to or higher than a predetermined speed, it may be determined as "moving".
Then, only the moving object presumed to be "moving" is used as the element group data.

FIG. 7 is a second explanatory diagram showing the processing at the estimation stage in the present invention.

[Event occurrence tendency estimation unit 131]
The event occurrence tendency estimation unit 131 has a plurality of first moving objects for the position group output from the position group generating means based on the position information of the first moving object and the second moving object to be estimated. The element group data of the progress direction and the progress direction of the plurality of second moving objects are input to the learning engine, and the occurrence tendency of a predetermined event in the position group is acquired as the output.

As another embodiment, when the position group generation unit 101 further divides the position group for each predetermined time zone, the event occurrence tendency estimation unit 131 determines the element group data in the position group based on the predetermined time zone to be estimated. Is input to the learning engine, and as its output, the tendency of occurrence of a predetermined event in the position group in the predetermined time zone may be acquired. The tendency of a predetermined event (for example, a traffic accident) to occur is a likelihood (probability of occurrence or score) output from the learning engine 100.

As described above, only the position information of the moving object is used, but it is also preferable to classify these position information into "every predetermined time zone" (early morning, night, etc.). As a result, the element group data can be accumulated for each predetermined time zone, and the position group can be generated to respond to the movement of the moving object according to the predetermined time zone.

Further, as another embodiment, when the position group generation unit 101 further divides the position group for each weather condition, the event occurrence tendency estimation unit 131 is the element group data in the position group based on the weather condition to be estimated. Is input to the learning engine, and as its output, the tendency of occurrence of a predetermined event in the position group of the weather condition may be acquired.

As described in detail above, according to the program, apparatus and method of the present invention, location information that can be collected only by an existing mobile phone communication system without requiring infrastructure equipment around a moving object. By using, it is possible to estimate the tendency of a predetermined event to occur in a geographical range.
In particular, according to the present invention, there is no need for commonality in the geographic range between the learning stage and the estimation stage, and an event occurrence tendency in the second geographic range is used by using a learning engine trained in the first geographic range. Can be estimated.

With respect to the various embodiments of the present invention described above, various changes, modifications and omissions within the scope of the technical idea and viewpoint of the present invention can be easily made by those skilled in the art. The above explanation is just an example and does not attempt to restrict anything. The present invention is limited only to the scope of claims and their equivalents.

1 Estimator 120 Event occurrence data storage unit 100 Learning engine 101 Position group generation unit 121 Teacher data input unit 131 Event occurrence tendency estimation unit 132 Position information collection unit 133 Estimated data selection unit 2 Mobile terminal

Claims (13)

A program that activates a computer mounted on a device that estimates the tendency of accident events to occur in multiple moving objects.
A position group generation means for generating a position group from the position information of each of a plurality of moving objects,
For each position group output from the position group generation means based on the position information of the moving object in which an accident event has occurred in the past, the element group data of a plurality of moving objects in the traveling direction is used as teacher data in the learning engine. Teacher data input means to input and learn the tendency of accident events,
For the position group output from the position group generation means based on the position information of the moving object to be estimated, element group data of a plurality of moving objects in the traveling direction is input to the learning engine, and the position is output as the output. A program characterized by operating a computer as an event occurrence tendency estimation means for acquiring the occurrence tendency of an accident event in a group. The occurrence of an accident event is based on a plurality of first moving objects and a plurality of second moving objects of different types.
The position group generation means generates a position group from the position information of the first moving object and the second moving object.
The teacher data input means has a plurality of first moving objects for each position group output from the position group generating means based on the position information of the first moving object and the second moving object in which an accident event has occurred in the past. The element group data of the traveling direction of one moving object and the traveling direction of a plurality of second moving objects are input to the learning engine as teacher data, and the data is input to the learning engine.
The event occurrence tendency estimating means has a plurality of first movements of the position group output from the position group generating means based on the position information of the first moving object and the second moving object to be estimated. The program according to claim 1, wherein the computer functions to input the element group data of the traveling direction of the object and the traveling direction of the plurality of second moving objects into the learning engine. The first moving object is a vehicle,
Second moving object, the program according to claim 2, characterized in Oh Ru <br/> that in humans. The program according to any one of claims 1 to 3, wherein the element group data causes a computer to function so as to further include a speed of a moving object. From claim 1, the position group is characterized in that the computer functions as a cluster in which the element group data is clustered by DBSCAN (Density-Based Spatial Clustering of Applications with Noise) based on the position information. The program according to any one of 4 . The position group according to any one of claims 1 to 4 , wherein the computer functions so as to be a geographical range divided based on latitude and longitude or a road range divided based on a road map. program. The teacher data input means inputs the total number of the total element group data counted for each position group and the representative value of each element into the learning engine.
Claim 1 is characterized in that the event occurrence tendency estimation means causes a computer to function so as to input the total number of all the element group data counted for each position group and the representative value of each element into the learning engine. The program according to any one of 6 to 6 . The representative value is an average value of each element of the element group data included in the location group, median, mode, in claim 7, characterized in that the functioning of the minimum value or computer so that the maximum value Described program. The position group generating means further divides the position group for each predetermined time zone.
Said event generating tendency estimation means, the element group data in the location group based on a predetermined time period to be estimated input to the learning engine, the occurrence tendency of the accident event at the position group of the predetermined time period as an output The program according to any one of claims 1 to 8 , wherein the computer functions to acquire the data. The position group generating means further classifies the position group according to the weather conditions.
The event occurrence tendency estimation means inputs element group data in the position group based on the weather condition to be estimated into the learning engine, and acquires the occurrence tendency of an accident event in the position group in the weather condition as its output. The program according to any one of claims 1 to 9 , wherein the computer functions as described above. A position information collecting means that collects position information for each time from each of a plurality of first moving objects and a plurality of second moving objects, and
For each of the first moving object and the second moving object, the movement / stay is estimated from the displacement of the position information, and only for the first moving object and the second moving object estimated to be "moving". the program according to claims 1 to 1 wherein one of 10 for causing the computer to further function as estimation data selecting means you output to the position group generation unit as the element group data. A device that estimates the tendency of accident events to occur in multiple moving objects.
A position group generation means for generating a position group from the position information of each of a plurality of moving objects,
For each position group output from the position group generation means based on the position information of the moving object in which an accident event has occurred in the past, the element group data of a plurality of moving objects in the traveling direction is used as teacher data in the learning engine. Teacher data input means to input and learn the tendency of accident events,
For the position group output from the position group generation means based on the position information of the moving object to be estimated, element group data of a plurality of moving objects in the traveling direction is input to the learning engine, and the position is output as the output. An apparatus characterized by having an event occurrence tendency estimation means for acquiring an accident event occurrence tendency in a group. It is a method of estimating the event occurrence tendency of the device in multiple moving objects.
The device
The first step of generating a position group from the position information of each of a plurality of moving objects,
For each position group output by the first step based on the position information of the moving object in which the accident event occurred in the past, the element group data of a plurality of moving objects in the traveling direction is input to the learning engine as teacher data. And the second step to learn the tendency of accident events,
For the position group output by the first step based on the position information of the moving object to be estimated, the element group data of a plurality of moving objects in the traveling direction is input to the learning engine, and the position group is output as the output. A method for estimating an event occurrence tendency of an apparatus, which comprises executing a third step of acquiring the occurrence tendency of an accident event in the above.

Images