JP2019127196A - Program, device and method for estimating driving characteristics of observed vehicle traveling in the periphery - Google Patents

Abstract

To provide a program, a device and a method that enable a driver of an own vehicle to recognize psychological characteristics of a driver of another vehicle traveling in the periphery of the own vehicle.SOLUTION: A device for estimating driving characteristics of a driver of an observed vehicle traveling in the periphery of an observing vehicle, includes: travel data storing means which stores speed of the observing vehicle, and a distance between the observing vehicle and the observed vehicle as seen from the observing vehicle, as travel data, in time series; feature amount data generating means for generating feature amount data containing as elements, the speed and the distance in the travel data; and a driving characteristics estimating engine for inputting multiple pieces of the feature amount data, and outputting a likelihood of a specific driving characteristic of the driver of the observed vehicle.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a technique for estimating a psychological characteristic of a driver in a vehicle.

Conventionally, a technique for estimating the driver's psychological characteristics from a driving signal (for example, steering angle and depression degree of accelerator and brake pedal) and a vehicle signal (for example, engine speed, vehicle speed, acceleration and angular velocity) in a traveling vehicle See, for example, non-patent documents 1 and 2. According to this technique, it comprises a learning step of constructing a learning model of an estimator using teacher data, and an estimation step of estimating a psychological characteristic using the estimator.
In the learning step, a learning model of the estimator is constructed in advance using, as teacher data, psychological characteristic data acquired by a paper test or the like and a driving signal and a vehicle signal acquired from a vehicle driven by the driver for each driver.
In the estimation step, by inputting the driving signal and the vehicle signal acquired from the traveling vehicle to the estimator, the psychological characteristics of the driver in the vehicle can be obtained.

Takuma Miyahara et al., “Estimation of Driver's Personal Characteristics Using Vehicle Signals”, Information Processing Society of Japan 78th National Convention, 3S-04, 3-147 to 3-148, [online], [January 2018 Search 18th], Internet <URL: file: /// C: / Users / hayah / Documents /% E6% A5% AD% E5% 8B% 99 / KDDI / 01% E6% 96% B0% E8% A6% 8F% E5% 87% BA% E9% A1% 98 / P2017-0382-P0906% E7% 9F% B3% E5% B7% 9D% E9% 9B% 84% E4% B8% 80 / IPSJ-Z78-3S- 04.pdf> Mitsunobu Kanuma et al., “Automatic Estimation of Driver Behavior and Personal Characteristics Using Driving Signals”, Automobile Engineering Society Academic Lecture Preprint, No. 128-13, pp.11-16, [online], [Heisei 30 Search January 18, 2009], Internet <URL: http: //jglobal.jst.go.jp/public/20090422/201302216010473907>

However, the driving signal and the vehicle signal obtained from the vehicle to be estimated may be affected by the driving of another vehicle traveling around the vehicle. For example, an accident or a driving problem may occur due to the dangerous driving of the driver of another vehicle.
In recent years, in particular, as driving behavior that causes harassment or intimidation of the other party, "driving" is a problem. This is an act of forcing a vehicle traveling ahead to give way, narrowing the inter-vehicle distance, making an abnormal approach, and chasing, high beam, passing, knurling, and side alignment.
In addition, if the driver of another vehicle traveling in the surrounding area has, for example, a decline in cognitive ability or lack of coordination, the driver's lack of attention or an operation error is the cause. , An accident may occur.

  On the other hand, the inventors of the present application can reduce the risk of an accident if the driver of the own vehicle can recognize the psychological characteristics of the driver of another vehicle traveling around the vehicle. I thought that.

  Therefore, the present invention estimates the psychological characteristics of the driver of the observed vehicle even when the driving signal / vehicle signal of the observed vehicle traveling around the observation vehicle cannot be obtained as viewed from the observation vehicle. It is an object to provide a program, an apparatus, and a method that can be used.

According to the present invention, there is provided a program that causes a computer mounted on a device for estimating the driving characteristic of the driver of the observed vehicle traveling around the observation vehicle to function.
Travel data storage means that stores in time series the speed of the observation vehicle and the distance between the observation vehicle and the observed vehicle as travel data;
Feature quantity data generating means for generating feature quantity data with speed and distance in the running data as elements;
The computer is caused to function as a driving characteristic estimation engine that inputs a plurality of feature amount data and outputs the likelihood of a predetermined driving characteristic in the driver of the observed vehicle.

According to another embodiment of the program of the invention:
Driving Characteristic Estimation The function of the computer is such that the predetermined driving characteristic of the engine is dangerous driving characteristic, emotional instability, noncoordination, sudden change of speed, rapid startability, quick stop, sudden stop, or meandering, or a combination thereof. It is also preferable that

According to another embodiment of the program of the invention:
The feature data is based on a predetermined time range,
The driving characteristic estimation engine inputs a plurality of feature quantity data of different predetermined time ranges,
Although the temporal change in the speed of the observation vehicle is small, the driving characteristic estimation engine that has input the feature amount data increases as the temporal change in the distance to the observed vehicle increases. It is also preferable to make the computer function so as to estimate the likelihood of the characteristic to be high.

According to another embodiment of the program of the invention:
About travel data storage means
The speed of the observation vehicle is measured by the temporal displacement of the position output from the positioning sensor installed on the observation vehicle,
It is also preferable to cause the computer to function such that the distance to the observed vehicle is measured by a distance sensor installed in the observed vehicle.

According to another embodiment of the program of the invention:
The traveling data storage means further stores the angular velocity with respect to the observation vehicle, and further stores the angle with respect to the observation vehicle,
It is also preferable to cause the computer to function so that the feature amount data of the feature amount data generation means further includes an angular velocity and an angle as elements.

According to another embodiment of the program of the invention:
Video data storage means for storing video data showing the observed vehicle viewed from the observation vehicle;
Causing a computer to function as a video change detection means for detecting a video change of the vehicle under observation using the video data storage means;
It is also preferable to cause the computer to function so that the feature amount data of the feature amount data generation means further includes an image change as an element.

According to another embodiment of the program of the invention:
It is also preferable to cause the computer to function so that the image change of the image change detection means is the blinking change of the lamp of the observation vehicle.

According to another embodiment of the program of the invention:
Sound data storage means storing sound data around the observation vehicle;
Causing the computer to function as sound change detection means for detecting sound change in a predetermined frequency band using sound data storage means;
It is also preferable to cause the computer to function so that the feature amount data of the feature amount data generation means further includes a sound change as an element.

According to another embodiment of the program of the invention:
The device is a smartphone or a mobile terminal and can be installed in an observation vehicle.
It is also preferable that the program is installed as an application.

According to another embodiment of the program of the invention:
When the likelihood output from the driving characteristic estimation engine exceeds a predetermined threshold value, the computer further functions as an alarm display means for displaying the driving characteristic of the driver of the observed vehicle as an alarm to the driver of the observation vehicle. It is also preferable that

According to another embodiment of the program of the invention:
The travel data storage means stores, for each of the plurality of observation vehicles, the speed of the observation vehicle and the distance between the observation vehicle viewed from the observation vehicle in time series.
Using the travel data storage means, for each vehicle to be observed, further function as data integration means for integrating travel data from a plurality of observation vehicles on the time axis,
It is also preferable that the feature quantity data generating means causes the computer to function to generate feature quantity data for each of the observed vehicles.

According to another embodiment of the program of the invention:
When the traveling data satisfies a predetermined condition based on the predetermined driving characteristics, it further includes learning data selecting means for selecting a plurality of traveling data as teacher data in a time series including the traveling data,
The feature amount data generation means generates feature amount data having the speed and the distance in the selected traveling data as elements,
The driving characteristic estimation engine constructs a learning model so as to output the likelihood of the predetermined driving characteristic in the driver of the observed vehicle by inputting a plurality of feature amount data as teacher data based on the predetermined driving characteristic. It is also preferable to operate the computer.

According to the present invention, there is provided a program for causing a computer mounted on an apparatus for constructing a learning model for estimating a driving characteristic of a driver of an observed vehicle running around an observation vehicle to function,
Travel data storage means for storing the speed of the vehicle and the distance between the observed vehicle as viewed from the observation vehicle as travel data in time series;
Learning data selection means for selecting a plurality of traveling data in time series including the traveling data when the traveling data satisfy a predetermined condition based on a predetermined driving characteristic;
Feature quantity data generating means for generating feature quantity data having elements of speed and distance in the selected traveling data;
The computer functions as a driving characteristic estimation engine that builds a learning model to output the likelihood of the predetermined driving characteristic of the driver of the observed vehicle by inputting a plurality of feature quantity data as teacher data based on the predetermined driving characteristic It is characterized by letting.

According to the present invention, there is provided an apparatus for estimating the driving characteristic of the driver of the observed vehicle while traveling around the observed vehicle,
Travel data storage means for storing the speed of the observation vehicle and the distance between the observed vehicle viewed from the observation vehicle as travel data in time series;
Feature quantity data generating means for generating feature quantity data with speed and distance in the running data as elements;
And a driving characteristic estimation engine for inputting a plurality of feature amount data and outputting a likelihood of a predetermined driving characteristic of a driver of the observed vehicle.

According to the present invention, there is provided a method of estimating driving characteristics of a device for estimating the driving characteristics of the driver of the observed vehicle traveling around the observation vehicle,
The apparatus has a travel data storage unit that stores, as travel data, the speed of the observation vehicle and the distance between the observed vehicle viewed from the observation vehicle in time series,
The device
A first step of generating feature data having elements of speed and distance in running data;
A second step of inputting a plurality of feature amount data and outputting a likelihood of a predetermined driving characteristic in the driver of the observed vehicle is executed.

  According to the program, the apparatus, and the method of the present invention, the driver of the observed vehicle can be obtained even when the driving signal / vehicle signal of the observed vehicle traveling around the observation vehicle cannot be obtained as viewed from the observation vehicle. Can be estimated.

It is the external view which looked at the observation vehicle from the driver of the observation vehicle. It is a functional block diagram of the driving | running characteristic estimation apparatus in this invention. It is a coordinate figure showing the track of the presumed vehicles to be observed. It is explanatory drawing showing the 1st feature-value data. It is explanatory drawing showing the input and output of a driving characteristic estimation engine. It is a functional block diagram of the portable terminal in this invention. It is explanatory drawing showing the 2nd feature-value data. It is an explanatory view showing integration of the run data which saw a specific observation vehicle from a plurality of observation vehicles. It is a functional block diagram which builds the learning model in a driving characteristic estimation engine.

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

  FIG. 1 is an external view of the observation vehicle as seen from the driver of the observation vehicle.

  The driver of the observation vehicle can see the observation vehicle traveling in the vicinity thereof. If the observed vehicle is driving dangerously, for example, the driver of the observation vehicle feels a strong accident risk and disturbs his / her driving operation.

  According to FIG. 1, for example, a mobile terminal 1 such as a smartphone is installed as a “driving characteristic estimation device” of the present invention on the dashboard of an observation vehicle. In this case, it is assumed that the application of the driving characteristic estimation function of the present invention is installed in the smartphone. Of course, it is not limited to a smart phone, and may be a dedicated device integrally mounted on a vehicle such as a navigation system, for example. Further, the vehicle to be observed is not necessarily traveling forward as viewed from the observation vehicle, and may be traveling laterally or backward.

The portable terminal 1 is equipped with a positioning sensor, and can acquire the current position by receiving positioning radio waves such as GPS (Global Positioning System).
In addition, the mobile terminal 1 is equipped with a camera, and can also capture an image of a vehicle under observation while traveling. The video data reflects the blinking state of the brake lamps, blinkers, and headlights of the observed vehicle, and the position of the observed vehicle viewed from the observation vehicle can also be detected.
Furthermore, the portable terminal 1 is equipped with a microphone, and can also pick up the sound of the vehicle under observation while traveling, such as the ringing of a horn.

As another embodiment, the mobile terminal 1 is connected to the observation vehicle in a CAN (Controller Area Network) via narrow-area wireless communication (for example, wireless LAN) or short-range wireless communication (for example, Bluetooth (registered trademark)). You may receive the vehicle information transmitted / received. CAN is a standard that is enhanced for noise resistance and is used for data transfer between interconnected devices.
In addition, the mobile terminal 1 holds map data, and can also acquire information such as a traveling road from the current position on the map.
Furthermore, the portable terminal 1 may also acquire road information and traffic information while traveling by receiving road traffic information (congested status, traffic restriction, etc.) from Vehicle Information and Communication System (VICS) (registered trademark). it can.

  FIG. 2 is a functional configuration diagram of the driving characteristic estimation apparatus according to the present invention.

The driving characteristic estimation device 1 of the present invention is for estimating the driving characteristic of the driver of the observed vehicle traveling around the observation vehicle. According to FIG. 2, the driving characteristic estimation device 1 has only the basic function of the present invention.
Of course, the driving characteristic estimation device 1 may be implemented as a server connected to the Internet. In that case, the server can receive travel data from the observation vehicle and estimate the driving characteristics of the observation vehicle.

  According to FIG. 2, the driving characteristic estimation device 1 has a traveling data storage unit 10, a feature amount data generating unit 11, and a driving characteristic estimation engine 12. Also, it optionally includes a data integration unit 17 (see FIG. 8 described later). These functional components are realized by executing a program that causes a computer installed in the device to function. Moreover, the flow of processing of these functional components can also be understood as a method of estimating operating characteristics of the device.

[Running data storage unit 10]
The travel data storage unit 10 stores the “speed” of the observed vehicle and the “distance” between the observed vehicle and the observed vehicle as travel data in time series.
An observation vehicle means a vehicle which acquires travel data.
An observed vehicle refers to another vehicle that is running around the observation vehicle.
The driving characteristic estimation device 1 of the present invention estimates driving characteristics of a driver of an observed vehicle using traveling data acquired by the observing vehicle.

According to FIG. 2, the travel data storage unit 10 stores travel data in which the “speed” of the observed vehicle and the “distance” with respect to the observed vehicle are associated with each other in time series.
The "velocity" of the observation vehicle is calculated based on the temporal displacement of the position output from the positioning sensor. As the positioning sensor, for example, there are GPS and base station positioning. Also, it may be acquired from an in-vehicle network such as CAN mounted on the observation vehicle. The "acceleration" of the observation vehicle can be calculated from the speed.

The “distance” of the vehicle to be observed is measured by a distance sensor installed on the observation vehicle.
As a distance sensor, for example, there is LiDAR (Light Detection and Ranging, Laser Imaging Detection and Ranging). In this method, the distance to the object is analyzed by irradiating a pulsed laser and measuring the time until the reflected light is received. For example, as a collision prevention system for automobiles, LiDAR is arranged around the front bumper or dashboard of an automobile so as to measure the distance between the vehicle ahead and a front object.
Moreover, as a distance sensor, there is also a sensor using a depth camera. The depth camera simultaneously acquires, for example, an RGB (Red-Green-Blue) image and a depth image (XYZ coordinates), and outputs a six-dimensional (RGBXYZ) point cloud (point cloud data) with color information. Thereby, the distance of the vehicle to be observed as seen from the observation vehicle can be acquired. Of course, even with video data taken by an RGB camera, the vehicle can be detected by video analysis, and the distance and angle between the vehicle and the observation vehicle can be calculated.

According to FIG. 2, the travel data storage unit 10 further stores “angular velocity” in addition to the speed for the observed vehicle, and further stores “angle” in addition to the distance for the observed vehicle. Yes.
The “angular velocity” of the observation vehicle can be calculated from an angular displacement measured by, for example, an acceleration sensor, a gyro sensor, or a geomagnetic sensor.
The “angle” of the observed vehicle can be detected from, for example, video data or point cloud data.

[Feature amount data generation unit 11]
The feature data generation unit 11 generates “feature data” having speed and distance as elements by dividing the traveling data into predetermined time ranges (windows). The feature amount data is a vector having travel data (at least speed and distance) as elements. The feature amount data for each window is output to the driving characteristic estimation engine 12.

  FIG. 3 is a coordinate diagram showing the estimated trajectory of the observed vehicle.

According to FIG. 3, the positional relationship between the observation vehicle and the observation vehicle is plotted on the geographical coordinate axis. The position of the observation vehicle is determined by the observation vehicle. On the other hand, the position of the observation vehicle is calculated by the distance and angle of the observation vehicle as viewed from the position of the observation vehicle. For example, at the timing (3) in FIG. 3, the position C of the vehicle to be observed is plotted from the distance and angle of the vehicle to be observed as viewed from the observation vehicle.
Moreover, the position shift is plotted as a trajectory with time. As apparent from FIG. 3, when the distance between the track of the observation vehicle and the track of the observation vehicle is not constant but changes suddenly, some sudden change in the observation vehicle causes the observation vehicle to change. It can be estimated that

  FIG. 4 is an explanatory diagram illustrating the first feature amount data.

According to FIG. 4, the change of the "velocity" of the observation vehicle and the change of the "distance" between the observation vehicle according to the elapsed time are shown. The scale of the vertical axis of speed and distance is set arbitrarily.
Then, the elapsed time is divided into windows (constant time intervals), and vectors having speed and distance as elements are generated as feature amount data for each window.
Note that the speed and distance that are the elements of the feature data are not limited to the values themselves, but the average value, maximum value, minimum value, median value, first and third quartile values, variance, FFT (Fast Fourier Transform) The amplitude of each subsequent frequency component may be used.

[Driving characteristic estimation engine 12]
The driving characteristic estimation engine 12 inputs a plurality of feature amount data of different predetermined time ranges (windows), and outputs the likelihood of the predetermined driving characteristic of the driver of the observed vehicle.

The driving characteristic estimation engine 12 may be a “machine learning engine” configured as feature data based on each likelihood of a predetermined driving characteristic as teacher data. The type of machine learning engine is not limited, but may be simple regression analysis, for example, Random Forest, Support Vector Machine, Naive Bayes, K-nearest neighbor, or the like.
In the case of using a machine learning engine, it is necessary to input in advance teacher data including a plurality of feature amount data serving as explanatory variables and driving characteristics serving as objective variables, and to construct a learning model.

  As an essential feature of the present invention, “despite the fact that the temporal change in the speed of the observation vehicle is small, the larger the temporal change in the distance to the observation vehicle, the stronger the driving characteristics of the observation vehicle It is in. That is, it can be assumed that the larger the movement of the observed vehicle relative to the movement of the observed vehicle, the more the observed vehicle moves. It is preferable that the driving characteristic estimation engine 12 to which such feature amount data is input estimates the likelihood of a negative predetermined driving characteristic in the observed vehicle to be high. As a result, the predetermined driving characteristic of the observed vehicle is clearly output.

  The predetermined driving characteristic of the driving characteristic estimation engine 12 may be any of dangerous driving characteristics, emotional instability, non-cooperation, rapid speed change, sudden startability, sudden stoppage, meandering property, or a combination thereof. .

  “Dangerous driving characteristics” include, for example, the possibility of traffic rule violations (overspeed / unreached or temporary stop violations, signal ignorance, pedestrian ignorance, lane changes in prohibited areas, etc.), traffic manners violations, etc. (Sudden interruption, sudden approach to intersection, etc.), possibility of other vehicle interference driving (spinning driving, widening, etc.) and other dangerous driving (sudden acceleration / deceleration, sudden steering, sudden stop, etc.) To do. In this manner, a large number of feature quantity data observed as dangerous driving are learned as teacher data.

  In addition, as emotional instability, non-cooperation, and other psychological characteristics of drivers that affect dangerous driving characteristics (as big 5 personalities, "non-extroversion", "non-workingness", "closeness of experience") There is. These psychological characteristics are previously obtained for each user by a questionnaire for a large number of users, and according to the psychological characteristics, the past feature amount data in which the vehicle driven by the user is observed is learned as teacher data Can do.

  Furthermore, as a driving operation that affects the dangerous driving characteristic, there are "speed rapid degeneration", "rapid startability" and "rapid stopability". In addition, there is “meandering property”, which is measured, for example, from video data acquired by a camera, by the movement of the vehicle with respect to a road lane marking. In these driving operations, a large number of past feature quantity data observed as dangerous driving can be learned as teacher data.

In addition, various indexes can be used as the predetermined operation characteristics of the operation characteristic estimation engine 12.
For example, there are DSQ (Driving Style Questionnaire) score and WSQ (Workload Sensitivity Questionnaire) score.
DSQ is an indicator of attitude, attitude, and attitude toward driving, and items such as “confidence in driving skills”, “reluctance”, and “sickness” are quantified with a score of 5 on the basis of a questionnaire.
WSQ is an index of what kind of driving load is felt strongly, and it is quantified by a score of 5 points based on a questionnaire about items such as "traffic situation grasp""road environment grasp""driving concentration inhibition" It becomes.
For example, according to the “driver aptitude diagnosis” implemented by the Automobile Accident Countermeasures Organization (NASVA), hazard sensitivity, attention distribution, judgment / operation timing, and movement accuracy for the “cognition and processing function” of the examinee Measure, and calculate a quantitative score. In addition, regarding "Psychology / Characteristics", we also calculate the score for the favor with others, the feeling of comfort, coordination, and the stability of emotion.
In addition, for example, a score calculated from an answer result and an answer result such as DBQ (Driving Behavior Questionnaire), DBI (Driving Behavior Inventory), DAS (Driving Anger Scale), etc., which are used to measure driver's psychological characteristics in Europe and the United States It may be.

  Further, the predetermined driving characteristic of the driving characteristic estimation engine 12 may be a general psychological characteristic not limited to the driving operation. For example, it may be a score calculated by an index such as an emotional empathy scale, an EAEQ (Everyday Attentional Experiences Questionnaire), an SSS (Sensation Seeking Scale), or an FFM (Five Factor Model).

  FIG. 5 is an explanatory view showing inputs and outputs of the driving characteristic estimation engine.

According to FIG. 5, the driving characteristic estimation engine 12 outputs the likelihood (probability and score) in the “emotional instability”. Here, any one of "low", "medium" and "high" is output as the likelihood of emotional instability. For example, a relative level in the likelihood distribution (for example, “high” if average + standard deviation or higher, “low” if average−standard deviation or lower, “medium” otherwise). May be.
For example, the driving characteristic estimation engine 12 outputs “high: 0.8”, “medium: 0.1”, and “low: 0.1” to the feature quantity data (1). After the likelihoods are calculated for all feature data, they are averaged and the label with the highest likelihood is output as the estimation result. According to FIG. 5, since the average value of the likelihood of “high” is the highest, the estimation result is output as “emotional instability: high”.

  FIG. 6 is a functional configuration diagram of the mobile terminal according to the present invention.

  According to FIG. 6, the portable terminal 1 includes the video data storage unit 130, the video change detection unit 131, the sound data storage unit 140, and the sound change in addition to the functional components of the driving characteristic estimation device of FIG. A detection unit 141, a vehicle information storage unit 150, a vehicle information change detection unit 151, an alarm display unit 16, and a data integration unit 17 are further included. These functional components are realized by executing a program that causes a computer installed in the device to function.

[Video data storage unit 130]
The video data storage unit 130 stores video data on which the observed vehicle viewed from the observation vehicle appears. The video data is taken by a camera (RGB camera or depth camera).

[Video Change Detection Unit 131]
The video change detection unit 131 detects a video change of the observation vehicle using the video data storage unit 130. As the image change, it is possible to detect the “flashing change of the lamp” of the observation vehicle as viewed from the observation vehicle. The lamps include brake lamps, blinkers and headlights. In particular, when the vehicle under observation is driving around, the lamp often blinks.
In addition, the “angular velocity” of the observed vehicle viewed from the observation vehicle can also be detected as the image change.
The “image change” detected here is included as an element in the feature amount data of the feature amount data generation unit 11.

[Sound data storage unit 140]
The sound data storage unit 140 stores sound data around the observation vehicle. Sound data is recorded by a microphone.

[Sound change detection unit 141]
The sound change detection unit 141 uses the sound data storage unit 140 to detect a sound change in a predetermined frequency band. As a sound change, a frequency band such as a horn or a human hoarseness is detected. Such a change in sound is often detected particularly when the vehicle under observation is driving around.
The “sound change” detected here is included as an element in the feature amount data of the feature amount data generation unit 11.

[Vehicle information storage unit 150]
The vehicle information storage unit 150 stores vehicle information transmitted and received in a CAN (Controller Area Network).
As vehicle information, there are various information as follows.
(Driving state information) Steering (lateral acceleration, front wheel turning angle, etc.), braking (braking distance to stop, forward acceleration, tire rotation speed, etc.), drive (vehicle speed, rear acceleration, tire rotation speed, etc.)
(Ambient environment information) Distance to front / rear / side obstacles, lane deviation situation (driving operation information) accelerator / brake pedal depression degree, steering wheel angle, steering angular velocity, various devices such as blinkers, wipers, power windows etc. Operation status

[Vehicle information change detection unit 151]
The vehicle information change detection unit 151 detects a change in predetermined vehicle information using the vehicle information storage unit 150. In particular, when the vehicle under observation is driving around, a sudden change is often detected in such vehicle information.

[Alarm display unit 16]
When the likelihood output from the driving characteristic estimation engine exceeds a predetermined threshold, the alarm display unit 16 displays the driving characteristic of the driver of the observed vehicle as an alarm to the driver of the observation vehicle. The alarm may be displayed on the display of the portable terminal 1 or may be used to call attention from the speaker by voice.
For example, in the case shown in FIG. 1 described above, the driver of the observation vehicle can drive at a distance while paying attention to the risk of an accident with respect to the observed vehicle seen in front.

  FIG. 7 is an explanatory diagram showing second feature amount data.

According to FIG. 7, compared with FIG. 4, the angular velocity and angle, and the flashing of the brake lamp are further included as elements of the feature amount data. The angular velocity can be measured by the acceleration sensor of the mobile terminal 1 disposed in the observation vehicle. The angle and the brake lamp can be detected from image data captured by the observed vehicle. The angular velocity, the angle, and the scale of the vertical axis of the brake lamp are each set arbitrarily.
Then, the elapsed time is divided into windows (fixed time intervals), and for each window, a vector having elements of speed and distance, angular speed, angle, and brake lamp is generated as feature amount data.
Of course, as in FIG. 4, the angular velocity and the angle, which are elements of the feature data, are not limited to the values themselves, and the average value, maximum value, minimum value, median, first and third quartiles, variance, It may be the amplitude for each frequency component after FFT (Fast Fourier Transform). In addition, the brake lamp may be a value based on not only ON / OFF but also the shortest, longest, average, etc. of the duration time as an element of the feature amount data.

  FIG. 8 is an explanatory view showing integration of travel data when a specific observed vehicle is viewed from a plurality of observation vehicles.

The travel data storage unit 10 of the driving characteristic estimation apparatus 1 for each of the plurality of observation vehicles, for each observation vehicle, the speed of the observation vehicle and the distance between the observation vehicle viewed from the observation vehicle in time series. I want to remember.
In this case, the data integration unit 17 integrates travel data from a plurality of observation vehicles on a time axis, using the travel data storage unit 10, for each observed vehicle. Thus, long-time travel data can be obtained for one vehicle to be observed. Moreover, even if the observed vehicle is a blind spot when viewed from one observation vehicle, the traveling data of the other observation vehicle can be supplemented.
And the feature-value data generation part 11 inputs the driving | running | working data for every to-be-observed vehicle integrated by the data integration part 17, and classify | categorizes into feature-value data for every to-be-observed vehicle.

According to FIG. 8, one observed vehicle A can be viewed from five observation vehicles a to e. That is, traveling data when looking at the observation vehicle A is obtained from each of the observation vehicles a to e.
At this time, the data integration unit 17 needs to specify that the travel data obtained by the plurality of observation vehicles a to e is a view of one observation vehicle A (identification of the observation vehicle). Method).
After that, the data integration unit 17 integrates the traveling data obtained from the respective observation vehicles a to e with the observed vehicle A (travel data integration method).

(First identification method of observed vehicle)
For example, the position at the acquisition time of the traveling data of the observed vehicle as viewed from the observation vehicle can be used.
Specifically, one observation vehicle a specifies the position of the observed vehicle A from the position of the observation vehicle a and the distance (and angle) between the observed vehicle A (see, for example, FIG. 3). ). Similarly, the other observation vehicle b also specifies the position of the observed vehicle A from the position of the observation vehicle b and the distance (and angle) between the observation vehicle b. If the two positions are within a predetermined distance, they can be identified as the same observed vehicle.
(Second identification method for observed vehicles)
For example, the image recognition result of the video data of the observed vehicle viewed from the observation vehicle can be used.
Specifically, vehicle characteristics such as a license plate, a shape, and a color are recognized from image data captured by a vehicle to be observed, and stored in association with travel data. The traveling data having the same vehicle characteristics can be specified as the same observed vehicle.

According to FIG. 8, it is divided into the following time zones according to the elapsed time.
Time zone t1: travel data for the observed vehicle A viewed from the observation vehicle a Time zone t2: travel data for the observed vehicle A viewed from both the observation vehicles a and b time zone t3: observed from the observation vehicle b Travel Data for Vehicle A In the time zones t1 and t3, since only travel data from one observation vehicle is obtained, the data integration unit 17 can output the travel data as it is.

(First integration method of driving data)
In the time zone t2, since the traveling data is obtained from the two observation vehicles a and b, either one of the traveling data may be output. In that case, it is preferable to select travel data obtained by the observation vehicle having a shorter (closer) distance to the observation vehicle. Further, as another embodiment, travel data with higher accuracy of the distance sensor (and camera) may be selected. In that case, it is also necessary to collect sensor accuracy information for each observation vehicle.
(The second integration method of driving data)
In time zone t2, travel data from two observation vehicles a and b may be integrated as a value. In this case, it is also preferable to calculate by an arbitrary calculation formula based on the amount of change in travel data.

  FIG. 9 is a functional configuration diagram for constructing a learning model in the driving characteristic estimation engine.

According to FIG. 9, when the traveling data satisfies a predetermined condition based on predetermined driving characteristics, the learning data selecting unit 18 further selects a plurality of traveling data as teacher data in a time series including the traveling data. The predetermined condition means, for example, a threshold value of travel data (an angle, an angular velocity, etc., as well as a speed and a distance) that is set in advance to determine a dangerous driving.
For example, when the relationship between the speed and the distance satisfies a predetermined condition (for example, acceleration is equal to or greater than a predetermined threshold), the learning data selection unit 18 selects such traveling data as, for example, dangerous driving. Here, not only the traveling data satisfying the predetermined condition but also a plurality of traveling data in a time series range including the traveling data are selected. That is, it is preferable that the traveling data of the front stage where the traveling data is detected and the traveling data of the rear stage are also included.
In addition, it is also preferable that such predetermined conditions differ according to the road information and traffic information obtained from CAN. This is because, for example, the type of dangerous driving is assumed to be different depending on road conditions.

  On the other hand, the feature amount data generation unit 11 generates feature amount data having speed and distance as elements by dividing the selected travel data into predetermined time ranges (windows). These feature quantity data can be, for example, feature quantity data based on dangerous operation.

  The driving characteristic estimation engine 12 inputs a plurality of feature amount data in different predetermined time ranges (windows) as teacher data based on the predetermined driving characteristics, so that the likelihood of the predetermined driving characteristics in the driver of the observed vehicle. A learning model is constructed to output. Thereby, a learning model can be constructed in actual operation.

  As described above in detail, according to the program, apparatus, and method of the present invention, it is impossible to obtain the driving signal and vehicle signal of the observed vehicle traveling around the observation vehicle as viewed from the observation vehicle. Also, it is possible to estimate the driver's psychological characteristics of the observed vehicle.

  For the various embodiments of the present invention described above, various modifications, corrections and omissions of the scope of the technical idea and aspect of the present invention can be easily made by those skilled in the art. The above description is merely an example and is not intended to be limiting in any way. The present invention is limited only as defined in the following claims and the equivalents thereto.

DESCRIPTION OF SYMBOLS 1 Driving characteristic estimation apparatus, portable terminal 10 Traveling data storage part 11 Feature-value data generation part 12 Driving characteristic estimation engine 130 Video data storage part 131 Video change detection part 140 Sound data storage part 141 Sound change detection part 150 Vehicle information storage part 151 Vehicle information change detection unit 16 Alarm display unit 17 Data integration unit 18 Learning data selection unit

Claims (15)

A program for causing a computer mounted on an apparatus for estimating a driver's driving characteristic of an observed vehicle traveling around the observation vehicle to function.
Travel data storage means that stores in time series the speed of the observation vehicle and the distance between the observation vehicle and the observed vehicle as travel data;
Feature quantity data generating means for generating feature quantity data with speed and distance in the travel data as elements;
A program that causes a computer to function as a driving characteristic estimation engine that inputs a plurality of feature amount data and outputs likelihood of a predetermined driving characteristic of a driver of an observed vehicle. The computer may be such that the predetermined driving characteristic of the driving characteristic estimation engine is a dangerous driving characteristic, emotional instability, noncoordination, rapid speed change, rapid startability, rapid stopness, meandering, or a combination thereof. The program according to claim 1, characterized in that The feature data is based on a predetermined time range,
The driving characteristic estimation engine inputs a plurality of feature amount data in different predetermined time ranges,
As the temporal change in the distance between the vehicle to be observed and the temporal change in the speed of the observation vehicle is small, the driving characteristic estimation engine, which has input the feature amount data, determines the predetermined value of the vehicle to be observed. The program according to claim 1 or 2, wherein the computer is made to function to estimate so as to increase the likelihood of driving characteristics. Regarding the travel data storage means,
The speed of the observation vehicle is measured by the temporal displacement of the position output from the positioning sensor installed on the observation vehicle,
4. The computer according to claim 1, wherein the computer is caused to function so that the distance to the observed vehicle is measured by a distance sensor installed in the observed vehicle. 5. program. The travel data storage means further stores the angular velocity with respect to the observation vehicle, and further stores the angle with respect to the observation vehicle.
The program according to any one of claims 1 to 4, wherein the feature amount data of the feature amount data generation unit causes a computer to further include the angular velocity and the angle as elements. A video data storage unit storing video data of the observed vehicle seen from the observed vehicle;
Causing the computer to function as a video change detection means for detecting a video change of the vehicle under observation using the video data storage means;
The program according to any one of claims 1 to 5, wherein the feature amount data of the feature amount data generation unit causes a computer to function so as to further include the video change as an element. The program according to claim 6, wherein the computer is caused to function so that the video change of the video change detection unit is a blinking change of a lamp of the observed vehicle. Sound data storage means storing sound data around the observation vehicle;
Causing the computer to function as a sound change detection unit that detects a sound change in a predetermined frequency band using the sound data storage unit;
The program according to any one of claims 1 to 7, wherein the feature amount data of the feature amount data generation means causes a computer to further include the sound change as an element. The device is a smartphone or a portable terminal, which can be installed in the observation vehicle.
The program according to any one of claims 1 to 8, wherein the program is installed as an application. A computer is further provided as an alarm display means for displaying the driving characteristics of the driver of the observed vehicle as an alarm when the likelihood output from the driving characteristics estimation engine exceeds a predetermined threshold. The program according to any one of claims 1 to 9, wherein the program is made to function. The travel data storage means stores, for each observation vehicle, the speed of the observation vehicle and the distance between the observation vehicle viewed from the observation vehicle in time series for each observation vehicle,
Using the travel data storage means, for each vehicle to be observed, further function as data integration means for integrating travel data from a plurality of observation vehicles on the time axis,
The program according to any one of claims 1 to 10, wherein the feature data generation unit causes a computer to function to generate feature data for each vehicle to be observed. When the traveling data satisfies a predetermined condition based on predetermined driving characteristics, further comprising learning data selecting means for selecting a plurality of traveling data as teacher data in a time series including the traveling data,
The feature amount data generating means generates feature amount data having the speed and the distance in the selected traveling data as elements;
The driving characteristic estimation engine constructs a learning model to output the likelihood of the predetermined driving characteristic in the driver of the observed vehicle by inputting a plurality of feature amount data as teacher data based on the predetermined driving characteristic. The program according to any one of claims 1 to 11, which causes the computer to function. A program for causing a computer mounted on a device for constructing a learning model for estimating a driver's driving characteristic of an observed vehicle traveling around the observation vehicle to function,
Travel data storage means for storing the speed of the vehicle and the distance between the observed vehicle as viewed from the observation vehicle as travel data in time series;
Learning data selection means for selecting a plurality of traveling data in time series including the traveling data when the traveling data satisfy a predetermined condition based on a predetermined driving characteristic;
Feature quantity data generating means for generating feature quantity data having elements of speed and distance in the selected traveling data;
The computer functions as a driving characteristic estimation engine that builds a learning model to output the likelihood of the predetermined driving characteristic of the driver of the observed vehicle by inputting a plurality of feature quantity data as teacher data based on the predetermined driving characteristic A program that features An apparatus for estimating driving characteristics of a driver of an observed vehicle traveling around the observation vehicle,
Travel data storage means that stores in time series the speed of the observation vehicle and the distance between the observation vehicle and the observed vehicle as travel data;
Feature quantity data generation means for generating feature quantity data having elements of speed and distance in the traveling data;
A driving characteristic estimation engine that inputs a plurality of feature amount data and outputs a likelihood of a predetermined driving characteristic for a driver of the observed vehicle. A method of estimating driving characteristics of a device for estimating driving characteristics of a driver of an observed vehicle traveling around the observation vehicle,
The apparatus includes a travel data storage unit that stores, as travel data, the speed of the observation vehicle and the distance between the observed vehicle viewed from the observation vehicle in time series,
The device
A first step of generating feature data having elements of speed and distance in the travel data;
A driving characteristic estimation method for an apparatus, comprising: inputting a plurality of characteristic amount data; and executing a second step of outputting a likelihood of a predetermined driving characteristic for a driver of the observed vehicle.

Images