Behavioral assessment system

Abstract

To provide a behavior evaluation system allowing for easy feedback to a driver with respect to safety confirmation behavior.SOLUTION: A behavior evaluation system 10 evaluates safety confirmation behavior of a driver of a vehicle. The behavior evaluation system 10 includes an acquisition unit 11, an observation data generation unit 12, and an evaluation unit 13. The acquisition unit 11 acquires a plurality of time-series data of measured values related to at least one of the driver condition and a vehicle condition. The observation data generation unit 12 calculates a plurality of types of feature quantities for each of the plurality of time-series data based on the time-series data, and generates observation-coordinate information including a plurality of types of feature quantities as components for each of the plurality of time-series data. The evaluation unit 13 calculates an evaluation score of the observed coordinates corresponding to the time-series data to be evaluated on the basis of a principal component axis calculated on the basis of variance of the observed coordinates corresponding to time-series data for learning.SELECTED DRAWING: Figure 5

Description

[Technical field]

[0001] The present invention relates to a behavior evaluation system for evaluating safety confirmation behavior of a driver who drives a vehicle. [Background technology]

[0002] In order to predict driving risks, it has been proposed to evaluate the driving behavior of a driver. In particular, it is known that evaluating the safety confirmation behavior of a driver is important in predicting driving risks. For example, Patent Document 1 discloses an automatic driving behavior evaluation system that evaluates driving behavior based on evaluation items including left and right confirmation actions defined as accident prevention actions. Patent Document 2 discloses a safety confirmation evaluation device that evaluates safety confirmation actions using a point-adding method based on evaluation conditions consisting of a combination of multiple evaluation items. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 5408572 [Patent Document 2] International Publication No. 2019 / 117004 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the driving behavior automatic evaluation system described in Patent Document 1 simply evaluates predefined physical quantities, and it is necessary to increase the number of evaluation items to improve the evaluation accuracy. Also, the safety confirmation evaluation device described in Patent Document 2 calculates the evaluation value using a point-addition method, and it is necessary to increase the number of evaluation branching conditions to improve the evaluation accuracy. However, there is a problem in that the more evaluation items and evaluation branching conditions are increased, the more complicated the feedback to the driver to encourage safer driving behavior becomes.

[0005] The present invention has been made to solve such problems, and has an object to provide a behavior evaluation system that can easily provide feedback to a driver regarding his or her safety confirmation behavior. [Means for solving the problem]

[0006] A behavior evaluation system according to one aspect of the present invention evaluates safety confirmation behavior of a vehicle driver. The behavior evaluation system includes an acquisition unit, an observation data generation unit, and an evaluation unit. The acquisition unit acquires a plurality of time series data of measurement values ​​related to at least one of a state of the driver and a state of the vehicle. The observation data generation unit calculates a plurality of types of feature amounts for each of the plurality of time series data based on the time series data, and generates information on observation coordinates including the plurality of types of feature amounts as components for each of the plurality of time series data. The evaluation unit calculates an evaluation score for the observation coordinates corresponding to the time series data to be evaluated based on a principal component axis calculated based on the variance of the observation coordinates corresponding to the learning time series data. Effect of the Invention

[0007] The present invention can provide a behavior evaluation system that can easily provide feedback to a driver regarding his or her safety confirmation behavior. [Brief description of the drawings]

[0008] [Figure 1] 1 is a schematic configuration diagram of a driving assistance system according to a first embodiment. [Diagram 2] 1 is a block diagram showing a configuration of a collection vehicle device according to a first embodiment. [Diagram 3] FIG. 2 is a diagram for explaining the head angle of a driver in the first embodiment. [Figure 4] 1 is a block diagram showing a configuration of a target vehicle device according to a first embodiment. [Diagram 5] FIG. 2 is a block diagram showing a configuration of a server according to the first embodiment. [Figure 6]FIG. 2 is a sequence diagram showing a processing procedure of the driving assistance system according to the first embodiment. [Figure 7] 11 is a flowchart showing a procedure of a learning process of the server according to the first embodiment. [Figure 8] FIG. 4 is a diagram showing time-series data of the head angle according to the first embodiment. [Figure 9] FIG. 4 is a diagram showing time series data of head velocity according to the first embodiment. [Figure 10] FIG. 4 is a diagram showing time-series data of vehicle speed according to the first embodiment. [Figure 11] FIG. 2 is a diagram showing principal component axes according to the first embodiment. [Figure 12] FIG. 2 is a diagram for explaining a first principal component axis according to the first embodiment. [Figure 13] FIG. 4 is a diagram for explaining a second principal component axis according to the first embodiment. [Figure 14] FIG. 4 is a diagram for explaining a third principal component axis according to the first embodiment. [Figure 15] 11 is a flowchart showing the procedure of evaluation processing of a server according to the first embodiment. [Figure 16] FIG. 11 is a diagram showing the correlation between the comprehensive evaluation score and the risk index according to the first embodiment. [Figure 17] FIG. 11 is a diagram for explaining a confirmation time shortage phase according to the second embodiment. [Figure 18] FIG. 11 is a block diagram showing a configuration of a target vehicle device according to a third embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] The present invention will be described below through embodiments, but the invention according to the claims is not limited to the following embodiments. In addition, not all of the configurations described in the embodiments are necessarily essential as means for solving the problems. For clarity of explanation, the following description and drawings are omitted and simplified as appropriate. In each drawing, the same elements are given the same reference numerals.

[0010] <Embodiment 1> First, a first embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a schematic diagram of a driving support system 1 according to a first embodiment. The driving support system 1 is a computer system that evaluates the safety confirmation behavior of a driver of a vehicle and supports driving by providing feedback on the safety confirmation behavior to the driver. In the first embodiment, the safety confirmation behavior refers to the driver moving his / her face from side to side and visually checking the situation around the vehicle. The driving support system 1 includes collection vehicle devices 20-1, 2, ... n, a target vehicle device 30, and a server 10, which are connected to each other via a network 4 so as to be able to communicate with each other. In the following, when the collection vehicle devices 20-1, 2, ... n are not distinguished from each other, the collection vehicle devices 20-1, 2, ... n are simply referred to as collection vehicle devices 20. In this figure, the number of collection vehicle devices 20 included in the driving support system 1 is three or more, but the number of collection vehicle devices 20 may be one or two.

[0011] The network 4 may be any of a variety of networks, such as the Internet, a wide area network (WAN), a local area network, or a dedicated line, or a combination of these.

[0012] The collection vehicle devices 20-1, 2, ...n are in-vehicle computer devices mounted on different collection vehicles 2-1, 2, ...n. The collection vehicle device 20 acquires information related to the driver's safety checking behavior through measurement. In the present embodiment 1, the information related to the safety checking behavior is information on measured values ​​related to the driver's condition and information on measured values ​​related to the vehicle condition. The collection vehicle device 20 then transmits the information related to the safety checking behavior to the server 10 as learning data for learning evaluation axes used to evaluate the safety checking behavior.

[0013] The target vehicle device 30 is an on-board computer device mounted on the target vehicle 3 driven by a driver who is to be evaluated. The target vehicle device 30 acquires information related to the driver's safety confirmation behavior through measurement. The target vehicle device 30 then transmits the information related to the safety confirmation behavior to the server 10 as evaluation target data. In response to receiving the evaluation result from the server 10, the target vehicle device 30 presents the evaluation result to the driver.

[0014] The server 10 is a computer device such as a server computer that evaluates the safety confirmation behavior of the driver of the target vehicle 3. The server 10 is also called a behavior evaluation system or a behavior evaluation device. The server 10 may be a cloud server built on a cloud. The server 10 learns an evaluation axis based on learning data, and evaluates the evaluation target data using the learned evaluation axis. The server 10 then transmits the evaluation result to the target vehicle device 30.

[0015] 2 is a block diagram showing the configuration of the collection vehicle device 20 according to the embodiment 1. The collection vehicle device 20 has a communication unit 21 and a measurement processing unit 22.

[0016] The communication unit 21 is connected to the vehicle-mounted camera 200 of the collection vehicle 2, and acquires image data of the driver, particularly the driver's head, from the vehicle-mounted camera 200. The vehicle-mounted camera 200 may be an infrared camera. The communication unit 21 then supplies the image data to the measurement processing unit 22. The communication unit 21 also acquires measurement values ​​related to the vehicle state from the ECU 210 (ECU = Electronic Control Unit) of the collection vehicle 2 via an in-vehicle communication bus such as a CAN (Controller Area Network). In the present embodiment 1, the measurement value related to the vehicle state is the measurement value of the vehicle's running speed (vehicle speed). However, without being limited thereto, the measurement value related to the vehicle state may indicate the accelerator opening, the steering angle, the operating state of the direction indicator, the position of the shift lever, etc. The communication unit 21 also transmits time-series data of the measurement values ​​generated by the measurement processing unit 22 to the server 10 via the network 4.

[0017] The measurement processing unit 22 calculates a measurement value related to the driver's state based on the image data, and generates time series data of the measurement value related to the driver's state. In the present embodiment 1, the measurement value related to the driver's state is the measurement value of the driver's head angle. The measurement processing unit 22 also generates time series data of the measurement value related to the vehicle state.

[0018] FIG. 3 is a diagram for explaining the head angle of the driver according to the first embodiment. FIG. 3 shows an image I of image data acquired by the collection vehicle device 20 from the vehicle-mounted camera 200. The image I captures the driver's head. The measurement processing unit 22 of the collection vehicle device 20 calculates the head angle of the driver by a known head posture estimation method. The head angle may be a yaw angle. For example, the measurement processing unit 22 detects feature points C of facial organs from the image I. The feature points C of the facial organs may include, for example, the left eye end point, the right eye end point, both cheeks, the right eyebrow end point, the left eyebrow end point, the left mouth end point, the right mouth end point, and the jaw center point. The measurement processing unit 22 calculates a rotation vector based on the position coordinates of each feature point, and calculates a yaw angle based on the rotation vector. In this way, the measurement processing unit 22 of the collection vehicle device 20 calculates the measurement value of the driver's head angle.

[0019] 4 is a block diagram showing the configuration of the target vehicle device 30 according to embodiment 1. The target vehicle device 30 has a communication unit 31, a measurement processing unit 32, and an output unit 33.

[0020] The communication unit 31 is connected to the vehicle-mounted camera 300 and ECU 310 of the target vehicle 3. The communication unit 31 also receives evaluation results from the server 10. The communication unit 31 and the measurement processing unit 32 basically perform the same processes as the communication unit 21 and the measurement processing unit 22 of the collection vehicle device 20, respectively. Therefore, in the explanation of Fig. 2, the collection vehicle device 20 will be replaced with the target vehicle device 30, and the collection vehicle 2 will be replaced with the target vehicle 3, and the explanation will be omitted.

[0021] The output unit 33 outputs the evaluation result received from the server 10 and notifies the driver of the evaluation result. The output unit 33 includes a display unit and an audio output unit . The display unit 34 is a display that displays the evaluation results. The audio output unit 35 is a speaker that outputs the evaluation result by voice.

[0022] 5 is a block diagram showing the configuration of the server 10 according to the embodiment 1. The server 10 includes an acquisition unit 11, an observation data generation unit 12, an evaluation unit 13, a learning processing unit 14, a transmission unit 15, and a storage unit 16.

[0023] The acquisition unit 11 acquires a plurality of pieces of time series data of measurement values ​​related to the driver's state and the vehicle state. The acquisition unit 11 supplies the acquired time series data to the observation data generation unit 12.

[0024] The observation data generating unit 12 calculates a plurality of types of feature amounts for each of the time series data based on the time series data. As an example, the feature amount includes at least one of the rotation speed of the driver's head (head speed), the driver's safety confirmation time (confirmation time), and the vehicle speed at the confirmation time. In the present embodiment 1, the feature amount is a value obtained by standardizing each of the head speed, confirmation time, and vehicle speed through a predetermined calculation process. In the following, for the sake of explanation, the number of types of feature amounts is set to three, but it may be four or more.

[0025] Then, the observation data generation unit 12 generates information on observation coordinates including multiple types of feature amounts as components for each time series data. In the present embodiment 1, the observation coordinates are three-dimensional coordinates having a standardized value of head velocity, a standardized value of confirmation time, and a standardized value of vehicle speed as components. The observation data generation unit 12 supplies the information on the observation coordinates to the evaluation unit 13 during evaluation processing, and supplies the information on the observation coordinates to the learning processing unit 14 during learning processing.

[0026] The evaluation unit 13 calculates an evaluation score of the observation coordinates corresponding to the time-series data to be evaluated based on the learned evaluation axis. The evaluation unit 13 supplies the calculated evaluation score to the transmission unit 15.

[0027] The learning processing unit 14 calculates the principal component axes based on the variance of the observation coordinates corresponding to the learning time-series data. Then, the learning processing unit 14 sets the calculated principal component axes as the evaluation axes, thereby learning the evaluation axes.

[0028] The transmission unit 15 transmits the evaluation result to the target vehicle device 30 .

[0029] The storage unit 16 is a storage medium that stores various information required for the evaluation process and the learning process.

[0030] FIG. 6 is a sequence diagram showing a procedure of the process of the driving assistance system 1 according to the first embodiment.

[0031] First, the communication unit 21 of the collection vehicle device 20 acquires image data of the driver's head from the vehicle-mounted camera 200 (step S10). Next, the measurement processing unit 22 of the collection vehicle device 20 calculates the head angle as the state of the driver (step S11). At this time, the measurement processing unit 22 may store the head angle information in a memory unit (not shown) of the collection vehicle device 20. Next, the communication unit 21 acquires vehicle speed information as vehicle state information from the ECU 210 (step S12). At this time, the communication unit 21 may store the vehicle speed information in a memory unit of the collection vehicle device 20. The measurement processing unit 22 generates time series data of the head angle and time series data of the vehicle speed. Then, the communication unit 21 transmits the time series data of these measurement values ​​to the server 10 as time series data for learning (step S13).

[0032] The server 10 executes a learning process (step S14) in response to receiving the time-series data from the plurality of collection vehicle devices 20. In this way, the server 10 learns the evaluation axis.

[0033] On the other hand, in steps S15 to 17, the target vehicle device 30 performs the same processes as in steps S10 to 12. Then, the communication unit 31 of the target vehicle device 30 transmits the time series data of the head angle and the time series data of the vehicle speed to the server 10 (step S18).

[0034] In response to receiving the time-series data from the target vehicle device 30, the server 10 executes an evaluation process using the learned evaluation axis (step S19). As a result, the server 10 calculates an evaluation score. The server 10 transmits the evaluation result to the communication unit 31 of the target vehicle device 30 (step S20). The evaluation result may be information simply indicating the evaluation score, or individual driving assistance information for the driver based on the evaluation score.

[0035] Then, the display unit 34 of the target vehicle device 30 displays the evaluation result, and the audio output unit 35 of the target vehicle device 30 outputs the evaluation result by audio (step S21).

[0036] Next, the learning process of the server 10 will be described using Fig. 7 with reference to Figs. 8 to 11. Fig. 7 is a flowchart showing the procedure of the learning process of the server 10 according to the first embodiment.

[0037] First, the acquisition unit 11 of the server 10 acquires time-series data for learning from the collection vehicle device 20 (step S30). Then, the observation data generation unit 12 identifies an event included in the time-series data of the head angle (step S31). Here, an event refers to one unit of a motion interval of the driver's safety confirmation behavior. In the present embodiment 1, an event refers to an interval from when the driver starts to turn his / her head left or right during safety confirmation, to when he / she stops turning his / her head and starts turning again.

[0038] FIG. 8 is a diagram showing time-series data of the head angle according to the first embodiment. In this figure, the solid line indicates the head angle, and the dashed line indicates the upper and lower thresholds for identifying an event. The observation data generating unit 12 identifies, as an event, a section from when the head angle transitions from the lower threshold to the upper threshold and then becomes less than the upper threshold, or a section from when the head angle transitions from the upper threshold to the lower threshold and then becomes equal to or greater than the lower threshold. Thus, an event is composed of a transition phase (dark colored portion) from when the head angle transitions from the lower (or upper) threshold to the upper (or lower) threshold, and a stop phase (light colored portion) from when the head angle transitions to when it next becomes less than the upper threshold (or equal to or greater than the lower threshold).

[0039] Returning to FIG. 7, the observation data generating unit 12 calculates the feature amounts, that is, the confirmation time, the head speed, and the vehicle speed, for each event (step S32). In the first embodiment, the observation data generating unit 12 regards the time of the stop phase for each event as the feature value of the confirmation time of the event.

[0040] The head velocity will be described with reference to Fig. 9. Fig. 9 is a diagram showing time series data of head velocity according to the first embodiment. The observation data generating unit 12 generates time series data of head velocity by differentiating the head angle with respect to time. For each event, the observation data generating unit 12 regards the average value of the head velocity in the transition phase as the feature amount of the head velocity of that event.

[0041] The vehicle speed will be described with reference to Fig. 10. Fig. 10 is a diagram showing time-series data of the vehicle speed according to the first embodiment. For each event, the observation data generating unit 12 sets the average value of the vehicle speed in the stop phase as the feature amount of the vehicle speed of the event.

[0042] When the head angle is used as a feature, the observation data generating unit 12 may use the average value of the head angle in the stop phase for each event as the feature of the head angle of that event.

[0043] 7, the observation data generating unit 12 calculates the feature amount of the time series data for each of the plurality of time series data based on the feature amount of the event, and generates the observation coordinates based on the feature amount of the time series data (step S33). Specifically, the observation data generating unit 12 calculates the feature amounts of the confirmation time, head speed, and vehicle speed of the time series data for each of the plurality of time series data as follows.

number

number

number

[0044] The observation data generating unit 12 generates an observation coordinate (Duration I ,Head speed I ,Vehicle speed I However, the observation data generating unit 12 may omit generating the observation coordinates and proceed to step S34.

[0045] When the observation data generating unit 12 uses the head angle as the feature amount, the feature amount of the time series data may be calculated as follows.

number

[0046] Then, the observation data generating unit 12 generates an observation value matrix Z including the observation coordinates of a plurality of time series data (step S34). Specifically, the observation data generating unit 12 generates an observation value column vector x 1 ,x 2 ,x 3 Generate.

number

[0047] The observed data generating unit 12 then converts each component of these observed value column vectors into a standardized value such that the average between the components is 0 and the variance is 1, as shown below.

number

number

number

[0048] Then, the observation data generation unit 12 generates an M×p observation matrix Z based on the observation column vector.

number

[0049] Next, the learning processing unit 14 calculates the principal component axis by principal component analysis (step S35). The principal component axis is expressed as a linear combination of unit vectors of the observation column vectors, that is, a linear combination of unit vectors determined corresponding to the type of feature. If the PCA coefficient vector, which is the coefficient vector of the linear combination, is a, the observation matrix projected onto the principal component axis is expressed as follows:

number

number

number

number

[0050] FIG. 11 is a diagram showing the principal component axes according to the first embodiment. As shown in the figure, the first principal component axis a (1) , the second principal component axis a (2) , and the third principal component axis a (3) are axes that maximize the variance when the observation coordinates are projected onto the respective axes, and are mutually orthogonal.

[0051] The number of principal component axes may be smaller than the number of types of feature quantities. For example, when the number of types of feature quantities is four or more and four or more eigenvectors are calculated, the learning processing unit 14 may select three eigenvectors. For example, the learning processing unit 14 may select the top three eigenvectors with the highest contribution rates. This allows complex high-dimensional information to be compressed into low-dimensional evaluation axes, so that the server 10 can easily perform evaluation even if the number of types of feature quantities increases, and can improve the evaluation accuracy.

[0052] 7, the learning processing unit 14 stores information about the calculated principal component axes in the storage unit 16 (step S36). The learning processing unit 14 ends the learning process.

[0053] FIG. 12 shows the first principal component axis a according to the first embodiment. (1) This is a diagram for explaining the first principal component axis a (1) The PCA coefficient a corresponding to the confirmation time, which is a component of 1 (1) , the PCA coefficient a corresponding to the confirmation speed 2 (1) and PCA coefficient a corresponding to vehicle speed 3 (1) For example, the PCA coefficient a 1 (1) is 0.46, and the PCA coefficient a 2 (1) is 0.49, and the PCA coefficient a 3 (1) is 0.75. In other words, the first principal component axis a (1) The PCA coefficients, which are the components of z, are all positive. 1 is large, z 2 is large, z 3 This shows that the axis is oriented in the direction in which the first principal component axis a is large. (1) is an axis that evaluates whether the confirmation time is sufficiently long, the confirmation speed is sufficiently fast, and the vehicle speed is sufficiently slow, and can be interpreted as an axis that indicates the consistency of safety confirmation behavior.

[0054] FIG. 13 shows the second principal component axis a according to the first embodiment. (2)This is a diagram for explaining the second principal component axis a. (2) The PCA coefficient a corresponding to the confirmation time, which is a component of 1 (2) , the PCA coefficient a corresponding to the confirmation speed 2 (2) and PCA coefficient a corresponding to vehicle speed 3 (2) For example, the PCA coefficient a 1 (2) is -0.72, and the PCA coefficient a 2 (2) is 0.69, and the PCA coefficient a 3 (2) is -0.01. Therefore, if we interpret it in the same way as in Figure 12, the second principal component axis a (2) is an axis that evaluates whether the confirmation time is short and the confirmation speed is fast, and can be interpreted as an axis that indicates the agility of safety confirmation actions.

[0055] FIG. 14 shows the third principal component axis a according to the first embodiment. (3) This figure shows the third principal component axis a (3) The PCA coefficient a corresponding to the confirmation time, which is a component of 1 (3) , the PCA coefficient a corresponding to the confirmation speed 2 (3) and PCA coefficient a corresponding to vehicle speed 3 (3) For example, the PCA coefficient a 1 (3) is -0.52, and the PCA coefficient a 2 (3) is -0.53, and the PCA coefficient a 3 (3) is 0.67. Here, the third principal component axis a (3) In the case of the third principal component axis a, the sign relationship of the PCA coefficients corresponding to the confirmation time and the vehicle speed is reversed. (3) is an axis that evaluates whether or not the driver is driving the vehicle when checking (i.e., performing the checking action without looking in the direction of travel), and can be interpreted as an axis that indicates the timeliness of safety checks.

[0056] In this way, by using an evaluation axis that compresses complex information derived from multiple feature quantities by principal component analysis, the server 10 can present more intuitive information to the driver as an evaluation result. Therefore, the server 10 can easily provide feedback to the driver to encourage safer driving behavior.

[0057] FIG. 15 is a flowchart showing the procedure of the evaluation process of the server 10 according to the first embodiment.

[0058] First, the acquisition unit 11 of the server 10 acquires time-series data for evaluation from the collection vehicle device 20 (step S40).

[0059] Next, in steps S41 to S44, the observation data generation unit 12 performs the same processes as steps S31 to S34 in Fig. 7 on the time series data for evaluation. As a result, the observation data generation unit 12 generates observation coordinates corresponding to the time series data for evaluation, and generates an observation value matrix Z'.

[0060] Next, the evaluation unit 13 acquires information on the principal component axes from the storage unit 16 (step S45). Next, the evaluation unit 13 calculates the evaluation score of the observation coordinates using the principal component axes as the learned evaluation axes (step S46). Specifically, the evaluation unit 13 calculates the observation value matrix Z′ and each principal component axis a (1)~(3) The first to third evaluation scores are calculated by taking the dot product with this.

number

number

[0061] Then, the evaluation unit 13 calculates a comprehensive evaluation score based on each evaluation score (step S47). For example, the evaluation unit 13 may determine the average value of the first to third evaluation scores as the comprehensive evaluation score, or may determine a weighted average value according to the contribution rate of the corresponding principal component axis as the comprehensive evaluation score.

[0062] FIG. 16 is a diagram showing the correlation between the overall evaluation score and the risk index according to the first embodiment. The horizontal axis of this diagram indicates the overall evaluation score, with the higher the value, the higher the evaluation. The vertical axis indicates the minimum TTC (Time to Collision), which is a risk index, with the higher the value, the higher the safety. Note that TTC is a value obtained by dividing the distance to an intersection of other vehicles that may cross the host vehicle's path by the relative speed to the intersection. As shown in this diagram, the overall evaluation score and the minimum TTC are determined by the coefficient of determination R 2 =0.49 positive correlation.

[0063] Furthermore, there was a significant correlation between the first evaluation score (contribution rate 42%) and the minimum TTC (R 2 =0.44, p-value <0.01). On the other hand, the correlation value of the second evaluation score (contribution rate 35%) with the minimum TTC was positive, and the p-value was “close to significant” (R 2 = 0.09, p value = 0.13), and the third evaluation score (contribution rate 23%) had a positive correlation with the minimum TTC, but the p value did not reach the significance level (R 2 = 0.04, p-value = 0.34). However, the contribution rate in all cases is not negligible, and it is clear that all principal component axes are meaningful in evaluating safety confirmation behavior.

[0064] As described above, according to the first embodiment, the server 10 performs evaluation using the principal component axis obtained based on the variance of the observation data as the evaluation axis, instead of the axis corresponding to each feature amount. This allows the server 10 to easily perform an intuitive and highly accurate evaluation. This makes it easier to provide feedback to the driver regarding the safety confirmation behavior.

[0065] The principal component axis is expressed in a simple form as a linear combination of unit vectors corresponding to each feature. In the first embodiment, the measurement values ​​on which the features are based relate to the rotational movement of the driver's head. The rotational movement of the head can be easily measured by imaging and image processing using an in-vehicle camera located away from the driver, so the driver does not need to wear a sensor or the like. In addition, the measurement of the rotational movement of the head is more robust than the measurement of the line of sight. Note that the use of the rotational movement of the head allows for earlier evaluation and risk prediction than the use of information related to driving operations.

[0066] It should be noted that the collection vehicle device 20 and the target vehicle device 30 may each have any sensor, such as an angular velocity sensor attached to the driver's head, in place of or in addition to the vehicle-mounted camera 200 and the vehicle-mounted camera 300, respectively.

[0067] In the first embodiment, the collection vehicle device 20 and the target vehicle device 30 transmit time series data of the measurement values ​​after the measurement process to the server 10. However, instead of this, the collection vehicle device 20 and the target vehicle device 30 may transmit the measurement values ​​after the measurement process to the server 10, and the server 10 may generate the time series data. Also, the collection vehicle device 20 and the target vehicle device 30 may transmit image data acquired from the vehicle-mounted camera 200 and the vehicle-mounted camera 300, respectively, to the server 10, and the server 10 may perform the measurement process and generate the time series data.

[0068] In the first embodiment, the server 10 acquires time-series data for learning from a plurality of collection vehicle devices 20. However, instead of this, the server 10 may acquire time-series data for learning from one collection vehicle device 20, or may acquire time-series data for learning from a target vehicle device 30.

[0069] In the first embodiment, the server 10 acquires time series data of measurement values ​​related to the driver's state and the vehicle state from the collection vehicle device 20 and the target vehicle device 30. However, instead of this, the server 10 may only acquire time series data of measurement values ​​of either the driver's state or the vehicle state from the collection vehicle device 20 and the target vehicle device 30.

[0070] In the first embodiment, the server 10 generates the observation coordinates for each time series data by using the average value of the feature amount for each event included in the time series data as the component of the observation coordinates. However, instead of this, the server 10 may generate the observation coordinates for each event by using the feature amount of the event as the component of the observation coordinates.

[0071] <Embodiment 2> Next, a second embodiment of the present invention will be described with reference to Fig. 17. The second embodiment is characterized in that the proportion of insufficient confirmation time is used as one of the feature quantities used for evaluation. The proportion of insufficient confirmation time indicates the proportion of events whose confirmation time is less than a threshold time among the events included in the time series data, and is expressed as follows:

number

[0072] FIG. 17 is a diagram for explaining the confirmation time shortage phase according to the second embodiment. The solid line indicates the head angle, the thick line indicates the average head angle in the stop phase on the vertical axis, and the horizontal axis indicates the length of the confirmation time. In this figure, th The phase where the confirmation time is insufficient is shown as a confirmation time insufficient phase.

[0073] By using the ratio of insufficient confirmation time as a feature, the server 10 can evaluate the safety confirmation behavior by taking into account information on the insufficient confirmation time that is lost in the feature of the above-mentioned confirmation time that is averaged over the duration of the event, thus improving the accuracy of the evaluation.

[0074] <Embodiment 3> Next, a third embodiment of the present invention will be described with reference to Fig. 18. The third embodiment is characterized in that a target vehicle device executes an evaluation process. 18 is a block diagram showing a configuration of a target vehicle device 30a according to embodiment 3. The target vehicle device 30a includes an observation data generating unit 36 ​​and an evaluating unit 37 in addition to the configuration of the target vehicle device 30.

[0075] The observation data generating unit 36 ​​uses the time-series data generated by the measurement processing unit 32 to execute processing similar to the evaluation processing performed by the observation data generating unit 12 of the server 10 according to the first embodiment (steps S41 to S44 in FIG. 15).

[0076] The evaluation unit 37 acquires the principal component axis information from the server 10 via the communication unit 31, and uses the principal component axis information to perform processing similar to the processing performed by the evaluation unit 13 of the server 10 in the first embodiment (steps S46 to S47 in FIG. 15).

[0077] As described above, according to the third embodiment, the target vehicle device 30a executes the evaluation process, so that the evaluation process can be performed in real time.

[0078] In the above embodiment, the present invention has been described as being configured as hardware, but the present invention is not limited to this. The present invention can also be realized by having a processor execute a computer program, for example, a behavior evaluation program, to execute various processes such as the learning process and evaluation process related to the behavior evaluation method.

[0079] In the above embodiment, the present invention has been described as being configured as hardware, but the present invention is not limited to this. Any process of the present invention can also be realized by causing a processor to execute a computer program.

[0080] In the above example, the program can be stored and provided to the computer using various types of non-transitory computer readable media. The non-transitory computer readable media includes various types of tangible storage media. Examples of the non-transitory computer readable media include magnetic recording media (e.g., flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (e.g., magneto-optical disks), CD-ROM (Read Only Memory), CD-R, CD-R / W, DVD (Digital Versatile Disc), BD (Blu-ray (registered trademark) Disc), and semiconductor memory (e.g., mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (Random Access Memory)). The program may also be provided to the computer by various types of transitory computer readable media. Examples of the transitory computer readable media include electric signals, optical signals, and electromagnetic waves. The temporary computer-readable medium can supply the program to the computer via a wired communication path such as an electric wire or an optical fiber, or via a wireless communication path.

[0081] In the above-mentioned embodiment, the computer is configured as a computer system including a personal computer, a word processor, etc. However, the present invention is not limited to this, and the computer can also be configured as a server of a LAN (local area network), a host for computer (personal computer) communication, a computer system connected to the Internet, etc. It is also possible to distribute functions to each device on the network and configure a computer as a whole network.

[0082] The order of execution of each process in the apparatus and method shown in the claims, specification, and drawings is not specifically stated as "before," "prior to," etc., and may be realized in any order as long as the output of a previous process is not used in a subsequent process. Even if the operational flow in the claims, specification, and drawings is explained using "first," "next," etc. for convenience, it does not mean that it is essential to perform the process in this order.

[0083] The present invention is not limited to the above-described embodiment, and can be modified as appropriate without departing from the spirit and scope of the present invention. In addition, a part or all of the above-described embodiment can be described as the following supplementary notes, but is not limited to the following. (Appendix 1) A behavior evaluation system for evaluating a safety confirmation behavior of a vehicle driver, comprising: an acquisition unit that acquires a plurality of time series data of measurement values ​​related to at least one of a state of the driver and a state of the vehicle; an observation data generating unit that calculates a plurality of types of feature amounts based on each of the plurality of time series data, and generates, for each of the plurality of time series data, information on observation coordinates including the plurality of types of feature amounts as components; an evaluation unit that calculates an evaluation score of the observation coordinates corresponding to the time-series data to be evaluated based on a principal component axis calculated based on the variance of the observation coordinates corresponding to the time-series data for learning; A behavioral evaluation system comprising: (Appendix 2) 2. The behavior evaluation system according to claim 1, wherein the number of principal component axes is smaller than the number of types of the features. (Appendix 3) 3. The behavior evaluation system according to claim 1, wherein the principal component axis is expressed by a linear combination of unit vectors determined corresponding to the types of the feature amounts. (Appendix 4) 4. The behavior evaluation system according to claim 1, wherein the measured values ​​are a measured value of a rotation angle of the driver's head and a measured value of a traveling speed of the vehicle. (Appendix 5) 5. The behavior evaluation system according to claim 1, wherein the plurality of types of features include at least one of the driver's head rotation speed, the driver's safety confirmation time, and the vehicle's traveling speed during the safety confirmation time. (Appendix 6) 6. The behavior evaluation system according to claim 1, wherein the plurality of types of features include a proportion of the driver's insufficient safety confirmation time. (Appendix 7) A behavior evaluation method for evaluating a safety confirmation behavior of a vehicle driver, comprising: acquiring a plurality of time series data of measurement values ​​related to at least one of a state of the driver and a state of the vehicle; an observation data generating step of calculating a plurality of types of feature amounts based on each of the plurality of time series data, and generating, for each of the plurality of time series data, information on observation coordinates including the plurality of types of feature amounts as components; an evaluation step of calculating an evaluation score of the observation coordinates corresponding to the time series data to be evaluated based on the principal component axis calculated based on the variance of the observation coordinates corresponding to the learning time series data; A behavioral evaluation method comprising: (Appendix 8) A behavioral evaluation program for evaluating a safety confirmation behavior of a vehicle driver, comprising: an acquisition process of acquiring a plurality of time series data of measurement values ​​related to at least one of a state of the driver and a state of the vehicle; an observation data generation process for calculating a plurality of types of feature amounts based on each of the plurality of time series data, and generating, for each of the plurality of time series data, information on observation coordinates including the plurality of types of feature amounts as components; an evaluation process of calculating an evaluation score of the observation coordinates corresponding to the time series data to be evaluated based on a principal component axis calculated based on the variance of the observation coordinates corresponding to the learning time series data; A behavioral evaluation program for running the above on a computer. [Explanation of symbols]

[0084] 1. Driving assistance system 2 Collection vehicles 3. Target vehicle 4. Network 10 Server (behavioral evaluation system, behavioral evaluation device) 11 Acquisition Department 12 Observation data generation section 13 Evaluation Section 14 Learning processing unit 15 Transmitter 16 Memory section 20 Collection vehicle equipment 30, 30a Target vehicle device 21,31 Communications Department 22,32 Measurement processing section 33 Output section 34 Display section 35 Audio output section 36 Observation Data Generation Unit 37 Evaluation Section 200,300 In-vehicle camera 210,310 ECU

Claims

1. A behavior evaluation system for evaluating a safety confirmation behavior of a vehicle driver, comprising: an acquisition unit that acquires a plurality of time series data of measurement values ​​related to a state of the driver and a state of the vehicle as learning data; an observation data generation unit that calculates a plurality of types of feature amounts based on each of the plurality of time series data, and generates, for each of the plurality of time series data, an observation coordinate including the plurality of types of feature amounts as components; an evaluation unit that calculates an evaluation score of the observation coordinate corresponding to the time-series data of the measurement value for learning, based on a principal component axis calculated so as to maximize the variance of the observation coordinate corresponding to the time-series data of the measurement value for learning acquired by the acquisition unit; Equipped with the number of the principal component axes is smaller than the number of types of the feature quantities; The plurality of types of feature amounts include a head rotation speed of the driver, a safety confirmation time of the driver, a traveling speed of the vehicle during the safety confirmation time, and a ratio of a safety confirmation time shortage of the driver. Behavioral assessment system.

2. The behavior evaluation system according to claim 1 , wherein the principal component axis is represented by a linear combination of unit vectors determined corresponding to the types of the feature quantities.

3. 2. The behavior evaluation system according to claim 1, wherein the measured values ​​are a measurement of a rotation angle of the driver's head and a measurement of a traveling speed of the vehicle.

Images