JP2016042298A - Accident information calculation apparatus and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to calculate the degree of occurrence of accidents using an obtainable amount of statistics.SOLUTION: A statistical-information acquisition section 30 acquires driving information and accident statistical information. A calculation section 32 calculates a parameter used in a predefined function for approximating an accident occurrence probability for a value of each feature quantity, on the basis of a constraint condition in relation to the probability that the value of each feature quantity is obtained under an accident occurrence condition as well as the driving information and the accident statistical information, and calculates a function using the calculated parameter. A feature-quantity acquisition section 28 acquires the value of each of the plurality of feature quantities related to the driving behavior of a vehicle driven by a driver of interest. A driving-behavior evaluation section 34 evaluates, as an accident-free probability, the product of the accident-free probabilities which are calculated to correspond to a plurality of types of accidents, respectively for the driving behavior of the vehicle driven by the driver of interest on the basis of the acquired value of each of the plurality of feature quantities and the function calculated for each of the plurality of types of accidents calculated by the calculation section 32.SELECTED DRAWING: Figure 2

Description

  The present invention relates to an accident information calculation apparatus and program, and more particularly, to an accident information calculation apparatus and program for calculating the degree of occurrence of an accident.

  Conventionally, in order to diagnose driving considering factors that are likely to lead to traffic accidents, there is a known method for obtaining traffic rules on the road and diagnosing driver driving characteristics using information on the rules and driving behavior. (Patent Document 1).

  In addition, the driver's past driving history (vehicle speed, brake and accelerator operation frequency, inter-vehicle distance, lane switching frequency, overtaking frequency, horn ringing frequency, etc.), driving status (weather, time zone, temperature, There is a method to estimate the degree of risk based on humidity, distinction between curve driving and straight driving, road type such as highway or ordinary road, etc.) and physical condition information of the driver (attention, body temperature, blood pressure, etc.) Known (Patent Document 2).

  Also, a method for diagnosing driving risk based on the degree of deviation between the driving behavior data serving as a reference and the driving behavior data of the host vehicle in each driving situation is known (Patent Document 3).

  Also, a method of calculating the probability of occurrence of an accident with respect to the driving state of the vehicle by a Bayesian network using past accident statistical information and an accident factor probability table for the occurrence of each accident is known (Patent Document 4). .

  In addition, by using the statistical driving information obtained from the driving operation information of a plurality of drivers including the evaluation subject and the driving information of the evaluation subject, by performing a relative evaluation with other drivers, A method for diagnosing driving is known (Patent Document 5).

JP 2006-209455 A JP 2006-277058 A JP 2006-343904 A JP 2009-282873 A JP 2012-247854 A

  However, the method of Patent Document 4 has difficulty in calculating the probability distribution of accident factors for accidents that have occurred in the past. For example, observable information such as the weather at the time of a traffic accident can be probability-distributed, but the driving behavior immediately before the occurrence of the accident is difficult to observe although it may correspond to the accident factor. Furthermore, in the case of continuous values such as acceleration and aside time rather than discrete information such as the presence or absence of excess speed, a large amount of case data is required, which makes it more difficult to obtain a probability distribution.

  In the present invention, there is provided an accident information calculation device and a program which are made to solve the above-described problems and which can calculate the degree of occurrence of an accident using an acquirable statistic. With the goal.

  In order to achieve the above object, the accident information calculation apparatus according to the present invention provides driving information indicating the degree to which each value of the characteristic amount related to the driving behavior of the vehicle is obtained, and the degree of occurrence of an accident caused by the characteristic amount. Statistical information acquisition means for acquiring accident statistical information representing, the driving information and the accident statistical information acquired by the statistical information acquisition means, and each value of the feature value under the condition where the accident occurred Calculation means for calculating the degree of occurrence of the accident for each value of the feature amount based on the constraint condition relating to the degree to be obtained.

  In addition, the program of the present invention, the computer, driving information representing the degree to which each value of the feature value related to the driving behavior of the vehicle is obtained, accident statistics information representing the degree of occurrence of the accident caused by the feature value, A statistical information acquiring means for acquiring the driving information and the accident statistical information acquired by the statistical information acquiring means, and a constraint on the degree to which each value of the feature value is obtained under the condition where the accident occurred And a program for functioning as a calculation means for calculating the degree of occurrence of the accident for each value of the feature amount.

  According to the present invention, the driving information representing the degree to which each value of the characteristic amount related to the driving behavior of the vehicle is obtained by the statistical information acquisition means, and the accident statistical information representing the degree of occurrence of an accident caused by the characteristic amount, To get.

  Then, based on the driving information and accident statistical information acquired by the statistical information acquisition unit and the constraint condition regarding the degree of each characteristic value obtained under the condition where the accident occurred, the calculation unit calculates the feature amount. The degree of accident occurrence is calculated for each value.

  As described above, the degree of occurrence of an accident for each value of the feature amount based on the driving information and the accident statistical information, and the constraint condition regarding the degree to which each value of the feature amount is obtained under the condition where the accident occurred. By calculating, the degree of occurrence of an accident can be calculated using the statistics that can be acquired.

  Further, the calculation means according to the present invention includes a restriction condition regarding the driving information and the accident statistical information acquired by the statistical information acquisition means, and a degree to which each value of the feature value under the condition where the accident occurs is obtained. Based on the above, a parameter used in a predetermined function approximating the degree of occurrence of the accident for each value of the feature value is calculated, and the function using the calculated parameter is calculated as the feature value. It is possible to calculate the degree of occurrence of the accident for each value.

  Further, the calculation means includes the driving information and the accident statistical information acquired by the statistical information acquisition means, and a constraint condition regarding a degree to which each value of the feature value is obtained under the condition where the accident occurs. Based on this, the parameter can be calculated using Bayes' theorem.

  Further, the present invention provides a feature amount acquisition unit that acquires a feature amount value related to the driving behavior of the vehicle of the target driver, the feature amount value acquired by the feature amount acquisition unit, and the calculation unit. Driving behavior evaluation means for evaluating the degree of occurrence of the accident or the degree of no occurrence of the accident with respect to the driving behavior of the vehicle of the target driver based on the degree of occurrence of the accident for each value of the feature amount; Can further be included.

  Further, the statistical information acquisition means includes the driving information representing a probability that each value of a feature amount related to driving behavior of a vehicle is obtained, and the accident statistical information representing a probability that an accident caused by the feature amount occurs. And the calculation means is the constraint on the driving information and the accident statistical information acquired by the statistical information acquisition means, and the probability that each value of the feature value is obtained under the condition where the accident occurred. A parameter used in a predetermined function approximating the probability of occurrence of the accident with respect to each value of the feature amount based on a condition, and calculating the function using the calculated parameter Can be. Thereby, the probability that an accident will occur can be calculated using the statistics that can be acquired.

  Further, the statistical information acquisition means includes the driving information indicating the probability that each value of the multi-dimensional feature value related to the driving behavior of the vehicle is obtained, and an accident caused by the multi-dimensional feature value. The accident statistical information representing the probability of occurrence is acquired, and the calculation means is the driving information and accident statistical information acquired by the statistical information acquisition means, and the dimensions of each of the dimensions of the plurality of dimensions. Based on the constraint condition regarding the probability that each value of the dimension is obtained under the condition where the accident occurred, a predetermined value approximating the probability that the accident will occur for each value of each dimension of the feature quantity is determined in advance. The parameter used in the function can be calculated, and the function using the calculated parameter can be calculated. Thereby, the probability that an accident will occur can be calculated using the statistics that can be acquired.

  Further, the statistical information acquisition means includes the driving information indicating the probability of obtaining each value for a plurality of feature values related to driving behavior of the vehicle, and a plurality of types of accidents caused by each of the plurality of feature values. The accident statistical information representing the probability of each occurrence, the calculation means, for each of the plurality of types of accidents, the driving information and accident statistical information acquired by the statistical information acquisition means, Based on the constraint condition regarding the probability that each value of the feature value corresponding to the accident under the condition where the accident occurred is obtained, the probability that the accident will occur is approximated for each value of the feature value A parameter used in a predetermined function can be calculated, and the function using the calculated parameter can be calculated. Thereby, the probability that an accident will occur can be calculated using the statistics that can be acquired.

  Further, the present invention provides a feature amount acquisition unit that acquires each value of the plurality of feature amounts related to the driving behavior of the vehicle of the target driver, and each value of the plurality of feature amounts acquired by the feature amount acquisition unit. And a probability that the accident does not occur for each of the plurality of types of accidents based on the function calculated for each of the plurality of types of accidents by the calculation means, and the driving behavior of the vehicle of the target driver And a driving behavior evaluation means for evaluating a product of the probability that the accident does not occur calculated for each of the plurality of types of accidents as an accident-free probability representing a probability that the accident does not occur. . Thus, the driving behavior of the driver can be evaluated using the statistics that can be acquired.

  In addition, the feature amount acquisition unit acquires the value of the feature amount related to the driving behavior of the vehicle of the target driver and the duration of the value of the feature amount, and the driving behavior evaluation unit acquires the feature amount acquisition unit. The accident occurs for each value of the feature amount by a predetermined duration calculated for each of the plurality of types of accidents by the calculating means and the duration value of the feature amount. On the basis of the function approximating the probability of performing, the product of the probability that the accident does not occur calculated for each of the plurality of types of accidents is evaluated as the accident-free probability for the driving behavior of the target driver's vehicle To be able to. As a result, it is possible to calculate the accident-free probability for the feature quantity that changes from moment to moment.

  Further, the feature amount acquisition means acquires the value of the feature amount related to the driving behavior of the vehicle of the target driver, and the travel distance when the value of the feature amount is obtained, and the driving behavior evaluation means The value of the feature amount and the travel distance acquired by the feature amount acquisition means, and each value of the feature amount by the predetermined travel distance calculated for each of the plurality of types of accidents by the calculation means. On the basis of the function that approximates the probability that the accident will occur, the product of the probability that the accident will not occur calculated for each of the plurality of types of accidents for the driving behavior of the target driver's vehicle. , And can be evaluated as the accident-free probability. As a result, it is possible to calculate the accident-free probability for the feature quantity that changes from moment to moment.

  Further, the function includes a predetermined value of the feature amount when the value of the function is maximum, or a predetermined value of the feature amount when the value of the function is minimum. can do.

  In addition, the calculation unit may include the driving information and the accident statistical information acquired by the statistical information acquisition unit, and each dimension of the dimensions under the condition where the accident has occurred for each dimension of the multi-dimensional feature amount. Based on the constraint on the probability of obtaining a value, the parameter is calculated using Bayes' theorem and assuming that the dimensions of the feature quantities of the plurality of dimensions are independent or dependent on each other. be able to.

  The function may be a Gaussian function that approximates the degree of occurrence of the accident for each value of the feature amount.

  The function may be a linear function that approximates the degree of occurrence of the accident with respect to each value of the feature amount.

  The function may be a sigmoid function that approximates the degree of occurrence of the accident for each value of the feature amount.

  As described above, according to the accident information calculation apparatus and program of the present invention, the driving information and the accident statistical information, and the constraint condition relating to the degree to which each value of the feature value under the condition where the accident occurred can be obtained. On the basis of this, by calculating the degree of occurrence of an accident for each value of the feature amount, an effect that the degree of occurrence of the accident can be calculated by using an acquirable statistic.

It is a figure which shows the concept which calculates an accident-free probability. It is a block diagram which shows the structure of the accident information calculation apparatus of embodiment of this invention. It is a figure which shows the concept which calculates a statistic applying Bayes' theorem. It is a flowchart which shows the content of the parameter calculation process routine in the accident information calculation apparatus of embodiment of this invention. It is a flowchart which shows the content of the driving | operation diagnosis processing routine in the accident information calculation apparatus of embodiment of this invention.

<Overview>
In order to maintain and improve the driving ability of the driver, continuous and appropriate measures are very important. However, at present, an evaluation index for diagnosing the driver's driving behavior has not been established. In the embodiment of the present invention, “accident-free probability (probability of not causing an accident)” is defined as one of evaluation indexes for diagnosing the driving behavior of the driver, and a method for calculating the accident-free probability will be described.

  FIG. 1 shows a conceptual diagram of an embodiment of the present invention. As shown in FIG. 1, in the embodiment of the present invention, an accident-free probability representing the probability that an accident will not occur is calculated based on the probability that an accident will occur with respect to the acquired feature value.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the embodiment of the present invention, the probability that an accident will occur is calculated for the value of the feature quantity that is mounted on the vehicle and acquired, and the probability that no accident will occur is calculated based on the calculated probability that the accident will occur. A case where the present invention is applied to an accident information calculation apparatus that calculates an accident-free probability is described as an example.

[First Embodiment]
<Configuration of accident information calculation device>
An accident information calculation apparatus according to the first embodiment of the present invention will be described. As shown in FIG. 2, the accident information calculation apparatus 10 according to the embodiment of the present invention includes a lateral acceleration sensor 12 that detects lateral acceleration of the vehicle, and a longitudinal acceleration sensor 14 that detects longitudinal acceleration of the vehicle. A vehicle speed sensor 16 that detects the speed of the vehicle, an inter-vehicle distance sensor 18 that detects an inter-vehicle distance between the host vehicle and a preceding vehicle, and sequentially images the front of the host vehicle including the road area and outputs a front image. The front camera 20, a driver camera 22 that sequentially captures and outputs a face image including the driver's eye area in the vehicle interior, a statistical information database 24 that stores accident statistical information and driving information, and an accident-free probability are calculated. A computer 26 and an output device 40 that outputs the accident-free probability calculated by the computer 26 are provided.

  The inter-vehicle distance sensor 18 detects the relative distance between the host vehicle and the preceding vehicle as the inter-vehicle distance and outputs it to the computer 26. As the inter-vehicle distance sensor 18, for example, a laser radar, a millimeter wave radar, or the like can be used.

  The front camera 20 is mounted on the vehicle, images the front of the host vehicle including the road area, and outputs an image. The front camera 20 images the front of the host vehicle and generates an image signal of the image (not shown), and an A / D converter (not shown) that A / D converts the image signal generated by the imaging unit. And an image memory (not shown) for temporarily storing the A / D converted image signal.

  The driver camera 22 captures and outputs a driver's face image. The driver camera 22 images a driver's face and generates an image signal of the image (not shown), and an A / D conversion unit (not shown) that A / D converts the image signal generated by the imaging unit. And an image memory (not shown) for temporarily storing the A / D converted image signal.

The statistical information database 24 is generated due to driving information representing the probability p (A _i ) that each value of the plurality of feature values A _i relating to the driving behavior of the vehicle is obtained, and each of the plurality of feature values A _i. Accident statistical information indicating the probability p (X _i ) of occurrence of each of a plurality of types of accidents X _i is stored.

Here, A _i represents a feature amount (speed, acceleration, inter-vehicle distance, side-viewing time, etc.) related to the driving behavior of the driver. X _i represents an accident (slip accident, rear-end collision, side-view accident, etc.) caused by the value of the feature value A _i . Further, p (X _i ) represents the probability that the accident X _i will occur, and p (A _i ) represents the probability that the value of the feature amount A _i will be obtained.

The accident statistical information is normalized as an accident rate per arbitrary amount, such as “accident rate per l [km] travel”, “accident rate per t [sec] travel”, and stored in the statistical information database 24. Is done. For example, in the case of an accident X _i is rear-end collisions, with the number N _i of 1 year rear-end collisions national _average mileage of 1 year nationwide L [miles], the information of car ownership number C nationwide, l [ km] The probability p (X _i ) of occurrence of a rear-end collision per driving is calculated in advance as in the following equation (1) and stored in the statistical information database 24. The probability p (X _i ) is calculated for each of a plurality of types of accidents.

In addition, the driving information probability p (A _i ) is preferably calculated in the same population as when the accident statistical information is calculated. For example, when the accident statistical information is obtained from the number of accidents in the whole country, the probability p (A _i ) of the driving information must be a distribution obtained when the traveling information of the vehicles in the whole country is statistics. Currently, it is difficult to obtain driving information, so it may actually be necessary to calculate from the driving data of a small number of drivers, but the driving data of a large number of drivers will be accumulated, by obtaining the probability p (a _i) of the operation information using the probability p (a _i) of the operation information is considered to gradually approaches the average distribution across the country.

  The computer 26 includes a CPU that controls the entire accident information calculation apparatus 10, a ROM as a storage medium that stores a program of a driving diagnosis processing routine that will be described later, a RAM that temporarily stores data as a work area, and a bus that connects these It is comprised including. In the case of such a configuration, a program for realizing the function of each component is stored in a storage medium such as a ROM or HDD, and each function is realized by executing the program by the CPU. To.

  When the computer 26 is described with function blocks divided for each function realizing means determined based on hardware and software, as shown in FIG. 2, the lateral acceleration of the vehicle detected by the lateral acceleration sensor 12, the longitudinal The longitudinal acceleration of the vehicle detected by the directional acceleration sensor 14, the vehicle speed detected by the vehicle speed sensor 16, the inter-vehicle distance between the host vehicle and the preceding vehicle detected by the inter-vehicle distance sensor 18, and imaged by the front camera 20. A feature amount acquisition unit 28 that acquires a plurality of feature amounts based on the forward image that has been captured and the driver's face image captured by the driver camera 22; accident statistical information and driving information stored in the statistical information database 24; A plurality of feature amounts acquired by the statistical information acquisition unit 30 and the feature amount acquisition unit 28; Based on accident statistics information and driving information acquired by the meter information acquisition unit 30, a calculation unit 32 that calculates parameters used in a predetermined function that approximates the probability of occurrence of an accident, and a calculation unit 32 And a driving behavior evaluation unit 34 that evaluates the driving of the target driver based on a function using the parameters.

  The feature amount acquisition unit 28 acquires each value of a plurality of feature amounts related to the driving behavior of the vehicle of the driver that is the target of the driving diagnosis.

  For example, the feature amount acquisition unit 28 detects a stop sign by image processing from the front image captured by the front camera 20, and detects the vehicle speed detected by the vehicle speed sensor 16 when the stop sign is detected. , Get as the passing speed at the time of pause.

  In addition, the feature amount acquisition unit 28 detects the driver's face by, for example, pattern matching processing based on the driver's face image captured by the driver camera 22. Next, the feature amount acquisition unit 28 detects the face direction of the driver by pattern matching processing learned based on a plurality of face images having different face directions based on the detected face of the driver. Then, the feature amount acquisition unit 28 calculates the time during which the aside operation is continued based on the driver's face direction, and the vehicle speed sensor 16 when the aside operation is continued. And the vehicle speed detected by. And the feature-value acquisition part 28 acquires the armpit driving distance which the armpit driving continued based on the acquired time and the vehicle speed.

  The feature amount acquisition unit 28 detects a red signal of a signal sign from the front image captured by the front camera 20 by image processing, and acquires the passing time of the vehicle at the intersection when the red signal is detected.

  The feature amount acquisition unit 28 then detects the lateral acceleration detected by the lateral acceleration sensor 12, the longitudinal acceleration detected by the longitudinal acceleration sensor 14, the inter-vehicle distance detected by the inter-vehicle distance sensor 18, The passing speed at the time of stoppage, the aside driving distance, and the intersection passing time at the time of a red signal are acquired as a plurality of feature amounts related to the driving behavior of the driver.

  The statistical information acquisition unit 30 acquires accident statistical information and driving information stored in the statistical information database 24.

  In the following, first, the definition of the accident-free probability calculated in the present embodiment and the formulation for estimating the accident-free probability will be described.

  In order to estimate the accident-free probability (probability that no accident will occur) from the driving behavior (driving data) of the driver of the vehicle, the acquisition of the feature value resulting from the accident and the formulation using the acquired feature value are issues. However, unless a true value of accident-free probability can be obtained, it is generally difficult to formulate by regression analysis or the like. Therefore, in the embodiment of the present invention, attention is focused on accident statistical information based on the actual number of accidents and statistical information such as driving information representing the distribution of driving data. In the embodiment of the present invention, the following approximations (A) and (B) are used.

(A) The probabilities that all accidents X _i occur are independent of each other.
(B) The accident X _i occurs only due to the feature value A _i .

Regarding the above approximation (A), since attention is paid to one vehicle (driving behavior of one driver), it is considered that a plurality of types of accidents hardly occur at the same time, so that it is considered appropriate. In addition, regarding the above approximation (B), an accident may occur due to the overlap of various factors. For example, the “accident due to wakimi” is mainly caused by “wakimi”. Alternatively, it is considered that a small number of features become dominant. Therefore, in the first embodiment, the accident-free probability is formulated after defining A _i as a one-dimensional feature value. By using the above approximations (A) and (B), the accident-free probability (probability of not causing an accident) p _safe is given by the following equation (2).

Note that p (X _i | A _i ) represents the probability that an accident X _i will occur with respect to the value of the feature value A _i .

Next, the probability p (X _i | A _i ) that an accident X _i will occur with respect to the value of the feature value A _i is defined. In the embodiment of the present invention, each probability p (X _i | A _i ) of occurrence of an accident X _i is approximated using a function for each value of the feature value A _i . p (X _i | A _i ) is approximated and defined using a Gaussian function as shown in the following formula (3) or formula (4).

A _max in the above equation (3) represents a value of the feature amount A _i at which the accident X _i occurs with a probability of 1 (100%), and is a predetermined value. Further, A _min in the above formula (4) represents a value of the feature value A _i at which the accident X _i does not occur with a probability of 1 (100%), and is a predetermined value. Σ _i represents a parameter of a Gaussian function.

Calculation unit 32, for each of a plurality of types of accidents, and operation information and accident statistics that are acquired by the statistical information acquiring unit 30, a feature quantity corresponding to accident X _i under conditions accident X _i occurs A _i The parameters used in the Gaussian function that approximates the probability of occurrence of the accident X _i for each value of the feature value A _i are calculated using the Bayes' theorem based on the constraint on the probability of obtaining each value of Then, a Gaussian function using the calculated parameters is calculated as the probability p (X _i | A _i ) that an accident will occur for each value of the feature quantity.

Specifically, the calculation unit 32 estimates the parameter σ _i (unknown number) included in the above formulas (3) and (4) by using knowledge about the feature amount A _i and the accident X _i . For example, the knowledge and the feature quantity A _i and accidents X _i, feature quantity value of the Gaussian function approximating becomes maximum, and the value of the Gaussian function approximating to preset at least one of smallest feature amount.

For example, in the case of “A _i : acceleration, X _i : slip accident”, it is assumed that a slip accident occurs when the acceleration exceeds the product of the static friction coefficient and the gravitational acceleration, and “A _max = static friction coefficient × gravity”. Approximate acceleration. (Alternatively, “A _min = 0 [m / s ² ]” may be set on the assumption that the accident probability when not accelerating is 0%.) Also, “A _i : inter-vehicle distance (or TTC: Time to Collision), X _i : rear-end collision due to lack of inter-vehicle distance ”, a rear-end collision always occurs when the inter-vehicle distance is 0 m. Therefore,“ A _max = 0 [m] ”can be obtained. As a result of the above approximation, the only unknown remaining in p (X _i | A _i ) is the parameter σ _i . Table 1 shows the relationship between the feature value A _i , the accident X _i , A _max , and A _min used in the embodiment of the present invention.

The parameter σ _i in the above formula (3) or (4) is a parameter that determines how safe the value of the feature value A _i is. By setting the parameter value well, It is thought that formulation with few contradictions is possible. Therefore, the approximated Gaussian function is a predetermined value of the feature value A _i when the value of the Gauss function is maximum, or a predetermined value of the feature value A _i when the value of the Gauss function is minimum. including.

As described above, the calculation unit 32 calculates the parameter σ _i of the approximated Gaussian function according to the Bayes' theorem using accident statistical information and driving information that are actual statistical information. Then, the calculation unit 32 calculates a Gaussian function using the calculated parameter σ _i .

First, p (X _i | A _i ) is transformed into the following equation (5) by applying Bayes' theorem.

Here, p (A _i | X _i ) represents the probability of obtaining the value of the feature value A _i under the condition where the accident X _i has occurred, and is an unknown value. Regarding the probability p (A _i ), the value of the two-sided infinite integral of p (A _i ) is 1 as shown in the following formula (6).

Further, a constraint condition regarding the probability p (A _i | X _i ) at which the value of the feature value A _i under the condition where the accident X _i occurs is shown in the following formula (7). Regarding the probability p (A _i | X _i ), it is difficult to obtain the distribution shape, but since the accident X _i is caused only by the feature amount A _i , as shown in the following equation (9), The value of the two-sided infinite integral of p (A _i | X _i ) is 1.

As shown in FIG. 3, the calculation unit 32 combines the above equations (1), (3), (4), (5), (6), and (7). The parameter σ _i can be calculated. If the calculation of the parameter σ _i is performed for all probabilities p (A _i | X _i ), the calculation formula for the accident-free probability p _{safe in} the above formula (2) can be applied.

In the following, the process for deriving σ _i , which is an unknown parameter of p (X _i | A _i ), will be described using the above formulas (3), (5), and (7) as an example.

When the value of the feature value A _i is defined as a continuous value, the above equation (3) is substituted into the above equation (5) and deformed, and as shown in the following equation (8), p (A _i | X _i ) is given.

Further, the above equation _p of (8) | a _(A i _{X i),} by substituting the above equation (7), the formula (9) below is obtained.

In the formula (9), the value p is unknown _| disappears (A i _{X i),} since the unknown is only parameter sigma _i, to obtain the parameter sigma _i by solving the above expression (9) it can. Here, when p (A _i ) acquired in advance is given as a function of A _i and the integration of the above equation (9) is analytically possible, it is actually integrated and parameter What is necessary is just to obtain | require (sigma) _i . If the integral of the equation (9) is difficult, or p if _{(A i)} are difficult defined as a function of _{A i is, A i = a i, k} (k = 1,2, ···, N _It is only necessary to discretize p (A _i ) as a _i ), and the above equations (7) and (9) can be approximated as the following equations (10) and (11), respectively.

Therefore, the unknown parameter σ _i can be obtained by solving the above equation (11). Similarly, when the value of the feature value A _i is defined as a discrete value A _i = a _{i, k} (k = 1, 2,..., N _ai ), Parameter σ _i is obtained. In the above description, unknown parameters are calculated by taking p (X _i | A _i ) of Equation (3) as an example. However, when Equation (4) is used, p (X) of Equation (9) or Equation (11) is also used. By substituting into X _i | A _i ) or p (X _i | a _{i, k} ), unknown parameters can be calculated in the same manner.

As described above, in the embodiment of the present invention, the feature amount A _i , the accident X _i , A _max , A _min , and p (X _i | A _i ) are defined based on knowledge and hypothesis, and are difficult to estimate. For the unknowns, the accident-free probability was formulated by using statistical information on accidents and driving data.

The driving behavior evaluation unit 34 is a probability approximated using a Gaussian function calculated for each of a plurality of types of accidents by the calculation unit 32 and each value of the plurality of feature amounts acquired by the feature amount acquisition unit 28. Based on p (X _i | A _i ), the probability (1-p (X _i | A _i )) that the accident X _i does not occur is calculated for each of a plurality of types of accidents. Then, the driving behavior evaluation unit 34 calculates the product of the probability of occurrence of an accident (1-p (X _i | A _i )) calculated for each of a plurality of types of accidents regarding the driving behavior of the target driver's vehicle. , The accident-free probability _psafe representing the probability that no accident will occur is calculated.

The output device 40 outputs the accident-free probability p _safe calculated by the driving behavior evaluation unit 34 as a result.

<Operation of the accident information calculation device 10>
Next, the operation of the accident information calculation apparatus 10 according to the first embodiment will be described. First, the computer 26 executes a parameter calculation processing routine shown in FIG.

  In step S <b> 100, the statistical information acquisition unit 30 acquires accident statistical information stored in the statistical information database 24.

  In step S <b> 102, the driving information stored in the statistical information database 24 is acquired by the statistical information acquisition unit 30.

In step S104, the calculation unit 32, an accident for each of the plurality of types of accidents, and accident statistics that are acquired by the operation information and the step S100 which is obtained in step S102, under conditions of an accident X _i occurs Based on the constraint condition relating to the probability that each value of the feature value A _i corresponding to X _i is obtained, the above formulas (1), (3), (4), (5) ), Equation (6), and Equation (7) are combined to approximate the probability of occurrence of accident X _i for each value of feature value A _i using a Gaussian function. The parameter used in the function shown in 4) is calculated, and the function using the calculated parameter is calculated as the probability p (X _i | A _i ) that an accident will occur for each value of the feature value A _i .

  Next, the vehicle travels by the driving operation of the driver to be diagnosed, the lateral acceleration of the vehicle is detected by the lateral acceleration sensor 12, the longitudinal acceleration of the vehicle is detected by the longitudinal acceleration sensor 14, and the vehicle speed sensor 16 The vehicle speed is detected, the inter-vehicle distance sensor 18 detects the inter-vehicle distance between the host vehicle and the preceding vehicle, the front camera 20 captures the front image, and the driver camera 22 captures the driver's face image. Sometimes, the computer 26 executes the driving diagnosis processing routine shown in FIG.

  In step S200, the lateral acceleration detected by the lateral acceleration sensor 12, the longitudinal acceleration detected by the longitudinal acceleration sensor 14, the passing speed at the time of pause, the inter-vehicle distance sensor, by the feature amount acquisition unit 28. The inter-vehicle distance detected by 18, the passing speed at the time of the temporary stop, the armpit driving distance, and the intersection passing time at the time of the red light are acquired as a plurality of feature amounts related to the driving behavior of the driver.

In step S202, the driving action evaluation unit 34, each value of the plurality of feature quantity A _i obtained in step S200, and for each of the plurality of types of accidents that have been calculated in step S104 of parameter calculation routine calculates Based on the Gaussian function, a probability (1-p (X _i | A _i )) that the accident X _i does not occur is calculated for each of a plurality of types of accidents.

In step S204, the probability (1−p (X) that the accident calculated for each of the plurality of types of accidents calculated in step S202 regarding the driving behavior of the target driver's vehicle by the driving behavior evaluation unit 34 does not occur. The product of _i | A _i )) is calculated as the accident-free probability p _safe representing the probability that no accident will occur.

In step S112, the output unit 40, and outputs the accident probability _{p safe} calculated in step S204 as a result.

  As described above, according to the accident information calculation apparatus of the first embodiment, for each of a plurality of types of accidents, the driving information and the accident statistical information, and the feature amount corresponding to the accident under the condition in which the accident has occurred. Based on the constraint on the probability that each value is obtained, the parameter used in a predetermined function that approximates the probability of occurrence of an accident for each value of the feature value is calculated, and the parameter using the calculated parameter The function is calculated, each value of the plurality of feature values related to the driving behavior of the vehicle of the target driver is obtained, and each of the plurality of feature values and the calculated function for each of the plurality of types of accidents are obtained. Based on this, the probability that no accident will occur for each of the multiple types of accidents is calculated, and the driving behavior of the target driver's vehicle is evaluated as an accident-free probability. It is possible to diagnose the operation of the driver.

  In addition, the driver's driving can be diagnosed without using statistics that are difficult to obtain. In addition, the operation of the driver can be diagnosed with high accuracy.

  In addition, by using statistics that can be easily observed without requiring large-scale information between accidents and factors, objective driving diagnosis and accident-free probability calculation can be performed in accordance with actual traffic conditions that can change over time. It becomes possible.

  In addition, by associating features such as acceleration and armpit driving time with the types of accidents such as slip accidents and armpit accidents, and based on theorems on conditional probabilities, statistics of driving information collected from many drivers and The accident-free probability for each driving state can be estimated from the accident statistics, and based on this, it is possible to diagnose whether driving is good or bad. Objective driving diagnosis according to actual traffic conditions is expected only with easily observable statistics, and it is possible to cope with changes in traffic conditions over time by sequentially updating only statistics.

[Second Embodiment]
Next, a second embodiment will be described. In addition, since the structure of the accident information calculation apparatus which concerns on 2nd Embodiment is the same as that of 1st Embodiment, it attaches | subjects the same code | symbol and abbreviate | omits description.

  The second embodiment is different from the first embodiment in that a function that approximates the probability that an accident will occur for each value of a feature value includes a plurality of parameters.

The function that approximates the probability p (X _i | A _i ) shown in the first embodiment can be arbitrarily selected. For example, the probability p (X _i | A _i ) of occurrence of an accident is approximated for each value of the feature quantity by a linear function, a sigmoid function, or the like as shown in the following formula (12) or formula (13). be able to.

Note that u _i and v _i in the above equations (12) and (13) represent parameters. In the first embodiment, the unknowns that can be calculated by combining the equations (1), (3), (4), (5), (6), and (7) are as follows: Since there is one, in the second embodiment in which the number of unknowns is two or more, another constraint condition is required when p (X _i | A _i ) is defined.

Therefore, when p (X _i | A _i ) of the above formula (12) or the above formula (13) is adopted instead of the above formula (3) or the above formula (4) of the first embodiment, the unknown one of u _i, v _i is a parameter, it is necessary to or determined based on or determined empirically, other information.

Alternatively, for example, when any one of A _max and A _min is known with respect to the above formula (12) and the above formula (13), one of the following formulas is established.

That is, by using the above formulas (14) and (15), the above formulas (12) and (13) are expressed by one unknown parameter as shown in the following formulas (16) and (17). It can be defined.

Here, Z _∞ is sufficiently large positive value, Z _-∞ is predefine as a sufficiently small negative value. As described above, it is possible to derive u _i and v _i which are a plurality of unknown parameters of p (X _i | A _i ).

  In addition, about the other structure and effect | action of the accident information calculation apparatus which concern on 2nd Embodiment, since it is the same as that of 1st Embodiment, description is abbreviate | omitted.

  As described above, according to the accident information calculation apparatus of the second embodiment, for each of a plurality of types of accidents, the accident statistics information and the driving information, and the feature amount corresponding to the accident under the condition in which the accident has occurred. Based on the constraint on the probability of obtaining each value, calculate a plurality of parameters to be used in a predetermined function that approximates the probability that an accident will occur for each value of the feature value, and calculate the plurality of parameters To calculate each function of the plurality of feature values related to the driving behavior of the vehicle of the target driver, and to calculate each value of the plurality of feature values and each of the calculated types of accidents. Statistics that can be obtained by calculating the probability that an accident will not occur for each of multiple types of accidents based on the calculated function and evaluating the driving behavior of the target driver's vehicle as an accident-free probability It is possible to diagnose operation of the driver used.

[Third Embodiment]
Next, a third embodiment will be described. In addition, since the structure of the accident information calculation apparatus which concerns on 3rd Embodiment is the same as that of 1st Embodiment, it attaches | subjects the same code | symbol and abbreviate | omits description.

The third embodiment is different from the first and second embodiments in that the feature quantity A _i is expanded to a multidimensional feature quantity and the feature quantity A _i is acquired as an n-dimensional vector.

The statistical information database 24 according to the third embodiment includes driving information representing the probability p (A _i ) that each value of the multi-dimensional feature value A _i relating to the driving behavior of the vehicle is obtained, and multi-dimensional And accident statistical information representing the probability of occurrence of an accident X _i caused by the feature amount A _i of the.

Further, in the computer of the accident information calculation apparatus according to the third embodiment, the feature amount acquisition unit 28 has a value of the multi-dimensional feature amount A _i regarding the driving behavior of the target driver for each of the plurality of types of accidents X _i. To get.
For example, the feature amount acquisition unit 28 is obtained from the vehicle speed detected by the vehicle speed sensor 16 and the front image captured by the front camera 20 in addition to the sideward driving distance for the accident caused by the side look in Table 1 above. The presence / absence of a pedestrian and the like is acquired as a value of each dimension of a multidimensional feature amount A _i (= (A _{i, 1} , A _{i, 2} ,..., A _{i, n} )) related to the driving behavior of the target driver. To do.
In addition, for the rear-end collision in Table 1, not only the inter-vehicle distance but also the degree of driver's eye closed detected from the driver image acquired from the driver camera 22, the multi-dimensional feature amount A _i ( = (A _{i, 1} , A _{i, 2} ,..., A _{i, n} )) may be acquired as values of each dimension.

The calculation unit 32 calculates the driving information and the accident statistical information acquired by the statistical information acquisition unit 30, and each value of the dimension under the condition in which the accident X _i occurs for each dimension of the multi-dimensional feature value A _i. Parameters used in a predetermined function approximating the probability of occurrence of an accident X _i for each value of each dimension of the feature value A _i are calculated on the basis of the constraint conditions relating to the obtained probability, and the calculated parameters The function using is calculated.

The calculation unit 32 calculates the parameters using Bayes' theorem and assuming that the dimensions of the multi-dimensional feature quantity A _i are independent or dependent from each other.
Specifically, it is assumed that the values of each dimension of the multi-dimensional feature quantity A _i are obtained independently or subordinate to each other. When the multi-dimensional feature quantity A _i acquired by the feature quantity acquisition unit 28 is n-dimensional such as A _i = (A _{i, 1} , A _{i, 2} ,..., A _{i, n} ). , each feature _{a i, 1, a i,} 2, ···, by assuming _{a i,} independently or dependent of _n, the above formula (7), the following equation (18) or (19) By using the following approximation and further using the following equation (20), n unknowns can be calculated. The following equation (20) is an example of a constraint condition regarding the probability that each value of the dimension is obtained under the condition in which the accident X _i has occurred for each dimension of the multi-dimensional feature value A _i .

Thus, _p may be defined to _{| | (A i X i)} , for example _p unknown parameters _(X i _{A i)} is _p such that n or less _{(X i} | _A i) the following equation ( A method of approximating with the function of the multivariate normal distribution of 21) is conceivable.

v _i represents a parameter, Σ _i represents a parameter matrix, and A _ref represents a preset vector. Since the number of parameters that can be obtained in the third embodiment is n, some parameters are determined empirically so that the number of parameters in the above equation (21) is n or less. Or predetermined based on other information. As described above, it is considered that it is possible to cope with an accident caused by a multidimensional feature amount.

The driving behavior evaluation unit 34 calculates the plurality of types of accidents based on the feature value acquired by the feature amount acquisition unit 28 and the Gaussian function calculated for each of the plurality of types of accidents calculated by the calculation unit 32. The probability (1-p (X _i | A _i )) that the accident X _i does not occur is calculated for each. Then, the driving behavior evaluation unit 34 calculates the product of the probability of occurrence of an accident (1-p (X _i | A _i )) calculated for each of a plurality of types of accidents regarding the driving behavior of the target driver's vehicle. , The accident-free probability _psafe representing the probability that no accident will occur is calculated.

  In addition, about the other structure and effect | action of the accident information calculation apparatus which concern on 3rd Embodiment, since it is the same as that of 1st Embodiment, description is abbreviate | omitted.

  As described above, according to the accident information calculation apparatus of the third embodiment, each value of the dimension under the condition in which the accident occurs with respect to each dimension of the driving information, the accident statistical information, and the multi-dimensional feature amount. Based on the constraint on the probability of obtaining a parameter, a parameter used in a predetermined function approximating the probability of occurrence of an accident is calculated for each value of each dimension of the feature value, and the calculated parameter is used. The function is calculated, each value of the plurality of feature values related to the driving behavior of the vehicle of the target driver is obtained, and each of the plurality of feature values and the calculated function for each of the plurality of types of accidents are obtained. Based on this, the probability that no accident will occur for each of the multiple types of accidents is calculated, and the driving behavior of the target driver's vehicle is evaluated as an accident-free probability. It is possible to diagnose the operation.

  In the third embodiment, instead of the function of the multivariate normal distribution of the above equation (21), approximation can be performed by a linear function of the following equation (22) and a sigmoid function of the equation (23).

v _i and w _{i, j} represent parameters. Since the number of parameters that can be obtained in the third embodiment is n as in the above equation (21), the number of parameters in the above equations (22) to (23) is n or less. Therefore, some parameters are determined empirically or in advance based on other information.

[Fourth Embodiment]
Next, a fourth embodiment will be described. In addition, since the structure of the accident information calculation apparatus which concerns on 4th Embodiment is the same as that of 1st Embodiment, it attaches | subjects the same code | symbol and abbreviate | omits description.

The fourth embodiment is different from the first to third embodiments in that the accident-free probability is calculated for the feature value A _i that changes every moment.

The p _safe (A ₁ , A ₂ ,...) Calculated in the first to third embodiments is obtained when the state (A ₁ , A ₂ ,. It means the probability of not causing an accident. For example, when p (X _i ) is “1 [km] accident probability per driving”, p _safe (A ₁ , A ₂ ,...) Is “state (A ₁ , A ₂ ,...)”. This means the accident probability when driving 1 [km] while maintaining
However, since it is rare that the same driving state is actually maintained, in the fourth embodiment, the accident-free probability is calculated for the driving state that changes every moment.

First, an arbitrary physical quantity L _i is defined. But simultaneously, conditional accident probability p for the feature amount A _i (X i _{| A i)} is defined as the "probability that remain with the operational values of certain characteristic amount A _i is an accident occurs until satisfy L _i" Is done. For example, _{L i} is applicable such as travel distance and travel time, when the travel distance _{L i p (X i | A} i) is the running of the "while the value of the feature amount _{A i} is constant _L i [km] The probability that an accident will occur before it is satisfied. Here, if the sum of the physical quantity _{L i} is the state of the feature quantity _{A i} relative section and _l i _{(A i),} accident _{X i} is not generated until the operation of the subject satisfies the _l i _{(A i)} When the probability (accident-free probability for accident X _i ) q _safe (X _i | A _i , l _i (A _i )) is defined, the following equation (24) is established.

Accordingly, while the value of the feature value A _i changes every moment, the accident-free probability q _safe until the target driving satisfies L _i is given by the following equation (25).

Here, _{S i} is a non-countable set of _{A i} (a set of all possible values of _{A i).}

For example, when the travel distance is used as the arbitrary physical quantity L _i , the feature quantity acquisition unit 28 obtains the value of the feature quantity A _i and the value of the feature quantity A _i regarding the driving behavior of the target driver's vehicle. The travel distance l _i (A _i ) is acquired.
The driving behavior evaluation unit 34 calculates the value of the feature amount A _i and the travel distance l _i (A _i ) acquired by the feature amount acquisition unit 28 and each of the plurality of types of accidents by the calculation unit 32. probability p accidents X _i is generated by the travel distance L _i which is predetermined for each value of the characteristic quantity a _{i |,} based on the function obtained by approximating (X i _{a i),} the operation of the vehicle target driver For the behavior, the product of the probability of occurrence of an accident (1-p (X _i | A _i )) calculated for each of a plurality of types of accidents is used as the accident-free probability q _safe (A ₁ , A ₂ ,... ) Is calculated.

  In addition, about the other structure and effect | action of the accident information calculation apparatus which concern on 3rd Embodiment, since it is the same as that of 1st-3rd Embodiment, description is abbreviate | omitted.

  According to the accident information calculation apparatus of the fourth embodiment, the acquired feature value, the duration or the running time, and each value of the feature amount up to a predetermined duration or the running time. Based on a function that approximates the probability that an accident will occur, the product of the probability that no accident will occur for each type of accident will be evaluated as the accident-free probability for the driving behavior of the target driver's vehicle Thus, it is possible to calculate the accident-free probability for the feature quantity that changes every moment.

In the fourth embodiment described above, the case of using the travel distance as any physical quantity L _i has been described as an example, may be used travel time as any physical quantity L _i. In this _{case, p (X i | A i} ) is "the probability that the travel time value of the feature amount A _i is continued remained constant accident occurs until meeting the travel of L _i time". The feature amount acquisition unit 28 acquires the value of the feature amount A _i related to the driving behavior of the vehicle of the target driver and the duration l _i (A _i ) of the value of the feature amount A _i . Then, the driving behavior evaluation unit 34 calculates the value of the feature amount A _i acquired by the feature amount acquisition unit 28 and the duration l _i (A _i ), and each of a plurality of types of accidents by the calculation unit 32. Based on a function that approximates the probability p (X _i | A _i ) of occurrence of an accident X _i for each value of the feature value A _i by a predetermined duration L _i , the driving of the target driver's vehicle For the behavior, the product of the probability of occurrence of an accident (1-p (X _i | A _i )) calculated for each of a plurality of types of accidents is used as the accident-free probability q _safe (A ₁ , A ₂ ,... ) Is calculated.

In the above embodiment, the case where the accident information calculation device is mounted on the vehicle has been described as an example. However, the present invention is not limited to this. For example, the accident information calculation device is realized by a server or the like. Also good. In this case, characteristic values relating to driving behaviors of a plurality of drivers are collected, and driving information representing the probability p (A _i ) at which each value for the plurality of characteristic values A _i relating to the driving behavior of the vehicle is obtained. You may make it ask. Further, the feature amount may be sequentially acquired in real time to evaluate the driving behavior of the target driver.

Further, in the above embodiment, the probability of occurrence of an accident is calculated for the acquired feature value, and the accident-free probability representing the probability that no accident will occur is calculated based on the calculated probability of occurrence of the accident. Although the case where the present invention is applied to the accident information calculation apparatus to be calculated has been described as an example, the present invention is not limited to this.
For example, the statistical information acquisition unit 30 acquires driving information indicating the degree to which each value of the characteristic amount related to the driving behavior of the vehicle is obtained, and accident statistical information indicating the degree of occurrence of an accident caused by the characteristic amount. The calculation unit 32 is based on the driving information and accident statistical information acquired by the statistical information acquisition unit 30 and the restriction condition regarding the degree to which each value of the characteristic value under the condition where the accident occurred is obtained. The degree of occurrence of an accident may be calculated for each value. In this case, the accident-free probability of the target driver is not calculated, but the degree of occurrence of an accident can be calculated for each feature value.
Further, when calculating the degree of occurrence of an accident for each value of the feature amount, the calculation unit 32 calculates the driving information and accident statistical information acquired by the statistical information acquisition unit 30 and the conditions under which the accident occurred. The parameters used in a predetermined function that approximates the degree of occurrence of an accident with respect to each value of the feature amount are calculated based on the restriction condition regarding the degree that each value of the feature amount is obtained at A function using parameters may be calculated as the degree of occurrence of the accident for each value of the feature value.

  Alternatively, the feature amount acquisition unit 28 acquires the value of the feature amount related to the driving behavior of the vehicle of the target driver, and the driving behavior evaluation unit 34 calculates the value of the feature amount acquired by the feature amount acquisition unit 28 and the calculation unit 32. The degree of occurrence of an accident or the degree of no occurrence of an accident may be evaluated for the driving behavior of the target driver's vehicle based on the degree of occurrence of an accident for each value of the feature amount calculated by the above. In this case, the probability is not calculated, but the degree at which an accident occurs or the degree at which no accident occurs can be calculated.

  As described above, when calculating the degree of occurrence of an accident or the degree of no occurrence of an accident for each value of the feature value, the probability is not calculated, so the Bayes' theorem shown in the above equation (7) is It is not used as a probability law but as a relationship between nonlinear functions.

  Further, in the above-described embodiment, the case where the accident-free probability is calculated has been described as an example. However, the present invention is not limited to this. For example, the statistical information acquisition unit 30 includes each value of the feature amount related to the driving behavior of the vehicle. Driving information indicating the probability of obtaining an accident and accident statistical information indicating the probability of occurrence of an accident caused by the feature amount, the calculation unit 32 acquires the driving information and the accident acquired by the statistical information acquisition unit 30 Predetermined function that approximates the probability that an accident will occur for each value of the feature based on statistical information and the constraint on the probability that each value of the feature will be obtained under the condition where the accident occurred It is also possible to calculate the parameters used in step 1 and calculate only the function using the calculated parameters.

  In the above embodiment, the case where the probability of occurrence of an accident is calculated for each value of the feature amount acquired by the feature amount acquisition unit 28 has been described as an example. The probability that an accident will not occur may be calculated for each value of the feature amount.

  In the above embodiment, the lateral acceleration detected by the lateral acceleration sensor 12, the longitudinal acceleration detected by the longitudinal acceleration sensor 14, the passing speed at the time of pause, and the inter-vehicle distance sensor 18 are used. Although the case where the detected inter-vehicle distance, the passing speed at the time of the temporary stop, the armpit driving distance, and the intersection passing time at the time of the red light are acquired as a plurality of feature amounts is described as an example, However, the present invention is not limited to this, and other feature amounts may be acquired.

  The program of the present invention can be provided by being stored in a recording medium.

DESCRIPTION OF SYMBOLS 10 Accident information calculation apparatus 12 Lateral direction acceleration sensor 14 Longitudinal direction acceleration sensor 16 Vehicle speed sensor 18 Inter-vehicle distance sensor 20 Front camera 22 Driver camera 24 Statistical information database 26 Computer 28 Feature-value acquisition part 30 Statistical information acquisition part 32 Calculation part 34 Driving behavior Evaluation unit 40 output device

Claims (16)

Statistical information acquisition means for acquiring driving information indicating the degree to which each value of the characteristic amount related to the driving behavior of the vehicle is obtained, and accident statistical information indicating the degree of occurrence of an accident caused by the characteristic amount;
Based on the driving information and the accident statistical information acquired by the statistical information acquisition means, and a constraint condition regarding the degree to which each value of the characteristic value under the condition in which the accident has occurred is obtained. Calculating means for calculating the degree of occurrence of the accident for each value;
Accident information calculation device.   The calculation means is based on the driving information and the accident statistical information acquired by the statistical information acquisition means, and a constraint condition relating to the degree to which each value of the feature value is obtained under the condition where the accident has occurred. Calculating a parameter used in a predetermined function approximating the degree of occurrence of the accident with respect to each value of the feature value, and converting the function using the calculated parameter to each value of the feature value. The accident information calculation apparatus according to claim 1, wherein the accident information calculation apparatus calculates the degree of occurrence of the accident.   The calculation means is based on the driving information and the accident statistical information acquired by the statistical information acquisition means, and a constraint condition relating to the degree to which each value of the feature value is obtained under the condition where the accident has occurred. The accident information calculation apparatus according to claim 2, wherein the parameter is calculated using Bayes' theorem. Feature amount acquisition means for acquiring a value of a feature amount related to the driving behavior of the vehicle of the target driver;
Based on the value of the feature quantity acquired by the feature quantity acquisition means and the degree of occurrence of the accident with respect to each value of the feature quantity calculated by the calculation means, the driving behavior of the vehicle of the target driver The accident information calculation device according to any one of claims 1 to 3, further comprising: a driving behavior evaluation unit that evaluates the degree of occurrence of the accident or the degree of occurrence of the accident. The statistical information acquisition means acquires the driving information indicating a probability that each value of a feature amount related to driving behavior of a vehicle is obtained, and the accident statistical information indicating a probability that an accident caused by the feature amount occurs. And
The calculation means is based on the driving information and the accident statistical information acquired by the statistical information acquisition means, and the constraint condition regarding the probability that each value of the feature value is obtained under the condition where the accident has occurred. And calculating a parameter to be used for a predetermined function approximating the probability of occurrence of the accident for each value of the feature amount, and calculating the function using the calculated parameter. 5. The accident information calculation apparatus according to any one of items 4. The statistical information acquisition means generates the driving information indicating the probability of obtaining each value of each dimension of the multi-dimensional feature quantity related to the driving behavior of the vehicle and an accident caused by the multi-dimensional feature quantity. The accident statistics information representing the probability,
The calculation means includes the driving information and the accident statistical information acquired by the statistical information acquisition means, and each value of the dimension under the condition in which the accident has occurred for each dimension of the multi-dimensional feature amount. Based on the constraint condition regarding the obtained probability, a parameter used in a predetermined function approximating the probability of occurrence of the accident is calculated for each value of each dimension of the feature value, and the calculated parameter The accident information calculation apparatus according to any one of claims 2 to 4, wherein the function is calculated using the function. The statistical information acquisition means includes the driving information indicating the probability that each value for a plurality of feature values related to driving behavior of a vehicle is obtained, and each of a plurality of types of accidents caused by each of the plurality of feature values. The accident statistics information representing the probability of occurrence,
The calculation means includes, for each of the plurality of types of accidents, the driving information and the accident statistical information acquired by the statistical information acquisition means, and the feature amount corresponding to the accident under the condition where the accident has occurred. Based on the constraint condition relating to the probability that each value is obtained, a parameter used in a predetermined function that approximates the probability that the accident will occur for each value of the feature amount is calculated, and the calculated The accident information calculation apparatus according to claim 5 or 6, wherein the function using a parameter is calculated. Feature quantity acquisition means for acquiring each value of the plurality of feature quantities relating to the driving behavior of the vehicle of the target driver;
Based on the value of each of the plurality of feature amounts acquired by the feature amount acquisition unit and the function calculated for each of the plurality of types of accidents by the calculation unit, for each of the plurality of types of accidents Calculate the probability that an accident will not occur, and for the driving behavior of the target driver's vehicle, the product of the probability that the accident will not occur calculated for each of the multiple types of accidents represents the probability that no accident will occur The accident information calculation device according to claim 7, further comprising driving behavior evaluation means for evaluating as a probability. The feature amount acquisition means acquires the value of the feature amount related to the driving behavior of the vehicle of the target driver, and the duration of the value of the feature amount,
The driving behavior evaluation means includes a value of the feature amount acquired by the feature amount acquisition means and the duration, and a predetermined duration calculated for each of the plurality of types of accidents by the calculation means. Based on the function that approximates the probability of occurrence of the accident for each value of the feature amount, the driving behavior of the vehicle of the target driver is calculated for each of the plurality of types of accidents The accident information calculation apparatus according to claim 8, wherein a product of probabilities that an accident does not occur is evaluated as the accident-free probability. The feature amount acquisition means acquires the value of the feature amount related to the driving behavior of the vehicle of the target driver, and the travel distance when the value of the feature amount is obtained,
The driving behavior evaluation unit is configured to obtain a predetermined travel distance calculated for each of the plurality of types of accidents by the value of the feature amount and the travel distance acquired by the feature amount acquisition unit and the calculation unit. Based on the function that approximates the probability of occurrence of the accident for each value of the feature amount, the driving behavior of the vehicle of the target driver is calculated for each of the plurality of types of accidents The accident information calculation apparatus according to claim 8, wherein a product of probabilities that an accident does not occur is evaluated as the accident-free probability. The function includes a predetermined value of the feature amount when the value of the function is maximum, or a predetermined value of the feature amount when the value of the function is minimum. The accident information calculation device according to claim 10.   The calculation means includes the driving information and the accident statistical information acquired by the statistical information acquisition means, and each value of the dimension under the condition in which the accident has occurred for each dimension of the multi-dimensional feature amount. 7. The parameter according to claim 6, wherein the parameter is calculated on the basis of the constraint on the obtained probability, using Bayes' theorem, and assuming that the dimensions of the feature quantities of the plurality of dimensions are independent or dependent from each other. Accident information calculation device.   4. The accident information calculation apparatus according to claim 2, wherein the function is a Gaussian function that approximates the degree of occurrence of the accident for each value of the feature amount.   4. The accident information calculation apparatus according to claim 2, wherein the function is a linear function that approximates the degree of occurrence of the accident for each value of the feature amount.   4. The accident information calculation apparatus according to claim 2, wherein the function is a sigmoid function that approximates the degree of occurrence of the accident for each value of the feature amount. Computer
Statistical information acquisition means for acquiring driving information indicating the degree to which each value of the characteristic amount related to driving behavior of the vehicle is obtained, and accident statistical information indicating the degree of occurrence of an accident caused by the characteristic amount, and the statistics Each value of the feature value based on the driving information and the accident statistical information acquired by the information acquisition means, and a constraint on the degree to which each value of the feature value is obtained under the condition where the accident occurred A program for functioning as a calculating means for calculating the degree of occurrence of the accident.