JP2020115306A - Assessment method and determination system of pedestrian accident risk by location - Google Patents

Abstract

To provide an assessment method of pedestrian accident risk by location focusing on the frequency of occurrence of situations similar to those in which an accident occurs, instead of simply measuring the traffic volume of vehicles of the number of pedestrians.SOLUTION: An assessment method of pedestrian accident risk by location includes the steps of: selecting a plurality of intersections in advance, and obtaining structure information, vehicle traffic volume information, pedestrian collision alert information and pedestrian accident information for each intersection; calculating, as a pedestrian collision alert occurrence rate, a ratio of the number of times the collision alert included in the pedestrian collision alert information is detected to the traffic volume included in the vehicle traffic information; creating a pedestrian accident count statistical model based on the correlation with the number of accidents in the pedestrian accident information with regards to each element and the pedestrian collision alert occurrence rate included in the structure information and the vehicle traffic volume information; and calculating the probability of occurrence of pedestrian accidents by checking each element and the pedestrian collision alert occurrence rate included in the structure information and the vehicle traffic volume information at an evaluation target point against the pedestrian accident count statistical model.SELECTED DRAWING: Figure 2

Description

The present invention relates to a method for objectively evaluating the risk of a vehicle causing an accident with a pedestrian, and a system for determining the risk.

In general, most of the accidents that occur when a vehicle travels are at intersections, and although it is certain that the accidents are often between vehicles, there are also many accidents between vehicles and pedestrians. Conventionally, in road design, the state of a dangerous intersection has been evaluated exclusively from the driver's viewpoint, considering that the incidence of accidents depends on whether or not the driver feels danger. For example, there is a traffic dynamics simulation device that is used when designing a new intersection or road while reflecting micro information by acquiring the degree of danger sensed by the driver in the simulation space. (See Patent Document 1).

However, even if we construct a simulation space with various conditions, collect various information, and design new intersections that reflect these information, we will construct new intersections based on the design. In that case, a phenomenon different from the simulation space may occur. In other words, it is possible to cause other dangers due to realistic factors due to changes in traffic volume, changes in pedestrian trends, unpredictable events, etc. even though the design is designed with safety in mind. It was

JP, 2002-140786, A JP, 2016-218862, A

Therefore, it has been proposed to detect speed information from an actually traveling vehicle (probe car) and evaluate an area with a high accident risk from a causal relationship with the accident (see Patent Document 2). This technology uses a causal relationship with the large number of accidents in an area that causes a large difference between the speed at which a general vehicle travels and the speed at which a commercial vehicle travels. For that reason, the general vehicle and the commercial vehicle are run as probe cars, respectively, and the data is used to evaluate the possibility of an accident.

However, it is certain that the change in speed according to the characteristics of the vehicle as described above is one of the causes of the causal relationship with the accident rate, but the increase or decrease in the accident rate is not due to only the speed change, In addition, the case generally referred to as an accident includes a case of a vehicle and a case of a pedestrian, and it is intended to integrally handle these cases.

By the way, among the accidents caused by the traveling of the vehicle, the ratio of the accident to the pedestrian is higher than that to the vehicle in the case of causing serious damage such as a fatal accident. This is because in a vehicle-to-vehicle accident, most of the time, when traveling at low speeds instead of traveling at high speeds, the amount of damage is small. This is because the frequency of serious damage is high.

However, the evaluation of which point in the road traffic network has a degree of danger has never been revealed until an actual accident, and from the viewpoint of preventing pedestrian accidents. , Potential hazards should be evaluated even at points where no accidents actually occur. It is also possible to evaluate the potential risk based on the road environmental conditions at a certain point on the road and the traffic volume of general vehicles, but this focuses exclusively on the traffic volume of vehicles. Therefore, the factor of the amount of pedestrians was neglected. However, it should not be judged that the danger is high even if the number of pedestrians is high. That is, it is empirically clear that an accident is unlikely to occur at an intersection where the vehicle is decelerated or stopped, even if the traffic volume and the number of pedestrians are enormous. It is also easy to understand from experience that the risk of an accident is high in a place where a moving vehicle and a pedestrian often come close to each other even if the number of people is small.

The present invention has been made in view of the above points, and an object thereof is not the method of simply measuring the traffic volume or the number of pedestrians of a vehicle, but the appearance frequency of a situation similar to the occurrence of an accident. It is intended to provide a method for evaluating a pedestrian accident risk level and a pedestrian risk level determination system for each point while paying attention to.

Therefore, the present invention according to the pedestrian accident risk evaluation method by point, a plurality of intersections are selected in advance, probe vehicle traffic information including a traffic volume at which a probe vehicle equipped with a collision warning device enters the intersections, Pedestrian collision warning information including the number of times the collision warning device mounted on the probe vehicle detected a collision warning with a pedestrian around the intersection, and walking including the number of accidents the pedestrian actually encountered. Person accident information is acquired, and the ratio of the number of times the collision warning included in the pedestrian collision warning information is detected to the traffic of the probe vehicle included in the probe vehicle traffic information is calculated as a pedestrian collision warning occurrence rate. Then, for the pedestrian collision warning occurrence rate, a pedestrian accident count statistical model is created based on the correlation with the number of accidents in the pedestrian accident information, and the pedestrian collision warning occurrence rate at the evaluation target point is calculated as the pedestrian accident count. The feature is that the probability of occurrence of a pedestrian accident is calculated by collating with a statistical model.

According to the above configuration, it is possible to obtain the probability of occurrence of an accident by using the statistical model as an accident occurrence situation at a plurality of preselected intersections and matching the statistical model with the statistical model. The probability of an accident at this time is such that it is possible to judge the degree of danger from the actual occurrence of the accident, depending on the magnitude of the probability of the accident. If the condition is close to the point, it is possible to determine that an accident may occur thereafter.

The present invention according to a pedestrian accident risk evaluation method by point, in the above configuration, for each intersection, the structure information including the number of branches and the high rise or low rise of buildings around the intersection, or the degree of congestion, Further, either or both of general vehicle traffic information including traffic volume at which an ordinary vehicle enters an intersection is acquired, and each of the elements included in the structure information or the general vehicle traffic information or both of them. Also, for the pedestrian collision warning occurrence rate, a pedestrian accident number statistical model is created based on the correlation with the number of accidents in the pedestrian accident information, and the structure information or the general vehicle traffic information at the evaluation target point or these It is characterized in that the probability of occurrence of a pedestrian accident is calculated by matching each element included in both and the pedestrian collision warning occurrence rate with the statistical model of the number of pedestrian accidents.

According to the above configuration, various information such as structure information and general vehicle traffic information will be reflected in the statistical model of the number of pedestrian accidents, so that not only mere traffic concentration but also the situation around the intersection can be adjusted. The risk of a potential accident can be evaluated. Here, as the structure information, the information such as the high-rise building or the low-rise building is mentioned, but the structure information is not limited to this, and information focusing on other conditions may be added. This information can include information about whether the road is a flat road, an uphill road, or a downhill road as conditions for the road itself. In the situation around the intersection, the planting conditions such as trees are included as information. be able to. However, if the amount of information to be added is huge, it may not have a direct relationship with pedestrian accidents, so focus on the quantitative conditions of pedestrians (whether residential area or commercial area). Preferably.

Further, the present invention according to a pedestrian accident risk evaluation method for each point, in the configuration described in any of the above invention, the passage of the probe vehicle included in the probe vehicle traffic information at all preselected intersections. Based on the ratio of the number of times the collision warning included in the pedestrian collision warning information is detected to the amount, a collision warning statistical model showing a detection tendency of an average collision warning is created, and the intersection at each point of the intersection is A model deviation residual is calculated from the collision warning detection tendency indicated by the collision warning statistical model, and the actual collision warning detection tendency at each point, and the pedestrian warning occurrence rate is calculated based on the model deviation residual. It is characterized by being modified.

In the case of such a configuration, an average relationship between the traffic volume of the probe vehicle and the number of collision warning detections is shown as a collision warning statistical model. The pedestrian warning occurrence rate can be corrected by calculating the model deviation residual due to the actual collision warning detection tendency at the intersection. Since this modification can be typified with other elements (for example, elements included in structure information), a detailed pedestrian accident count statistical model can be created and the accuracy of the accident probability can be improved. Can be improved. When calculating the accident probability based on the average relationship between the traffic volume of the probe vehicle and the number of collision warning detections, the increase in the traffic volume of the probe vehicle and the increase in the collision warning detection number must change linearly. However, since the model deviation residual is non-linear, the suitability for probability calculation is improved.

In addition, in the invention of each of the above configurations, the prediction of the occurrence of the pedestrian accident is performed by using the elements included in the structure information and the vehicle traffic information at the evaluation target point and the pedestrian collision warning occurrence rate as the pedestrian accident number. Derivation of the pre-prediction value by matching it with a statistical model, and deriving the post-prediction value by correcting the pre-prediction value based on the number of accidents and the number of accidents that pedestrians have subsequently encountered at the evaluation target point. However, the posterior predicted value having a high accident occurrence probability can be used as the final predicted value.

In the case of the above configuration, it is possible to reflect the posterior prediction value that reflects the actual number of accidents to the prior prediction value based on the statistical model of the number of pedestrian accidents. It is possible to evaluate with high accuracy. The pre-predicted value in this case is the predicted value when it is matched with the statistical model of the number of pedestrian accidents, and the posterior predicted value is the predicted value that is updated by the actual number of accidents at each evaluation target point. .. Therefore, ex-post includes the case where the actual number of accidents in the past is present, and further includes the case where the number of accidents increases due to the occurrence of accidents after calculation of the predicted value. When updating with the number of accidents after the predicted value is calculated, it is possible to deal with traffic conditions that change every moment.

Further, in the invention of each of the above configurations, the collision warning device determines whether the image processing means capable of recognizing a person, the speed recognizing means, and a vehicle traveling at a predetermined speed or more collide with a person within a predetermined time. The detection means includes a processing means and an output means for outputting an alarm when a collision is determined by the determination processing means, and the number of collision warning detections included in the pedestrian collision warning information is the output means. The configuration may be based on the number of alarms output by.

In the case of the above configuration, the collision warning device can cause the person (pedestrian) recognized by the image processing to judge the possibility of a collision from the relationship between the distance to the vehicle and the traveling speed of the vehicle. In addition, it is possible to determine the possibility of collision only when the vehicle travels at a predetermined speed exceeding the walking speed (for example, 7 km/h or more). The possibility of collision is determined within a predetermined time (for example, within 2 seconds) based on the relationship with the distance to the recognized person (pedestrian) on the assumption that the vehicle speed is maintained, for example. There is a method of determining that there is a possibility of a collision when it is determined that the vehicle arrives at. The output means for outputting the warning includes visual and audible warning means for the vehicle driver, and at the same time, it is possible to separately provide a means for notifying the server etc. by radio or the like. The pedestrian collision warning information can be acquired by measuring the location and the number of outputs.

On the other hand, the present invention related to a pedestrian accident risk determination system according to location stores a plurality of intersections in advance and stores probe vehicle traffic information including a traffic volume at which a probe vehicle equipped with a collision warning device enters the intersection. And a pedestrian that stores pedestrian collision warning information including the number of times a collision warning with a pedestrian around the intersection is detected by the collision warning device mounted on the probe vehicle A collision warning information database, a pedestrian accident information database that stores pedestrian accident information that includes the number of accidents that pedestrians actually encountered, and the information stored in each database is processed to predict the occurrence of pedestrian accidents. And a processing means for calculating, wherein the processing means determines a ratio of the number of times the collision warning included in the pedestrian collision warning information is detected with respect to the traffic of the probe vehicle included in the probe vehicle traffic information. Calculated as an alarm occurrence rate, for the pedestrian collision alarm occurrence rate, create a pedestrian accident number statistical model based on the correlation with the number of accidents in the pedestrian accident information, the pedestrian collision alarm occurrence rate at the evaluation target point The probability of occurrence of a pedestrian accident is calculated by collating with the statistical model of the number of pedestrian accidents.

According to the above configuration, the probe vehicle traffic information, the pedestrian collision warning information, and the pedestrian accident information for each of the plurality of intersections are stored in the respective databases, and the processing means is used by using the information stored in these databases. Can be processed by. The processing means finally creates a statistical model of the number of pedestrian accidents, and concrete information on each element included in the structure information and vehicle traffic information at the evaluation target point and the pedestrian collision warning occurrence rate. By collating with the pedestrian accident count statistical model, it is possible to identify a point with a potentially high accident rate. Note that the structure information, the vehicle traffic information, and the like may be stored in a database, and the pedestrian accident count statistical model can reflect the relationship between these information and the accident count in the pedestrian accident information.

Further, as another invention relating to a pedestrian accident risk determination system for each point, a plurality of intersections are selected in advance, and probe vehicle traffic information including a traffic volume at which a probe vehicle equipped with a collision warning device enters the intersections. And a pedestrian collision warning information including the number of times the collision warning device mounted on the probe vehicle detects a collision warning with a pedestrian around the intersection around the intersection. Pedestrian collision warning information database, pedestrian accident information database that stores pedestrian accident information including the number of accidents that pedestrians actually encountered, and pedestrian accident occurrence by processing the information stored in each database A collision warning included in the pedestrian collision warning information with respect to the traffic amount of the probe vehicle included in the probe vehicle traffic information at all preselected intersections. A collision warning statistical model showing an average collision warning detection tendency based on a ratio of the number of times of detection, and a collision warning detection tendency indicated by the collision warning statistical model at each intersection point, and , The model deviation residual is calculated from the detection tendency of the actual collision warning at each point, and the pedestrian accident statistical model is created for the model deviation residual based on the correlation with the number of accidents in the pedestrian accident information. The probability of occurrence of a pedestrian accident is calculated by matching the model deviation residual at the evaluation target point with the statistical model of the number of pedestrian accidents.

According to the above configuration, since it is possible to obtain the model deviation residual for each intersection under individual conditions, it is based on a specific pedestrian warning detection tendency as compared with the case of an average relationship between traffic volume and warning detection times. A pedestrian accident count statistical model can be created. In the present invention as well, the structure information, the vehicle traffic information, etc. can be stored in the database and reflected in the statistical model of the number of pedestrian accidents, as described above.

In addition, in the invention related to the pedestrian accident risk determination system for each point, it may be configured to include an input unit for inputting individual information and an output unit for outputting the calculated probability of occurrence of a pedestrian accident. In the output unit in this case, there may be a method of displaying road information (map data) on a monitor and changing the color according to the probability of occurrence of a pedestrian accident.

According to the present invention related to the pedestrian accident risk evaluation method for each point, the risk is evaluated not by the risk that is biased to the traffic volume of the vehicle but by the appearance frequency of the situation that is close to the actual occurrence of the accident. can do. In addition, in the case of the invention in which the pre-predicted value is updated to the post-predicted value, the predicted value can be set with the highest probability, and the effect of the measure is verified even after some accident prevention measures are taken. can do. In the verification in this case, since a quantitative relationship between the frequency at which the pedestrian collision warning is detected and the number of accidents including actual pedestrians is calculated, it is possible to make a quantitative comparison after the fact. Is.

Further, by widely publicizing the results evaluated by the above-mentioned evaluation method, it is possible to call attention to a high-risk point and to prevent accidents from occurring. In particular, as a result of being announced by an administrative agency, it can be made known to local residents, and as a result, it is possible to call attention to the residents who frequently use it as a road for living.

On the other hand, according to the present invention of the pedestrian accident risk determination system for each point, various information can be processed based on the above-described evaluation method, and processing using data such as pedestrian collision warning information can be realized. It can be anything. Further, it is possible to perform processing including a huge amount of data such as structure information and vehicle traffic information. The individual information included in the structure information, vehicle traffic information, and pedestrian collision warning information includes primary data acquired from an external information agency as well as uniquely processed secondary data. With respect to the processing, the processing means also makes it possible to easily process the huge amount of data.

It is explanatory drawing which shows embodiment of the pedestrian accident risk determination system classified by point. It is explanatory drawing which shows an example of the output state in an output part (monitor etc.). It is explanatory drawing which shows 1st Embodiment of the pedestrian accident risk determination method classified by point. It is explanatory drawing which shows 2nd Embodiment of the pedestrian accident risk determination method classified by point.

Embodiments of the present invention will be described below with reference to the drawings. For convenience of explanation, the point-by-point pedestrian accident risk determination system will be described, and then the point-based pedestrian accident risk determination method will be described.

<Pedestrian accident risk assessment system by location>
FIG. 1 is a schematic diagram showing an embodiment of a pedestrian accident risk determination system according to location. A pedestrian accident risk degree determination system (hereinafter, sometimes simply referred to as a system) 1 according to the present embodiment includes various databases 11 to 15 and a processing device (processing means) 16. For details of the various databases 11 to 15, the structure information database 11 that stores structure information, the vehicle traffic information database 12 that stores vehicle traffic information, the traffic information of the probe vehicle equipped with the collision warning device (probe) The vehicle traffic information database 13 for storing (car traffic information), the pedestrian collision warning information database 14 for storing pedestrian collision warning information, and the pedestrian accident information database 15 for storing pedestrian accident information. The information stored in these databases 11 to 15 is obtained from the external information 2, and includes the primary data of the external information 2 and the processed secondary data.

As the external information 2, digital road network data (for example, road network data provided by Zenrin Co., Ltd.) 21, national land numerical information urban area land use subdivision mesh data (land use subdivision mesh data and Omitted) 22, traffic performance data (for example, probe data provided by Pioneer Corporation) 23, probe car traffic data 24 and warning data 25 by a probe car equipped with a collision warning device, and vehicles and pedestrians at each point. There is accident data (number of accidents) 26 and the like that have occurred between and.

The structure information stored in the structure information database 11 is, for example, referring to the digital road network data 21, while referring to the number of branches at intersections, the digital road network data 21 and national land numerical information urban area land use subdivision. With reference to the mesh data 22, it is possible to include types such as a high-rise building land use type mesh, a low-rise building land use type mesh, and a low-rise building (dense land) land use type mesh.

The number of branches at the intersection is a condition depending on the number of road connections, and is three at a three-branch intersection and four at a four-branch intersection. The high-rise building land use type mesh is a mesh consisting of commercial/commercial buildings or condominiums with 4 or more floors where buildings are densely populated in residential areas and urban areas. The usage type mesh is a mesh that is a land in which residential buildings of three floors or less are collectively distributed. The low-rise building (dense area) land-use-type mesh is a mesh in which residential buildings with three or less stories are densely packed.

The vehicle traffic information stored in the vehicle traffic information database 12 can be a total value of the traffic for one year that a general vehicle flows into at each intersection. For example, the information of the traffic record data 23 can be processed by map matching with the digital road network data 21. The traffic performance data 23 is, for example, a collection of data of car navigation systems sold by Pioneer Corporation, and can exclusively reflect the running substance of a general vehicle.

The probe car traffic information stored in the probe car traffic information database 13 is for independently collecting the traffic flow of each probe vehicle (probe car) equipped with a collision warning device into each intersection. It measures the volume of traffic to an intersection through unintentional traffic by (a large number of) probe cars. This traffic amount can be a total value during a period in which a plurality (a large number) of probe cars travel for a predetermined period. The total amount can be updated as long as the probe car continues to run. The traffic volume at each intersection is processed by, for example, acquiring the movement state based on the position information by the collision warning device and performing map matching of the aggregated information (probe car traffic data 24) to the digital road network data 21. be able to.

The pedestrian collision warning information stored in the pedestrian collision warning information database 14 is particularly collected in order to obtain the number of times of collision warning detection by the collision warning device mounted on the probe car described above. The information of the probe car warning data 25 obtained by aggregating the detection position and the number of times of detection of the pedestrian collision warning in the probe car while the car is traveling is used. For the matching with the position where the collision warning is detected, by performing map matching with the digital road network data 21, it is possible to integrate the number of times of warning detection in the area around the intersection based on the position of the intersection. The peripheral area may be within a radius of 30 m from the center of the intersection, for example.

The pedestrian accident information stored in the pedestrian accident information database 15 uses the information of the accident data 26, and the position of the pedestrian accident that has occurred in the past several years (for example, 7 years) (intersection and its vicinity) ) And the number of cases. The area around the intersection can be defined as within 15 m in radius from the center of the intersection. If this area is made large, distinction from different intersections in the vicinity becomes insufficient, and if it is made small, it is excluded from accidents around the intersections, so it is preferable to set it within the range of 10 m to 20 m.

As described above, since various information is stored in each of the databases 11 to 15 as secondary data processed in addition to the primary data, by processing these information by the processing device 16, a pedestrian accident The occurrence rate can be calculated. Specifically, the pedestrian collision warning occurrence rate is calculated by dividing the traffic information (traffic volume) by the probe vehicle (probe car) for each point by the pedestrian collision warning information (warning detection frequency), This is classified by the correlation with the structure information and the vehicle traffic information. A statistical model of the number of pedestrian accidents is created by putting together a plurality of classified correlation states. Since this statistical model has multiple types of correlation states, if the above information (classified) at other points can be obtained, it is possible to infer the probability of a pedestrian accident by collating with the relevant statistical model. It becomes.

The statistical model of the number of pedestrian accidents may be a negative binomial regression model with structure information, general vehicle traffic information, pedestrian collision warning information as predictive variables, and pedestrian accident information as result variables. it can. In addition, the risk of pedestrian accidents can be calculated by Bayesian updating using the parameters estimated by the negative binomial regression model as a priori parameter and the number of pedestrian accidents included in the pedestrian accident information as new data. The posterior distribution of (statistical expected value of the number of pedestrian accidents) is calculated, and its maximum posterior probability (MAP) estimated value is an index of potential pedestrian accident risk (final pedestrian accident number statistics Expected value).

In this case, the prior distribution is a gamma distribution (conjugate prior distribution of Poisson distribution), and the Shape parameter (α: φ in the negative binomial distribution) and the Rate parameter (β: φλ ⁻¹ in the negative binomial distribution) are negative. By using the binomial regression model of, the Bayesian update can be performed by an update model formula such as the following formula with these as prior parameters, and the final statistical expected value can be obtained.

As mentioned above, the posterior distribution of the statistical expected value of the number of pedestrian accidents due to Bayesian update is the maximum posterior probability when the above pedestrian accident number statistical model is calculated by the actual pedestrian accident occurrence situation. The (MAP) estimated value is used for calculation of an index of the potential risk of pedestrian accidents (statistical expected value of the final number of pedestrian accidents). Apart from this, when confirming the improvement status due to pedestrian accident prevention measures (prevention work, etc.), the posterior distribution can be calculated by the same Bayes update, which can also be used to verify the effect of the measures. ..

The individual specific information stored in each of the databases 11 to 15 may be stored as secondary data by causing the arithmetic unit 17 of the processing device 16 to process various data 21 to 26 of the external information 2. it can. Further, the processing device 16 is provided with a storage unit 18 for temporarily storing the data read from each of the databases 11 to 15 in the calculation by the calculation unit 17.

By connecting an output unit (monitor) 3 to the system 1, it is possible to display a calculation result or information stored after processing. The output unit 3 may display specific points (intersections) in the information (map information) of the digital road network data 21 and may be displayed in different colors in descending order of pedestrian accident probability after processing. Good. The input unit 4 may be connected to the system 1, and the display information of the output unit 3 may be appropriately changed by inputting a command from the input unit 4.

The output state of the output unit (monitor or the like) 3 may be graphic information mapped to map information as shown in FIG. The circles shown in this figure are color-coded when they are actually displayed.The parts with a high degree of risk are colored red, and the risk is chromatically changed by gradually changing from yellow to green. It is displayed. The size of the circle may be changed in addition to or in addition to the color. In addition, a white small circle “•” is displayed in the circle at the same time to indicate that it is an intersection where a pedestrian accident actually occurred, and a flag is written in the white small circle to indicate the accident. It can also be classified to indicate that the number of occurrences is large.

It should be noted that although intersections based on the digital road network data are displayed in the map information on the output unit (monitor, etc.) 3, no particular display is made for those intersections where the probability of a pedestrian accident is not high. I am supposed to. This is to make an attention map for displaying an intersection having a risk, and it is important to pay attention to an intersection having a higher risk than an intersection having a lower risk. If the local government publishes the map information of the entire area of a specific local government by such a display method, it can be useful for alerting residents to a pedestrian accident.

<Pedestrian accident risk determination method according to location (first embodiment)>
Next, a first embodiment of a pedestrian accident risk level determination method for each point will be described. FIG. 3A conceptually shows the pedestrian accident risk degree determination method (hereinafter, may be simply referred to as a determination method) according to the first embodiment. This determination method can be roughly classified into three blocks. Each block is indicated by a quadrilateral with a chain line in the figure. That is, a pedestrian accident count statistical model estimation (creation) block, a pedestrian collision warning occurrence rate calculation block, and a pedestrian accident occurrence probability calculation (evaluation) block.

The pedestrian accident count statistical model estimation block is a block that creates a pedestrian accident count statistical model based on the general vehicle traffic volume, structure information, pedestrian accident count, and pedestrian collision warning occurrence rate for each point. As shown in the figure, the general vehicle traffic volume uses the general vehicle probe data (the above-mentioned traffic record data 23) and the digital road network (the above-mentioned digital road network data 21), and the structure information is as described above. Numerical information of national land Urban area land use The information obtained by classifying the subdivision mesh data 22 is used. The number of pedestrian accidents uses the accident data 26 described above.

Since the pedestrian collision warning is detected even at intersections where pedestrian accidents have not occurred, it is assumed that the pedestrian collision warning occurrence rate greatly affects the pedestrian accident count statistical model. That is, the pedestrian collision warning occurrence rate is calculated from the traffic volume at each point by the probe car equipped with the collision warning device and the number of times of detection of the warning at each point, and the occurrence rate and other conditions (general vehicle traffic volume, structure The product information and the number of accidents) are associated with each other. Specifically, the negative binomial regression model is applied with the general vehicle traffic volume, the structure information, and the pedestrian collision warning occurrence rate as the prediction variables, and the pedestrian accident count as the reaction variable. With the distribution obtained by applying this negative binomial regression model, the probability of occurrence of pedestrian accidents at points under the same conditions can be obtained.

As shown in Fig. 3(b), the probability of occurrence of a pedestrian accident based on the negative binomial regression model is used as the prior distribution of the predicted values, and the predicted values are ex You may ask for the distribution. The posterior distribution can be obtained, for example, by performing Bayesian update by the method of Bayesian statistics. In this case, the maximum posterior probability (MAP) estimated value from the posterior distribution can be used as the final predicted value of the number of pedestrian accidents at the determination target point.

The prior distribution in this case can be a predicted value based on a statistical model of the number of pedestrian accidents, and the posterior distribution can be a predicted value updated with the actual number of accidents at each evaluation target point. If the posterior distribution is updated with the number of new accidents after the predicted value calculated earlier is confirmed, the predicted value corresponding to the ever-changing traffic situation can be obtained, and measures for accident prevention were taken. In this case, it is possible to verify the effect of that measure.

<Pedestrian accident risk assessment method by location (second embodiment)>
Next, a second embodiment of the pedestrian accident risk level determination method for each point will be described. FIG. 4A conceptually shows the determination method of the second embodiment. This determination method is also the same as in the first embodiment, and includes three blocks: a pedestrian accident number statistical model estimation (creation) block, a pedestrian collision warning occurrence rate calculation block, and a pedestrian accident occurrence probability calculation (evaluation) block. It is divided into blocks.

The part different from the first embodiment is a pedestrian collision warning occurrence rate calculation block. As mentioned above, since the pedestrian collision warning occurrence rate is supposed to have a large effect on the pedestrian accident warning statistical model, it is reflected in the pedestrian accident warning statistical model after correcting this pedestrian collision warning occurrence rate. It was decided.

That is, the first statistical model (collision warning statistical model) is created by using the traffic volume and the number of times of pedestrian collision warning detection by the probe car for each point, and the collision indicated by the collision warning statistical model at each point at the intersection. A model deviation residual is calculated from the warning detection tendency and the actual collision warning detection tendency at each point, and is used in place of the pedestrian collision warning occurrence rate in the first embodiment. .. Therefore, the model deviation residual will be reflected in the pedestrian accident count statistical model, and a pedestrian accident statistical model will be created based on this model deviation residual based on the correlation with the accident count in the pedestrian accident information. Is.

When calculating the accident probability based on the average relationship between the traffic volume and the number of collision warning detections, the increase in traffic volume and the increase in the number of collision warning detections will change linearly. Since the non-linearity is used when the residual is used, it also has the effect of improving the suitability of probability calculation.

Also in this embodiment, the final predicted value of the pedestrian accident rate may be calculated by Bayes updating the prior distribution as shown in FIG. 4B to the posterior distribution.

Specific examples of calculating the predicted value of the pedestrian accident occurrence rate according to the above-described embodiment will be described below.

<Example 1>
The pedestrian's accident risk level at each point was evaluated using actual information. The evaluation target is 5462 residential road intersections in Toyohashi City, Aichi Prefecture. Digital road network data uses road network data provided by Zenrin Co., Ltd., and land use subdivision mesh data is used by the Ministry of Land, Infrastructure, Transport and Tourism. Numerical information on national land provided Urban area land use subdivided mesh data was used, and for actual traffic data of general vehicles, probe data provided by Pioneer Corporation was used.

As the collision warning device, Mobile Eye Vision Technologies Limited Mobile Eye (registered trademark) is used, and when a collision occurs within 2.0 seconds against a person recognized when traveling at a speed of about 7 km/h or more. If it is judged, a collision warning should be given. A vehicle of a business operator (taxi company) was used as a probe vehicle (probe car), and 39 of the vehicles were equipped with the collision warning device, which was obtained during one year from September 1, 2017 to August 31, 2018. Used information. Since the position information of the probe car can be grasped by the position information of the collision warning device, as described later, based on the position information of the collision warning device, by performing map matching with the digital road network data, the probe You can calculate the traffic of cars.

The structure information, vehicle traffic information, probe car information, and pedestrian collision warning information are classified or specified as follows. As the structure information, first, information on the number of road connections was specified from the road network data. Specifically, the information is three at a three-branch intersection and four at a four-branch intersection. Second, based on the national land numerical information urban area land use subdivision mesh data, the high-rise building land use type mesh, the low-rise building land use type mesh, and the low-rise building (dense land) land use type mesh were specified as the surrounding environment.

The vehicle traffic information is based on probe data (general vehicle data) provided by Pioneer Corporation, and is a value obtained by summing up one year's worth of traffic that general vehicles flow into at each intersection. Specifically, the position data (point data) indicating the position of each vehicle every 3 seconds is map-matched with the digital road network data. More specifically, by dividing the vehicle position data (point data) for each trip and calculating the closest road link (neighboring link) from the position data (point data), the link cost of the neighboring link is reduced to 1/10. It was decided to reduce to. In addition, the shortest route search from the first departure point to the last arrival point of each trip divided by trip is performed by the Dijkstra method, and the neighboring links limited to the road links on the extracted shortest route are calculated again. did. The link means a road section between the nodes, and the node means a node such as an intersection (including other points on the road network expression).

The probe car information is data relating to the traffic volume of the probe car (probe vehicle), and this traffic volume data is based on the position information of the collision warning device and uses the latitude/longitude data of the probe car about every 10 seconds, The information obtained during one year (September 1, 2017 to August 31, 2018) was calculated by map matching with the digital road network data, and the information was obtained by summing them. In this case, the traffic volume is calculated based on the position data (point data) obtained by the probe car by the same method as the above-described traffic volume calculation method for a general vehicle.

The pedestrian collision warning information is the information that the collision warning device mounted on the probe car (probe vehicle) is activated, that is, the mobile eye (registered trademark of the same company) is used, and the speed is about 7 km/h or more. The collision warning given to the person recognized during traveling is judged in a situation where it is judged that the collision will occur within 2.0 seconds. The location where the collision warning device operates can be known from the position information of the collision warning device. Therefore, the number of times this collision warning device was activated was counted and used as the number of pedestrian collision warnings at each intersection. In addition, since it is counted as a collision warning related to the intersection, the number of cases at the warning intersection is to be counted within the radius of 30 m from the center of the intersection.

Here, the pedestrian collision warning occurrence rate was calculated based on the information shown above. That is, the pedestrian collision warning occurrence rate is the number of pedestrian collision warning occurrences per traffic of the probe car, and can be calculated by dividing the number of pedestrian collision warning occurrences at each intersection by the traffic of the probe car. It is possible.

The pedestrian accident information was calculated based on the data of pedestrian accidents that actually occurred in the past. Specifically, in order to limit to pedestrian accidents related to intersections, seven years from 2008 to 2015 (excluding 2012) that occurred within a radius of 15 m from the center of each intersection were totaled, and each intersection was collected. The number of pedestrian accidents was defined as

A statistical model of the number of pedestrian accidents was created using the above information. As a statistical model of the number of pedestrian accidents, a negative binomial regression model was applied with structure information, general vehicle traffic information, pedestrian collision warning information as prediction variables, and pedestrian accident information as result variables. Table 1 shows the coefficient parameter and the variance parameter of the predictor variable estimated by the negative binomial regression model.

The estimated values of the coefficient parameters and variance parameters of the predictor variables are as shown in Table 1. However, the evaluation of the calculation results shows that the pedestrian collision warning occurrence rate is 3.31, and thus the negative binomial According to the characteristics of the regression model, if the pedestrian collision warning occurrence rate decreases by 10%, the statistical expected value of the number of accidents (the risk of accident occurrence) is 1.0-exp ^{(-1 x 3.31)} = It has been shown to decrease by 28.18 (about 30)%.

Here, based on the above statistical model, the conditions that belong to the top of the distribution (meaning that the pedestrian collision warning occurrence rate is highest) 1% or less are the highest risk, and the conditions that belong to the range of 1% to 2% High risk, conditions belonging to the range of 2% to 5% are medium risk, conditions belonging to the range of 5% to 10% are low risk, conditions belonging to the range of 10% to 20% are general risk, and Other than that was classified as no risk. By comparing each condition with each condition for each intersection, the risk level for each intersection is evaluated (determined) and mapped on the map data, so that the points and the risk levels can be visualized as illustrated in FIG. Can be grasped in.

Further, in this embodiment, in order to grasp the potential pedestrian accident risk at each intersection, the posterior distribution is calculated by the Bayesian statistical method (bayes update), and the maximum posterior probability (MAP) estimated value is obtained. I decided. Specifically, the distribution of occurrence probability (statistical expected value) of the number of pedestrian accidents by the negative binomial regression model is set as a prior distribution, and this prior distribution is set as a gamma distribution (conjugate prior distribution of Poisson distribution). The prior Shape parameter (α: φ in the negative binomial distribution) and the prior Rate parameter (β: φ λ ⁻¹ in the negative binomial distribution) corresponding to the intersection were calculated using the negative binomial regression model, respectively. Further, Bayes update is performed by the update model formula described in the above-mentioned Equation 1 using these prior parameters and the actual number of pedestrian accidents at each intersection as new data, and the posterior distribution of the statistical expected value of the number of pedestrian accidents is calculated. Then, the maximum posterior probability (MAP) estimated value was used as the index of the potential risk of pedestrian accidents (statistical expected value of the final number of pedestrian accidents). In the posterior distribution in this case, 1% or less of the top (meaning the highest incidence of pedestrian collision warning), 1% to 2%, 2% to 5%, 5% to 10%, 10% to 20% It was possible to evaluate the degree of danger of each intersection in a range other than that range and to make it visually recognizable in the state illustrated in FIG.

Although creating the prior distribution is not essential, by updating the actual number of pedestrian accidents as new data by Bayesian updating, it is possible to reflect the actual accident occurrence status in the statistical values. Then, of the intersections under the same condition, the risk level of the place where the pedestrian accident actually occurs is updated to a high evaluation, and secondly, the actual pedestrian accident occurs at the intersection under the same condition. It is possible to assess the potential risk of accidents in places that are not.

<Example 2>
In addition, as a second example, as in the above case, the degree of accident risk of a pedestrian at each point was evaluated using actual information. The evaluation target is 5463 points of living road intersections in Toyohashi City, Aichi Prefecture (one more point than in the case of Example 1), and each of the digital road network data, land use subdivision mesh data, and general vehicle traffic record data The same as in Example 1. In addition, the pedestrian warning device was also the same, and 39 probe cars (probe vehicles) were also selected. Various kinds of information (structure information, vehicle traffic information, probe car information, pedestrian collision warning information, and pedestrian accident information) used for processing are also the same as in the first embodiment.

In this embodiment, a collision warning statistical model is created in place of the pedestrian collision warning occurrence probability index, and the model deviation residual is used as the index. Specifically, the traffic volume of the probe car at each intersection was used as a predictive variable, and the number of pedestrian collision warnings at each intersection was used as a reaction variable to create a statistical model applied to the coefficient data. More specifically, a negative binomial regression model was applied to the above-mentioned prediction coefficient and reaction coefficient to estimate the coefficient parameter and variance parameter of the prediction variable. Then, for the number of pedestrian collision warnings at each intersection, the model deviation residual from the estimated negative binomial regression model was calculated.

A statistical model of the number of pedestrian accidents was created including the model deviation residuals calculated in this way. For the statistical model of the number of pedestrian accidents, a negative binomial regression model was applied with the structure information, the traffic information of general vehicles and the above model deviation residual as the predictive variables and the pedestrian accident information as the result variable. Table 2 shows coefficient parameters and variance parameters of predictor variables estimated by the negative binomial regression model.

Estimated values such as coefficient parameters and variance parameters of predictor variables are as shown in Table 2, but the evaluation of the calculation results is as long as it is compared with AIC (Akaike's Information Criterion) of Example 1. Since the AIC of this embodiment is smaller, it can be said that the compatibility is preferable. This AIC is an index showing the suitability of the statistical model, and the smaller the number, the better the model. Such a difference is that the pedestrian collision warning occurrence probability of the first embodiment is an average value (linear) of the number of pedestrian collision warning cases with respect to the traffic of the probe car, but by using the model deviation residual, It is considered that the relationship between the traffic volume of the probe car and the number of pedestrian collision warnings is non-linear.

It should be noted that points with high model deviation residuals mean points where a large number of pedestrian collision warnings occur relative to the average relationship between the traffic volume of the probe car and the number of pedestrian collision warnings. Since it is considered that the number of pedestrians in a situation close to an accident is large, it is possible to improve the suitability by using the model deviation residual in the pedestrian accident statistical model. Become.

The evaluation after creating the statistical model of the number of pedestrian accidents is the same as in the case of the first embodiment, and is 1% or less of the top of the distribution (meaning that the pedestrian collision warning occurrence rate is highest). % To 2%, 2% to 5%, 5% to 10%, 10% to 20%, and other ranges are evaluated as the risk level of each intersection, and visually grasped in the state illustrated in FIG. I was able to do so. The Bayesian update also made it possible to evaluate the above risk based on the posterior distribution.

The embodiments and examples according to the present invention are as described above, but these are examples of the present invention and the present invention is not limited to the above-described embodiments and the like. Therefore, it is possible to appropriately change each requirement according to the above-described embodiment or add other elements. For example, the collision warning device can be from another manufacturer, and the number of samples can be increased by more probe cars. Further, when outputting to the output unit, the highest risk level to the general risk level is 1% or less, 1% to 2%, 2% to 5%, 5% to 10%, and 10% to 20% of the top of the distribution. Although the above is classified and the others are classified as having no risk, this range may be appropriately changed and a range considered to be appropriate may be set for each region.

1 Pedestrian accident risk level determination system 2 External information 3 Output unit 4 Input unit 11 Structure information database 12 Vehicle traffic information database 13 Probe car traffic information database 14 Pedestrian collision warning information database 15 Pedestrian accident information database 16 Processing device (processing means)
17 Calculation Unit 18 Storage Unit 21 Digital Road Network Data 22 Land Use Subdivision Mesh Data 23 Passing Track Data 24 Probe Car Passing Data 25 Probe Car Warning Data 26 Accident Data

Claims (8)

A plurality of intersections are selected in advance, and probe vehicle traffic information including the traffic volume at which a probe vehicle equipped with a collision warning device enters the intersection, and the collision warning device mounted on the probe vehicle centers the intersection. Acquiring pedestrian collision warning information including the number of times a collision warning with a pedestrian in the vicinity was detected, and pedestrian accident information including the number of accidents that the pedestrian actually encountered,
Calculating the ratio of the number of times the collision warning included in the pedestrian collision warning information is detected as the pedestrian collision warning occurrence rate, with respect to the traffic of the probe vehicle included in the probe vehicle traffic information.
About the pedestrian collision warning occurrence rate, a pedestrian accident warning statistical model is created based on the correlation with the number of accidents in the pedestrian accident information, and the pedestrian collision warning occurrence rate at the evaluation target point is the pedestrian accident warning statistical model. A pedestrian accident risk assessment method for each point, characterized in that the probability of occurrence of a pedestrian accident is calculated by collating with. The pedestrian accident risk evaluation method according to claim 1, wherein for each of the intersections, structure information including the number of branches and high or low rises or density of buildings around the intersection, and the intersection. Further, either or both of the general vehicle traffic information including the traffic volume of the general vehicle approaching is acquired,
About each element included in the structure information or the general vehicle traffic information or both and the pedestrian collision warning occurrence rate, a pedestrian accident number statistical model is created based on the correlation with the number of accidents in the pedestrian accident information. However, each element included in the structure information or the general vehicle traffic information or both of them at the evaluation target point and the pedestrian collision warning occurrence rate are collated with the pedestrian accident count statistical model to thereby detect a pedestrian accident. A pedestrian accident risk assessment method for each location, which is characterized by calculating the occurrence probability of In the pedestrian accident risk evaluation method according to claim 1 or 2,
An average collision based on the ratio of the number of times the collision warning included in the pedestrian collision warning information is detected to the traffic of the probe vehicle included in the probe vehicle traffic information at all preselected intersections. Create a collision warning statistical model that shows warning detection trends,
From the collision warning detection tendency indicated by the collision warning statistical model at each intersection point, and from the actual collision warning detection tendency at each point, a model deviation residual is calculated,
The pedestrian warning risk rate is corrected by using the model deviation residual, and the pedestrian accident risk evaluation method according to point is characterized. The prediction of the occurrence of the pedestrian accident is
Preliminary predicted value is derived by matching each element and pedestrian collision warning occurrence rate included in the structure information and vehicle traffic information at the evaluation target point with the pedestrian accident count statistical model, and at the evaluation target point Derivation of the posterior predictive value by correcting the prior predictive value based on the number of accidents and the number of accidents pedestrians encountered after the fact,
The pedestrian accident risk evaluation method according to any one of claims 1 to 3, wherein one of the posterior predicted values having a high accident occurrence probability is used as a final predicted value. The collision warning device includes an image processing means capable of recognizing a person, a speed recognizing means, a determination processing means for determining that a vehicle traveling at a predetermined speed or more collides with a person within a predetermined time, and the determination processing means. And output means for outputting an alarm when it is determined that a collision occurs,
The number of times of collision warning detection included in the pedestrian collision warning information is based on the number of times of warnings output by the output means. Method. A plurality of intersections are selected in advance, and a probe vehicle traffic information database that stores probe vehicle traffic information including traffic volume at which a probe vehicle equipped with a collision warning device enters the intersection, and a collision mounted on the probe vehicle. A pedestrian collision warning information database that stores pedestrian collision warning information including the number of times a collision warning with a pedestrian around the intersection around the intersection is detected by an alarm device, and the number of accidents actually encountered by pedestrians A pedestrian accident information database that stores pedestrian accident information, and a processing unit that processes the information stored in each of the databases to calculate a pedestrian accident occurrence prediction,
The processing means is
Calculating the ratio of the number of times the collision warning included in the pedestrian collision warning information is detected as the pedestrian collision warning occurrence rate, with respect to the traffic of the probe vehicle included in the probe vehicle traffic information.
About the pedestrian collision warning occurrence rate, a pedestrian accident warning statistical model is created based on the correlation with the number of accidents in the pedestrian accident information, and the pedestrian collision warning occurrence rate at the evaluation target point is the pedestrian accident warning statistical model. A pedestrian accident risk judgment system according to location, which is characterized in that the probability of occurrence of a pedestrian accident is calculated by collating with. A plurality of intersections are selected in advance, and a probe vehicle traffic information database that stores probe vehicle traffic information including traffic volume at which a probe vehicle equipped with a collision warning device enters the intersection, and a collision mounted on the probe vehicle. A pedestrian collision warning information database that stores pedestrian collision warning information including the number of times a collision warning with a pedestrian around the intersection around the intersection is detected by an alarm device, and the number of accidents actually encountered by pedestrians A pedestrian accident information database that stores pedestrian accident information, and a processing unit that processes the information stored in each of the databases to calculate a pedestrian accident occurrence prediction,
The processing means is
An average collision based on the ratio of the number of times the collision warning included in the pedestrian collision warning information is detected to the traffic of the probe vehicle included in the probe vehicle traffic information at all preselected intersections. Create a collision warning statistical model that shows warning detection trends,
From the collision warning detection tendency indicated by the collision warning statistical model at each intersection point, and from the actual collision warning detection tendency at each point, a model deviation residual is calculated,
Regarding the model deviation residual, create a statistical model of pedestrian accidents based on the correlation with the number of accidents in the pedestrian accident information, and collate the model deviation residual at the evaluation target point with the statistical model of pedestrian accidents. A pedestrian accident risk judgment system for each point, which is characterized by calculating the probability of occurrence of a pedestrian accident. For each intersection, a structure information database that stores structure information including the number of branches and high-rise or low-rise or denseness of buildings in the vicinity of the intersection, and a vehicle including the amount of traffic that a vehicle enters at the intersection It further comprises either one or both of a vehicle traffic information database that stores traffic information,
The statistical model of the number of pedestrian accidents in the processing means is created based on the correlation with the number of accidents in the pedestrian accident information together with each element included in the structure information or the general vehicle traffic information or both of them. And
The calculation of the probability of occurrence of the pedestrian accident is by collating with the pedestrian accident number statistical model together with each element included in the structure information or the general vehicle traffic information or both at the evaluation target point. A certain pedestrian accident risk determination system according to claim 6 or 7.