Risk prediction device and risk prediction method

Abstract

To generate a highly accurate and versatile risk map taking account of passing states of a pedestrian and the like predicted to exist at a blind spot by time zone.SOLUTION: A risk prediction device is configured to: acquire blind spot information which is information on an area being a blind spot when viewed from a driver of a vehicle, and passage information which is information on an area where at least one of a pedestrian and a bicycle has passed, by time zone in a predetermined area; set a first risk level indicating a degree of risk to the driver to each position within the predetermined area by time zone based on the acquired blind spot information, and set a second risk level indicating a degree of risk for the driver to each position within the predetermined area by time zone based on the acquired passage information; and generate a risk map indicating a third risk level at each position in the predetermined area for each time zone by superimposing the first risk level and the second risk level set to each position within the predetermined area.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present disclosure relates to a risk prediction device and a risk prediction method. [Background technology]

[0002] Conventionally, a device has been proposed in which, when pedestrian movement trajectory data is acquired from an external source by a trajectory acquisition means, cautionary point prediction means predicts pedestrian-to-pedestrian accident warning points on a map stored in a map storage means based on the acquired movement trajectory data (Patent Document 1).

[0003] In addition, a system has been proposed that includes a determination unit that determines the range of the blind spot of a subject passing through a target area, an estimation unit that estimates the behavior of pedestrians passing through the target area, and an output unit that outputs blind spot information that indicates the behavior of pedestrians estimated by the estimation unit within the range of the blind spot determined by the determination unit (Patent Document 2). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2006-330822 [Patent Document 2] Japanese Patent Publication No. 2020-091613 Summary of the Invention [Problem to be solved by the invention]

[0005] However, the technology described in Patent Document 1 focuses only on pedestrian movement information, and does not use information on the environment in which pedestrians and vehicles pass when making predictions, so there is room for improvement in the accuracy of the prediction information.

[0006] In addition, the technology described in Patent Document 2 determines the range of blind spots as information on the environment in which the target person passes and estimates the behavior of passersby, but there is a problem in that if a passerby in the blind spot cannot be recognized, information indicating the behavior of passersby cannot be provided.

[0007] The present disclosure has been made in consideration of the above points, and aims to provide a risk prediction device that can generate a highly accurate risk map that takes into account the traffic conditions by time of day for pedestrians and other people who are predicted to be in blind spots. [Means for solving the problem]

[0008] A first aspect of the risk prediction device includes an acquisition unit that acquires blind spot information, which is information on areas that are blind spots from the perspective of a vehicle driver in a specified area, and traffic information, which is information on areas where at least one of pedestrians and bicycles have passed, by time period; a setting unit that sets a first risk level indicating the degree of risk to the driver for each position in the specified area by time period based on the acquired blind spot information, and sets a second risk level indicating the degree of risk to the driver for each position in the specified area by time period based on the acquired traffic information; and a generation unit that superimposes the first risk level and the second risk level set for each position in the specified area to generate a risk map for each time period that indicates a third risk level for each position in the specified area.

[0009] The second aspect is a risk prediction method in which a computer executes the following processing: blind spot information, which is information on areas that are blind spots from the perspective of a vehicle driver, and traffic information, which is information on areas where at least one of pedestrians and cyclists have passed, in a predetermined area, by time period; based on the acquired blind spot information, set a first risk level indicating the degree of risk to the driver for each position in the predetermined area, by time period; and based on the acquired traffic information, set a second risk level, which indicates the degree of risk to the driver for each position in the predetermined area, by time period; and superimpose the first risk level and the second risk level set for each position in the predetermined area to generate a risk map for each time period that indicates a third risk level for each position in the predetermined area. [Effects of the Invention]

[0010] According to the present disclosure, it is possible to generate a highly accurate risk map that takes into account the traffic conditions by time of day, such as pedestrians who are predicted to be in blind spots. [Brief explanation of the drawings]

[0011] [Figure 1] 1 is an explanatory diagram showing the overall configuration of a risk prediction device and peripheral devices according to an embodiment of the present invention; [Figure 2] 1 is a block diagram showing a hardware configuration of a risk prediction device according to an embodiment of the present invention. [Figure 3] 1 is a block diagram showing a functional configuration of a risk prediction device according to an embodiment of the present invention; [Figure 4] FIG. 2 is an explanatory diagram showing an area that can be recognized by an infrastructure camera. [Figure 5] FIG. 10 is an explanatory diagram showing sensing by a moving object. [Figure 6] FIG. 1 is an explanatory diagram showing the route of pedestrians and the location of vehicles. [Figure 7] FIG. 2 is an explanatory diagram showing a blind spot area as seen by a driver of a vehicle. [Figure 8] FIG. 2 is an explanatory diagram showing an area where pedestrians and the like pass through. [Figure 9] This is an explanatory diagram showing the routes taken by pedestrians and the location of vehicles between 7:00 AM and 8:00 AM, as well as a micro-risk map. [Figure 10] This is an explanatory diagram showing the routes taken by pedestrians and the location of vehicles between 10:00 AM and 11:00 AM, as well as a micro-risk map. [Figure 11] This is an explanatory diagram showing the routes taken by pedestrians and the location of vehicles between 3:00 PM and 4:00 PM, as well as a micro-risk map. [Figure 12] This is an explanatory diagram showing the routes taken by pedestrians and bicycles, the location of vehicles, and a micro-risk map from 6:00 PM to 8:00 PM. [Figure 13] FIG. 10 is an explanatory diagram showing a three-dimensional risk map. [Figure 14] FIG. 10 is an explanatory diagram showing the road network of an area with links and nodes, and showing the fourth risk level of each link. [Figure 15] FIG. 10 is an explanatory diagram showing a fourth risk level for each route from a start point to an end point. [Figure 16] FIG. 1 is an explanatory diagram showing real-time recognition by an infrastructure camera, a micro risk map from 3:00 PM to 4:00 PM, and a one-minute prediction of the information recognized in real time. [Figure 17] 10 is a flowchart showing the flow of a risk map generation process according to the present embodiment. [Figure 18] 10 is a flowchart showing a flow of a risk prediction process according to the present embodiment. [Figure 19] 10 is a flowchart showing the flow of a search process according to the present embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0012] An example of an embodiment of the present disclosure will be described below with reference to the drawings. The same or equivalent components and parts in each drawing are designated by the same reference numerals. The dimensional proportions in the drawings are exaggerated for the sake of explanation and may differ from the actual proportions.

[0013] Fig. 1 is a diagram showing the overall configuration of a risk prediction device 10 and peripheral devices according to this embodiment. The relationship between the risk prediction device 10 and peripheral devices will be described below using Fig. 1. Note that the number of each device is not limited to the example in Fig. 1, and there may be multiple devices.

[0014] The risk prediction device 10 according to this embodiment is a computer such as an on-premise server, a cloud server, a notebook computer, a desktop computer, etc. In the example of Fig. 1, the risk prediction device 10, an infrastructure camera 70, a mobile terminal 80, a sensing mobile object 90, and a vehicle 110 are communicably connected to each other via a communication line N.

[0015] The infrastructure camera 70 is a video camera installed in a predetermined area. The infrastructure camera 70 has a shooting function for shooting the surroundings, a processing function for performing video and audio recognition on the captured video, and a communication function. For example, multiple infrastructure cameras 70 are installed in locations with a high volume of traffic participants and capture video of the surroundings. Note that traffic participants include, for example, pedestrians, bicycles, motorbikes, kick scooters, vehicles 110, and autonomous robots present in the predetermined area. The infrastructure camera 70 uses its processing function to recognize information related to the passage of at least one of pedestrians and bicycles (hereinafter referred to as "pedestrians, etc.") from the captured video, including position information, speed vectors, attribute information (details of which will be described later), etc. Specifically, the infrastructure camera 70 recognizes the position information of pedestrians, etc. based on the position of the pedestrians, etc. in the image, the installation position, installation angle, and angle of view of the infrastructure camera 70, etc. Furthermore, the infrastructure camera 70 recognizes the speed vector from the difference in position information of pedestrians, etc. between multiple frames. In addition, the infrastructure camera 70 recognizes attribute information such as the classification of pedestrians or bicycles, and the gender, age, and bone structure of people in the image, by using a trained machine learning model, etc.

[0016] In this embodiment, traffic participants whose traffic information is to be recognized are pedestrians and the like. However, the target may include other traffic participants who may cross the road in addition to pedestrians and bicycles. The infrastructure camera 70 adds time information about the time the information was acquired to the recognized traffic information and transmits it to the risk prediction device 10 via the communication line N. The transmitted traffic information is stored in a storage unit 20 of the risk prediction device 10, which will be described later. Note that the processing function may be provided on the risk prediction device 10 side, rather than on the infrastructure camera 70 side. Furthermore, in addition to the infrastructure camera 70, a sensor that detects intrusions using far-infrared rays may be installed within the area, and traffic information may be recognized based on information detected by the sensor, and the recognized traffic information may be transmitted to the risk prediction device 10 and stored in the storage unit 20.

[0017] The mobile terminal 80 is an electronic device carried by pedestrians and riders of bicycles passing through a predetermined area, and is a portable gadget such as a smartphone, smart watch, tablet terminal, smart glasses, etc. The mobile terminal 80 receives a notification from a notification unit 22 of the danger prediction device 10 (described later) via a communication line N.

[0018] The sensing mobile object 90 may be a vehicle, bicycle, motorcycle, kick scooter, autonomous robot, gadget worn by a pedestrian, drone, or other object that travels within a specified area and can sense while moving. Note that autonomous robots include, for example, delivery robots and cleaning robots, as well as service robots that are expected to be used in cities in the future. Hereinafter, the sensing mobile object 90 will be referred to as the "mobile object 90." The mobile object 90 is equipped with a sensor and senses its surroundings while traveling within a specified area. Examples of sensors include a 3D depth sensor, a 2D camera, and a distance sensor. A 3D depth sensor detects the depth of an object by continuously irradiating an infrared pattern and analyzing the infrared pattern from infrared images continuously captured by an infrared camera. A 2D camera is an example of an image sensor. A 2D camera captures images using visible light and generates visible light video information. A distance sensor detects the distance to an object by irradiating, for example, a laser or ultrasound.

[0019] The sensor also estimates the self-position of the moving object 90 based on the information obtained by sensing. The self-position of the moving object 90 may be determined using information measured by a GPS (Global Positioning System) or the like. The sensor also recognizes relative position information of pedestrians, etc. within a predetermined area based on the information obtained by sensing, and recognizes the position information and movement speed of pedestrians, etc. within the predetermined area based on the relative position information of the recognized pedestrians, etc. and the self-position of the moving object 90. The sensor also recognizes attribute information of pedestrians, etc. based on the information obtained by sensing. As a result, information related to passage is recognized, similar to the case of the infrastructure camera 70 described above.

[0020] Furthermore, the sensor recognizes environmental information such as buildings and parked vehicles around the mobile object 90 based on the information obtained by sensing. If the mobile object 90 is a vehicle, the sensor is equipped with a camera that captures images in front of the vehicle, and recognizes environmental information that can acquire blind spot areas (described below). The sensor attaches time information when the information was acquired to each of the recognized information related to passage and environmental information, and transmits the information to the risk prediction device 10 via communication line N. The transmitted information related to passage and environmental information is stored in a storage unit 20 of the risk prediction device 10 (described below).

[0021] Vehicle 110 represents a vehicle that travels in a predetermined area, such as a regular automobile, an autonomous vehicle, or a motorcycle, and travels at a speed faster than that of a pedestrian. Vehicle 110 receives a notification from a notification unit 22 of risk prediction device 10, which will be described later, via a communication line N from risk prediction device 10. Vehicle 110 is also controlled by a control unit 24 of risk prediction device 10, which will be described later, via the communication line N from risk prediction device 10.

[0022] Fig. 2 is a block diagram showing the hardware configuration of the risk prediction device 10. As shown in Fig. 2, the risk prediction device 10 has a CPU (Central Processing Unit) 51, a ROM (Read Only Memory) 52, a RAM (Random Access Memory) 53, an input device 54, an output device 55, a storage medium reader 56, and a communication I / F 57. Each component is connected to each other via a bus 58 so as to be able to communicate with each other.

[0023] The CPU 51 is a central processing unit that executes various programs and controls each component. That is, the CPU 51 reads programs from the ROM 52 and executes the programs using the RAM 53 as a work area. The CPU 51 controls each component and performs various arithmetic processing in accordance with the programs stored in the ROM 52.

[0024] The storage device configured by the ROM 52 stores various programs including an operating system and various data. The ROM 52 stores a processing program 11 for executing a generation process, a risk prediction process, and a search process, which will be described later.

[0025] The memory configured by the RAM 53 temporarily stores programs and data as a working area.

[0026] The input device 54 includes a pointing device such as a mouse and a keyboard, and is used to make various inputs.

[0027] The output device 55 is, for example, a liquid crystal display, and is a device for outputting various types of information. The output device 55 may employ a touch panel system and function as the input device 54. The output device 55 may also include a speaker or the like as an audio output means.

[0028] The storage medium reader 56 reads data stored in various storage media such as CD (Compact Disc)-ROM, DVD (Digital Versatile Disc)-ROM, Blu-ray Disc, and USB (Universal Serial Bus) memory, and writes data to the storage media.

[0029] The communication I / F 57 is an interface for communicating with the infrastructure camera 70, the mobile terminal 80, the sensing mobile body 90, and the vehicle 110, which are external to the danger prediction device 10. For this communication, a wireless communication standard such as 4G, 5G, or Wi-Fi (registered trademark) is used.

[0030] Next, the functional configuration of the risk prediction device 10 according to this embodiment will be described with reference to Fig. 3. As shown in Fig. 3, the risk prediction device 10 according to this embodiment includes an acquisition unit 12, a setting unit 14, a generation unit 16, a prediction unit 18, a notification unit 22, a control unit 24, and a search unit 26. A storage unit 20 is also provided in a predetermined storage area of ​​the ROM 52 or the RAM 53. The CPU 51 executes a processing program 11 stored in the ROM 52, thereby functioning as the acquisition unit 12, the setting unit 14, the generation unit 16, the prediction unit 18, the notification unit 22, the control unit 24, and the search unit 26.

[0031] The acquisition unit 12 acquires information related to passage recognized and transmitted by the infrastructure camera 70 at each time, and stores the information in the storage unit 20. The acquisition unit 12 also acquires information related to passage recognized and transmitted by a sensor mounted on the mobile object 90 at each time and environmental information, and stores the information in the storage unit 20. The storage unit 20 is configured with a database or the like, and may be built inside the risk prediction device 10, or may be built externally and accessed by the risk prediction device 10 via a network.

[0032] FIG. 4 illustrates an infrastructure camera management area (the shaded area in FIG. 4 ) that can be recognized by installed infrastructure cameras 71-78 in a predetermined area, and an area (the diagonally shaded area in FIG. 4 ) in which the infrastructure cameras 71-78 cannot observe the behavior and situation of pedestrians and other people, i.e., information related to traffic. Note that infrastructure camera 70 is a general term for infrastructure cameras 71-78. In the example of FIG. 4 , infrastructure cameras 70 are installed at major intersections 61A and 61B, but infrastructure cameras 70 are not installed on the central road 62 between intersections 61A and 61B or on a residential road 63 branching off from the central road 62. As an example of a predetermined area, the area illustrated in FIG. 4 has an elementary school 100 in the upper left of the page, a dry cleaning store 101 above the central road 62, a residence 102 across from it, a cram school 103 diagonally across the central road 62 from the residence 102, and a convenience store 104 across from that. The following explanation will be given using the area in FIG. 4 . Within the area described above, the infrastructure camera 70 takes photographs at multiple locations at different times for a predetermined period of time, and thus the information relating to passage that is temporally comprehensive for the predetermined period of time is stored in the storage unit 20.

[0033] FIG. 5 is a diagram showing how the behavior and situation of pedestrians and the like within the area, i.e., information related to traffic, and environmental information around the mobile object 90 are recognized through sensing by a mobile object 90 equipped with various sensors in the area of ​​FIG. 4 . As described above, the mobile object 90 performs sensing using the on-board sensors while moving on the road, and the sensors recognize information related to traffic of pedestrians and the like within the area and environmental information. Here, in FIG. 4 , there are areas where the behavior and situation of pedestrians and the like cannot be observed by the infrastructure camera 70, but as shown in FIG. 5 , sensing by the mobile object 90 makes it possible to recognize information related to traffic throughout the entire area of ​​FIG. 4 . In this way, the mobile object 90 performs sensing for a predetermined period at each time while moving through the area, and thus, temporally and spatially comprehensive traffic-related information and environmental information are stored in the storage unit 20.

[0034] The acquisition unit 12 acquires traffic information and environmental information for a predetermined period of time from the storage unit 20. Based on the acquired information, the acquisition unit 12 acquires traffic information and blind spot information by time period. Specifically, the acquisition unit 12 acquires traffic information by time period based on the acquired traffic information. More specifically, the acquisition unit 12 identifies the trajectories of pedestrians and the like within an area from the position information of pedestrians and the like at each time, and acquires traffic information by accumulating the trajectories for each time period. The acquisition unit 12 identifies, for example, an area that includes a collection of trajectories included in each time period, an area where trajectories are accumulated at a predetermined frequency or more, as a traffic area, and acquires the range (position information) of the traffic area within the area as traffic information. In addition, the acquisition unit 12 assigns attribute information of pedestrians and the like included in the original traffic information to the traffic information. Furthermore, the acquisition unit 12 identifies the number of trajectories accumulated in identifying the traffic area as the number of pedestrians or the like passing through the traffic area, and also assigns the identified number of people to the traffic information as attribute information.

[0035] Specifically, the acquisition unit 12 statistically processes the acquired environmental information by time period based on the time information attached to the information to acquire blind spot information. More specifically, the acquisition unit 12 uses environmental information recognized by a sensor mounted on a vehicle, which is an example of a moving object 90, to identify blind spot areas, which are areas that are blind spots from the vehicle's perspective. A blind spot is a position that may be invisible at a specific time and in a specific location, such as a blind spot caused by a stationary object such as a building or sign, a blind spot caused by a regularly parked vehicle, or a blind spot caused by a temporarily stopped vehicle. For example, the acquisition unit 12 accumulates positions that are blind spots at each time within an area by time period based on the environmental information recognized at each time, and identifies a frequency distribution in which each position within the area is a blind spot. In the identified frequency distribution, the acquisition unit 12 identifies a range of a predetermined variance value centered on a peak value such as the mean value or median as a blind spot area, and acquires the range (position information) of the blind spot area within the area as blind spot information. Note that the size of each blind spot area may vary.

[0036] The acquisition unit 12 stores the acquired blind spot information and traffic information by time period in the storage unit 20. Furthermore, the acquisition unit 12 updates the blind spot information and traffic information stored in the storage unit 20 at a predetermined timing based on the traffic-related information and environmental information that are successively acquired.

[0037] In the above description, the case where blind spot information and traffic information are acquired by statistically processing the information related to past traffic and environmental information acquired by the acquisition unit 12 has been described, but the present invention is not limited to this. For example, when it is desired to extract only the essence of information when the amount of information increases, the acquisition unit 12 may input the information related to traffic and environmental information into a machine learning model such as a CNN (Convolutional Neural Network) to extract feature quantities of each of the blind spot information and traffic information. Furthermore, the acquisition unit 12 may input information on past traffic accidents, etc., into the machine learning model, but the risk prediction device 10 according to this embodiment can perform highly accurate risk prediction even without inputting information on past traffic accidents, etc.

[0038] Furthermore, the acquisition unit 12 acquires real-time movement information including position information and speed vectors of pedestrians, etc. moving within a predetermined area, as well as attribute information of the pedestrians, etc. For example, the acquisition unit 12 reads out information related to passage for a fixed period of time stored in the storage unit 20, and acquires the position information, speed vectors, and attribute information of pedestrians, etc. included in the read out information related to passage as real-time movement information and attribute information of pedestrians, etc. The acquisition unit 12 passes the acquired movement information and attribute information associated with the movement information to the prediction unit 18.

[0039] Fig. 6 is a diagram showing the routes taken by pedestrians and the placement of vehicles during a specific time period (7:00 AM to 8:00 AM in the example of Fig. 6) in the area of ​​Fig. 4. The example of Fig. 6 shows an example in which elementary school students 121 to 125, who are pedestrians, are going to elementary school 100, with elementary school students 122 and 123 starting from house 102, elementary school student 121 passing between vehicles 112 and 113, joining group 124 who are already walking, and group 125 going to elementary school 100. Specific examples of blind spot information and traffic information will be described using Fig. 7 and Fig. 8, respectively, for the example of Fig. 6.

[0040] Fig. 7 is a diagram showing blind spot areas seen from a vehicle in the example of Fig. 6. For example, suppose vehicles 111 to 113 are parked on the central road. Note that the above-mentioned vehicle 110 is a general term that includes vehicles 111 to 113. When viewed from a vehicle traveling on the central road, areas 211, 212, and 213 that are blocked by parked vehicles 111, 112, and 113 are blind spots. Also, when viewed from a vehicle traveling on the central road, areas 214 and 215 that are blocked by a building near the entrance to a residential road that branches off from the central road are blind spots.

[0041] FIG. 8 is a diagram showing the traffic areas of pedestrians and the like in the example of FIG. 6. The trajectories of each person in the group of elementary school students walking to school in the example of FIG. 6 are identified from traffic information, accumulated by time period, and statistically processed, thereby identifying a traffic area as shown in FIG. 8. By adding attribute information such as the number of people and their ages to this traffic area, it becomes possible to determine that the traffic information indicating this traffic area represents a group of elementary school students walking to school. In this way, by accumulating the trajectories of pedestrians and the like identified from traffic information by time period and statistically processing them to identify a traffic area, and adding attribute information to obtain the traffic information, it becomes possible to understand the situation of regular movements such as commuting to school or work, as well as random movements.

[0042] 9 to 12, the setting unit 14 sets a danger level indicating the degree of danger to the driver of the vehicle 110 for each position within a predetermined area by time period. The setting unit 14 sets a first danger level based on blind spot information, and sets a second danger level based on traffic information.

[0043] For example, the setting unit 14 may set a first danger level of "1" for each position included in an area that has a predetermined relationship with the blind spot area indicated by the blind spot information stored in the storage unit 20, and may set a first danger level of "0" for each position included in other areas. The area that has a predetermined relationship with the blind spot area may be an area where a pedestrian or the like is expected to jump out from the blind spot area, such as an area sandwiched between blind spot areas or an area near the blind spot area. Furthermore, depending on the geometric shape and size of the blind spot area, the time or distance from the blind spot area at which the driver of the vehicle 110 can recognize a pedestrian or the like coming out of the blind spot may vary. Therefore, the setting unit 14 may set a first danger level that is higher for a position that takes less time to reach the blind spot or is shorter distance from the blind spot area, for example, depending on the geometric shape and size of the blind spot area.

[0044] Furthermore, for example, if the traffic area indicated by the traffic information stored in the storage unit 20 is in a dangerous situation, such as a location without a traffic light or a crosswalk, the setting unit 14 may set a second risk level of "2" for each location included in the traffic area, and a second risk level of "0" for each location that is not a traffic area or is included in a traffic area with a traffic light or a crosswalk. This makes it possible to increase the second risk level for locations that are not crosswalks but where pedestrians frequently cross. Furthermore, for example, if the traffic area indicated by the traffic information assigned attribute information indicating multiple people and an age in their early teens extends off the sidewalk, if the attribute information assigned to the traffic information indicates an age in their 60s or older, or if the traffic area indicated by the traffic information assigned attribute information of a bicycle extends onto the roadway, the setting unit 14 may set the second risk level of each location included in the traffic area higher than the second risk level of locations that do not fall into such cases. This makes it possible to set a high second risk level for locations that are likely to cause traffic accidents or near misses, such as places where children walk off the sidewalk while walking to or from school in groups, places where bicycles step onto the road to avoid frequently parked vehicles, and places where elderly people cross intersections diagonally.

[0045] As shown in the examples of FIGS. 9 to 12, the generation unit 16 generates a risk map for each time period indicating a third risk level by superimposing the first risk level and the second risk level set for each location within a predetermined area. By superimposing the first risk level based on blind spot information and the second risk level based on traffic information in this manner, the risk map can indicate locations where pedestrians or other vehicles are crossing from blind spots. By checking the risk map, the driver can identify locations with high risk levels and understand that the frequency of risk occurrence is high depending on the risk level. Specifically, the generation unit 16 projects the blind spot information and the traffic information onto one area, calculates a third risk level by adding the first risk level and the second risk level set by the setting unit 14 for each location within the area, and generates a risk map in which each location is color-coded according to the third risk level. An example of a risk map generated by the generation unit 16 is shown below.

[0046] FIG. 9 shows a risk map in which FIG. 9(A) shows the routes of pedestrians 121-125 and the locations of vehicles 111-113 taken from 7:00 AM to 8:00 AM as shown in FIG. 6, and FIG. 9(B) shows a third risk level in which traffic information and blind spot information are superimposed. Here, the third risk level is displayed in three levels. Specifically, third risk level 1 simply indicates a blind spot area. Third risk level 2 indicates a location where pedestrians or the like cross in a dangerous situation. Third risk level 3 indicates a location that is a blind spot area and also where pedestrians or the like cross in a dangerous situation. Note that the criteria for the third risk level are merely examples, and other criteria may be used. For example, the third risk level may be set to three levels obtained by multiplying the first risk level by the second risk level.

[0047] 9, as a specific example, since region 311 is a region near blind spot region 211, setting unit 14 sets a first danger level to 1. Furthermore, since region 311 is not a passage area, setting unit 14 sets a second danger level to 0. Generation unit 16 superimposes these and sets a third danger level of 1 for region 311. Furthermore, since region 312 is a region near blind spot region 212, setting unit 14 sets a first danger level to 1. Furthermore, since region 312 is a passage area where there is no traffic light or crosswalk, setting unit 14 sets a second danger level to 2. Generation unit 16 superimposes these and sets a third danger level of 3 for region 312. Furthermore, generation unit 16 sets a third danger level of 3 for region 313 for the same reason as region 312. The region 314 is an area sandwiched between the blind spot region 214 and the blind spot region 215, but is not a traffic area, so the generation unit 16 sets the third danger level of 1 for the region 314.

[0048] 10A shows the layout of the route taken by pedestrian 126 from 10:00 AM to 11:00 AM, and FIG. 10B shows a risk map showing a third risk level obtained by superimposing traffic information and blind spot information. Specifically, area 315 on the central road where pedestrian 126 is heading toward dry cleaning shop 101 is not near a blind spot area, so setting unit 14 sets the first risk level to 0. Also, area 315 is a traffic area without a traffic light or crosswalk, so setting unit 14 sets the second risk level to 2. Generation unit 16 superimposes these and sets area 315 as a third risk level of 2. Furthermore, area 314 sandwiched between blind spot area 214 and blind spot area 215 is the same as the time period in FIG. 9, and is near a blind spot area but not a traffic area, so generation unit 16 sets area 314 as a third risk level of 1.

[0049] 11A shows the routes of pedestrians 121-125 and the location of vehicles 110 from 3:00 PM to 4:00 PM, and FIG. 11B shows a risk map in which traffic information and blind spot information are superimposed to create a third risk level. As a specific example, since region 316 is an area near blind spot region 216 and blind spot region 217, setting unit 14 sets a first risk level of 1 for region 316. Furthermore, since region 316 is also a traffic area without a traffic light or a crosswalk, setting unit 14 sets a second risk level of 2 for region 316. Generation unit 16 superimposes these and sets a third risk level of 3 for region 316. Furthermore, for region 317 near blind spot region 218, generation unit 16 sets a third risk level of 3 for region 316 for the same reason. The area 314 sandwiched between the blind spot area 214 and the blind spot area 215 is the same as the time period in Figure 10, and is near the blind spot area but is not a traffic area, so the generation unit 16 sets the third danger level to 1 for the area 314.

[0050] 12A shows the route of pedestrians 127 and bicycles 131 and the location of vehicles 110 from 6:00 PM to 8:00 PM, and FIG. 12B shows a risk map showing a third risk level obtained by superimposing traffic information and blind spot information. As a specific example, the setting unit 14 sets a first risk level of 1 for an area 318 near a blind spot area 217, and because area 318 is also a traffic area without a traffic light or a crosswalk, the setting unit 14 sets a second risk level of 2, and the generation unit 16 superimposes these and sets a third risk level of 3 for area 318. Furthermore, the setting unit 14 sets a first danger level of 1 for an area 319 near the blind spot area 219, and because area 319 is also a traffic area where bicycles may step onto the road, the setting unit 14 sets a second danger level of 2 for the area 319, and the generation unit 16 superimposes these and sets a third danger level of 3 for the area 319. For the same reason as for area 319, the generation unit 16 sets a third danger level of 3 for area 320 sandwiched between blind spot area 220 and blind spot area 215.

[0051] As described above, the generation unit 16 generates a risk map for each time period by setting a third risk level for each location within the area. FIG. 13 is a diagram showing how the risk maps generated by the generation unit 16 are stored in the storage unit 20 with time periods as the axis. This is hereinafter referred to as a "three-dimensional risk map." The three-dimensional risk map is displayed, for example, on the screen of a navigation system installed in the vehicle 110 or on a map application of the mobile terminal 80, and the risk map for a specific time period is displayed when the driver or the like selects a time period. Note that the storage order of the risk maps in FIG. 13 is not limited to the example of FIG. 13. The three-dimensional risk map can also be used to take measures, such as installing sensors on the infrastructure side or installing warning devices, for locations with high risk levels displayed on the three-dimensional risk map.

[0052] Furthermore, the generating unit 16 generates a risk map including a fourth risk level obtained by aggregating the third risk levels for each road segment of the road network within the predetermined area. In FIG. 14, FIG. 14(A) shows the road network of a given area using links and nodes, and FIG. 14(B) shows the fourth risk level of each link. As shown in FIG. 14, a risk map in which the road network of an area is represented using links and nodes and a fourth risk level is assigned to each link is hereinafter referred to as a "macro risk map." The risk maps described in FIG. 9 to FIG. 13 above are hereinafter referred to as "micro risk maps." In the macro risk map of FIG. 14, road sections are represented by links, and the fourth risk level is the sum of the third risk levels set for each position between nodes (i.e., links). Hereinafter, the third risk level is referred to as the "micro risk level," and the fourth risk level is referred to as the "macro risk level." For example, in FIG. 14(B), the macro risk level of the link between node a and node b is 3. As another example, the macro risk level of the link between node d and node g in FIG. 14(B) is 4.

[0053] FIG. 15 is a diagram showing the total macro risk level shown in FIG. 14(B) for each route from start point a to end point k shown in FIG. 14(A). Note that the black circles in FIG. 15 indicate the node numbers that vehicle 110 passes through. Specifically, in case 1 of FIG. 15, the route passes through node a, node f, node i, node j, and node k. Furthermore, from FIG. 14(B), the macro risk level between node a and node f is 1, the macro risk level between node f and node i is 1, the macro risk level between node i and node j is 2, and the macro risk level between node j and node k is 1. In case 1, the total macro risk level from start point a to end point k is 5. As another example, in case 5 of FIG. 15, the route passes through node a, node b, node d, node g, node h, and node k. Furthermore, from Figure 14(B), the macro risk level between node a and node b is 3, the macro risk level between node b and node d is 2, the macro risk level between node d and node g is 4, the macro risk level between node g and node h is 2, and the macro risk level between node h and node k is 1, so in case 5, the total macro risk level from start point a to end point k is 12. Therefore, Figure 15 shows that, between the routes in case 1 and case 5, case 5 is more risky.

[0054] The search unit 26 searches for one or more routes from a start point to an end point in a predetermined area on a road network, and searches for and presents an optimal route from the one or more searched routes based on the sum of macro risk levels, which is a fourth risk level, for road segments included in the route. Specifically, in the examples of FIGS. 14 and 15, the search unit 26 presents case 1, which has the lowest sum of macro risk levels (the sum of macro risk levels is 5), as the optimal route. However, the optimal route is presented as a candidate, and route determination is left to the driver of the vehicle 110, for example. Note that the driver may determine the route by taking into consideration, for example, the pedestrian passage frequency distribution, road width, and distance as additional conditions when determining the route.

[0055] Based on the real-time movement information of pedestrians and the like received from the acquisition unit 12 and the micro danger map, the prediction unit 18 predicts the time at which a pedestrian or the like will pass a danger position where the danger level on the micro danger map is equal to or greater than a predetermined value. As described above, the acquisition unit 12 also acquires attribute information of pedestrians and the like using image recognition or the like. The prediction unit 18 predicts the time at which a pedestrian or the like will pass a danger position based on previously acquired traffic information that has been acquired together with the same attribute information as the attribute information acquired together with the movement information. Specifically, when comparing the traffic information, which is past information, with the movement information, which is current information, the prediction unit 18 compares past information indicating that a person will be at a certain location at a certain time with current information indicating that a person has actually arrived at that location. Furthermore, the prediction unit 18 predicts the movement position and time based on the current position information of the pedestrian or the like, the walking speed vector, and the passage area indicated by the traffic information. In the aforementioned comparison, since attribute information is already associated with each of the traffic information and movement information, the prediction unit 18 can perform more accurate comparison by matching the attribute information. For example, if the recognition time period is 2:00 PM to 3:00 PM and the attribute information of the pedestrians recognized by the infrastructure camera 70 is that there are multiple pedestrians and they are in their early teens, the prediction unit 18 predicts their subsequent trajectory based on past traffic information of similar cases, such as a group of pedestrians walking home from school. Furthermore, for example, based on the skeletal information recognized by the infrastructure camera 70 as pedestrian attribute information, the prediction may reflect the pedestrian's walking style characteristics, such as stride length and likelihood of straying onto the roadway. Furthermore, since the traffic information and movement information are data associated with attribute information in the prediction, the driver of the vehicle 110 can specify the attribute information and check the travel trajectory of that attribute. Note that linking by attribute information is not essential. Linking attribute information in this way contributes to the accuracy of the prediction by the prediction unit 18.

[0056] FIG. 16 shows, in FIG. 16(A), the movement information and attribute information of pedestrians and the like being recognized in real time by infrastructure camera 70, in FIG. 16(B) a micro risk map previously generated for the relevant time period (similar to FIG. 11(B)), and in FIG. 16(C) the prediction unit 18 predicting the state about one minute after FIG. 16(A). Specifically, when the movement information of pedestrian 125 in FIG. 16(A) is first recognized in real time by infrastructure camera 72, area 411 in FIG. 16(C) is an area that cannot be recognized by infrastructure cameras 71 to 78, as shown in FIG. 4. However, by using the information on the risk map in FIG. 16(B) in conjunction with the real-time movement information, prediction unit 18 predicts the time at which pedestrian 125 will arrive at area 411. Specifically, for example, infrastructure camera 72 recognizes that pedestrian 125 is walking at a walking speed of 2 km / h, and prediction unit 18 predicts that pedestrian 125 will arrive at area 411, which is 30 m away and has a micro danger level of 3 (see area 316 in Figure 16(B)), and predicts that pedestrian 125 will cross area 411, which is in the blind spot of parked vehicle 110, approximately one minute later.

[0057] Furthermore, the prediction unit 18 predicts the time at which a pedestrian or the like will pass a hazardous position based on the micro hazard level for each position included in a road segment through which the vehicle 110 will pass while traveling along the optimal route presented by the search unit 26. This means that the search unit 26 searches for and presents an optimal route based on the sum of macro hazard levels, and then the prediction unit 18 makes a prediction based on the micro hazard level for each position included in the optimal route. Note that the positions predicted by the prediction unit 18 are not limited to positions included in the optimal route, but the prediction unit 18 predicts each position included in a route selected by the driver of the vehicle 110 from among multiple routes presented by the search unit 26.

[0058] The notification unit 22 issues a warning to at least one of the pedestrian, etc. who is predicted to pass through the hazardous position and the vehicle 110 around the hazardous position where the pedestrian, etc. is predicted to pass, a predetermined time before the predicted time. Specifically, when notifying the pedestrian, etc. of a warning, the notification unit 22 issues the warning to a gadget such as a mobile terminal 80 carried by the pedestrian, etc., by an alert sound, flashing a specific warning pattern, vibration, or the like. Note that the warning is not limited to an object held by the pedestrian, etc., and the notification unit 22 may install a device on the infrastructure side that emits light or sound, and notify the pedestrian, etc. who is predicted to pass through the hazardous position to emit light or sound from the device. Note that the notification unit 22 may also issue a warning as a preventative measure before the pedestrian, etc. crosses the hazardous position.

[0059] Furthermore, when notifying a vehicle 110 near a hazardous position where a pedestrian or the like is predicted to pass, the notification unit 22 notifies the vehicle 110 by displaying an alert on the screen of a navigation system mounted on the vehicle 110 or by outputting a sound from the navigation system. As another example, the notification unit 22 may notify the vehicle 110 by displaying an alert on a head-up display projected on the windshield of the vehicle 110. Note that, although various methods for the notification unit 22 to notify the vehicle 110 have been exemplified, the method is not limited to one type of notification, and these methods may be combined as appropriate.

[0060] The control unit 24 controls the traveling of the vehicle 110 around a danger position where a pedestrian or the like is predicted to pass. Specifically, the control unit 24 controls the vehicle 110 to reduce its speed and to change the traveling direction of the vehicle 110 to avoid contact with the pedestrian or the like. Note that the control unit 24 may not directly control the vehicle 110, but may instead control the vehicle 110 by displaying a speed limit such as "slow down to 20 km / h" on an electronic signboard or the like at the side of the road on which the vehicle 110 is traveling.

[0061] Next, the operation of the risk prediction device 10 according to this embodiment will be described. Fig. 17 is a flowchart showing the flow of risk map generation processing. Fig. 18 is a flowchart showing the flow of risk prediction processing. Fig. 19 is a flowchart showing the flow of search processing. The risk prediction device 10 executes the risk map generation processing shown in Fig. 17, the risk prediction processing shown in Fig. 18, and the search processing shown in Fig. 19. Each process in the risk prediction device 10 is executed by the CPU 51 functioning as the acquisition unit 12, setting unit 14, generation unit 16, prediction unit 18, notification unit 22, control unit 24, and search unit 26. The risk map generation processing is an example of a risk prediction method of the present invention.

[0062] First, the risk map generation process of FIG. 17 will be described. 17, the CPU 51 acquires information from each sensor. Specifically, the CPU 51 acquires information related to passage recognized and transmitted by the infrastructure camera 70, and information related to passage recognized and transmitted by the sensors provided in the mobile object 90 and environmental information, and stores the information in the storage unit 20.

[0063] In step S103, the CPU 51 acquires blind spot information. Specifically, the CPU 51 acquires blind spot information by aggregating the environmental information stored in the storage unit 20 by time period, performing statistical processing, etc.

[0064] In step S105, the CPU 51 sets a first danger level based on the blind spot information. Specifically, the CPU 51 sets the first danger level "1" to each position included in an area that has a predetermined relationship with the blind spot area indicated by the blind spot information (for example, an area sandwiched between blind spot areas, an area near the blind spot area, etc.), and sets the first danger level "0" to each position included in other areas. Note that the setting criteria for the first danger level are merely examples and are not limited to these.

[0065] In step S107, the CPU 51 acquires traffic information. Specifically, the CPU 51 accumulates the trajectories of pedestrians and the like identified from the traffic information stored in the storage unit 20 by time period, performs statistical processing, and the like to identify a traffic area, and acquires traffic information indicating the traffic area.

[0066] In step S109, the CPU 51 sets a second danger level based on the traffic information. Specifically, if the predetermined area is a traffic area in a dangerous situation, the CPU 51 sets the second danger level to 2. Furthermore, if the predetermined area is a non-traffic area or a traffic area in which the situation is not dangerous, the CPU 51 sets the second danger level to 0. Note that the setting criteria for the second danger level are merely examples and are not limited to these.

[0067] Note that step S103 and step S107 are parallel processes, and either may be executed first, or they may be executed simultaneously.

[0068] In step S111, the CPU 51 generates a micro risk map by superimposing the first risk level and the second risk level. Specifically, if the first risk level at a certain position is 1 and the second risk level is 2, the CPU 51 sets the micro risk level at that position to 3, and similarly sets micro risk levels for each position to generate a micro risk map. In step S113, the CPU 51 stores the micro risk map in the storage unit 20.

[0069] In step S115, the CPU 51 generates a macro risk map. Specifically, the CPU 51 generates a macro risk map including a macro risk level obtained by aggregating the micro risk levels for each road segment of the road network within a predetermined area. In step S117, the CPU 51 stores the macro risk map in the storage unit 20. Then, the process ends.

[0070] Step S101 is executed sequentially at each time interval when information is acquired by the infrastructure camera 70 and the sensors mounted on the mobile object 90, and the processes from step S103 onward are executed at predetermined timings. As a result, the travel-related information and environmental information stored in the storage unit 20 are updated (added) sequentially, and the micro risk map and macro risk map are updated at predetermined timings.

[0071] Next, the risk prediction process of FIG. 18 will be described. In step S201 of FIG. 18, the CPU 51 calls up the micro risk map stored in the storage unit 20.

[0072] In step S203, the CPU 51 acquires real-time movement information. Specifically, the CPU 51 acquires real-time movement information of pedestrians and the like recognized by the infrastructure camera 70 or the like.

[0073] In step S205, the CPU 51 predicts the future passage time of the pedestrian, etc. Specifically, the CPU 51 predicts the movement position of the pedestrian, etc. several seconds to several minutes in the future and the time of arrival at that position, based on the acquired real-time movement information of the pedestrian, etc.

[0074] Note that step S201 and step S203 are parallel processes, and either may be executed first, or they may be executed simultaneously.

[0075] In step S207, the CPU 51 determines whether the risk level at the predicted time is equal to or greater than a predetermined value. That is, the CPU 51 determines whether the risk level of the area is high. If the CPU 51 determines that the risk level at the predicted time is equal to or greater than a predetermined value (step S207: YES), the process proceeds to at least one of step S209 and step S211. On the other hand, if the CPU 51 determines that the risk level at the predicted time is not equal to or greater than a predetermined value (step S207: NO), the process ends.

[0076] In step S209, the CPU 51 notifies the driver of the vehicle 110 of a warning. Specifically, the CPU 51 notifies the driver by displaying an alert on the screen of the navigation system of the vehicle 110. Note that instead of or in addition to the warning notice, the CPU 51 may control the vehicle 110. Then, the processing ends.

[0077] In step S211, the CPU 51 issues a warning to the pedestrian, etc. Specifically, the CPU 51 issues an alert to the mobile terminal 80 carried by the pedestrian, etc. Then, the process ends.

[0078] When both step S209 and step S211 are executed, either one may be executed first, or they may be executed simultaneously.

[0079] Next, the search process of FIG. 19 will be described. In step S301 of FIG. 19, the CPU 51 receives information on the start point and end point in a predetermined area.

[0080] In step S303, the CPU 51 searches the road network for route candidates from the start point to the end point in a predetermined area.

[0081] In step 305, the CPU 51 calls up the macro risk map for a predetermined area stored in the storage unit 20.

[0082] In step S307, the CPU 51 refers to the macro risk map and calculates the sum of the macro risk levels assigned to the links included in each of the searched route candidates.

[0083] In step S309, CPU 51 presents the candidate with the lowest total macro risk level as the optimum route. The presenting method may be, for example, displaying the optimum route on the screen of the navigation system of vehicle 110, or announcing the optimum route by voice from a speaker of vehicle 110. Then, the process ends.

[0084] Note that instead of generating the macro risk map in step S115 of the generation process, the macro risk map may be generated during execution of the search process, for example, instead of being called in step S305. In this case, the macro risk map may be deleted when the calculation of the sum of macro risks in step S307 is completed, when the search process is completed, or the like, without being stored in the storage unit 20. Note that after the route is determined, the risk prediction process of Fig. 18 is executed using the micro risk map.

[0085] As described above, the risk prediction device 10 according to the present embodiment includes an acquisition unit 12 that acquires, by time period, blind spot information, which is information on areas that are blind spots from the driver of the vehicle 110 in a predetermined area, and traffic information, which is information on areas where at least one of pedestrians and cyclists have passed; a setting unit 14 that sets, by time period, a first risk level indicating the degree of danger to the driver for each location in the predetermined area based on the acquired blind spot information, and a second risk level indicating the degree of danger to the driver for each location in the predetermined area based on the acquired traffic information; and a generation unit 16 that superimposes the first risk level and the second risk level set for each location in the predetermined area to generate a risk map for each time period indicating a third risk level for each location in the predetermined area. Thus, the risk prediction device 10 according to the present embodiment can generate a highly accurate risk map that takes into account the traffic conditions of pedestrians and the like predicted to be in the blind spot by time period.

[0086] Furthermore, according to the risk prediction device 10 of this embodiment, the generated risk map can present serious danger locations to the driver of the vehicle 110. For example, even for locations where no traffic accidents or near misses have occurred in the past, potentially serious danger locations are predicted and presented on the risk map, so the risk prediction device 10 of this embodiment has the effect of preventing and avoiding traffic accidents.

[0087] Furthermore, the risk prediction device 10 according to this embodiment is configured to further include a prediction unit 18 that acquires movement information, which is real-time information on at least one of pedestrians and cyclists moving in a predetermined area, and predicts, based on the acquired movement information, the time when the pedestrian or cyclist will pass a hazardous position on the hazard map where the third hazard level is equal to or greater than a predetermined value. Therefore, the risk prediction device 10 according to this embodiment compares past information, such as blind spot information and traffic information, with current movement information recognized in real time to predict the time when the pedestrian, etc. will pass the hazardous position, which is future information, and therefore can increase the accuracy of the prediction.

[0088] Furthermore, the risk prediction device 10 according to this embodiment can predict the situation several seconds to several minutes into the future based on the generated risk map and real-time movement information of the surrounding area, even when pedestrians or other objects in blind spots cannot be recognized in real time. It is difficult to predict the situation several seconds to several minutes into the future using real-time movement information alone, so the risk prediction device 10 can predict the situation by comparing it with past information, such as a risk map. Furthermore, simply learning information about actual traffic accidents and near misses may not reflect potential dangers. Furthermore, information about actual accidents or other events may not be enough to fully capture constantly changing road environment information. However, the risk prediction device 10 according to this embodiment overcomes these issues and achieves high versatility in preventing and avoiding traffic accidents.

[0089] Furthermore, the risk prediction device 10 according to this embodiment is configured such that the acquisition unit 12 acquires attribute information of at least one of pedestrians and bicycles along with traffic information, and the prediction unit 18 predicts the time at which at least one of the pedestrian and bicycle will pass through the danger position based on traffic information acquired in the past that has been acquired together with the same attribute information as the attribute information acquired together with the movement information. Thus, the risk prediction device 10 according to this embodiment links past traffic information and real-time movement information with attribute information to predict future trajectories, thereby increasing the accuracy of predictions.

[0090] Furthermore, for example, the risk prediction device 10 according to this embodiment not only predicts danger in advance and presents a risk map in the event of an accident such as a child suddenly running out into the road, which the driver of the vehicle 110 or an autonomous vehicle recognizes in real time and applies the brakes too late, but also notifies pedestrians and the vehicle 110 to warn them and controls them, thereby achieving comprehensive prevention and avoidance of traffic accidents.

[0091] Furthermore, the risk prediction device 10 according to this embodiment is further configured to include a search unit that generates a risk map including a fourth risk level obtained by aggregating the third risk levels for each road segment of a road network within a predetermined area, searches for one or more routes from a start point to an end point within the predetermined area on the road network, and searches for and presents an optimal route from the one or more routes found based on the sum of the fourth risk levels for the road segments included in the route. Therefore, the risk prediction device 10 according to this embodiment presents not only the risk levels for individual locations but also the risk levels for each segment, allowing the driver of the vehicle 110 to select a route that avoids risk and enabling traffic accident prevention on a route-by-route basis.

[0092] In this embodiment, the area shown in FIG. 4 has been described as an example of the predetermined area, but the present invention is not limited to this. For example, the risk prediction device 10 of this embodiment can also be used in areas such as an office district lined with companies, or a busy shopping district with many restaurants and retail stores. In addition, in this embodiment, the infrastructure camera 70 is installed at an intersection of a central priority road, but the present invention is not limited to this. By installing the infrastructure camera 70 in narrow alleys, the range that can be recognized in real time can be expanded, and the accuracy of risk prediction by the risk prediction device 10 for the next few seconds to minutes can be further improved.

[0093] In addition, the generation process, risk prediction process, and search process executed by the CPU through software (programs) loaded in the above-described embodiment may be executed by various processors other than the CPU. Examples of such processors include dedicated electrical circuits, such as programmable logic devices (PLDs) (such as field-programmable gate arrays (FPGAs)) whose circuit configuration can be changed after manufacture, and application-specific integrated circuits (ASICs) that are processors with circuit configurations specifically designed to execute specific processes. The generation process, risk prediction process, and search process may be executed by one of these various processors, or by a combination of two or more processors of the same or different types (e.g., multiple FPGAs, or a combination of a CPU and an FPGA). The hardware structure of these various processors is, more specifically, an electrical circuit that combines circuit elements such as semiconductor devices.

[0094] In the above embodiment, the processing program 11 is pre-stored (installed) in a storage device, but the present invention is not limited to this. The processing program 11 may be provided in a form recorded on a recording medium such as a CD-ROM, a DVD-ROM (Digital Versatile Disc Read Only Memory), or a USB (Universal Serial Bus) memory. The processing program 11 may also be downloaded from an external device via a network.

[0095] (Additional note 1) a setting unit that sets a first risk level indicating the degree of risk to the driver for each position in the predetermined area for each time period based on the acquired blind spot information, and sets a second risk level indicating the degree of risk to the driver for each position in the predetermined area for each time period based on the acquired traffic information; and a generation unit that generates a risk map for each time period by superimposing the first risk level and the second risk level set for each position in the predetermined area, the third risk level being indicated for each position in the predetermined area.

[0096] (Additional note 2) 2. The risk prediction device according to claim 1, wherein the acquisition unit acquires movement information, which is real-time information on at least one of pedestrians and bicycles moving in the specified area, and further includes a prediction unit that predicts, based on the acquired movement information, a time when the pedestrian and bicycle will pass a risk position on the risk map where the third risk level is equal to or greater than a specified value.

[0097] (Additional note 3) 3. The risk prediction device according to claim 2, further comprising a notification unit that issues a warning to at least one of the pedestrian and bicycle predicted to pass through the risk position a predetermined time before the time, and at least one of a vehicle in the vicinity of the risk position predicted to be passed by the pedestrian and / or bicycle.

[0098] (Additional note 4) 4. The risk prediction device according to claim 2, further comprising a control unit that controls the traveling of vehicles around the risk position where at least one of the pedestrian and the bicycle is predicted to pass.

[0099] (Additional note 5) 5. The risk prediction device according to claim 2, wherein the acquisition unit acquires the movement information including the position and speed vector of at least one of a pedestrian and a bicycle moving in the specified area, and acquires attribute information of at least one of the pedestrian and the bicycle, and the prediction unit predicts the time at which at least one of the pedestrian and the bicycle will pass through the risk position based on the movement information and the attribute information.

[0100] (Additional note 6) 6. The risk prediction device according to claim 5, wherein the acquisition unit acquires attribute information of at least one of the pedestrian and the bicycle along with the traffic information, and the prediction unit predicts the time at which at least one of the pedestrian and the bicycle will pass the risk location based on traffic information acquired in the past that has been acquired together with the same attribute information as the attribute information acquired together with the movement information.

[0101] (Additional note 7) The danger prediction device according to any one of claims 1 to 6, wherein the acquisition unit stores the acquired blind spot information and traffic information in a storage unit, and sequentially updates the blind spot information and traffic information stored in the storage unit using information acquired by sensors installed in the specified area and sensors equipped on moving bodies moving in the specified area.

[0102] (Additional note 8) The risk prediction device according to any one of claims 2 to 6, wherein the generation unit generates the risk map including a fourth risk level obtained by aggregating the third risk levels for each road segment of a road network within the specified area, searches for one or more routes from a start point to an end point in the specified area on the road network, and searches for and presents an optimal route from the one or more routes found based on the sum of the fourth risk levels for the road segments included in the route.

[0103] (Additional note 9) 9. The risk prediction device according to claim 8, wherein the prediction unit predicts a time at which at least one of the pedestrian and the bicycle will pass through the risk position, based on the third risk level for each position included in the road segment through which a vehicle traveling along the optimal route presented by the search unit will pass.

[0104] (Additional note 10) In a predetermined area, blind spot information, which is information on areas that are blind spots as seen by a vehicle driver, and traffic information, which is information on areas where at least one of pedestrians and bicycles have passed, are acquired by time period; Based on the acquired blind spot information, a first danger level is set for each time period for each position within the predetermined area, indicating a degree of danger to the driver; and based on the acquired traffic information, a second danger level is set for each time period for each position within the predetermined area, indicating a degree of danger to the driver; generating a risk map for each time period showing a third risk level for each position in the predetermined area by superimposing the first risk level and the second risk level set for each position in the predetermined area; A risk prediction program that causes a computer to execute processing. [Explanation of symbols]

[0105] 10. Hazard prediction device 12 Acquisition Department 14 Setting section 16 Generation part 18 Prediction Department 20 Storage area 22 Notification Department 24 Control Unit 26 Exploration Department 51 CPU 52 ROM 53 RAM 54 Input Device 55 Output Device 56 Storage media reader 57 Communication I / F 58 Bus 70 Infrastructure Camera 80 Mobile Devices 90 Mobile 110 vehicles N communication line

Claims

1. an acquisition unit that acquires, by time period, blind spot information, which is information on areas that are blind spots as seen by a vehicle driver in a predetermined area, and traffic information, which is information on areas where at least one of pedestrians and bicycles have passed; a setting unit that sets a first danger level indicating a degree of danger to the driver for each position within the predetermined area for each time period based on the acquired blind spot information, and that sets a second danger level indicating a degree of danger to the driver for each position within the predetermined area for each time period based on the acquired traffic information; a generating unit that generates a risk map for each time period, which indicates a third risk level at each position in the predetermined area by superimposing the first risk level and the second risk level set for each position in the predetermined area; A danger prediction device comprising:

2. the acquisition unit acquires movement information that is real-time information on at least one of pedestrians and bicycles moving in the predetermined area; a prediction unit that predicts, based on the acquired movement information, a time when the pedestrian and the bicycle will pass through a danger position in the danger map where the third danger level is equal to or greater than a predetermined value; The risk prediction device according to claim 1 .

3. a notification unit that issues a warning to at least one of the pedestrian and the bicycle who are predicted to pass through the danger position a predetermined time before the time, and a vehicle in the vicinity of the danger position through which the pedestrian and the bicycle are predicted to pass, The risk prediction device according to claim 2 .

4. a control unit that controls the traveling of a vehicle around the danger position where at least one of the pedestrian and the bicycle is predicted to pass; The risk prediction device according to claim 2 or 3.

5. the acquisition unit acquires the movement information including a position and a speed vector of at least one of a pedestrian and a bicycle moving in the predetermined area, and acquires attribute information of at least one of the pedestrian and the bicycle; the prediction unit predicts a time when at least one of the pedestrian and the bicycle will pass through the danger position based on the movement information and the attribute information; The risk prediction device according to claim 4.

6. the acquisition unit acquires attribute information of at least one of the pedestrian and the bicycle together with the traffic information; the prediction unit predicts a time at which at least one of the pedestrian and the bicycle will pass through the danger position based on the traffic information acquired together with the same attribute information as the attribute information acquired together with the movement information, among the traffic information acquired in the past; The risk prediction device according to claim 5 .

7. The acquisition unit stores the acquired blind spot information and the traffic information in a storage unit, and sequentially updates the blind spot information and the traffic information stored in the storage unit based on information acquired by sensors installed in the predetermined area and sensors provided on moving bodies moving in the predetermined area. The risk prediction device according to claim 6.

8. the generation unit generates the risk map including a fourth risk level obtained by aggregating the third risk levels for each road segment of a road network within the predetermined area; a search unit that searches for one or more routes from a start point to an end point in the predetermined area on the road network, and searches for and presents an optimal route from the one or more searched routes based on the sum of the fourth risk levels of the road segments included in the route; The risk prediction device according to claim 6.

9. the prediction unit predicts a time at which at least one of the pedestrian and the bicycle will pass through the danger position based on the third danger level for each position included in the road segment through which a vehicle traveling along the optimal route presented by the search unit will pass. The risk prediction device according to claim 8.

10. In a predetermined area, blind spot information, which is information on areas that are blind spots as seen by a vehicle driver, and traffic information, which is information on areas where at least one of pedestrians and bicycles have passed, are acquired by time period; based on the acquired blind spot information, a first danger level indicating a degree of danger to the driver for each position within the predetermined area for each time period, and based on the acquired traffic information, a second danger level indicating a degree of danger to the driver for each position within the predetermined area for each time period, generating a risk map for each time period showing a third risk level at each position in the predetermined area by superimposing the first risk level and the second risk level set for each position in the predetermined area; A risk prediction method in which processing is performed by a computer.

Images