Intersection warning system

Abstract

To provide an intersection warning system capable of improving safety of a traffic participant when a mobile body in an intersection is moving toward a warning area.SOLUTION: A warning device 10 of an intersection warning system performs: acquiring a position of a mobile body 8 from a traffic environment image 2; generating a movement vector Vm of the mobile body 8 based on time series of a position of the mobile body 8; generating direction vectors Vd1, Vd2 toward the warning area Aw; determining the level of a possibility that the mobile body 8 enters the warning area Aw based on an angle θm of the movement vector Vm of the mobile body 8 with respect to the direction vectors Vd1, Vd2; and executing output processing to output the warning information expressing the possibility in order to notify a traffic participant 9 of the information in a case with the higher possibility that the mobile body 8 enters the warning area Aw.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to an intersection warning system for improving the safety of traffic participants at intersections. [Background technology]

[0002] A conventional system for improving intersection safety is described in Patent Document 1. In this system, when a traveling vehicle is present within a predetermined area before an intersection in the traveling lane and the traffic light is expected to be red when the traveling vehicle reaches the intersection, the system extends the red light duration and simultaneously delays the time it takes for the pedestrian signal for a crosswalk across the traveling lane to change from red to green.

[0003] This improves the safety of traffic participants moving at and near crosswalks, even if, for example, a driver of a moving vehicle is late in noticing a red light. In this specification, the term "traffic participants" includes pedestrians, bicycles, light electric vehicles, wheelchairs, animals, and the like. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2023-97155 Summary of the Invention [Problem to be solved by the invention]

[0005] The above-mentioned conventional systems have the problem that, although they are effective when a vehicle is traveling in a lane before an intersection, they are not effective when the vehicle is traveling within the intersection. In other words, when a moving object such as a vehicle within an intersection is moving toward a security area such as a crosswalk, the safety of traffic participants in and around the security area may be reduced.

[0006] The present invention has been made to solve the above-mentioned problems, and aims to provide an intersection warning system that can improve the safety of traffic participants when a moving object within an intersection is moving toward a warning area. [Means for solving the problem]

[0007] In order to achieve the above object, the intersection warning system of claim 1 is characterized by comprising: a mobile body position acquisition unit that acquires the position of a mobile body by a predetermined image recognition process from a traffic environment image including a mobile body moving in an intersection area, which is the area of ​​the intersection, and one of a plurality of road areas adjacent to the intersection area; a movement vector generation unit that generates a movement vector of the mobile body based on a time series of the position of the mobile body; a direction vector generation unit that generates a direction vector from the intersection area toward a warning area, which is at least one of the plurality of road areas; a judgment unit that judges the likelihood of the mobile body entering the warning area based on the angle of the movement vector of the mobile body relative to the direction vector; and an output processing unit that executes output processing to output warning information representing the judgment result to notify traffic participants when the judgment unit judges that there is a high likelihood that the mobile body will enter the warning area.

[0008] According to this intersection warning system, it is determined that there is a high possibility that a moving object will enter a predetermined warning area based on the angle of the moving object's movement vector relative to a direction vector, and when such a determination is made, an output process is executed to output warning information representing the determination result to notify traffic participants, thereby making it possible to notify traffic participants of the warning information.

[0009] Here, the movement vector of the moving object is generated based on a time series of the moving object's position acquired from a traffic environment image by a predetermined image recognition process, and the direction vector is generated as a direction from the intersection area to a warning area, which is at least one area among multiple road areas. Therefore, the likelihood of the moving object entering the predetermined warning area can be appropriately determined based on the angle of the moving object's movement vector relative to the direction vector. Then, when it is determined that there is a high possibility that the moving object will enter the warning area, warning information representing the determination result is output to notify traffic participants. When a traffic participant is present in or near the predetermined warning area, the traffic participant can be notified that there is a high possibility that the moving object will enter the predetermined warning area, thereby improving the safety of the traffic participants. Note that the term "system" in this specification may be any term that is composed of one or more devices.

[0010] In the present invention, it is preferable that the direction vector generating unit determines the orientation of the direction vector based on the display state of a traffic signal at the intersection.

[0011] According to this intersection warning system, the direction of the direction vector is determined based on the display status of the traffic lights at the intersection, so that the direction of the direction vector can be appropriately set at intersections where traffic lights are present.

[0012] In the present invention, when each of four rectangular straight line sections surrounding an intersection area in a traffic environment image is divided into a plurality of areas, a counting unit counts the number of intersections between a straight line extending along a movement vector and each straight line section that are located in any of the plurality of areas; an inflow / outflow point setting unit sets the area of ​​each straight line section that has the largest number of intersections counted by the counting unit as an inflow point and an outflow point of a moving object based on the direction of the movement vector; a driving line setting unit sets a straight line connecting the inflow point and the outflow point of each of two opposing straight line sections as a driving line of a moving object in a road area; and a method of dividing the plurality of road areas into four straight line sections. The vehicle navigation system further includes a center line generation unit that generates a center line of the first road area and a center line of the second road area based on an enlarged area defined by four intersections where the driving lines of a first road area, which is one of the road areas, intersect with the driving lines of a second road area that intersect with the first road area, and it is preferable that the direction vector generation unit generates a direction vector that extends from a center point, which is the intersection point of the center line of the first road area and the center line of the second road area, toward the warning area along one of the center lines of the first road area and the second road area.

[0013] According to this intersection warning system, when each of four rectangular straight sections surrounding an intersection area in a traffic environment image is divided into multiple areas, the number of intersections between each straight section and a line extending along a movement vector and the line sections is counted, and the area with the largest number of intersections counted by the counting unit is set as the entry point and exit point of the moving object based on the direction of the movement vector. Then, the straight lines connecting the entry point and exit point on each of the two opposing straight sections are set as the traveling line of the moving object in the road area.

[0014] When multiple moving objects move through an intersection, the area of ​​each straight line with the most intersections with lines extending along the movement vectors of the moving objects is the area where moving objects most frequently enter and exit the intersection. Therefore, by setting a line connecting such entry and exit points as the driving line of a moving object in one lane of a road, this driving line can be considered to be an average of the driving lines of multiple moving objects. Furthermore, by generating the center line of a first road and the center line of a second road based on an enlarged area defined by four intersections between the driving lines of both lanes of a first road and the driving lines of both lanes of a second road intersecting with the first road, these center lines can be generated with high accuracy.

[0015] In addition, a direction vector is generated so that it extends along either the center line of the first road or the center line of the second road, with the intersection of the center line of the first road and the center line of the second road as its center point, so that the direction vector can be appropriately generated so that it points from the center point of the intersection toward the surveillance area.

[0016] In the present invention, the system further comprises a first traffic area setting unit that sets a first traffic area consisting of an intersection area and a plurality of road areas based on the enlarged area, and a second traffic area acquisition unit that executes one of a first acquisition process that acquires the movement area of ​​the moving body on the road surface based on a time series of the position of the moving body by a predetermined first image recognition process, and acquires a second traffic area consisting of the intersection area and a plurality of road areas based on the time series of the movement area, and a second acquisition process that acquires the second traffic area from a traffic environment image by a predetermined second image recognition process, and it is preferable that the first traffic area setting unit sets the first traffic area so that the first traffic area fits within the second traffic area.

[0017] According to this intersection warning system, a first traffic area is set based on the enlarged area. Furthermore, a predetermined first image recognition process is used to acquire the moving area on the road surface of the moving object based on a time series of the moving object's position, and the second traffic area is acquired by one of a first acquisition process that acquires a second traffic area consisting of an intersection area and multiple road areas based on the time series of the moving area, and a second acquisition process that acquires the second traffic area from a traffic environment image by a predetermined second image recognition process. By acquiring the second traffic area in this manner, the second traffic area is acquired as something close to the actual road area.

[0018] By setting the first traffic area so that it fits within such a second traffic area, it is possible to prevent the first traffic area from becoming an area larger than the actual road area, thereby improving the accuracy of setting the first traffic area.

[0019] In the present invention, the system further includes a first traffic area setting unit that sets a first traffic area consisting of an intersection area and multiple road areas based on an enlarged area; a monitoring area setting unit that sets multiple rectangular monitoring areas in the first traffic area based on the direction of the direction vector, the center line of the first road area, and the center line of the second road area; a participant position acquisition unit that acquires the positions of traffic participants other than moving bodies from a traffic environment image by a predetermined third image recognition process; and an alert level setting unit that sets an alert level based on the relationship between the positions of the traffic participants, the position of the moving body, and the multiple monitoring areas, and it is preferable that the alert information is configured to include information corresponding to the alert level.

[0020] According to this intersection warning system, a first road area consisting of an intersection area and multiple road areas is set based on the enlarged area, and multiple monitoring areas in the first traffic area are set in a rectangular shape based on the direction of the direction vector, the center line of the first road, and the center line of the second road. Furthermore, the positions of traffic participants are acquired from the traffic environment image by a predetermined third image recognition process, and an alert level is set based on the relationship between the positions of the traffic participants, the position of the moving object, and the multiple monitoring areas. Then, alert information including information corresponding to the alert level is reported from the alarm device, so that an appropriate level of alert information can be reported to traffic participants based on the relationship between the positions of the traffic participants, the position of the moving object, and the multiple monitoring areas.

[0021] In the present invention, it is preferable that the alert level setting unit sets the alert level based on the display status of traffic signals at the intersection in addition to the positions of traffic participants, the positions of moving objects, and the relationship between the multiple monitoring areas.

[0022] This intersection warning system can notify traffic participants of an appropriate level of warning information based on the display status of traffic signals at the intersection in addition to the relationship between the positions of traffic participants, the positions of moving objects, and multiple monitoring areas.

[0023] In the present invention, it is preferable to further include a notification device that notifies traffic participants of the warning information as at least one of audio information and visual information.

[0024] According to this intersection warning system, the warning information is notified to traffic participants by the alarm device as at least one of audio information and visual information, which allows traffic participants to properly recognize that there is a high possibility that a moving object will enter the warning area, thereby further improving the safety of traffic participants. [Brief explanation of the drawings]

[0025] [Figure 1]1 is a diagram showing a schematic configuration of an intersection warning system according to an embodiment of the present invention; [Figure 2] FIG. 1 is a diagram showing an image of an intersection and its vicinity captured by a camera. [Figure 3] FIG. 2 is a block diagram showing the functional configuration of the alarm device. [Figure 4] 3 is a diagram for explaining the functions of a moving object position acquisition unit and a movement vector generation unit. FIG. [Figure 5] 10 is a diagram for explaining the functions of a counting unit, an inflow / outflow point setting unit, and a driving line setting unit. FIG. [Figure 6] 10A and 10B are diagrams for explaining the functions of a center line generating unit and a direction vector generating unit. [Figure 7] 10A and 10B are explanatory diagrams of a determination method performed by a determination unit when a moving object turns left at an intersection. [Figure 8A] FIG. 10 is a diagram showing a state before a moving object starts to turn left at an intersection. [Figure 8B] FIG. 10 is a diagram showing a state in which a moving object starts to turn left at an intersection. [Figure 8C] FIG. 8C is a diagram showing a state in which the moving body has moved further forward than the position in FIG. 8B. [Figure 8D] FIG. 8D is a diagram showing a state in which the moving body has moved further forward than the position in FIG. 8C. [Figure 9] 10A and 10B are explanatory diagrams of a determination method performed by a determination unit when a moving object turns right at an intersection. [Figure 10] 10A and 10B are explanatory diagrams of a determination method of a determination unit when a moving object goes straight through an intersection. [Figure 11] 10 is an explanatory diagram of a determination method used by a determination unit when a moving object turns right at an intersection in a case where the direction vector is switched. FIG. [Figure 12] 10 is a diagram showing an example of a result of acquisition of a second traffic area by a second traffic area acquisition unit. FIG. [Figure 13] 10 is a diagram showing another example of the result of acquisition of the second traffic area by the second traffic area acquisition unit. FIG. [Figure 14] FIG. 10 is a diagram showing a correction state when the first traffic area does not fit into the second traffic area. [Figure 15]FIG. 10 is a diagram illustrating an example of a monitoring area set. [Figure 16] FIG. 10 is a diagram illustrating another example of a monitoring area set. DETAILED DESCRIPTION OF THE INVENTION

[0026] An intersection warning system according to one embodiment of the present invention will be described below with reference to the drawings. The intersection warning system 1 of this embodiment is intended to improve the safety of traffic participants at an intersection 5, and as shown in Figure 1, includes a camera 3, a speaker 4, and a warning device 10.

[0027] The alarm device 10 is composed of a computer equipped with a CPU, storage, memory, and I / O interface, and is capable of communicating with the camera 3 and speaker 4. As will be described later, the alarm device 10 executes various control processes based on image signals from the camera 3, and in response to these processes, transmits an alarm information signal to the speaker 4.

[0028] In this embodiment, the intersection 5 is configured in the shape of a crossroad, where a first road 5a and a second road 5b intersect in a crisscross pattern. Pedestrian crossings 5c, 5c are provided in a portion of the first road 5a near the intersection 5, and pedestrian crossings 5d, 5d are provided in a portion of the second road 5b near the intersection 5. Furthermore, as shown in FIG. 2, the intersection 5 is provided with a plurality of traffic signals 6 for pedestrians (hereinafter referred to as "pedestrian signals 6") and a plurality of traffic signals 7a, 7b for vehicles (only two are shown). Hereinafter, these traffic signals 7a, 7b for vehicles will be collectively referred to as "vehicle signals 7" where appropriate.

[0029] Camera 3 is placed diagonally above intersection 5 and has a communication function. Camera 3 captures image 2 (hereinafter referred to as "traffic environment image 2") including intersection 5, part of first road 5a, and part of second road 5b at an angle as shown in Fig. 2, and transmits the image signal to warning device 10. In this case, traffic environment image 2 is captured as a video.

[0030] In the following explanation, the area in the traffic environment image 2 where two roads 5a, 5b intersect with each other is referred to as the "intersection area," the area of ​​the road extending from the intersection area is referred to as the "road area," and the area consisting of the "intersection area" and the "road area" is referred to as the "traffic area."

[0031] Furthermore, the speaker 4 (alert device) is placed near the intersection 5, and when it receives an alert information signal from the alert device 10, it outputs audio information corresponding to the alert information signal, as will be described later.

[0032] Next, the functional configuration of the alert device 10 will be described with reference to Fig. 3. As shown in the figure, the alert device 10 includes a mobile object position acquisition unit 11, a movement vector generation unit 12, a counting unit 13, an inflow / outflow point setting unit 14, a driving line setting unit 15, a center line generation unit 16, a direction vector generation unit 17, a determination unit 18, a first traffic area setting unit 19, a second traffic area acquisition unit 20, a monitoring area setting unit 21, a participant position acquisition unit 22, an alert level setting unit 23, and an output processing unit 24.

[0033] First, when the alert device 10 receives an image signal from the camera 3, the moving object position acquisition unit 11 acquires the position of the moving object from the traffic environment image 2. Specifically, as shown in FIG. 4, a predetermined image recognition process is performed to acquire the position of a moving object 8, such as a four-wheeled vehicle, in each frame of the traffic environment image 2 using a bounding box 30. In this case, the predetermined image recognition process is an object recognition process that applies a machine learning model (for example, OpenVino: a registered trademark).

[0034] Next, the movement vector generation unit 12 generates a movement vector Vm of the moving object 8 based on the position of the moving object 8 in the current frame of the traffic environment image 2 and the position of the moving object 8 in the immediately previous frame. As shown in Fig. 4, the movement vector Vm is generated as a vector that starts at the midpoint of the bottom side of the bounding box 30 and extends in the traveling direction of the moving object 8. Note that Fig. 4 shows, as an example, a state in which the moving object 8 is traveling only on the first road 5a side.

[0035] Furthermore, when each of the straight line portions 2a to 2d (hereinafter referred to as "edge straight line portions 2a to 2d") along the top, bottom, left, and right edges of the traffic environment image 2 is divided into a large number of regions of a predetermined width, the counting unit 13 counts the number of intersections between a straight line (shown by a dashed line in FIG. 4) extending along the movement vector Vm and the edge straight line portions 2a to 2d in each region for each frame. Then, a bar graph (histogram) shown in FIG. 5 is created as a graph representing the integrated values ​​of these counting results. In this case, the counting by the counting unit 13 is performed for a predetermined time (for example, 10 minutes).

[0036] 5, dotted bar graph 31 represents the cumulative value of the number of intersections between a line extending in the direction of movement vector Vm and edge straight line portions 2a-2d, while hatched bar graph 32 represents the cumulative value of the number of intersections between a line extending in the opposite direction to movement vector Vm and edge straight line portions 2a-2d.

[0037] Furthermore, the inflow / outflow point setting unit 14 sets the inflow point Pin and the outflow point Pout of the moving object 8 as described below based on the histograms 31 and 32 created by the counting unit 13. That is, among the multiple regions in each of the edge straight line portions 2a to 2d of the traffic environment image 2, the region with the largest number of intersections counted by the counting unit 13, i.e., the midpoint of the longest region of the histograms 31 and 32, is set as the inflow point Pin and the outflow point Pout of the moving object 8 based on the direction of the movement vector Vm.

[0038] On the other hand, the driving line setting unit 15 sets the driving lines L1a, L1b of the moving body 8 on the first road 5a and the driving lines L2a, L2b of the moving body 8 on the second road 5b using the method described below based on the inflow point Pin and outflow point Pout of the moving body 8 set by the inflow / outflow point setting unit 14 (see Figure 5).

[0039] That is, the straight lines connecting the inflow point Pin and the outflow point Pout on the left and right straight edge portions 2c and 2d of the traffic environment image 2 are set as the driving lines L1a and L1b of the lanes on both sides of the moving object 8 in the road area of ​​the first road 5a (hereinafter referred to as the "first road area"). Similarly, the straight lines connecting the inflow point Pin and the outflow point Pout on the top and bottom straight edge portions 2a and 2b of the traffic environment image 2 are set as the driving lines L2a and L2b of the lanes on both sides of the moving object 8 in the road area of ​​the second road 5b (hereinafter referred to as the "second road area"). As a result, the driving lines L1a and L1b and the driving lines L2a and L2b intersect with each other at four intersection points Px1 to Px4.

[0040] Furthermore, the center line generation unit 16 generates a center line Lc1 of the first road area and a center line Lc2 of the second road area based on the four driving lines L1a, L1b, L2a, and L2b set by the driving line setting unit 15 using the method described below.

[0041] That is, as shown in FIG. 6, a rectangular area is created defined by four intersection points Px1 to Px4 where four driving lines L1a, L1b, L2a, and L2b intersect with each other, and by enlarging this rectangular area, a rectangular enlarged area 41a (the area shown by dotted lines in the figure) defined by four points Py1 to Py4 is generated.

[0042] In this case, the ratio at which the rectangular area defined by the four intersection points Px1 to Px4 is enlarged is set to a ratio (for example, 1.2 to 1.5 times) that makes the rectangular area as similar as possible, and these ratios are set in advance through preliminary tests.

[0043] Next, a straight line (shown as a dashed line in the drawing) extending through the midpoint P14 between the two points Py1 and Py4 and the midpoint P23 between the two points Py2 and Py3 is generated as the center line L1c of the first road area, and a straight line (shown as a dashed line in the drawing) extending through the midpoint P12 between the two points Py1 and Py2 and the midpoint P34 between the two points Py3 and Py4 is generated as the center line L2c of the second road area. Accordingly, the intersection of the two center lines L1c and L2c is generated as the center point Pc of the intersection 5.

[0044] The direction vector generator 17 generates direction vectors using the following method: First, as shown in Fig. 6, two direction vectors Vd1 and Vd3 extending from the center point Pc of the intersection 5 along the center line L2c to two midpoints P34 and P12, and two direction vectors Vd2 and Vd4 extending from the center point Pc of the intersection 5 along the center line L1c to two midpoints P14 and P23 are set as direction vectors.

[0045] Furthermore, by a predetermined image recognition process (machine learning model), the display state of the vehicle traffic signal 7 is acquired from the traffic environment image 2. In this case, for example, a CNN (Convolutional Neural Network) is used as the predetermined image recognition process.

[0046] Next, one of the two direction vectors Vd1 and Vd2 is generated (selected) based on the display state of the vehicle traffic light 7. For example, when the vehicle traffic light 7a for the first road 5a is displaying a green light, the direction vector Vd1 is generated, and when the vehicle traffic light 7b for the second road 5b is displaying a green light, the direction vector Vd2 is generated.

[0047] This is because, due to the angle of the traffic environment image 2, when the vehicle traffic light 7a for the first road 5a is displaying a green signal, the area near the crosswalk 5d on the near side of the second road 5b (lower side of Figure 2) is set as the alert area Aw, as shown in Figure 2, and when the vehicle traffic light 7b for the second road 5b is displaying a green signal, the area near the crosswalk 5c on the left side of the first road 5a is set as the alert area Aw.

[0048] Next, the judgment unit 18 determines whether or not there is a high possibility that the moving body 8 will enter the surveillance area Aw based on the relationship between the angle of one of the two direction vectors Vd1, Vd2 generated by the direction vector generation unit 17 and the angle of the movement vector Vm, as described below.

[0049] The determination method in the determination unit 18 will be described below. First, with reference to FIG. 7, a description will be given of an example in which the direction vector Vd1 is generated by the direction vector generation unit 17 and the moving object 8 traveling on the first road 5a turns left at the intersection 5.

[0050] 7, the angle of the direction vector Vd1 relative to the reference line Lx (hereinafter referred to as the "direction vector angle") is θd1, and the angle of the movement vector Vm relative to the reference line Lx (hereinafter referred to as the "movement vector angle") is θm. In this case, the reference line Lx is a line extending in the left-right direction of the traffic environment image 2, and the two angles θd1 and θm have positive values ​​for clockwise angles.

[0051] Furthermore, for ease of understanding, the vector Vmx in the figure is obtained by shifting the movement vector Vm so that the starting point of the movement vector Vm coincides with the starting point of the direction vector Vd, but it is not necessary to shift it in mathematical calculations. Furthermore, dθ is a predetermined angle that defines the angular region (the angular region between the two lines Lj, Lj) for determining the possibility that the moving object 8 will enter the surveillance region Aw.

[0052] In the judgment unit 18, when θd1-dθ<θm<θd1+dθ is satisfied for the movement vector angle θm, it is judged that there is a high possibility that the moving body 8 will enter the alert area Aw, and when this condition is not satisfied, it is judged that there is a low possibility that the moving body 8 will enter the alert area Aw.

[0053] For example, as shown in Figures 8A to 8D, when a moving object 8a traveling on a first road 5a turns left at an intersection 5, θd1+dθ<θm holds while the moving object 8a moves from the position shown in Figure 8A to the position shown in Figure 8C, and therefore the determination unit 18 determines that there is a low possibility that the moving object 8a will enter the alert area Aw (i.e., the area near the crosswalk 5d). Then, when the moving object 8a moves from the position shown in Figure 8C to the position shown in Figure 8D, θm<θd1+dθ holds, and therefore it is determined that there is a high possibility that the moving object 8a will enter the alert area Aw.

[0054] 9, when a moving object 8 traveling on the first road 5a turns right at an intersection 5, the movement vector angle θm changes from a negative value to a positive value as the moving object 8 turns. Until θd1-dθ<θm is established, it is determined that there is a low possibility that the moving object 8 will enter the alert area Aw, and once θd1-dθ<θm is established, it is determined that there is a high possibility that the moving object 8 will enter the alert area Aw.

[0055] 10, when the moving object 8 travels straight through the intersection 5, θd1-dθ<θm<θd1+dθ always holds, and it is determined that there is a high possibility that the moving object 8 will enter the alert area Aw. In this case, an event in which the moving object 8 travels straight through the intersection 5 despite the vehicular traffic light 7b on the second road 5b having a red light occurs, for example, when the second road 5b changes from a green light to a red light while the moving object 8 following the right-turning vehicle remains in the intersection 5.

[0056] Note that in the determination unit 18, when the moving body 8 turns right at the intersection 5, when the moving body 8 turns left at the intersection 5, and when the moving body 8 goes straight at the intersection 5, the value of dθ may be set to different values. This is because when the moving body 8 turns left at the intersection 5, at the timing when -dθ + θd1 < θm < dθ + θd1 holds, the distance between the moving body 8 and the warning area Aw is likely to be smaller compared to the case where the moving body 8 turns right at the intersection 5 and so on.

[0057] Next, with reference to FIG. 11, the determination method of the determination unit 18 when the direction vector Vd2 is generated by the direction vector generation unit 17 will be described. As shown in FIG. 11, when the direction vector Vd2 is generated, the angle of the direction vector Vd2 with respect to the reference line Lx (hereinafter referred to as "direction vector angle") is θd2.

[0058] And when θd2 - dθ < θm < θd2 + dθ holds at the moving vector angle θm, it is determined that the moving body 8 is likely to enter the warning area Aw, and when this condition is not satisfied, it is determined that the moving body 8 is unlikely to enter the warning area Aw. Note that FIG. 11 is an example when the moving body 8 traveling on the second road 5b turns right at the intersection 5, but when the moving body 8 traveling on the second road 5b turns left at the intersection 5 or when the moving body 8 goes straight at the intersection 5, the likelihood of the moving body 8 entering the warning area Aw is also determined by the same method.

[0059] Next, the second traffic area acquisition unit 20 will be described. In this second traffic area acquisition unit 20, the second traffic area is acquired from the traffic environment image 2 by one of the first acquisition process and the second acquisition process described below. This second traffic area corresponds to an estimated traffic area composed of the intersection area and the road area in the traffic environment image 2.

[0060] In the first acquisition process, a second traffic area 43 as shown by dotted lines in Fig. 12 is acquired by a predetermined first image recognition process described below. First, for any one moving object (hereinafter referred to as "moving object A"), the bounding box in the current frame and the bounding box from several frames earlier are acquired, and a line segment connecting the left and right points of the bases of both boxes is generated to generate a rectangle (for example, rectangle 42 in Fig. 12), and the inside of the rectangle is painted, for example, yellow while the background image is colored black.

[0061] By performing this process for several frames in which moving object A is recognized, the trajectory of moving object A sweeping across the road is displayed as a yellow-filled image. The above process is also performed for multiple vehicles other than moving object A over a period of several minutes. Finally, the yellow-filled area is extracted by image processing to obtain a second traffic area 43 as shown by dotted lines in FIG. 12. In this case, a process using a machine learning model (for example, OpenVino: a registered trademark) is used as the predetermined first image recognition process.

[0062] Meanwhile, in the second acquisition process, a second traffic area 44 as shown by dotted lines in Fig. 13 is acquired from the traffic environment image 2 by a predetermined second image recognition process having an area recognition function. Specifically, the area recognition function classifies the traffic environment image 2 into characteristic areas, and extracts road areas from the classified areas to acquire the second traffic area 44. In this case, a process applying a machine learning model (for example, OpenVino: registered trademark) is used as the predetermined second image recognition process.

[0063] Hereinafter, a case will be described in which the second traffic area 43 shown by dotted lines in FIG. 12 is acquired by the second traffic area acquisition unit 20 through the first acquisition process.

[0064] Next, a description will be given of the first traffic area setting unit 19. In this first traffic area setting unit 19, a first traffic area 41 is set based on the enlarged area 41a generated by the center line generating unit 16.

[0065] 6 is set as the intersection area 41a. In addition, the area defined by a straight line L1e that passes through two points Py1 and Py2 and extends to the left and right straight edge portions 2c and 2d of the traffic environment image 2, a straight line L1d that passes through two points Py3 and Py4 and extends to the left and right straight edge portions 2c and 2d of the traffic environment image 2, and the left and right straight edge portions 2c and 2d of the traffic environment image 2, excluding the intersection area 41a (the area shown by hatching), is set as two road areas 41b, 41b corresponding to the first road 5a.

[0066] Furthermore, of the area defined by a straight line L2d that passes through two points Py1 and Py4 and extends to the upper and lower edge straight line portions 2a and 2b of the traffic environment image 2, a straight line L2e that passes through two points Py2 and Py3 and extends to the left and right edge straight line portions 2a and 2b, and the left and right edge straight line portions 2a and 2b, the area (shown by hatching) excluding the intersection area 41a is set as two road areas 41b, 41b corresponding to the second road 5b.The first traffic area 41 is set as an area consisting of the intersection area 41a shown by dotted lines in Figure 6 and four road areas 41b shown by hatching in Figure 6.

[0067] Then, the first traffic area setting unit 19 compares the first traffic area 41 set as described above with the second traffic area 43 acquired as described above by the second traffic area acquisition unit 20. Then, if the first traffic area 41 is contained within the second traffic area 43, the first traffic area 41 is maintained as is.

[0068] 14, when the first traffic area 41 does not fit within the second traffic area 43, a corrected first traffic area 41x is generated by correcting the first traffic area 41 so that it fits within the second traffic area 43. That is, the straight lines L1d and L1e of the first traffic area 41 are corrected to dashed lines L1dx and L1ex, and the straight lines L2d and L2e are corrected to dashed lines L2dx and L2ex, thereby generating the corrected first traffic area 41x. Then, this corrected first traffic area 41x is set as the first traffic area.

[0069] In the following description, a case will be described as an example in which the first traffic area 41 initially set by the first traffic area setting unit 19 is maintained.

[0070] Next, the monitoring area setting unit 21 will be described. When either the vehicular traffic light 7a for the first road 5a or the vehicular traffic light 7b for the second road 5b is displaying a green signal, the monitoring area setting unit 21 sets a combination of a monitoring area for a moving object and a monitoring area for a traffic participant, as described below. In the following description, the monitoring area for a moving object will be referred to as a "moving object monitoring area," the monitoring area for a traffic participant will be referred to as a "participant monitoring area," and the combination of the moving object monitoring area and the participant monitoring area will be referred to as a "monitoring area set."

[0071] First, when the vehicle traffic light 7a for the first road 5a is displaying a green signal, the monitoring area setting unit 21 sets a combination of four moving object monitoring areas Ar0 to Ar3 (areas shown in dotted lines) and four participant monitoring areas At0 to At3 (areas shown in hatched lines) as a monitoring area set, as shown in Figure 15.

[0072] That is, moving object monitoring area Ar0 is set as an area defined by three straight lines L1e, L2c, and L2e and the upper straight line portion 2a of the traffic environment image 2, and moving object monitoring area Ar1 is set as an area defined by four points P12, Py2, P23, and Pc. Furthermore, moving object monitoring area Ar2 is set as an area defined by four points Pc, P23, Py3, and P34, and moving object monitoring area Ar3 is set as an area defined by three straight lines L1d, L2c, and L2e and the lower straight line portion 2b of the traffic environment image 2.

[0073] Meanwhile, participant monitoring area At0 is set as an area defined by two straight lines L1d, L2d and the left straight line portion 2c and the lower straight line portion 2b of the traffic environment image 2, and participant monitoring area At1 is set as an area defined by three straight lines L1d, L2d, L2c and the lower straight line portion 2b of the traffic environment image 2. Furthermore, participant monitoring area At2 is set as the same area as moving object monitoring area Ar3, and participant monitoring area At3 is set as an area defined by two straight lines L1d, L2e and the right straight line portion 2d and the lower straight line portion 2b of the traffic environment image 2.

[0074] On the other hand, when the vehicle traffic light 7b for the second road 5b is displaying a green signal, the monitoring area setting unit 21 sets a combination of four moving object monitoring areas Ar4 to Ar7 (areas shown in dotted lines) and four participant monitoring areas At4 to At7 (areas shown in hatched lines) as a monitoring area set, as shown in Figure 16.

[0075] That is, moving object monitoring area Ar4 is set as an area defined by three straight lines L1c, L1d, and L2e and right edge straight line portion 2d of traffic environment image 2, and moving object monitoring area Ar5 is set as an area defined by four points Pc, P23, Py3, and P34. Furthermore, moving object monitoring area Ar6 is set as an area defined by four points P14, Pc, P34, and Py4, and moving object monitoring area Ar7 is set as an area defined by three straight lines L1c, L1d, and L2d and left edge straight line portion 2c of traffic environment image 2.

[0076] Meanwhile, participant monitoring area At4 is set as an area defined by two straight lines L1d, L2d and the left straight line portion 2c and lower straight line portion 2b of the traffic environment image 2, and participant monitoring area At5 is set as the same area as moving object monitoring area Ar7. Furthermore, participant monitoring area At6 is set as an area defined by three straight lines L1c, L1e, L2d and the left straight line portion 2c of the traffic environment image 2, and participant monitoring area At7 is set as an area defined by two straight lines L1e, L2d and the left straight line portion 2c and upper straight line portion 2a of the traffic environment image 2.

[0077] Furthermore, the participant position acquisition unit 22 acquires the positions of traffic participants in the traffic environment image 2 by the method described below. Specifically, as shown in Fig. 15, a predetermined third image recognition process (machine learning model) uses a bounding box 30 to acquire the positions of traffic participants 9, such as pedestrians, in each frame of the traffic environment image 2. In this case, for example, R-CNN (Region-Convolutional Neural Networks) is used as the predetermined third image recognition process.

[0078] Furthermore, the alert level setting unit 23 sets the alert level to one of four levels LV0 to LV3 based on a combination of the results of the following condition determinations (f1) to (f6). In this case, level LV3 is set as the highest level of danger, and level LV0 is set as the lowest level of danger. (f1) Whether the monitoring area set set in the monitoring area setting unit 21 is a combination of monitoring areas Ar0 to Ar3 and At0 to At3, or a combination of monitoring areas Ar4 to Ar7 and At4 to At7. (f2) Whether or not the moving object 8 is present within the moving object monitoring area set in the monitoring area setting unit 21. (f3) If the moving object 8 is present within a moving object monitoring area, which moving object monitoring area does the moving object 8 exist in? In this case, if at least a portion of the moving object 8 is located within any moving object monitoring area, the moving object 8 is determined to exist in that moving object monitoring area. (f4) If the moving object 8 is present in the moving object monitoring area, whether or not the moving speed of the moving object 8 is equal to or greater than a predetermined value. (f5) Whether or not a traffic participant 9 such as a pedestrian is present within the participant monitoring area set in the monitoring area setting unit 21. (f6) If the traffic participant 9 is present in a participant monitoring area, which participant monitoring area is the traffic participant 9 present in? In this case, if at least a part of the traffic participant 9 is located within any participant monitoring area, the traffic participant 9 is determined to be present in that participant monitoring area.

[0079] An example of setting the alert level in the alert level setting unit 23 will be described below, taking as an example a case where the monitoring area set is set to a combination of monitoring areas Ar0 to Ar3 and At0 to At3. <Setting example 1> If no traffic participant 9 is present in the participant monitoring areas At0 to At3, the alert level is set to level LV0 regardless of whether the moving object 8 is present in the moving object monitoring areas Ar0 to Ar3. <Setting example 2> When the moving object 8 is not present in the moving object monitoring areas Ar0 to Ar3, the alert level is set to level LV0 regardless of whether the traffic participant 9 is present in the participant monitoring areas At0 to At3. <Setting example 3> When it is determined that the moving object 8 is unlikely to enter the alert area Aw, if the traffic participant 9 is present in any of the monitoring areas At0 to At3, the alert level is set to level LV0. <Setting example 4> When a traffic participant 9 is present in the monitoring area At0 or At3 and a moving object 8 is present in the monitoring area Ar1 or Ar2, the alert level is set to level LV1 regardless of the moving speed of the moving object 8. <Setting example 5> When a traffic participant 9 is present in the monitoring area At1 or At2 and a moving body 8 is present in the monitoring area Ar1, if the moving speed of the moving body 8 is below a predetermined value, the alert level is set to level LV2, and if the moving speed of the moving body 8 exceeds the predetermined value, the alert level is set to level LV3. <Setting example 6> When a traffic participant 9 is present in the monitoring area At1 or At2 and the moving object 8 is present in the monitoring area Ar2, the alert level is set to level LV3 regardless of the moving speed of the moving object 8.

[0080] <Setting Example 7> The alert level setting unit 23 may set the alert level as follows: First, the participant position acquisition unit 22 uses a predetermined machine learning algorithm (for example, a deep neural network (DNN)) to learn the change patterns of the display states of the vehicle traffic lights 7a for the first road 5a and the vehicle traffic lights 7b for the second road 5b.

[0081] Furthermore, just before the vehicular traffic light 7b for the second road 5b changes from green to red, the monitoring area setting unit 21 sets the monitoring area set to a combination of the monitoring areas Ar0-Ar3 and At0-At3, and both the monitoring areas Ar4-Ar7 and At4-At7. In this state, if a moving object 8 is present in either the moving object monitoring area Ar1 or Ar2 and a traffic participant 9 is present in either the monitoring area At0 or At3, the alert level is set to level LV3 before the vehicular traffic light 7a for the first road 5a changes from red to green.

[0082] Next, we will explain the output processing unit 24. In this output processing unit 24, based on the determination result of the determination unit 18 and the alert level set by the monitoring area setting unit 21, an alert information output process is executed as described below.

[0083] In this alarm information output process, if both of the following conditions (f10) and (f11) are met, an alarm information signal is output to the speaker 4. This alarm information signal is configured to include audio information such as "vehicle approaching." (f10) The determination unit 18 determines that there is a high possibility that the moving object 8 will enter the alert area Aw. (f11) In the alert level setting unit 23, the alert level is set to level LV2 or LV3.

[0084] When the alarm information signal is received by the speaker 4, the speaker 4 outputs audio information (alarm information) such as "vehicle approaching."

[0085] For example, in the alert level setting unit 23, when the alert level is set to level LV2, audio information "Vehicle approaching" is output as a gentle warning, and when the alert level is set to level LV3, audio information "Vehicle approaching, watch out!" is output to notify that a state requiring caution is in progress.

[0086] Furthermore, in the case of the above-mentioned <Setting Example 7>, audio information is output saying, "Vehicle approaching, watch out for vehicles jumping out at the green light!" This allows the traffic participants 9 to be appropriately notified that there is a high possibility that the moving object 8 will enter the alert area Aw.

[0087] In addition, in the warning information output process, if the above-mentioned condition (f10) is met and the traffic participant 9 is present in the warning area Aw or its vicinity, the warning information signal is output, and if the above-mentioned condition (f10) is met and the traffic participant 9 is not present in the warning area Aw or its vicinity, the warning information signal may not be output.

[0088] As described above, the intersection warning system 1 of this embodiment determines whether or not there is a high possibility that the moving body 8 will enter the predetermined warning area Aw, based on the movement vector angle θm of the moving body 8 with respect to one of the direction vector angles θd1, θd2 of the direction vectors Vd1, Vd2. When it is determined that there is a high possibility that the moving body 8 will enter the predetermined warning area Aw, an alert information signal is output to the speaker 4 to notify the traffic participants 9 of this.

[0089] As a result, by outputting a voice such as "Vehicle approaching, please be careful" from the speaker 4, if a traffic participant 9 is in or near a specified alert area Aw, the traffic participant 9 can be notified that there is a high possibility that the moving body 8 will enter the specified alert area Aw, thereby improving the safety of the traffic participant 9.

[0090] In addition, by determining the direction vector Vd1 or the direction vector Vd2 based on the state of the vehicle traffic light 7 at the intersection 5, the direction vector Vd1 or the direction vector Vd2 can be appropriately set at the intersection 5 where the vehicle traffic light 7 is located.

[0091] Furthermore, the number of intersections between straight lines extending along the movement vector Vm and the edge straight line portions 2a-2d of the traffic environment image 2 is counted for each frame, and histograms 31 and 32 are created as graphs representing the integrated values ​​of these counting results. The midpoints of the longest areas of the histograms 31 and 32 are set as the entry point Pin and exit point Pout of the moving object 8. Furthermore, based on these entry point Pin and exit point Pout, the traveling lines L1a and L1b of the moving object 8 on the first road 5a and the traveling lines L2a and L2b of the moving object 8 on the second road 5b are set, and a center line L1c of the first road area and a center line L2c of the second road area are generated based on the four traveling lines L1a, L1b, L2a, and L2b. This allows these center lines L1c and L2c to be generated with high accuracy.

[0092] Furthermore, the center point Pc of the intersection 5 is set as the intersection of the two center lines L1c, L2c, and the direction vectors Vd1, Vd2 are generated as vectors that start at the center point Pc of the intersection 5 and extend along the two center lines L1c, L2c. This allows the direction vectors Vd1, Vd2 to be appropriately generated so as to point from the center point of the intersection 5 toward the warning area Aw.

[0093] Furthermore, by one of the first acquisition process and the second acquisition process described above, the second traffic area 43 is acquired from the traffic environment image 2. By acquiring the second traffic area 43 in the above manner, the second traffic area 43 can be acquired as being close to an actual road area.

[0094] Meanwhile, an expanded area 41a is set based on the four driving lines L1a, L1b, L2a, and L2b, and a first traffic area 41 is set based on this expanded area 41a. The first traffic area 41 is compared with the second traffic area 43, and if the first traffic area 41 fits within the second traffic area 43, the first traffic area 41 is maintained as is. On the other hand, if the first traffic area 41 does not fit within the second traffic area 43, the first traffic area 41 is corrected to fit within the second traffic area 43, thereby generating a corrected first traffic area 41x, and this corrected first traffic area 41x is set as the first traffic area. As described above, if the first traffic area 41 does not fit within the second traffic area 43, the first traffic area 41 is corrected to fit within the second road area 43, thereby preventing the first traffic area 41 from becoming an area larger than the actual road area and improving the setting accuracy of the first traffic area 41.

[0095] Furthermore, the alert level is set based on the relationship between the position of the traffic participant 9 and the monitoring area At0 to At3 or the monitoring area At4 to At7, as well as the relationship between the position of the moving body 8 and the monitoring area Ar0 to Ar3 or the monitoring area Ar4 to Ar7, so that an appropriate level of alert information can be notified to the traffic participant 9.

[0096] The method for generating a direction vector in the direction vector generating unit 17 is not limited to the method in the embodiment, and the following method may also be used. For example, the direction vector generating unit 17 may be configured to generate a direction vector that is approximately parallel to the center line L2c instead of the direction vector Vd1 when the vehicle traffic light 7 on the first road 5a is displaying a green light. That is, the direction vector that is generated may be a direction vector extending from point Px2 to point Px3, a direction vector extending from midpoint P23 to point Py3, or the like.

[0097] Also, for example, the direction vector generating unit 17 may be configured to receive a control signal from the vehicular traffic signal 7 and generate one of the direction vectors Vd1 and Vd2 based on this control signal. Furthermore, the direction vector generating unit 17 may be configured to determine the signal display state of the vehicular traffic signal 7 before generating the four direction vectors Vd1 to Vd4, and to generate one of the direction vectors Vd1 and Vd2 based on the determination result.

[0098] In addition, for example, when the vehicular traffic light 7 on the first road 5a is displaying a green signal, the direction vector generation unit 17 may generate a direction vector Vd3 in addition to the direction vector Vd1. In this case, the area near the crosswalk 5d on the far side of the second road 5b (upper side in FIG. 2) may be set as a security area, and whether or not there is a high possibility that the moving object 8 will enter the security area may be determined based on the angle between the direction vector Vd3 and the movement vector Vm. Similarly to the above, for example, when the vehicular traffic light 7 on the second road 5b is displaying a green signal, the direction vector generation unit 17 may generate a direction vector Vd4 in addition to the direction vector Vd2.

[0099] In addition, at intersections without traffic lights, traffic participants, intersection areas, and road areas may be recognized by image recognition processing, and the direction vector generation unit 17 may generate a direction vector that points from near the center of the intersection area toward the nearby road area where the traffic participants are located.

[0100] In the embodiment, the counting unit 13 counts the number of intersections between a straight line extending along the movement vector Vm and the edge straight line portions 2a to 2d of the traffic environment image 2. Alternatively, four rectangular straight line portions (line segments) may be arranged to surround the intersection area 41a in the traffic environment image 2, and the counting unit 13 may count the number of intersections between a straight line extending along the movement vector Vm and each of the four straight line portions. In this case, the four rectangular straight line portions may be each parallel to a side of the intersection area 41a, or each parallel to the edge straight line portions 2a to 2d of the traffic environment image 2.

[0101] In the embodiment, a computer is used as the alarm device 10, but a server or a cloud server may be used instead. In that case, the alarm device 10 may be configured to communicate with the camera 3 and the speaker 4 via a communication network such as the Internet or a LAN.

[0102] In the embodiment, the speaker 4 is used as the notification device, but the notification device of the present invention is not limited to this, and may be any device that notifies traffic participants of warning information as at least one of audio information and visual information. For example, a warning light or the like may be used as the notification device, or a warning light and a speaker 4 may be used together. Furthermore, two speakers 4 may be provided near the alert area of ​​the first road 5a and near the alert area of ​​the second road 5b, respectively.

[0103] The embodiment is an example in which a video captured by a camera 3 is used as the traffic environment image 2, but instead, still images may be captured continuously by the camera 3 at extremely short time intervals, and these still images may be used as the traffic environment image 2.

[0104] In the embodiment, the traffic environment image 2 is taken by the camera 3, but an imaging device other than the camera 3 that can take the traffic environment image 2 may be used.

[0105] The embodiment is an example in which the intersection warning system 1 is applied to a crossroad-shaped intersection 5, but the intersection warning system of the present invention is not limited to this and can be applied to intersections of various shapes. For example, the intersection warning system of the present invention may be applied to a four-way intersection where one of two roads intersects with the other at an angle, or to a T-shaped or "T"-shaped three-way intersection.

[0106] In the embodiment, a four-wheeled vehicle is shown as the moving object 8, but the moving object of the present invention is not limited to this and may be anything that moves within an intersection. For example, a motorcycle or the like may also be used as the moving object.

[0107] In the embodiment, the intersection warning system 1 is configured to include the camera 3, the speaker 4, and the warning device 10, but the intersection warning system may also be configured to include only the warning device 10.

[0108] In the intersection warning system 1, the second traffic area acquisition unit 20 may be omitted, in which case the first traffic area may be set by the first traffic area setting unit 19 alone. [Explanation of symbols]

[0109] 1. Intersection Warning System 2. Traffic environment images 2a~2d Straight section 4 Speaker (alarm device) 5 Intersection 7 Traffic lights 8,8a Mobile 9 Transportation participants 10 Warning device 11 Mobile object position acquisition section 12. Motion vector generation unit 13 Counting Section 14. Inflow / Outflow Point Setting Section 15 Driving line setting section 16 Center line generation unit 17 Direction vector generator 18 Judgment section 19 1st traffic area setting section 20 Second traffic area setting section 21 Monitoring area setting section 22 Participant location acquisition section 23 Alert Level Setting Unit 24 Output Processing Section 41 1st traffic area 41a Intersection area, expansion area 41b Road area 43 Second traffic area 44 Second traffic area Vm movement vector θm Angle of the movement vector Vd1,Vd2 direction vector θd1,θd2 Angle of direction vector Aw warning area Pin Inflow point Pout Outlet point L1a, L1b driving lines L2a, L2b driving lines Px1~Px4 intersection L1c, L2c center line PC center point Ar0~Ar7 Mobile monitoring area At0~At7 Participant monitoring area

Claims

1. a mobile object position acquisition unit that acquires a position of a mobile object by a predetermined image recognition process from a traffic environment image including the mobile object moving in an intersection area that is an area of ​​the intersection and any of a plurality of road areas that are continuous with the intersection area; a motion vector generation unit that generates a motion vector of the moving object based on a time series of positions of the moving object; a direction vector generation unit that generates a direction vector from the intersection area toward a security area that is at least one area of ​​the plurality of road areas; a determination unit that determines the likelihood of the moving object entering the security area based on an angle of the movement vector of the moving object with respect to the direction vector; an output processing unit that, when the determination unit determines that there is a high possibility that the moving object will enter the alert area, executes an output process to output alert information representing the determination result to notify traffic participants; An intersection warning system comprising:

2. 2. The intersection warning system according to claim 1, An intersection warning system, wherein the direction vector generation unit determines the orientation of the direction vector based on the display state of a traffic signal at the intersection.

3. 2. The intersection warning system according to claim 1, a counting unit configured to count the number of intersections between a line extending along the movement vector and each of the four rectangular straight line sections surrounding the intersection area in the traffic environment image, when each of the four rectangular straight line sections is divided into a plurality of areas; and an inflow / outflow point setting unit that sets the region with the largest number of intersections counted by the counting unit among the plurality of regions in each of the straight line sections as an inflow point and an outflow point of the moving body based on the direction of the movement vector; a travel line setting unit that sets a straight line connecting the inflow point and the outflow point in each of the two straight line portions that are opposed to each other as a travel line of the moving object in the road area; a centerline generating unit that generates a centerline of a first road area and a centerline of a second road area based on an enlarged area obtained by enlarging a rectangular area defined by four intersections between the driving line of a first road area, which is one of the plurality of road areas, and the driving line of a second road area that intersects with the first road area; Furthermore, An intersection warning system characterized in that the direction vector generation unit generates the direction vector so that it extends from a center point which is the intersection of the center line of the first road area and the center line of the second road area toward the warning area while following one of the center line of the first road area and the center line of the second road area.

4. 4. The intersection warning system according to claim 3, a first traffic area setting unit that sets a first traffic area including the intersection area and the plurality of road areas based on the enlarged area; a second traffic area acquisition unit that executes one of a first acquisition process that acquires a movement area of ​​the moving object on a road surface based on a time series of positions of the moving object by a predetermined first image recognition process, and acquires a second traffic area consisting of the intersection area and the plurality of road areas based on the time series of the movement area, and a second acquisition process that acquires the second traffic area from the traffic environment image by a predetermined second image recognition process; Furthermore, An intersection warning system, wherein the first traffic area setting unit sets the first traffic area so that the first traffic area is contained within the second traffic area.

5. 4. The intersection warning system according to claim 3, a first traffic area setting unit that sets a first traffic area including the intersection area and the plurality of road areas based on the enlarged area; a monitoring area setting unit that sets a plurality of rectangular monitoring areas in the first traffic area based on the direction of the direction vector, the center line of the first road area, and the center line of the second road area; a participant position acquisition unit that acquires positions of traffic participants other than the moving object from the traffic environment image by a predetermined third image recognition process; an alert level setting unit that sets an alert level based on the relationship between the positions of the traffic participants, the positions of the moving objects, and the plurality of monitoring areas; Furthermore, An intersection warning system, wherein the warning information is configured to include information corresponding to the warning level.

6. 6. The intersection warning system according to claim 5, An intersection warning system characterized in that the warning level setting unit sets the warning level based on the display status of traffic signals at the intersection in addition to the relationship between the positions of the traffic participants, the positions of the moving bodies, and the multiple monitoring areas.

7. The intersection warning system according to any one of claims 1 to 6, The intersection warning system further comprises a notification device that notifies the traffic participants of the warning information as at least one of audio information and visual information.

Images