JP2012014527A - Pedestrian detection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a pedestrian detection device capable of accurately detecting a pedestrian for a vehicle taking a right or left turn, the pedestrian who is crossing a pedestrian crosswalk which a vehicle will cross after taking the right or left turn.SOLUTION: A pedestrian detection device comprises: image acquisition means to acquire an image of the surroundings of a vehicle; right/left turn prediction means to predict that the vehicle is taking a right or left turn; search mode setting means to select, based on right/left turn prediction information, either a search area-equalized mode to search inside an image region to be acquired with pre-set predetermined accuracy or a search area-biased mode to divide an image region into a plurality of search areas and search the areas with respectively different accuracy levels based on the right/left turn prediction information; pedestrian search means to search for a pedestrian based on the search area mode; and result report means to report the search result. The search mode setting means sets the search area mode based on a distance between the vehicle and an intersection where the vehicle intends to take the right or left turn.

Description

  The present invention relates to a pedestrian detection device mounted on a vehicle, and more particularly to a pedestrian detection device that detects a pedestrian while setting an optimal pedestrian search area.

  Conventionally, a pedestrian detection device mounted on a vehicle detects a pedestrian existing in front of the vehicle or the like from an image acquired by a camera. And if there is a risk of collision between the vehicle and the detected pedestrian, a pre-crash safety system has been developed that warns the driver and controls the traveling speed of the vehicle to avoid the collision. Yes.

  Furthermore, although obstacles are usually detected in the entire image area that can be acquired by a camera or the like, the obstacle detection device disclosed in Patent Document 1 is expected to have a high possibility of collision between the vehicle and the obstacles. The obstacle detection area corresponding to the road condition is set at the location.

  Specifically, the obstacle detection device disclosed in Patent Literature 1 sets an obstacle detection region based on map information recorded in the car navigation device and image information acquired by the camera, Furthermore, an obstacle detection area corresponding to the predicted course of the vehicle is set according to the driver's steering operation (steer angle), vehicle speed, and the like while the vehicle is traveling.

JP 2009-271766 A

  However, in the obstacle detection device disclosed in Patent Document 1, when a vehicle turns right or left, an obstacle detection region is specially considered for a pedestrian crossing a pedestrian crossing immediately after the vehicle turns right or left. Since it is not set, the pedestrian cannot be detected accurately.

  Therefore, an object of the present invention is to provide a pedestrian detection device capable of accurately detecting a pedestrian crossing a pedestrian crossing immediately after the vehicle turns right or left when the vehicle turns right or left. is there.

In order to achieve the above object, a pedestrian detection device of the present invention is a pedestrian detection device mounted on a vehicle, wherein image acquisition means for acquiring an image around the vehicle, and the vehicle turns right or left. A left / right turn prediction means for predicting, a search range equalization mode for searching in an image area acquired by the image acquisition means with a predetermined accuracy, and an image area acquired by the image acquisition means as a plurality of search areas The search range mode setting means for setting one of the search range modes of the search range bias mode for dividing and searching with different accuracy based on the prediction information by the right / left turn prediction means and the search range mode setting means are set. Based on the search range mode, the pedestrian search means for searching for a pedestrian from the image acquired by the image acquisition means, and the search by the pedestrian search means. And a result notification means for notifying a result, the search range mode setting means, a vehicle, based on the distance between the intersection which the vehicle turns right or left turn, and sets the search range mode.
With this configuration, the search range equalization mode for searching the image area acquired by the image acquisition means with a predetermined accuracy set in advance, and the image area acquired by the image acquisition means are divided into a plurality of search areas, In order to set one of the search range modes with the search range bias mode to search with different accuracy based on the prediction information by the right / left turn prediction means, when the vehicle turns right or left, immediately after the vehicle turns right or left It is possible to accurately detect a pedestrian crossing a pedestrian crossing.

A preferable search range bias mode is characterized in that a specific region for searching at least one of a plurality of divided search regions with high accuracy is set.
With this configuration, since at least one of the divided search areas is searched with high accuracy, a pedestrian can be detected with high efficiency.

A preferable search range bias mode is characterized in that a region other than the specific region is searched with low accuracy among a plurality of divided search regions.
With this configuration, since a specific area to be searched with high accuracy and an area other than the specific area to be searched with low accuracy are set, it is possible to focus on searching for pedestrians in areas where there is a high risk of collision with the vehicle. it can.

The preferred search range mode setting means sets the search range bias mode as the search range mode at a first preset timing close to an intersection where the vehicle is predicted to turn right or left by the right / left turn prediction means. It is characterized by.
With this configuration, the search range bias mode is set as the search range mode at the first preset timing close to the intersection where the vehicle is predicted to turn right or left, so the pedestrian crossing immediately after the vehicle turns right or left A pedestrian crossing the road can be detected with high accuracy.

More preferably, the search range mode setting means sets the search range equal mode to the search range mode at the second timing set in advance after the vehicle starts turning right or left after the vehicle is predicted to turn right or left by the right / left turn prediction means. It is characterized by setting as.
With this configuration, the pedestrian crossing the pedestrian crossing immediately after the vehicle has made a right or left turn in order to set the search range equal mode as the search range mode at the second timing set in advance after the vehicle starts turning right or left. It can be detected in a wide range with high accuracy.

Further, the more preferable search range mode setting means includes a vehicle position calculation means for calculating a distance between the vehicle and an intersection predicted to turn right or left by the right / left turn prediction means, and a distance calculated by the vehicle position calculation means. If the distance is determined to be less than the threshold by the vehicle position determination means and the vehicle position determination means to determine whether or not is less than a preset threshold, the search range bias mode is selected as the search range mode, And a search range mode selection unit that selects the search range equal mode as the search range mode when the vehicle position determination unit determines that the distance is equal to or greater than the threshold value.
With this configuration, the vehicle at the intersection is selected in order to select one of the search range mode between the search range bias mode and the search range equal mode based on the distance between the vehicle and the intersection where the vehicle is predicted to turn right or left. A pedestrian can be detected with high accuracy according to the traveling situation.

More preferably, after the left turn of the vehicle is started at the intersection where the vehicle is predicted to turn left by the right / left turn prediction means, the vehicle position determination means has a distance calculated by the vehicle position calculation means that is less than the travel lane width after the left turn of the vehicle. The search range mode selection means sets the search range equal mode as the search range mode when the vehicle position determination means determines that the distance is less than the travel lane width. To do.
With this configuration, the left turn state of the vehicle is selected in order to select one of the search range bias mode and the search range equal mode based on the distance between the vehicle and the intersection where the vehicle is predicted to turn left. Accordingly, it is possible to accurately detect a pedestrian in a wide range.

More preferably, after the vehicle starts to turn right at the intersection predicted to turn right by the right / left turn prediction means, the vehicle position determination means determines whether the vehicle has passed the intersection, and the search range mode selection means is When the vehicle position determination means determines that the vehicle has passed the intersection, the search range equalization mode is set as the search range mode.
With this configuration, a pedestrian is selected according to the right turn state of the vehicle in order to select one of the search range bias mode and the search range equal mode based on whether or not the vehicle has passed the intersection. Can be accurately detected over a wide range.

Preferably, the vehicle further includes a winker control means for controlling a winker indicating a right turn or a left turn of the vehicle, and the second timing is a timing at which the winker indicating a right turn or a left turn of the vehicle is canceled by the winker control means. It is characterized by being.
With this configuration, the search range mode is selected from the search range bias mode and the search range equal mode based on the blinker cancellation, so that the pedestrian can be accurately widened according to the right or left turn state of the vehicle. Can be detected.

Preferably, if the vehicle is predicted to turn left by the right / left turn prediction means, the left area in the image area is searched with higher accuracy than the right area in the image area, and the vehicle is detected by the right / left turn prediction means. Is predicted to turn right, the right area in the image area is searched with higher accuracy than the left area in the image area.
With this configuration, based on the prediction that the vehicle will turn left or right, a region to be searched with high accuracy is set in the image region, so that a pedestrian with a high risk of collision with the vehicle can be detected with high accuracy. it can.

Further, a preferable search range bias mode is characterized in that a specific area is searched with high accuracy among a plurality of divided search areas, and an area other than the specific area is not searched.
With such a configuration, it is possible to intensively search for a pedestrian who has a high risk of collision with the vehicle.

Preferably, the high accuracy is characterized by high load and / or high definition.
With this configuration, a pedestrian can be detected more accurately.

  In order to achieve the above object, a pedestrian detection method of the present invention is a pedestrian detection method executed by a pedestrian detection device mounted on a vehicle, an image acquisition step for acquiring an image around the vehicle, and a vehicle Obtained in a right / left turn prediction step for predicting that the vehicle will turn right or left, a vehicle position calculation step for calculating the distance between the vehicle and the intersection where the vehicle is predicted to turn right or left in the right / left turn prediction step, and an image acquisition step Search range equalization mode for searching within a predetermined image area with a predetermined accuracy, and search range bias mode for dividing the image area acquired at the image acquisition step into a plurality of search areas and searching with different accuracy A search range mode setting step for setting any one of the search range modes based on the distance between the vehicle and the intersection calculated in the vehicle position calculation step; Based on the search range mode set in the search range mode setting step, a pedestrian search step for searching for a pedestrian from the image acquired in the image acquisition step, and a result notification for notifying a result searched for in the pedestrian search step Steps.

  Moreover, in order to achieve the said objective, each process which each structure of the pedestrian detection apparatus of this invention mentioned above can be regarded as a pedestrian detection method which gives a series of processing procedures. This method is provided in the form of a program for causing a computer to execute a series of processing procedures. This program may be installed in a computer in a form recorded on a computer-readable recording medium.

  As described above, according to the pedestrian detection device of the present invention, when a vehicle turns right or left, a pedestrian search area that predicts that the vehicle turns right or left and searches for pedestrians based on the prediction. Therefore, it is possible to accurately detect a pedestrian who crosses the pedestrian crossing immediately after the vehicle turns right or left. Furthermore, in order to broaden the pedestrian search area for searching for pedestrians according to the right or left turn state of the vehicle, a wide range of pedestrians can be accurately detected when the vehicle makes a right or left turn and immediately after a right or left turn. Can be detected.

The figure which shows the pedestrian detection apparatus 100 which concerns on the 1st Embodiment of this invention. The block diagram which shows functionally the pedestrian detection apparatus 100 which concerns on the 1st Embodiment of this invention. The flowchart which shows the pedestrian detection method (left turn) 300 which the pedestrian detection apparatus 100 which concerns on the 1st Embodiment of this invention performs. The figure which shows a mode that the camera mounted in the vehicle images the front periphery of the said vehicle The figure which shows the front periphery image of the said vehicle imaged with the camera mounted in the vehicle as shown in FIG. The figure which shows a mode that a vehicle is predicted to turn left by the route guidance in a car navigation system The figure which shows 1st distance P1 The figure which shows the method of calculating 1st distance P1 based on GPS information and map information The figure which shows a mode that the vehicle approaches the intersection predicted to turn left and gazes at the left side of the vehicle The figure which shows a mode that the left turn state of a vehicle progresses in the intersection where a vehicle is predicted to turn left, and it gazes at the left side and the right side of a vehicle The figure which shows the calculation load ratio in the left side area | region of a vehicle, and the right side area | region of a vehicle The figure which shows a mode that the calculation load ratio in the left side area | region of a vehicle and the right side area | region of a vehicle is changed gradually. The flowchart which shows the pedestrian detection method (right turn) 400 which the pedestrian detection apparatus 100 which concerns on the 1st Embodiment of this invention performs. The figure which shows a mode that a vehicle is predicted to turn right by route guidance in a car navigation system The figure which shows 1st distance P1 The figure which shows a mode that a vehicle approaches the intersection predicted to turn right and gazes at the right side of the vehicle The figure which shows a mode that vehicles do not pass the intersection The figure which shows a mode that the right turn state of a vehicle progresses in the intersection where a vehicle is predicted to turn right, and it gazes at the left side and the right side of the vehicle The figure which shows the pedestrian detection apparatus 500 which concerns on the 2nd Embodiment of this invention. The block diagram which shows functionally the pedestrian detection apparatus 500 which concerns on the 2nd Embodiment of this invention. The flowchart which shows the pedestrian detection method (left turn) 700 which the pedestrian detection apparatus 500 which concerns on the 2nd Embodiment of this invention performs.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
<First Embodiment>
FIG. 1 is a diagram showing a pedestrian detection device 100 according to the first embodiment of the present invention. 1, a pedestrian detection device 100 includes a pedestrian detection camera 110, a pedestrian detection determination ECU (Electronic Control Unit) 120, a search range control ECU 130, a display display control ECU 140, a display display 150, And a car navigation ECU 160.

  In FIG. 1, a pedestrian detection camera 110 is mounted, for example, in the front part of a vehicle, and mainly captures an image for detecting a pedestrian existing in front of the vehicle. Specifically, an infrared camera is used as the pedestrian detection camera 110.

  Pedestrian detection judgment ECU120 searches for a pedestrian from the image (imaging information) imaged by the pedestrian detection camera 110, and judges whether a pedestrian exists. The pedestrian detection determination ECU 120 searches for a pedestrian in an area where attention should be paid to the presence of the pedestrian based on the search range information from the search range control ECU 130.

  The search range control ECU 130 determines a region where attention should be paid to the presence of a pedestrian based on the traveling road information and planned travel route information from the car navigation ECU 160, and controls the search range for searching for the pedestrian. It should be noted that the region that should be noted of the presence of a pedestrian and the search range are controlled based on the road information on travel from the car navigation ECU 160 and the planned travel route information. A detailed description of the control method will be described later.

  The display control ECU 140 controls the screen displayed on the display 150 based on the display display information from the pedestrian detection determination ECU 120 and the pedestrian detection information. For example, the display control ECU 140 may display an image captured by the pedestrian detection camera 110 on the display 150 and display the pedestrian detected in the image area with emphasis. Note that the display control ECU 140 for display and the pedestrian detection determination ECU 120 communicate via, for example, a digital signal line. Further, when a pedestrian is detected in the image area, the detected pedestrian may be displayed on the display for display, or may be notified to the driver, for example, by sounding a warning sound.

  The display for display 150 displays a screen based on the screen information from the display control ECU 140 for display.

  The car navigation ECU 160 is based on GPS (Global Positioning System) information and map information recorded in a memory or the like, information on the road on which the vehicle is currently traveling (traveling road information), and from the current location to the destination. The search range control ECU 130 is notified of the route information (scheduled travel route information). The car navigation ECU 160 and the search range control ECU 130 communicate via, for example, a CAN (Controller Area Network) communication line.

  FIG. 2 is a block diagram functionally showing the pedestrian detection apparatus 100 according to the first embodiment of the present invention. In FIG. 2, the pedestrian detection apparatus 100 includes an image acquisition unit 210, a pedestrian search unit 220, a search range mode setting unit 230, a result notification unit 240, a right / left turn prediction unit 250, and a recording unit 260. Prepare.

  The image acquisition unit 210 captures an image of the vehicle periphery using the pedestrian detection camera 110 and acquires an image of the vehicle periphery.

  The pedestrian search means 220 searches for a pedestrian from the image acquired by the image acquisition means 210 using the pedestrian detection determination ECU 120. Note that the pedestrian search means 220 searches the image area acquired by the image acquisition means 210 with a constant calculation load, or searches a specific area in the image area intensively.

  The search range mode setting unit 230 uses the search range control ECU 130 to determine a region where attention should be paid to the presence of a pedestrian based on the traveling road information and the planned traveling route information from the car navigation ECU 160, and image acquisition unit A search range mode for searching for a pedestrian from the image acquired by 210 is set.

  More specifically, search range mode setting means 230 includes vehicle position calculation means 231, vehicle position determination means 232, and search range mode selection means 233.

  The vehicle position calculation means 231 calculates the distance between the vehicle and the intersection where the vehicle turns right or left. Specifically, the vehicle position calculation means 231 uses the car navigation ECU 160 to acquire the position of the vehicle based on GPS information and map information in the car navigation system, and calculates the distance from the intersection.

  The vehicle position determination unit 232 determines whether the distance between the vehicle and the intersection calculated by the vehicle position calculation unit 231 is less than a preset threshold value. In other words, the vehicle position determination unit 232 determines whether or not the vehicle is approaching the intersection. Furthermore, the vehicle position determination means 232 determines whether the vehicle has made a right turn or a left turn at the intersection (whether the right turn or the left turn state has advanced).

  When the vehicle position determination unit 232 determines that the vehicle is approaching the intersection, the search range mode selection unit 233 selects the search range mode as the search range mode. Then, the search range mode selection means 233 selects the search range equal mode as the search range mode when the vehicle position determination means 232 determines that the vehicle has turned right or left at the intersection (the right turn or left turn state has advanced). To do. Here, the search range equal mode is a search within the image area with a predetermined constant calculation load, and the search range bias mode is a first search area searched with a high calculation load in the image area. It means setting the second search area to be searched with a low calculation load.

  In addition, the pedestrian search means 220 mentioned above searches a pedestrian from the image acquired by the image acquisition means 210 with the search range and calculation load based on the search range mode selected by the search range mode selection means 233. Details of the search range and calculation load corresponding to the search range mode will be described later.

  The result notification means 240 notifies the result searched by the pedestrian search means 220. Specifically, the result notification unit 240 may display the detected pedestrian on the display 150, and if there is a risk of collision between the vehicle and the detected pedestrian, avoid the collision. The driver may be warned to do so.

  The right / left turn predicting means 250 is typically a car navigation system, which obtains a planned travel route to the destination based on GPS information and map information, and the vehicle turns right or left from the planned travel route. Predict that.

  The recording unit 260 is typically a memory or the like, and the recording unit 260 records map information in the car navigation system, specific setting contents of the search range mode described above, and the like.

  Next, the flow of processing executed by the pedestrian detection device 100 according to the first embodiment of the present invention will be described.

  FIG. 3 is a flowchart showing a pedestrian detection method (left turn) 300 executed by the pedestrian detection apparatus 100 according to the first embodiment of the present invention. In FIG. 3, the pedestrian detection method (left turn) 300 includes search range equal mode setting step S310, left turn prediction step S320, first distance calculation step S330, first distance determination step S340, and search range bias mode setting. It includes step S350, second distance calculation step S360, second distance determination step S370, and search range equal mode setting step S380.

  Usually, a pedestrian detection device mounted on a vehicle searches for a pedestrian with a certain calculation load in an entire image area that can be acquired by a camera or the like. Furthermore, here, it is assumed that route guidance to the destination is carried out by a car navigation system mounted on the vehicle.

  First, in the search range equal mode setting step S310, the search range mode setting means 230 selects the search range equal mode as the search range mode. At the same time, the image acquisition unit 210 images the front periphery of the vehicle and acquires the front periphery image of the vehicle. FIG. 4 is a diagram illustrating a state in which the camera mounted on the vehicle images the front periphery of the vehicle. In FIG. 4, the image acquisition unit 210 images the entire area that can be acquired by the camera. FIG. 5 is a view showing a front peripheral image of the vehicle imaged by the camera mounted on the vehicle as shown in FIG. The pedestrian search means 220 searches for a pedestrian from the image acquired by the image acquisition means 210 based on the search range equal mode selected by the search range mode setting means 230, and as shown in FIG. A pedestrian is detected from the front peripheral image.

  As described above, the pedestrian detection device 100 normally mounted on a vehicle captures an image with a camera mounted on the vehicle, and searches for the pedestrian in the front peripheral image area of the acquired vehicle with a certain calculation load. ing.

  Next, in the left turn prediction step S320, the right / left turn prediction means 250 predicts that the vehicle will turn left. For example, it can be predicted that the vehicle will turn left by route guidance in the car navigation system. FIG. 6 is a diagram illustrating a state in which the vehicle is predicted to turn left by route guidance in the car navigation system. In FIG. 6, the vehicle is approaching an intersection where the vehicle is predicted to turn left by the right / left turn prediction means 250.

  In the first distance calculation step S330, the vehicle position calculation means 231 calculates a first distance P1 that is the distance between the vehicle and the intersection where the vehicle turns to the left. FIG. 7 is a diagram illustrating the first distance P1. As shown in FIG. 7, the first distance P1 is a distance along the traveling lane direction of the vehicle, and is a distance between the center point of the vehicle and the center of the intersection where the vehicle is predicted to turn left.

Specifically, the first distance P1 is calculated based on GPS information and map information. FIG. 8 is a diagram illustrating a method of calculating the first distance P1 based on the GPS information and the map information. In FIG. 8, the map information is composed of nodes and links. As shown in FIG. 8, the link length A between the node a and the node b, the link length B between the node b and the node c, the link length C between the node c and the node d, and the link length between the node d and the node e. E. Further, when a vehicle exists between the node a and the node b and at a distance X from the node a, the first distance P1 from the vehicle to the center node e of the intersection is expressed by the following (Equation 1). Is calculated by
P1 = (A−X) + B + C + D (Equation 1)
The distance X from the node a may be calculated and estimated based on the elapsed time after the vehicle passes through the node a and the average speed after the vehicle passes through the node a.

  In the first distance determination step S340, the vehicle position determination means 232 determines whether or not the first distance P1 calculated in the first distance calculation step S330 is less than the first threshold value L1. Here, the first threshold L1 is a preset distance. In other words, the vehicle position determination unit 232 determines whether or not the vehicle is approaching the intersection based on the first distance P1.

  When it is determined in the first distance determination step S340 that the first distance P1 is less than the first threshold L1, the process proceeds to the search range deviation mode setting step S350 (Yes in the first distance determination step S340).

  On the other hand, when it is determined in the first distance determination step S340 that the first distance P1 is greater than or equal to the first threshold L1, the process returns to the first distance calculation step S330 (No in the first distance determination step S340).

  In other words, the first distance calculation step S330 and the first distance determination step S340 are repeated at predetermined time intervals to monitor whether or not the vehicle is approaching the intersection.

  Note that the first threshold value L1 is set in consideration of a range to be watched by a pedestrian on the left side of the vehicle when approaching an intersection where the vehicle is predicted to turn left and when the vehicle subsequently turns left. Specifically, when the vehicle turns left, there is a risk of involving a pedestrian on the left side of the vehicle, so that the pedestrian is appropriately detected by gazing in advance before the vehicle turns left. Moreover, about the pedestrian who crosses the pedestrian crossing immediately after a vehicle turns left, it observes beforehand before a vehicle turns left, and detects a pedestrian appropriately.

  The first threshold L1 may be recorded in advance in the recording unit 260, and a plurality of first threshold values L1 may be recorded for each intersection based on the travel lane width of the vehicle and the travel lane width after the vehicle turns left. Furthermore, since the timing at which the vehicle enters the intersection varies depending on the vehicle speed, a plurality of first threshold values L1 may be recorded according to the vehicle speed and / or acceleration.

  In search range bias mode setting step S350, the search range mode selection means 233 selects the search range bias mode as the search range mode. FIG. 9 is a diagram illustrating a state in which the vehicle approaches the intersection where the vehicle is predicted to turn to the left and pays attention to the left side of the vehicle. In FIG. 9, when the vehicle approaches an intersection predicted to turn left (first distance P1 <first threshold L1), the search range mode selection means 233 selects the search range bias mode as the search range mode and concentrates Specifically, it is set to search for a pedestrian in the left region of the vehicle.

  Considering the road conditions, it is unlikely that there are pedestrians in the right area of the vehicle, so no pedestrians are searched in the right area and no pedestrians are searched in the right area. Search for pedestrians by increasing computational load. Specifically, in the front peripheral image of the vehicle imaged by the camera mounted on the vehicle shown in FIG. 5, the search process is not performed in the right region of the image region, Is set so that the search process is performed with high accuracy in the left side area.

  Here, since it is unlikely that there is a pedestrian in the right region of the vehicle, it is set not to search for a pedestrian in the right region but to search for a pedestrian only in the left region of the vehicle. However, the present invention is not limited to this. For example, the range to be watched by a pedestrian on the left side of the vehicle is not the left half region of the front peripheral image of the vehicle imaged by the camera mounted on the vehicle, but the left three minutes of the entire image region. It may be set to one area, or the left two-third area of the entire image area. Moreover, you may search a pedestrian by reducing calculation load in the right side area | region of the vehicle which is not the range which should pay attention to a pedestrian.

  Furthermore, the detailed setting contents of the search range bias mode are recorded in advance in the recording means 260, and a plurality of records are recorded for each intersection based on the travel lane width of the vehicle and the travel lane width after the vehicle turns left. It doesn't matter. Furthermore, a plurality of detailed setting contents of the search range bias mode may be recorded according to the speed and / or acceleration of the vehicle.

  And the pedestrian search means 220 searches a pedestrian from the image acquired by the image acquisition means 210 based on the search range bias mode selected by the search range mode selection means 233.

  In the second distance calculation step S360, after setting the search range deviation mode as the search range mode, the vehicle position calculation unit 231 calculates a second distance P2 that is the distance between the vehicle and the intersection where the vehicle turns left. The second distance P2 is a distance along the traveling lane direction of the vehicle, and is a distance between the center point of the vehicle and the center of the intersection where the vehicle is predicted to turn left. Note that the second distance P2 is calculated based on GPS information and map information, and a specific calculation method is the same as the above-described calculation method of the first distance P1, and thus detailed description thereof is omitted.

  In the second distance determination step S370, the vehicle position determination means 232 determines whether or not the second distance P2 calculated in the second distance calculation step S360 is less than the second threshold L2.

  When it is determined in the second distance determination step S370 that the second distance P2 is less than the second threshold L2, the process proceeds to a search range equal mode setting step S380 (Yes in the second distance determination step S370).

  On the other hand, when it is determined in the second distance determination step S370 that the second distance P2 is greater than or equal to the second threshold L2, the process returns to the second distance calculation step S360 (No in the second distance determination step S370).

  In other words, after the search range bias mode is set as the search range mode, the second distance calculation step S360 and the second distance determination step S370 are repeated at predetermined time intervals to determine whether the vehicle is turning left. Monitoring.

  The second threshold L2 is set in consideration of a range to be watched by a pedestrian crossing a pedestrian crossing immediately after the vehicle turns left when turning left at an intersection where the vehicle is predicted to turn left. Specifically, when the vehicle turns to the left, the search range bias mode is selected as the search range mode in the search range bias mode setting step S350, and the vehicle is set to search pedestrians intensively in the left region of the vehicle. However, when the left turn state progresses, it is necessary to search for a pedestrian also in the right region of the vehicle.

  Further, the second threshold value L2 may be recorded in advance in the recording unit 260, and a plurality of second threshold values L2 may be recorded for each intersection based on the travel lane width of the vehicle and the travel lane width after the left turn of the vehicle. Furthermore, since the timing at which the vehicle enters the intersection and the timing at which the vehicle turns to the left depending on the speed of the vehicle fluctuates, a plurality of second threshold values L2 are recorded according to the speed and / or acceleration of the vehicle. It doesn't matter.

  In search range equal mode setting step S380, search range mode setting means 230 selects the search range equal mode as the search range mode. FIG. 10 is a diagram illustrating a state in which the left turn state of the vehicle proceeds at an intersection where the vehicle is predicted to turn left, and attention is paid to the left side and the right side of the vehicle. In FIG. 10, when the left turn state of the vehicle advances at the intersection where the vehicle is predicted to turn left (second distance P2 <second threshold L2), the search range mode selection unit 233 sets the search range equal mode to the search range. Select as mode. In other words, from the state where the search process is being performed at a high load in the left side region of the front peripheral image region of the vehicle imaged by the camera mounted on the vehicle, the entire front peripheral image region of the vehicle is Switch to a state where search processing is performed at a constant load.

  Here, the second threshold value L2 is the travel lane width after the vehicle turns left, and when the second distance P2 is less than the second threshold value L2, it is determined that the vehicle is turning left. The second threshold L2 is not limited to the travel lane width after the vehicle turns left, and may be any other value as long as it can be determined whether the vehicle is making a left turn. You may set according to the speed and / or acceleration of a vehicle for every intersection.

  As described above, when the vehicle makes a left turn, the pedestrian detection device 100 switches a region for searching for a pedestrian according to the distance between the vehicle and the intersection. Thereby, the pedestrian is appropriately detected in consideration of the range to be watched by the pedestrian and the calculation amount in the search process according to the state of the vehicle traveling at the intersection.

  FIG. 11 is a diagram illustrating a calculation load ratio in the left side region of the vehicle and the right side region of the vehicle. FIG. 11A is a diagram showing the calculation load ratio in the left region of the vehicle, and FIG. 11B is a diagram showing the calculation load ratio in the right region of the vehicle.

  First, when the first distance P1 between the vehicle and the intersection where the vehicle turns to the left is equal to or greater than the first threshold (when the vehicle is not approaching the intersection), the search range equalization mode is selected as the search range mode. That is, the calculation load ratio in the left side area of the vehicle and the calculation load ratio in the right side area of the vehicle are equal to 50 [%].

  Next, when the first distance P1 between the vehicle and the intersection where the vehicle turns to the left is less than the first threshold value (when the vehicle is approaching the intersection), the search range bias mode is selected as the search range mode. That is, the calculation load ratio in the left area of the vehicle is 50 + ω [%], and the calculation load ratio in the right area of the vehicle is 50−ω [%]. In the left area of the vehicle, which should be watched by the pedestrian, the calculation load is increased by ω [%], and in the right area of the vehicle where there is a low possibility that a pedestrian is present, the calculation load is increased by ω [%]. It is lowered. Thereby, highly accurate and highly efficient pedestrian search processing according to the situation is realized.

  When the second distance P2 between the vehicle and the intersection where the vehicle turns to the left is less than the second threshold value (when the vehicle is turning left), the search range equalization mode is selected as the search range mode.

  Note that ω, which is an increase / decrease range of the calculation load ratio, is 0 to 50 [%], and ω is set to 50 [%] when the search process of the right region of the vehicle is not performed in the search range bias mode. do it.

  Further, while the search range bias mode is selected, ω is variable. For example, as the vehicle approaches the intersection, the calculation load is gradually increased in the left side region of the vehicle, and the calculation load is gradually decreased in the right side region of the vehicle. .

  FIG. 12 is a diagram illustrating a state in which the calculation load ratios in the left side region and the right side region of the vehicle are gradually changed. FIG. 12A is a diagram showing the calculation load ratio in the left region of the vehicle, and FIG. 12B is a diagram showing the calculation load ratio in the right region of the vehicle.

  When the first distance P1 between the vehicle and the intersection where the vehicle turns to the left is equal to or greater than the first threshold (when the vehicle is not approaching the intersection), the second distance P2 between the vehicle and the intersection where the vehicle turns to the left is the second distance P2. When it is less than the threshold value (when the vehicle is turning left), the search range equal mode is selected as the search range mode. In this case, similarly to FIG. 11, the calculation load ratio in the left side area of the vehicle and the calculation load ratio in the right side area of the vehicle are equal to 50 [%].

  Next, when the first distance P1 between the vehicle and the intersection where the vehicle turns to the left is less than the first threshold (when the vehicle is approaching the intersection), the search range bias mode is selected as the search range mode, and the left side of the vehicle When the calculation load ratio in the region is gradually increased from 50 [%] and the second distance P2 between the vehicle and the intersection where the vehicle turns to the left is less than the second threshold (when the vehicle turns left), The calculation load ratio is 50 + ω [%]. At the same time, the calculation load ratio in the right region of the vehicle gradually decreases from 50 [%] to 50-ω [%].

  Although an example in which the calculation load ratio in the left side region of the vehicle and the right side region of the vehicle is gradually changed is shown here, the present invention is not limited to this. For example, the calculation load ratio may be changed according to a function (a broken line or a curve), and the image area for searching for a pedestrian is divided into three or more divisions instead of the left and right divisions, or the upper and lower divisions in consideration of perspective. The calculation load ratio may be set and changed for each region.

  As described above, according to the pedestrian detection device 100 according to the first embodiment of the present invention, when the vehicle turns left at the intersection, when the vehicle approaches the intersection, the left side of the vehicle is determined based on the search range bias mode. The pedestrian is searched by increasing the calculation load for the area, and then the pedestrian is searched by equalizing the calculation load for the left area and the right area of the vehicle based on the search range equal mode when the left turn state of the vehicle advances. Thereby, a highly accurate and highly efficient pedestrian search process can be realized according to the situation where the vehicle turns left at the intersection.

Next, a case where the vehicle turns right will be described.
FIG. 13 is a flowchart showing a pedestrian detection method (right turn) 400 executed by the pedestrian detection apparatus 100 according to the first embodiment of the present invention. In FIG. 13, the pedestrian detection method 400 (right turn) includes a search range equal mode setting step S310, a right turn prediction step S410, a first distance calculation step S330, a first distance determination step S340, and a search range bias mode setting. It includes step S350, intersection relative position calculation step S420, intersection passage determination step S430, and search range equalization mode setting step S380. In FIG. 13, the same steps (processes) shown in FIG. 3 are denoted by the same reference numerals, and detailed description thereof is omitted.

  First, in the search range equal mode setting step S310, the search range equal mode is selected, and based on the search range equal mode, the pedestrians are searched with equal calculation load for the left and right regions of the vehicle.

  In the right turn prediction step S410, the right / left turn prediction means 250 predicts that the vehicle will turn right. For example, it can be predicted that the vehicle will turn right by route guidance in the car navigation system. FIG. 14 is a diagram illustrating a state in which the vehicle is predicted to turn right by route guidance in the car navigation system. In FIG. 14, the vehicle is approaching an intersection where the vehicle is predicted to turn left by the right / left turn prediction means 250.

  In the first distance calculation step S330, the vehicle position calculation means 231 calculates a first distance P1 that is the distance between the vehicle and the intersection where the vehicle turns to the right. FIG. 15 is a diagram illustrating the first distance P1. As shown in FIG. 15, the first distance P1 is a distance along the traveling lane direction of the vehicle, and is a distance between the center point of the vehicle and the center of the intersection where the vehicle is predicted to turn right.

  In the first distance determination step S340, the vehicle position determination means 232 determines whether or not the first distance P1 calculated in the first distance calculation step S330 is less than the first threshold value L1.

  The first threshold L1 is set in consideration of the risk of a collision with a pedestrian on the left side of the vehicle when approaching an intersection where the vehicle is predicted to turn right and when the vehicle subsequently turns right. Specifically, when the vehicle turns to the right, even if there is a pedestrian on the left side of the vehicle, it is unlikely that the pedestrian will enter the traveling direction of the vehicle. About intensively searching for pedestrians.

  Further, the first threshold value L1 when the vehicle turns right may be set to a value different from the first threshold value L1 when the vehicle turns left.

  In search range bias mode setting step S350, the search range mode selection means 233 selects the search range bias mode as the search range mode. FIG. 16 is a diagram illustrating a state in which the vehicle approaches the intersection where the vehicle is predicted to make a right turn and is focused on the right side of the vehicle. In FIG. 16, when the vehicle approaches an intersection predicted to turn right (first distance P1 <first threshold L1), the search range mode selection unit 233 selects the search range bias mode as the search range mode, and concentrates Specifically, the pedestrian is searched for in the right region of the vehicle.

  Considering the road conditions, it is unlikely that there are pedestrians in the left area of the vehicle, so no pedestrians are searched in the left area and no pedestrians are searched in the left area. Search for pedestrians by increasing computational load.

  Here, since it is unlikely that a pedestrian is present in the left area of the vehicle, the pedestrian is not searched for in the left area, but is searched for only in the right area of the vehicle. However, the present invention is not limited to this. For example, the range to be watched by a pedestrian on the right side of the vehicle is not the right half region of the front peripheral image of the vehicle imaged by the camera mounted on the vehicle, but the right three minutes of the entire image region. It may be set to one area, or to the right two-third area of the entire image area. Moreover, you may search a pedestrian by reducing calculation load in the left side area | region of the vehicle which is not the range which should pay attention to a pedestrian.

  And the pedestrian search means 220 searches a pedestrian from the image acquired by the image acquisition means 210 based on the search range bias mode selected by the search range mode selection means 233.

  Next, as described above, when the vehicle turns left, the second distance P2 is calculated in the second distance calculation step S360, and the second distance P2 is compared with the second threshold value L2 in the second distance determination step S370. The timing for returning the search range mode to the search range equal mode is monitored by monitoring whether or not the vehicle is turning left.

  On the other hand, when the vehicle turns to the right, the relative position between the vehicle and the intersection is calculated in the intersection relative position calculation step S420, and whether or not the vehicle has passed the intersection in the intersection passage determination step S430 is monitored. The timing is set to return the mode to the search range uniform mode.

  In the intersection relative position calculation step S420, the vehicle position calculation means 231 calculates the relative position between the vehicle and the intersection where the vehicle turns right after setting the search range deviation mode as the search range mode. The relative position is calculated based on GPS information and map information.

  In the intersection passage determination step S430, the vehicle position determination means 232 determines whether or not the vehicle has passed the intersection based on the relative position calculated in the intersection relative position calculation step S420.

  When it is determined in the intersection passage determination step S430 that the vehicle has not passed the intersection, the process returns to the intersection relative position calculation step S420 (No in the intersection passage determination step S430). FIG. 17 is a diagram illustrating a state in which the vehicle does not pass through the intersection. In FIG. 17, it is determined that the vehicle has not passed the intersection based on the relative position between the vehicle and the intersection where the vehicle turns to the right. In this case, since it is unlikely that there is a pedestrian in the left area of the vehicle, the search range bias mode is selected in which the pedestrian is not searched for in the left area but only the right area of the vehicle is searched. There is.

  On the other hand, when it is determined that the vehicle has passed the intersection in the intersection passage determination step S430, the process proceeds to the search range equal mode setting step S380 (Yes in the intersection passage determination step S430).

  In other words, after setting the search range bias mode as the search range mode, the intersection relative position calculation step S420 and the intersection passage determination step S430 are repeated at predetermined time intervals to monitor whether the vehicle is turning right. is doing.

  In search range equal mode setting step S380, search range mode setting means 230 selects the search range equal mode as the search range mode. FIG. 18 is a diagram illustrating a state in which a right turn state of the vehicle progresses at an intersection where the vehicle is predicted to turn right, and attention is paid to the left side and the right side of the vehicle. In FIG. 18, when the vehicle turns right at the intersection where the vehicle is predicted to turn right (when the vehicle passes the intersection), the search range mode selection unit 233 sets the search range equal mode as the search range mode. select.

  Here, when the vehicle passes the intersection, it is determined that the vehicle is turning right. However, the present invention is not limited to this, and it can be determined whether the vehicle is turning right. Any other method may be used.

  Even if the vehicle does not pass through the intersection, the calculation load ratio in the left side region of the vehicle and the right side region of the vehicle are gradually increased according to the relative position between the vehicle and the intersection where the vehicle turns right. The calculation load ratio at may be changed.

  As described above, when the vehicle turns to the right, the pedestrian detection device 100 switches a region for searching for a pedestrian according to the distance between the vehicle and the intersection. Thereby, the pedestrian is appropriately detected in consideration of the range to be watched by the pedestrian and the calculation amount in the search process according to the state of the vehicle traveling at the intersection.

  As described above, according to the pedestrian detection device 100 according to the first embodiment of the present invention, when the vehicle turns right at the intersection, when the vehicle approaches the intersection, the right side of the vehicle is determined based on the search range bias mode. The pedestrian is searched by increasing the calculation load for the area, and then the pedestrian is searched by equalizing the calculation load for the left area and the right area of the vehicle based on the search range equal mode when the vehicle turns right. Thereby, a highly accurate and highly efficient pedestrian search process can be realized according to the situation where the vehicle turns right at the intersection.

<Second Embodiment>
FIG. 19 is a diagram showing a pedestrian detection device 500 according to the second embodiment of the present invention. In FIG. 19, a pedestrian detection device 500 is different in that a body control ECU 510 is added to the pedestrian detection device 100 according to the first embodiment of the present invention shown in FIG.

  FIG. 20 is a block diagram functionally showing a pedestrian detection apparatus 500 according to the second embodiment of the present invention. 20, a pedestrian detection device 500 is different in that a blinker control means 610 is added to the pedestrian detection device 100 according to the first embodiment of the present invention shown in FIG.

  In the pedestrian detection apparatus 500 shown in FIGS. 19 and 20, the same reference numerals are assigned to the same components as those of the pedestrian detection apparatus 100 according to the first embodiment of the present invention shown in FIGS. Detailed description will be omitted. In the present embodiment, differences from the first embodiment of the present invention will be described in detail.

  The body control ECU 510 of the pedestrian detection device 500 shown in FIG. 19 controls a winker or the like when the vehicle turns right or left. The search range control ECU 130 controls the search range for searching for pedestrians based on the blinker information from the body control ECU 510 in addition to the traveling road information and planned traveling route information from the car navigation ECU 160.

  The blinker control means 610 of the pedestrian detection device 500 shown in FIG. 20 controls the blinker when the vehicle turns right or left. Specifically, when the vehicle turns right or left, the blinker blinks based on the driver's operation, and when the vehicle turns right or left, the blinker blinking is cancelled.

  FIG. 21 is a flowchart showing a pedestrian detection method (left turn) 700 executed by the pedestrian detection apparatus 500 according to the second embodiment of the present invention. In FIG. 21, the pedestrian detection method (left turn) 700 includes search range equal mode setting step S310, left turn prediction step S320, first distance calculation step S330, first distance determination step S340, and search range bias mode setting. Step S350, turn signal information acquisition step S710, turn signal cancellation determination step S720, and search range equal mode setting step S380 are included. The same steps (processes) illustrated in FIG. 21 are denoted by the same reference numerals, and detailed description thereof is omitted.

  Compared with the pedestrian detection method (left turn) 300 according to the first embodiment of the present invention, the pedestrian detection method (left turn) 700 according to this embodiment monitors and determines whether the vehicle is turning left. The process for judging is different.

  In the pedestrian detection method (left turn) 700, the processing from the search range equal mode setting step S310 to the search range bias mode setting step S350 is the same as the pedestrian detection method (left turn) 300 according to the first embodiment of the present invention. Here, the pedestrian search means 220 searches for a pedestrian from the image acquired by the image acquisition means 210 based on the search range bias mode selected by the search range mode selection means 233. And

  Next, in the winker information acquisition step S710, the winker control means 610 acquires winker information when the vehicle turns left. Specifically, information indicating whether or not the left turn winker is blinking is acquired.

  In the winker cancel determination step S720, the winker control means 610 determines whether or not the left turn winker is canceled based on the winker information acquired in the winker information acquisition step S710.

  When it is determined in the winker cancel determination step S720 that the left turn winker is canceled, the process proceeds to the search range equal mode setting step S380 (Yes in the winker cancel determination step S720).

  On the other hand, when it is determined in the winker cancel determination step S720 that the left turn winker has not been canceled, the process returns to the winker information acquisition step S710 (No in the winker cancel determination step S720).

  In other words, after setting the search range bias mode as the search range mode, the winker information acquisition step S710 and the winker cancel determination step S720 are repeated at predetermined time intervals to monitor whether the vehicle is turning left. ing.

  In search range equal mode setting step S380, search range mode setting means 230 selects the search range equal mode as the search range mode. When the vehicle turns left at an intersection where the vehicle is predicted to turn left (when the left turn blinker is canceled), the search range mode selection unit 233 selects the search range equal mode as the search range mode. In other words, from the state where the search process is being performed at a high load in the left side region of the front peripheral image region of the vehicle imaged by the camera mounted on the vehicle, the entire front peripheral image region of the vehicle is Switch to a state where search processing is performed at a constant load.

  As described above, according to the pedestrian detection device 500 according to the second embodiment of the present invention, when the vehicle turns left at the intersection, when the vehicle approaches the intersection, the left side of the vehicle is determined based on the search range bias mode. The pedestrian is searched by increasing the calculation load for the area, and then the pedestrian is searched by equalizing the calculation load for the left area and the right area of the vehicle based on the search range equal mode when the left turn state of the vehicle advances. Thereby, a highly accurate and highly efficient pedestrian search process can be realized according to the situation where the vehicle turns left at the intersection.

  In the present embodiment, the search range mode is switched from the search range bias mode to the search range equal mode based on the blinker information when the vehicle turns left. It goes without saying that the same effect can be obtained by switching the search range mode from the search range bias mode to the search range equal mode based on the right turn blinker information even when the vehicle turns right.

  Furthermore, the timing at which the search range mode is switched from the search range bias mode to the search range equal mode is not limited to the blinker information. For example, the steering angle information of the vehicle is detected by detecting the steering angle of the vehicle using a steering sensor or the like. It may be determined based on the above. In other words, based on the steering angle, it is determined whether the vehicle is turning right or left.

  The present invention reduces the amount of computation for detecting a pedestrian, and when a vehicle turns right or left, detects a pedestrian that accurately crosses a pedestrian crossing immediately after the vehicle turns right or left. Useful for devices and the like.

100, 500 Pedestrian detection device 110 Pedestrian detection camera 120 Pedestrian detection judgment ECU
130 Search range control ECU
140 Display Control ECU for Display
150 Display 160 Car Navigation ECU
210 Image acquisition means 220 Pedestrian search means 230 Search range mode setting means 231 Vehicle position calculation means 232 Vehicle position determination means 233 Search range mode selection means 240 Result notification means 250 Right / left turn prediction means 260 Recording means 300, 700 Pedestrian detection method (Left turn)
400 Pedestrian detection method (right turn)
510 Body Control ECU
610 Winker control means S310 Search range equal mode setting step S320 Left turn prediction step S330 First distance calculation step S340 First distance determination step S350 Search range bias mode setting step S360 Second distance calculation step S370 Second distance determination step S380 Search range equality Mode setting step S410 Right turn prediction step S420 Intersection relative position calculation step S430 Intersection passage determination step S710 Winker information acquisition step S720 Winker cancellation determination step

Claims (13)

A pedestrian detection device mounted on a vehicle,
Image acquisition means for acquiring an image around the vehicle;
A right / left turn predicting means for predicting that the vehicle makes a right turn or a left turn;
A search range equalization mode for searching within an image area acquired by the image acquisition means with a predetermined accuracy, and an image area acquired by the image acquisition means are divided into a plurality of search areas, each having different accuracy Search range mode setting means for setting any search range mode of search range bias mode to be searched based on prediction information by the right / left turn prediction means,
Based on the search range mode set by the search range mode setting means, pedestrian search means for searching for pedestrians from the image acquired by the image acquisition means,
A result notifying means for notifying the result searched by the pedestrian searching means,
The said search range mode setting means sets the said search range mode based on the distance of the said vehicle and the intersection where the said vehicle turns right or left, The pedestrian detection apparatus characterized by the above-mentioned.   2. The pedestrian detection device according to claim 1, wherein the search range bias mode sets a specific region for searching at least one of the plurality of divided search regions with high accuracy.   The pedestrian detection device according to claim 2, wherein the search range bias mode searches a region other than the specific region among the plurality of divided search regions with low accuracy. The search range mode setting means includes:
The vehicle sets the search range bias mode as a search range mode at a first preset timing close to an intersection at which the vehicle is predicted to turn right or left by the right / left turn prediction means. The pedestrian detection device according to any one of claims 1 to 3. The search range mode setting means includes:
The intersection range where the vehicle is predicted to turn right or left by the right / left turn prediction means is set as the search range mode at the second timing preset after the vehicle starts turning right or left. The pedestrian detection device according to claim 4, wherein the pedestrian detection device is characterized. The search range mode setting means includes:
Vehicle position calculating means for calculating a distance between the vehicle and an intersection at which the vehicle is predicted to turn right or left by the right / left turn prediction means;
Vehicle position determination means for determining whether or not the distance calculated by the vehicle position calculation means is less than a preset threshold;
When the vehicle position determination means determines that the distance is less than the threshold, the search range bias mode is selected as the search range mode, and the vehicle position determination means determines that the distance is greater than or equal to the threshold. The pedestrian detection device according to claim 4, further comprising: search range mode selection means for selecting the search range equal mode as the search range mode. After the start of the left turn of the vehicle, the intersection where the vehicle is predicted to turn left by the right / left turn prediction means,
The vehicle position determination means determines whether the distance calculated by the vehicle position calculation means is less than a travel lane width after the vehicle turns left;
The search range mode selection unit sets the search range equal mode as a search range mode when the vehicle position determination unit determines that the distance is less than the travel lane width. 6. The pedestrian detection device according to 6. After the start of a right turn of the vehicle, an intersection where the vehicle is predicted to turn right by the right / left turn prediction means,
The vehicle position determination means determines whether the vehicle has passed the intersection;
The search range mode selection unit sets the search range equal mode as a search range mode when the vehicle position determination unit determines that the vehicle has passed the intersection. The pedestrian detection device according to any one of 7. A winker control means for controlling a winker indicating a right turn or a left turn of the vehicle,
The pedestrian detection apparatus according to claim 5, wherein the second timing is a timing at which a winker indicating a right turn or a left turn of the vehicle is canceled by the winker control means. When the vehicle is predicted to turn left by the right / left turn predicting means, the left area in the image area is searched with higher accuracy than the right area in the image area,
The right region in the image region is searched with higher accuracy than the left region in the image region when the vehicle predicts that the vehicle will turn right. Item 10. The pedestrian detection device according to any one of Items 1 to 9.   The walking according to claim 2, wherein the search range bias mode searches the specific area with high accuracy among the plurality of divided search areas and does not search any area other than the specific area. Person detection device.   The pedestrian detection device according to claim 1, wherein the high accuracy is high load and / or high definition. A pedestrian detection method executed by a pedestrian detection device mounted on a vehicle,
An image acquisition step of acquiring an image around the vehicle;
A left-right turn prediction step for predicting that the vehicle will turn right or left; and
A vehicle position calculating step for calculating a distance between the vehicle and an intersection at which the vehicle is predicted to turn right or left in the right / left turn prediction step;
The search range equalization mode for searching the image area acquired in the image acquisition step with a predetermined accuracy set in advance, and the image area acquired in the image acquisition step is divided into a plurality of search areas, each having different accuracy A search range mode setting step for setting any one of the search range modes of the search range bias mode to be searched based on the distance between the vehicle and the intersection calculated in the vehicle position calculation step;
Based on the search range mode set in the search range mode setting step, a pedestrian search step for searching for a pedestrian from the image acquired in the image acquisition step;
A pedestrian detection method including a result notification step of notifying a result searched for by the pedestrian search step.

Images