JP2007264778A - Apparatus for recognizing pedestrian - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To accurately predict a future moving state of a pedestrian. <P>SOLUTION: A pedestrian shape acquiring part 22 creates an edge image from image data input from an external field sensor 14; detects the width W of the right and left legs of a pedestrian candidate; estimates the head of the pedestrian candidate; and estimates the height H of the pedestrian candidate according to the position of the head. Based on the height H of the pedestrian candidate acquired by the acquiring part 22 and the width W of the legs, a crossing possibility determining part 23 determines whether or not the ratio (W/H) of the width W of the legs to the height H is a predetermined value α or more, thereby determining whether the pedestrian candidate will cross the course of one's own vehicle. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a pedestrian recognition device.

2. Description of the Related Art Conventionally, for example, a control device that detects a moving speed and a moving direction of an external collision object detected by a radar device mounted on the host vehicle and executes a collision avoidance process is known (for example, Patent Document 1). reference).
Conventionally, a method for detecting the moving direction of a pedestrian outside the host vehicle from an image obtained by, for example, photographing with a camera is known (see, for example, Non-Patent Document 1).
JP 2001-28050 A Hiroaki Shimizu and Tomaso Poggio, Direction Estimation of Pedestrian from Multiple Still Images, 2004 IEEE Intelligent Vehicles Symposium, pp.596-600

By the way, according to the control device according to an example of the above prior art, the moving speed and the moving direction are detected from the change in the position of the collision target. However, it is difficult to accurately detect the moving speed and moving direction of a collision object that takes a relatively complicated behavior such as a pedestrian from the change in position. It is not possible to accurately determine the possibility.
In addition, according to the method according to the above prior art, there is a problem in that it is only possible to detect the movement direction of the pedestrian and not accurately predict the movement state of the pedestrian taking a complicated behavior. Arise.
The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a pedestrian recognition device that can accurately determine the possibility that a pedestrian will hinder the traveling of the host vehicle.

  In order to solve the above problems and achieve the object, the pedestrian recognition device according to the present invention according to claim 1 is a pedestrian detection means for detecting a pedestrian existing in the outside world of the own vehicle (for example, implementation). The external sensor 14 and the pedestrian candidate extraction unit 21) in the form, and pedestrian state detection means for detecting the pedestrian state detected by the pedestrian detection means (for example, pedestrian shape acquisition in the embodiment) Part 22) and whether or not the pedestrian enters the course of the own vehicle based on the state of the leg of the pedestrian among the states of the pedestrian detected by the pedestrian state detection means. It is characterized by comprising determination means for determining (for example, a crossing possibility determining unit 23 in the embodiment).

  According to the pedestrian recognition device described above, whether or not a pedestrian enters the course of the host vehicle based on the state of the pedestrian's legs among the pedestrian states detected by the pedestrian state detection means. By determining the above, it is possible to accurately predict the pedestrian's moving state (that is, moving walking and moving speed) from the state of the leg that appropriately reflects the will of the pedestrian taking a relatively complicated behavior. It is possible to accurately determine the possibility that the pedestrian will interfere with the traveling of the host vehicle.

  Further, in the pedestrian recognition apparatus according to the present invention as set forth in claim 2, the determination means is configured such that the pedestrian enters the course of the own vehicle based on an opening degree of the leg of the pedestrian and a change in the opening degree. It is characterized by determining whether to do.

  According to the pedestrian recognition device described above, it is possible to accurately detect the movement state of the pedestrian based on the state of the pedestrian's leg, in particular, the opening degree of the leg part and the change in the opening degree. Whether or not a pedestrian enters the course can be accurately determined.

As described above, according to the pedestrian recognition device of the present invention, the movement state of the pedestrian is accurately predicted from the state of the leg portion appropriately reflecting the will of the pedestrian taking a relatively complicated behavior. Therefore, it is possible to accurately determine the possibility that the pedestrian will hinder the traveling of the host vehicle.
Furthermore, according to the pedestrian recognition device of the present invention as set forth in claim 2, the movement state of the pedestrian can be accurately detected based on the opening degree of the leg portion and the change in the opening degree. Whether or not a pedestrian enters can be accurately determined.

Hereinafter, a pedestrian recognition device according to an embodiment of the present invention will be described with reference to the accompanying drawings.
A vehicle including a pedestrian recognition apparatus 10 according to the present embodiment, for example, as shown in FIG. 1, transmits a driving force of an internal combustion engine 11 to a transmission (T) such as an automatic transmission (AT) or a continuously variable automatic transmission (CVT). / M) is mounted on a vehicle that transmits to the driving wheels of the vehicle via 12, and includes a processing device 13, an external sensor 14, a host vehicle state sensor 15, a brake actuator 16, an EPS actuator 17, and an alarm device 18. It is configured with.
Further, the processing device 13 includes, for example, a pedestrian candidate extraction unit 21, a pedestrian shape acquisition unit 22, a crossing possibility determination unit 23, and a travel control unit 24.

The external sensor 14 includes a pair of cameras 14La and 14Rb and image processing units 14Lb and 14Rb each including a CCD camera, a CMOS camera, or the like that can capture an image in a visible light region or an infrared region, for example.
Each image processing unit 14Lb, 14Rb applies a predetermined image such as filtering or binarization processing to each image of the external world in the traveling direction of the host vehicle obtained by photographing with a pair of cameras 14La, 14Rb. Processing is performed to generate a pair of image data composed of pixels of a two-dimensional array and output the generated image data to the processing device 13.

  The host vehicle state sensor 15 includes, as vehicle information of the host vehicle, for example, a vehicle speed sensor that detects the speed (vehicle speed) of the host vehicle, a yaw angle (rotation angle around the vertical axis of the vehicle center of gravity), and a yaw rate (up and down of the vehicle center of gravity). A yaw rate sensor that detects the rotational angular velocity around the direction axis, and a steering angle sensor that detects the steering angle (direction and size of the steering angle input by the driver) and the actual steering angle (steering angle) according to the steering angle Or a steering torque sensor for detecting steering torque, a positioning signal such as a GPS (Global Positioning System) signal for measuring the position of the vehicle using an artificial satellite, or an information transmission device outside the host vehicle. Position sensor that detects the current position and traveling direction of the host vehicle based on the detection result of an appropriate gyro sensor, acceleration sensor, etc., and a turn indicator and brake on / off Each sensor is configured to detect a state.

  The pedestrian candidate extraction unit 21 of the processing device 13 is based on the image data input from the external sensor 14, and among the objects existing in the detection areas of the pair of cameras 14La and 14Rb in the traveling direction of the host vehicle, A pedestrian candidate is detected, and the position of the pedestrian candidate is calculated. For example, the pedestrian candidate extraction unit 21 performs predetermined recognition processing on a pair of image data input from the image processing units 14Lb and 14Rb of the external sensor 14, and is installed in the vehicle interior with a predetermined interval. The distance to the pedestrian candidate is detected by a triangulation method or the like based on the distance between the paired cameras 14La and 14Rb and the parallax of the pedestrian candidate on the pair of image data obtained by photographing. .

The pedestrian shape acquisition unit 22 adjusts the brightness of the pixels constituting the image data with respect to the data of the area including the pedestrian candidate extracted by the pedestrian candidate extraction unit 21 on the image data input from the external sensor 14. An edge image including the photographed pedestrian outline point sequence is generated by the corresponding edge extraction. Then, for example, pattern matching processing using a model image or the like for a human body stored in advance is performed, the contour point sequence included in the edge image is recognized by a straight line, etc., and the outer shape of the leg part of the pedestrian candidate is identified, The opening W of the leg is detected.
The pedestrian shape acquisition unit 22 estimates the pedestrian candidate's head extracted by the pedestrian candidate extraction unit 21 on the image data input from the external sensor 14, and walks according to the position of the head. The height H of the candidate is estimated.

  Based on the height H of the pedestrian candidate acquired by the pedestrian shape acquisition unit 22 and the opening W of the leg, the crossing possibility determination unit 23, for example, the ratio of the opening W of the leg to the height H (W / H) ) Is greater than or equal to a predetermined value α (for example, α = 0.45), it is determined whether there is a possibility that the pedestrian candidate may cross the course of the host vehicle.

Depending on the determination result by the crossing possibility determination unit 23, for example, when the pedestrian candidate may cross the course of the own vehicle, the traveling control unit 24 generates contact between the pedestrian candidate and the own vehicle. The traveling state of the host vehicle is controlled so as to avoid the above. For example, the travel control unit 24 controls the driving force of the internal combustion engine 11, the control signal for controlling the speed change operation of the transmission 12, the control signal for controlling the deceleration operation by the brake actuator 16, and the vehicle by the EPS actuator 17. At least one of the control signals for controlling the steering operation of the steering mechanism (not shown) is output, and deceleration control or steering control of the host vehicle is executed as the contact avoiding operation.
Further, the traveling control unit 24 controls at least one of the alarm output timing and the output content by the alarm device 18 according to the determination result by the crossing possibility determination unit 23.

The pedestrian recognition device 10 according to the present embodiment has the above-described configuration. Next, the operation of the pedestrian recognition device 10 will be described.
First, for example, in step S01 shown in FIG. 2, the image data output from the external sensor 14 is acquired.
Next, in step S02, for example, as shown in FIG. 3A, a region P including pedestrian candidates is extracted from the acquired image data.
In step S03, for example, as shown in FIG. 3 (b), a walk taken by edge extraction according to the brightness of pixels constituting the image data for the data of the region P including the pedestrian candidate. An edge image including the contour point sequence of the person is generated.

In step S04, for example, as shown in FIG. 3C, pattern matching processing is performed using a model image or the like for a human body stored in advance, and the contour point sequence included in the edge image is recognized by a straight line or the like. The external shape LE of the leg part of a pedestrian candidate is identified.
And in step S05, as shown, for example in FIG.3 (d), the opening degree W of the right and left leg part of a pedestrian candidate is detected.

In step S06, for example, as shown in FIG. 3E, the head HE of the pedestrian candidate extracted by the pedestrian candidate extraction unit 21 on the image data input from the external sensor 14 is estimated.
In step S07, for example, as shown in FIG. 3F, the height H of the pedestrian candidate is estimated according to the estimated position of the head HE of the estimated pedestrian candidate.

In step S08, based on the height H of the pedestrian candidate and the opening W of the leg, for example, the ratio (W / H) of the opening W of the leg to the height H is a predetermined value α (for example, α = 0). .45 etc.) or not.
If this determination is “YES”, the flow proceeds to step S 10 described later.
On the other hand, if this determination is “NO”, the flow proceeds to step S 09.
In step S09, it is determined that there is no possibility that the pedestrian candidate crosses the course of the host vehicle, and the series of processes is terminated.

Further, in step S10, it is determined that the pedestrian candidate may cross the course of the own vehicle.
And in step S11, based on the state (for example, a position, a vehicle speed, etc.) of the own vehicle output from the vehicle state sensor 15, the contact avoidance which avoids contact generation with a pedestrian candidate and the own vehicle is actuated, Terminate the process.

  As a result, for example, as shown in FIG. 4, when the temporal change in the opening W of the leg portion of the pedestrian candidate is relatively small (the opening W1 shown in FIG. 4), the pedestrian candidate becomes the course of the own vehicle. When the time change of the opening degree W of the leg part of the pedestrian candidate is relatively large (the opening degree W2 shown in FIG. 4), the pedestrian candidate is the course of the own vehicle. It is judged that there is a possibility of crossing.

For example, as shown in FIG. 5A, when the pedestrian moves along a direction substantially parallel to the traveling direction of the host vehicle, the opening W of the leg estimated from the image data is relative. Therefore, it is determined that there is no possibility that the flighter crosses the course of the own vehicle.
On the other hand, for example, as shown in FIG. 5B, when the pedestrian moves along a direction intersecting the traveling direction of the host vehicle, the leg opening W estimated from the image data is relatively Since it becomes large, it is judged that there is a possibility that the pedestrian crosses the course of the own vehicle.

As described above, according to the pedestrian recognition device 10 according to the present embodiment, based on the state of the leg of the pedestrian (particularly, the opening W of the leg and the change in the opening) By determining whether or not a pedestrian enters, the pedestrian's movement state (that is, moving gait and movement speed) is determined from the state of the leg that appropriately reflects the will of the pedestrian taking a relatively complicated behavior. ) Can be predicted with high accuracy, and it is possible to accurately determine the possibility that a pedestrian will hinder the traveling of the host vehicle.
Moreover, by determining according to the ratio (W / H) of the leg opening W to the height H of each pedestrian, it is possible to accurately predict the moving state regardless of the pedestrian's body shape. .

It is a lineblock diagram of the pedestrian recognition device concerning an embodiment of the invention. It is a flowchart which shows operation | movement of the pedestrian recognition apparatus which concerns on embodiment of this invention. FIGS. 3A to 3F are diagrams illustrating an example of each image processing for image data output from an external sensor. It is a graph which shows the relationship between the change of the opening degree of a leg part of a pedestrian, and the presence or absence of crossing possibility. FIGS. 5A and 5B are views showing the opening degree of the leg of the pedestrian on the image data output from the external sensor.

Explanation of symbols

10 Pedestrian recognition device 14 External sensor (pedestrian detection means)
21 Pedestrian candidate extraction unit (pedestrian detection means)
22 Pedestrian shape acquisition unit (pedestrian state detection means)
23 Crossing possibility determination unit (determination means)

Claims (2)

Pedestrian detection means for detecting pedestrians present in the outside world of the vehicle,
Pedestrian state detection means for detecting the state of the pedestrian detected by the pedestrian detection means;
Determining means for determining whether or not the pedestrian enters the course of the own vehicle based on the state of the leg of the pedestrian among the states of the pedestrian detected by the pedestrian state detecting means; A pedestrian recognition device comprising: The said determination means determines whether the said pedestrian enters into the course of the own vehicle based on the opening degree of the leg part of the said pedestrian, and the change of this opening degree. Pedestrian recognition device.

Images