JP2010066810A - Pedestrian detecting apparatus for vehicle - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a pedestrian detecting apparatus for vehicle, capable of accurately deciding the start of movement of a pedestrian, from a standstill. <P>SOLUTION: With regard to a pedestrian behavior (vehicle-path crossing action), by use of a knee portion movement velocity Vkt and a shoulder portion movement velocity Vst on which the pedestrian will is reflected, the conditional expression in the pedestrian movement start is defined by the form of the kinetic energy of the knee portion and the shoulder portion as: (Vst/α)<SP>2</SP>+(Vkt/α)<SP>2</SP>>1 (where Vst>0 and Vkt>0, α: normalized coefficient). Thus, based on the squared value of the speed in the kinetic energy, the degree of reflection of the knee portion movement velocity Vkt and the shoulder portion movement speed Vst affecting the behavior prediction of a pedestrian 8 is increased, and decision accuracy is enhanced. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a vehicle pedestrian detection device.

As shown in Patent Document 1, a vehicle pedestrian detection device that detects that a pedestrian interferes with running is detected by detecting the pedestrian's height and leg opening, and the leg opening relative to the height. In some cases, it is determined whether or not the ratio of the pedestrian is greater than or equal to a predetermined value to determine the pedestrian crossing possibility.
JP 2007-264778 A

  However, the vehicle pedestrian detection device detects that the leg opening is wide after the pedestrian has already started to move. Under such a configuration, the determination is delayed and collision avoidance is detected. There is a problem that is delayed.

  This invention is made | formed in view of such a situation, The technical subject is providing the pedestrian detection apparatus for vehicles which can determine accurately about the pedestrian's movement start from the stationary state.

In order to achieve the technical problem, the present invention (the invention according to claim 1)
A pedestrian detection device for a vehicle for determining movement of a pedestrian from a stationary state,
Imaging means for photographing the periphery of the vehicle;
Pedestrian information detection means for detecting a pedestrian based on photographing information from the imaging means and detecting pedestrian information and estimating a pedestrian's knee position and upper body position based on the pedestrian information. When,
Based on information from the pedestrian information detecting means, a moving speed detecting means for detecting a knee position moving speed of the knee position and an upper body position moving speed of the upper body part;
Based on the knee position movement speed and the upper body position movement speed from the movement speed detection means, the kinetic energy by the knee movement and the kinetic energy by the upper body movement are calculated, and the kinetic energy total amount of both kinetic energy Energy calculating means for calculating
A determination means for determining that the pedestrian has started to move when the total amount of kinetic energy exceeds a predetermined value based on the total amount of kinetic energy from the energy calculating means;
Is provided as a configuration. A preferred embodiment of claim 1 is as described in claim 2 and the following.

  According to the invention of claim 1, the pedestrian's behavior (traveling path crossing behavior) is determined by the knee position moving speed of the knee position reflecting the intention of the pedestrian and the upper body position moving speed of the upper body position. Based on the square value of the velocity in the kinetic energy by not only accurately predicting it but also using the knee position and upper body position movement speed to form the kinetic energy of the knee and upper body. The degree of reflection of the knee position movement speed and the upper body position movement speed, which affect the pedestrian's behavior prediction, can be increased. For this reason, it is possible to accurately determine the movement of the pedestrian from the stationary state without a determination delay.

  According to the invention of claim 2, the pedestrian information detecting means is a trapezoidal pedestrian plane having a pedestrian's knee width as a lower base and a pedestrian's upper body width as an upper base from photographing information of an imaging means. The model is set so as to estimate, and the moving speed detecting means detects the spreading speed of the lower base estimated by the pedestrian information detecting means as the knee position moving speed, and the pedestrian information as the upper body position moving speed. Since the detection means is set to detect the estimated movement speed of the upper base, information on the knee and upper body of the pedestrian is simplified to calculate the knee position movement speed and the upper body position movement speed. And the processing burden can be reduced. For this reason, the movement start determination from a pedestrian's stop state can be accelerated.

  According to the invention of claim 3, the energy calculating means is set to calculate the total energy amount of the knee position moving speed and the upper body position moving speed normalized by the normalization coefficient. Therefore, the pedestrian's movement start determination can be performed without depending on the pedestrian's physique.

  According to the invention of claim 4, the energy calculating means is set so as to change the normalization coefficient based on the height information in the pedestrian information detected by the pedestrian information detecting means. The reflected normalization coefficient can be used, and the movement start determination accuracy from the pedestrian stop state can be further increased.

  According to the invention of claim 5, the pedestrian information detecting means estimates the knee position as a ratio position peculiar to the knee position with respect to the height information in the pedestrian information detected by the pedestrian information detecting means. In addition, since the upper body part position is set to be estimated as a ratio position specific to the upper body part position with respect to the height information in the pedestrian information, the knee part position and the upper body part position are simply and accurately determined. Can be estimated well.

  According to the invention of claim 6, since the output means for outputting warning information to the driver when the determination means determines the movement start of the pedestrian, the driver detects the movement start of the pedestrian based on the warning information. It is possible to know at an early stage and avoid collision between the pedestrian and the host vehicle.

  According to the invention of claim 7, since the upper body part is the shoulder part, the behavior of the upper body part can be accurately grasped by the shoulder part.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
In FIG. 1, the code | symbol 1 is a vehicle (own vehicle), The vehicle pedestrian detection apparatus 2 which concerns on this embodiment is mounted in this vehicle 1. FIG. As shown in FIGS. 1 and 2, the vehicle pedestrian detection device 2 includes a stereo camera 3 as an imaging unit, a brake control actuator 4, an alarm display device 5, and a processing device 6. In addition, the photographing information from the stereo camera 3 is input to the processing device 6, while the control signal is output from the processing device 6 to the brake control actuator 4 and the alarm display device 5.

  In the present embodiment, the stereo camera 3 is attached to the upper part of the front window and photographs the entire front side of the vehicle 1 in the traveling direction. Accordingly, the stereo camera 3 also captures the vicinity of the traveling road 7 on the front side of the vehicle 1, and when there is a pedestrian 8 as shown in FIGS. 3 and 4, The pedestrian 8 is also photographed. As is well known, this stereo camera 3 can record information in the depth direction by simultaneously photographing an object from a plurality of different directions. Basically, two or more cameras can be recorded. The lens mechanism and shutter mechanism are combined into one unit. The stereo camera 3 is provided with an image processing unit (not shown), and shooting information shot by the stereo camera 3 is input to the processing device 6 after undergoing stereo image processing or the like in the image processing unit. It is supposed to be.

  The brake control actuator 4 adjusts the operation of the brake, and the brake control actuator 4 operates the brake when a brake operation instruction signal (control signal) is input from the processing device 6. It has become.

  The alarm display device 5 is a display device that displays an alarm display and notifies the driver when an alarm signal (control signal) from the processing device 6 is input. In the present embodiment, the alarm display device 5 A display device of a navigation device is used.

  The processing device 6 determines whether or not there is a pedestrian 8 that moves (starts walking) from the stationary state to the traveling path 7 based on the shooting information from the stereo camera 3. 2, a pedestrian detection processing unit 9, a pedestrian extraction processing unit 10, a pedestrian movement start determination processing unit 11, and an output processing unit 12 are provided. The pedestrian detection processing unit 9 detects pedestrians 8 present in the shooting information based on shooting information (image information) from the stereo camera 3. For this reason, the pedestrian detection processing unit 9 detects the distance and size to the target object with a stereo camera to detect the pedestrian 8, pattern matching by shape, or an IC card carried by the pedestrian It is determined whether or not the object is a pedestrian by receiving a signal from a telephone or radio wave.

  The pedestrian extraction processing unit 10 extracts a pedestrian 8 that is present within a predetermined distance from the vehicle 1 and is stationary from the pedestrians 8 detected by the pedestrian detection processing unit 9 (FIG. 3). FIG. 4). For this reason, in the pedestrian extraction processing unit 10, a matching process between the stored data of the pedestrian 8 stored in advance and the pedestrian 8 data detected by the pedestrian detection processing unit 9 is performed, and the target pedestrian 8 is chosen.

  The pedestrian movement start determination processing unit 11 determines whether or not the pedestrian 8 extracted by the pedestrian extraction processing unit 10 starts moving from the stationary state toward the travel path 7. This pedestrian movement start determination processing unit 11, a pedestrian information detection unit 13, a movement speed detection unit 14, an energy calculation unit 15, and a determination unit 16 are provided to perform the determination, as shown in FIG. The pedestrian information detection unit 13 obtains the distance from the vehicle 1 to the pedestrian 8 and the height thereof based on the shooting information from the stereo camera 3 and, as shown in FIG. Of calculating the height h of the pedestrian and using the apparent height h to calculate the shoulder height (shoulder position) and the knee height (knee position) indicating the representative position of the upper body position of the pedestrian 8 have. Specifically, the shoulder height is obtained from shoulder height = 0.80 × h, and the knee height is obtained from knee height = 0.27 × h. Here, the coefficients “0.80” and “0.27” are obtained in advance as optimum values for deriving the shoulder height and knee height from the apparent height.

  The moving speed detector 14 detects the shoulder position moving speed and the knee position moving speed of the target pedestrian 8. In the moving speed detector 14, the shoulder position detected by the pedestrian information detector 13. The apparent movement speeds Vst and Vkt are calculated for the knee position based on information from the stereo camera 3 (change information with time) (see FIG. 7). Here, the apparent speed is a displacement speed on the image of the pedestrian 8 photographed by the stereo camera 3, and the speed in the front direction of the pedestrian 8 is positive.

Based on the apparent moving speeds Vst and Vkt detected by the moving speed detector 14, the energy calculator 15 determines an energy equivalent value (Vst / α) ² + (Vkt / α) ² is calculated. Although this energy equivalent value will be described later, α used here is a normalization coefficient and changes according to the apparent height h of the pedestrian 8. FIG. 8 shows a specific example of a normalization coefficient that changes in accordance with the apparent height of the pedestrian 8, and the normalization coefficient α corresponding to the apparent height of the pedestrian 8 is used in calculating the energy equivalent value. It is done.

  The determination unit 16 determines whether or not the energy equivalent value calculated by the energy calculation unit 15 exceeds 1. That is, when the determination unit 16 determines that the energy equivalent value exceeds 1, it determines that the condition for the pedestrian 8 to start moving is satisfied, and when the energy equivalent value is determined to be 1 or less, the pedestrian 8 starts to move. It is determined that the above is not satisfied.

The processes of the energy calculation unit 15 and the determination unit 16 are based on the experimental conditional expression (Equation 1) found by the inventors.

In this experimental conditional expression (Equation 1), the energy equivalent value (Vst / α) ² + (Vkt / α) ² is 1, as is apparent from the description of the processing of the energy calculation unit 15 and the determination unit 16 described above. When it exceeds, the pedestrian 8 satisfies the start condition. This is because the total amount of kinetic energy due to knee movement and kinetic energy due to shoulder movement exceeds a predetermined value. The apparent moving speed Vst and the apparent moving speed Vkt of the knee position are normalized by the normalization coefficient α as described above. Specifically, in the normal walking start operation, the foot is swung out while the body is tilted by the moment of gravity generated by shifting the action point of the floor reaction force, and walking is started. For this reason, the said experimental conditional expression (Formula 1) is obtained by assuming that the apparent moving speed required at the start of walking is proportional to the height. In other words, for a person with a height of 150 cm, both Vst and Vkt are 150/170 times that of a person with a height of 170 cm, and the normalization factor is 150/170 times that of a person with a height of 170 cm (in the experiment, the height was about 170 cm and α = 400). The reason for normalizing in this way is to determine the movement of the pedestrian without depending on the pedestrian's physique.

  The apparent movement speed Vst of the shoulder position and the apparent movement speed Vkt of the knee position in the experimental conditional expression (Equation 1) are as follows. This is because the apparent movement speeds Vst and Vkt accurately reflect the intention of the pedestrian 8 and the movement of the pedestrian 8 to the travel path 7 can be predicted by these. The movement velocities Vkt and Vst are used to form the kinetic energy of the knee and shoulder, based on the square value of the velocities in the kinetic energy, and the apparent movement speed that affects the behavior prediction of the pedestrian 8. This is because the degree of reflection of Vkt and Vst can be increased.

Specifically, taking the case of the normalization coefficient 400 (height 170 cm) as an example, if the above experimental conditional expression is illustrated, a circular boundary line (radius = 400) is shown as shown in FIG. In Fig. 9, the inner side in the first quadrant indicates the non-moving region of the pedestrian 8, and the outer side of the circular boundary indicates the moving region of the pedestrian 8 (shaded region in Fig. 9). On the other hand, in FIG. 9, the alternate long and short dash line indicates the boundary line with the moving speed (Vst, Vkt) instead of the kinetic energy, and when the boundary line is indicated with the moving speed (Vst, Vkt). The outside of the boundary line indicates the movement start area of the pedestrian 8. For this reason, in the case of the experimental conditional expression (Equation 1), the movement start area of the pedestrian 8 is expanded to the inside of the boundary line (one-dot chain line) indicated by the moving speed (Vst, Vkt) ( The region surrounded by the circle and the alternate long and short dash line), and using the experimental conditional expression (Equation 1), the pedestrian 8 is compared with the case where the boundary line (the alternate long and short dash line) is indicated by the apparent moving speed (Vst, Vkt). It is possible to detect the movement from the stationary state with high accuracy without a determination delay.
In this embodiment, the normalization coefficient used for each term of the apparent moving speeds Vst and Vkt is the same as α. However, weighting is performed by changing the normalization coefficient according to conditions and circumstances. May be.

  The output processing unit 12 outputs an alarm signal to the alarm display device 5 when the determination unit 16 determines that the pedestrian 8 movement start condition is satisfied, and also operates the brake to the brake control actuator 4. It has a function of outputting an instruction signal.

Such a vehicle pedestrian 8 detection device 2 operates as follows. A specific description will be given centering on the flowchart of FIG. 10 showing the processing contents of the processing device 6.
First, when image information (image processed) is input from the stereo camera 3 to the processing device 6, a detection process of the pedestrian 8 is performed based on the image data in S1 (S indicates a step). This is for detecting the pedestrian 8 present in the image data. In the next S2, among the pedestrians 8, pedestrians 8 that are within a predetermined distance from the vehicle 1 and are stationary are detected. The detection process in this case is performed by a matching process between the image data of the pedestrian 8 in S1 and the image data of the pedestrian 8 stored in advance. Next, in S3, it is determined whether or not the target pedestrian 8 exists based on the result of the detection process in S2. This is because it is possible to know early whether or not a pedestrian has a possibility of collision. When the S3 is NO, the process is returned and monitoring of the target pedestrian 8 is continued, while when the S3 is YES, a movement start determination process for the pedestrian 8 is performed at S4.

  The movement start determination process of the pedestrian 8 is performed according to the flowchart shown in FIG. 11 in this embodiment. In this determination process, first, in S4-1, the apparent height h of the pedestrian 8 is calculated from the distance and height from the own vehicle 1 to the pedestrian 8, and in the next S4-2, the apparent height h is calculated. From the height h, the shoulder height (shoulder position) and knee height (knee position) of the pedestrian 8 are calculated (shoulder height = 0.80 × h, knee height = 0.27 × h) as described above.

Next, in S4-3, the apparent movement speed Vst of the shoulder position obtained in S4-2 is calculated, and in S4-4, the apparent movement speed Vkt of the knee position obtained in S4-2 is calculated. In S4-5, a normalization coefficient α corresponding to the height h obtained in S4-1 is obtained from the map data (see FIG. 8). In S4-6, an energy equivalent value (Vst / α) ² + ( Vst / α) ² is calculated. Then, in the next S4-7, it is determined whether or not the energy equivalent value of S4-6 is greater than 1, and if it is determined that the energy equivalent value of S4-6 exceeds 1, then S4- 8, it is determined that there is a pedestrian 8 that starts to move to the road 7. On the other hand, when it is determined in S4-7 that the energy equivalent value in S4-6 is smaller than 1, it is determined in S4-9 that there is no pedestrian 8 that moves to the travel path 7.

  When the movement start determination process for the pedestrian 8 is completed, in S5, it is determined whether or not there is a pedestrian 8 that moves to the travel path 7 based on the result of the movement determination process for the pedestrian 8 in S4. When S5 is NO, the process returns to S1 and monitoring of the pedestrian 8 is continued. When the determination of S5 is YES, an alarm signal and brake operation control are output. Thereby, the driver can know the pedestrian 8 starting to move to the travel path 7 with a warning display, and can prevent a collision between the pedestrian 8 and the vehicle 1 by the brake operation. In this case, not only the warning display but also a warning sound may be output alone or together with the warning display.

  12 to 14 show a second embodiment. In this 2nd Embodiment, in order to reduce the calculation load of the dynamic prediction determination of the pedestrian 8, the content which replaced the pedestrian 8 of 1st Embodiment with the plane model (pedestrian link model) is shown. . That is, in this second embodiment, after calculating the shoulder height (shoulder position) and knee height (knee position) of the pedestrian 8 from the apparent height h of the pedestrian, as in the first embodiment (shoulder Height = 0.80 × h, knee height = 0.27 × h), the shoulder width at the shoulder height of the pedestrian (between ● marks) is the upper bottom, the width of the knee at the pedestrian's knee height A trapezoidal pedestrian plane model with the bottom of (● between) is estimated. FIG. 12 shows a case where a trapezoidal pedestrian plane model is estimated for a pedestrian in front view, and FIG. 13 shows a case where a trapezoidal pedestrian plane model is estimated for a pedestrian in side view.

In the pedestrian plane model, the upper base of the trapezoidal shape (both shoulder width at the shoulder height) has the following properties.
(1) Because of the forward tilt, the width of the shoulder relative to height (upper base) hardly changes.
(2) The upper base moves parallel to the traveling direction because it tilts forward.
(3) The moving speed of the upper base increases as the speed after the start of walking increases.

Further, the lower base (inter-knee width) of the trapezoidal shape has the following properties.
(1) The width of the knee portion with respect to the height increases because the foot on the first step moves forward and the other foot does not move.
(2) Since the foot on one side is not moving, the width of the knee is expanded by moving only one side of the lower bottom (see the state in FIG. 6).
(3) The faster the walking speed is, the faster the bottom spreads.

ここで、Ｖu：上底の平行移動速度 Ｖｄ：下底の広がる速度
尚、実験条件式（数２）において、上底の移動方向と下底の広がる方向は同一方向となる。また、αは、前記第１実施形態同様、正規化係数を示す。 For this reason, the inventors of the present invention derived an experimental conditional expression (Formula 2) for the movement of the pedestrian from the properties of the upper and lower bases of the trapezoidal shape.

Here, Vu: parallel movement speed of the upper base Vd: speed of spreading of the lower base In the experimental conditional expression (Equation 2), the moving direction of the upper base and the expanding direction of the lower base are the same direction. Α represents a normalization coefficient as in the first embodiment.

  This experimental conditional expression (Equation 1) is also used for the pedestrian movement start determination process as in the first embodiment, and the details thereof will be specifically described based on the flowchart shown in FIG.

Also in the second embodiment, as in the first embodiment, first, the apparent height h of the pedestrian 8, the shoulder height (shoulder position), and the knee height (knee position) of the pedestrian 8 are calculated ( S4-1 ′, S4-2 ′). Next, in S4-3 ′, a trapezoidal shape with the pedestrian 8 as a plane model (pedestrian 8 link model) is calculated from the shoulder position and knee position determined in S4-2 ′, and then, In S4-4 ′, the trapezoidal upper base parallel movement speed Vu in S4-3 ′ is calculated, and in S4-5 ′, the trapezoidal lower base spreading speed Vd in S4-3 ′ is calculated. In S4-6 ′, a normalization coefficient α corresponding to the height h obtained in S4-1 is obtained from the map data (see FIG. 8). In S4-7 ′, an energy equivalent value (Vu / α) ^2. + (Vd / α) ² is calculated. Then, in the next S4-8 ′, it is determined whether or not the energy equivalent value of S4-7 ′ is greater than 1. If it is determined that the energy equivalent value of S4-7 ′ exceeds 1, In S4-9 ′, it is determined that there is a pedestrian 8 that starts moving. On the other hand, when it is determined in S4-8 ′ that the energy equivalent value in S4-7 ′ is smaller than 1, in S4-10 ′, it is determined that there is no moving pedestrian 8. Thereby, in this 2nd Embodiment, not only the movement start of the pedestrian 8 can be determined accurately and early, but the calculation load of the dynamic prediction determination of the pedestrian 8 can be reduced.

The figure which shows the vehicle carrying the pedestrian detection apparatus for vehicles which concerns on 1st Embodiment. The block diagram which shows each process process in the pedestrian detection apparatus for vehicles which concerns on 1st Embodiment. The figure which showed the state from which the pedestrian begins to move toward a driving path crossing from the vehicle side. The figure which showed in plan the state where a pedestrian starts moving toward a running path crossing. The figure which shows the specific content of the pedestrian movement start determination processing part which concerns on 1st Embodiment. Explanatory drawing explaining the relationship between a pedestrian's apparent height, shoulder height, and knee height. Explanatory drawing explaining the apparent moving speed Vst of a shoulder position, and the moving speed Vkt of a knee position. Map data showing the relationship between the apparent height of pedestrians and the normalization factor. Explanatory drawing explaining the movement start area | region of the pedestrian under the experimental conditional expression which concerns on 1st Embodiment. The flowchart which shows the example of control which concerns on 1st Embodiment. The flowchart which shows the example of control of the pedestrian movement start determination process which concerns on 1st Embodiment. Explanatory drawing explaining the process which estimates the plane model (front view) used in 2nd Embodiment. Explanatory drawing explaining the process which estimates the plane model (side view) used in 2nd Embodiment. The flowchart which shows the example of control of the pedestrian movement start determination process which concerns on 2nd Embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Vehicle 2 Pedestrian detection apparatus for vehicles 3 Stereo camera (imaging means)
4 Actuator for brake control 5 Warning display device 6 Processing device 8 Pedestrian 11 Pedestrian movement start determination processing unit 13 Pedestrian information detection unit (pedestrian information detection means)
14 Moving speed detector (moving speed detector)
15 Energy calculation part (energy calculation means)
16 Determination part (determination means)

Claims (7)

A pedestrian detection device for a vehicle for determining movement of a pedestrian from a stationary state,
Imaging means for photographing the periphery of the vehicle;
Pedestrian information detection means for detecting a pedestrian based on photographing information from the imaging means and detecting pedestrian information and estimating a pedestrian's knee position and upper body position based on the pedestrian information. When,
Based on information from the pedestrian information detecting means, a moving speed detecting means for detecting a knee position moving speed of the knee position and an upper body position moving speed of the upper body part;
Based on the knee position movement speed and the upper body position movement speed from the movement speed detection means, the kinetic energy by the knee movement and the kinetic energy by the upper body movement are calculated, and the kinetic energy total amount of the both kinetic energy Energy calculating means for calculating
Determination means for determining that the pedestrian has started to move when the total amount of kinetic energy exceeds a predetermined value based on the total amount of kinetic energy from the energy calculating means;
Equipped with,
A vehicle pedestrian detection device characterized by the above. In claim 1,
The pedestrian information detection means is set to estimate a trapezoidal pedestrian plane model with the pedestrian width between the knees as the bottom and the pedestrian's upper body width as the bottom from the imaging information of the imaging means. And
The moving speed detecting means detects the lower base spreading speed estimated by the pedestrian information detecting means as the knee position moving speed, and the pedestrian information detecting means as the upper body position moving speed. Set to detect the estimated moving speed of the upper base,
A vehicle pedestrian detection device characterized by the above. In claim 1 or 2,
The energy calculating means is set to calculate the total energy amount of the knee position moving speed and the upper body position moving speed normalized by a normalization coefficient.
A vehicle pedestrian detection device characterized by the above. In claim 3,
The energy calculating means is set to change the normalization coefficient based on height information of pedestrian information detected by the pedestrian information detecting means.
A vehicle pedestrian detection device characterized by the above. In claim 1,
The pedestrian information detection means estimates the knee position as a ratio position specific to the knee position with respect to height information in the pedestrian information detected by the pedestrian information detection means, and the upper body position Is set to be estimated as a ratio position specific to the upper body position with respect to the height information in the pedestrian information,
A vehicle pedestrian detection device characterized by the above. In claim 1,
Output means for outputting warning information to the driver when the determination means determines the movement of the pedestrian is provided,
A vehicle pedestrian detection device characterized by the above. In any one of Claims 1-6,
The upper body is a shoulder;
A vehicle pedestrian detection device characterized by the above.

Images