JP2017010560A - Vehicle driving support device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a vehicle driving support device capable of avoiding contact between a pedestrian or a bicycle passenger who is operating a portable terminal and a vehicle.SOLUTION: A vehicle driving support device includes a detection part for detecting a pedestrian or a bicycle passenger existing in front of a vehicle with the use of an on-vehicle camera mounted on the vehicle. It is determined whether a portable terminal is being operated or not by image processing based on the operation posture of the portable terminal by noting the positions of a head and a hand with respect to the pedestrian or the bicycle passenger. When it is determined that a failure is to occur in the driving of the vehicle, brake alarm is issued to a driver and also automatic brake is actuated.SELECTED DRAWING: Figure 2

Description

The present invention
The present invention relates to a driving support apparatus for a vehicle provided with a detection unit that detects a pedestrian or a bicycle occupant existing in front of the vehicle by an in-vehicle camera mounted on the vehicle.

  Conventionally, there has been a technology disclosed in Patent Literature 1 as the above-described vehicle driving support device. In this technology, three-dimensional objects (cars, pedestrians, bicycles, etc.) on the traveling path of the vehicle are photographed and detected by the in-vehicle camera of the detection unit, and after the detection, the possibility of collision with the vehicle is determined and attention is paid to the driver. It used to alert, warn, and finally activate the automatic brake to avoid contact with the vehicle.

JP 2008-126957 A

In recent years, among pedestrians, there are pedestrians walking while looking at mobile terminals such as mobile phones and smartphones. In addition, among bicycle riders, there are bicycle riders who drive a bicycle with one hand while looking at a mobile terminal. These pedestrians are obsessed with the operation of mobile terminals, distracting attention, and are not aware of the approach of cars, which is one of the traffic issues.
However, the conventional technology described above does not distinguish between a pedestrian or a bicycle occupant who is moving while looking at a mobile terminal, and a pedestrian or a bicycle occupant who is moving without looking at a mobile terminal. As a normal pedestrian or bicycle occupant who has the same possibility of contact, the vehicle is controlled so that the vehicle does not contact them. Therefore, according to the above-described conventional technology, there is a possibility that a pedestrian or a bicycle occupant who is moving while looking at the mobile terminal is in contact with the vehicle.
An object of the present invention is to provide a vehicle driving support device capable of avoiding contact between a vehicle and a pedestrian or bicycle rider who is operating a portable terminal.

The feature of the present invention is that
In a vehicle driving support device including a detection unit that detects a pedestrian or a bicycle occupant existing in front of a vehicle by an in-vehicle camera mounted on the vehicle,
For the pedestrian or bicycle occupant, it is determined whether or not the mobile terminal is being operated by image processing based on the operation attitude of the mobile terminal that focuses on the position of the head and hand, and an obstacle occurs in driving the vehicle. In this case, the brake warning is given to the driver and the automatic brake is activated. (Claim 1)

Whether the pedestrian or the bicycle occupant is operating the mobile terminal by the image processing based on the operation posture of the mobile terminal focusing on the position of the head and the hand with respect to the pedestrian or the bicycle occupant with the above configuration. Determine.
For example, when it is determined that the vehicle is in contact with a pedestrian or bicycle rider who is operating the mobile terminal and the vehicle is operating, a brake warning is issued to the driver and an automatic brake is activated.
Thereby, the contact of the vehicle with the pedestrian or bicycle rider who is operating the portable terminal can be avoided.
In addition, it is determined whether or not the pedestrian or the bicycle occupant is operating the mobile terminal by image processing based on the operation posture of the mobile terminal focusing on the position of the head and the hand. Therefore, the accuracy of determination can be improved. (Claim 1)

In the present invention,
If the brake alarm is made to be two types of the pedestrian and the bicycle occupant, that is, the case where the portable terminal is operated and the case where the portable terminal is not operated, the following effects can be obtained. (Claim 2)

  Contact between a normal pedestrian or bicycle rider who is not operating the portable terminal and the vehicle can be avoided. (Claim 2)

In the present invention,
If the image processing is characterized by pattern matching having a learning function, the following effects can be obtained. (Claim 3)

  Since such a learning function is provided, the performance of the present apparatus can be improved. (Claim 3)

In the present invention,
The in-vehicle camera is a stereo camera, and calculates the feature amount from the image to determine whether the pedestrian or the bicycle occupant is operating the portable terminal, and has the following effects. Can do. (Claim 4)

  The three-dimensional position of the hand and head of the pedestrian or bicycle rider can be obtained from the distance image, and a pedestrian who may come into contact with the vehicle can be easily determined, and there is no fear of omission of determination. In addition, for example, the Hog feature value normalizes a histogram for obtaining a luminance gradient, and thus is resistant to changes in brightness such as lighting. Therefore, it is easy to detect a pedestrian who may be in contact with a vehicle in any scene. There are advantages. (Claim 4)

According to the present invention,
It has been possible to provide a driving support device for a vehicle that can avoid contact between a pedestrian operating a portable terminal or a bicycle pedestrian and a vehicle.

System configuration diagram of vehicle driving support device Overall flowchart of vehicle driving support device Judgment flowchart (A) is a schematic view of a normal pedestrian walking without looking at the mobile terminal, and (b) is a pedestrian walking while looking at the mobile terminal (a pedestrian operating the mobile terminal) , The same shall apply hereinafter) viewed from the side, (c) is a schematic view of a pedestrian walking from the front while looking at the mobile terminal. Schematic view of a bicycle occupant driving a bicycle while looking at the mobile device (bicycle occupant operating the mobile device, the same applies hereinafter) seen from the side. It is a figure which shows the characteristic of the pedestrian who walks while looking at a portable terminal, (a) is a schematic diagram of the pedestrian who is transversely (or diagonally) with respect to the vehicle, (b) is directly behind the vehicle, Schematic diagram of a pedestrian facing straight ahead (A)-(d) is a schematic diagram which shows the attitude | position of the pedestrian who is looking at a portable terminal, (e) is a schematic diagram which shows the attitude | position of the pedestrian who does not look at a portable terminal The figure which shows the means to calculate the angle of the head and hand with respect to the trunk of the pedestrian who is operating the portable terminal (A) is a figure which shows the timing which act | operates a brake alarm and an automatic brake when the pedestrian who is looking at a portable terminal exists, (b) is the brake when the pedestrian who does not look at a portable terminal exists Diagram showing the timing to activate the alarm / automatic brake Schematic diagram showing the positional relationship between the vehicle and the pedestrian before contact Overall Flowchart of Vehicle Driving Support Device of Second Embodiment Determination flowchart according to the second embodiment

Hereinafter, embodiments for carrying out the present invention will be described with reference to the drawings.
FIG. 1 shows a system configuration diagram of a vehicle driving support device 9. The driving support device 9 includes a pedestrian detection unit 1 (corresponding to a detection unit) that detects a pedestrian 5 (see FIG. 7, FIG. 10, etc.) in front of the vehicle by photographing with the in-vehicle camera 11,
With respect to the pedestrian 5, the pedestrian 5 can use the mobile terminal 50 (see FIGS. 4B and 4C) by image processing based on the operation posture of the mobile terminal 50 focusing on the positions of the head 5 </ b> A and the hand 5 </ b> B. A pedestrian determination unit 2 that determines whether or not the user is operating,
A vehicle control unit 3. The portable terminal 50 is a general term for a mobile phone, a smartphone, and the like.

  The vehicle control unit 3 includes an alarm unit 31 and an automatic brake unit 32. When the pedestrian determination unit 2 determines that a failure occurs in the driving of the vehicle 100, that is, the pedestrian 5 may contact the vehicle 100 (see FIG. 10), and the pedestrian 5 is a portable terminal. When the contact possibility determination means 22 of the pedestrian determination unit 2 determines that the vehicle is moving while looking at 50 (while operating the portable terminal 50), a brake warning is issued to the driver and an automatic brake is activated. As described above, the vehicle control unit 3 controls the vehicle 100.

  There are two types of brake alarms for the pedestrian 5, the case where the portable terminal 50 is operated and the case where the portable terminal 50 is not operated, and the vehicle control unit 3 may cause the pedestrian 5 to contact the vehicle 100. Also, when the pedestrian determination unit 2 determines that the pedestrian 5 is not moving while looking at the portable terminal 50, the vehicle 100 is configured so that the brake warning is given to the driver and the automatic brake is activated. To control.

  When the pedestrian 5 is in a state of moving while looking at the mobile terminal 50, the vehicle control unit 3 automatically gives a brake warning at an earlier timing than when the pedestrian 5 is in a state of moving while looking at the mobile terminal 50. The vehicle 100 is controlled so that the brake is activated.

  As shown in FIG. 2, in step 1 (step S1), it is determined whether or not the detected pedestrian 5 exists outside the own vehicle traveling path. If the vehicle exists outside the traveling path of the vehicle, the vehicle is controlled based on the TTC (predicted contact time for the pedestrian 5) obtained from the distance to the pedestrian 5 and the vehicle speed.

  In the next step 2, it is determined for t seconds whether the pedestrian existing outside the own vehicle traveling path is the pedestrian 5 with the possibility of contact looking at the portable terminal 50 or the normal pedestrian 5 (for details, see FIG. 3 (refer to the flowchart of the pedestrian determination unit 2).

  In the final step 3, vehicle control is performed on the determined pedestrian 5 with a possibility of contact and the normal pedestrian 5.

  As described above, the timing for issuing the brake alarm and the timing for operating the automatic brake are determined by moving while looking at the portable terminal 50 and the pedestrian 5 with a possibility of contact while moving while looking at the portable terminal 50. Unlike the normal pedestrian 5 who is not in the state of being, it is as shown in FIGS. 9 (a) and 9 (b). 9 (a) and 9 (b) will be described in detail later.

[Details of Step 2 (Part 1)]
The determination by the pedestrian determination unit 2 includes a learning stage and a practical stage. In the learning stage, as shown in FIGS. 4 (a) to 4 (c) and FIGS. 7 (a) to 7 (e), the images D1 and D1 in various directions of the pedestrian 5 looking at the portable terminal 50 are shown. D2, D3, D4,... (See FIG. 8A) are collected. FIG. 7E is a schematic diagram showing the posture of the pedestrian 5 who is not looking at the mobile terminal 50.

  4A is a schematic view of a normal pedestrian 5 walking without looking at the portable terminal 50, and FIG. 4B is a schematic view of the pedestrian 5 walking while looking at the portable terminal 50. FIG. FIG. 4C is a schematic diagram viewed from the lateral direction, and is a schematic diagram of the pedestrian 5 walking from the front while looking at the mobile terminal 50.

  Then, the feature amount calculation means 21 (see FIG. 1) of the pedestrian determination unit 2 obtains a Hog (Histograms of Oriented Gradients) feature amount X for each image and, as shown in FIG. A learning database is constructed in which the angle (y1) between the head 5A and the torso 5C of the person 5 and the angle (y2) between the hand 5B and the torso 5C are constructed.

Codes A1 and A2 in FIG. 8B are (nx1) matrices for converting Hog feature amounts X into angles, and learning types from n learning data sets {(yi, xi) | i = 1... N}. It is determined using a pattern recognition method.
Examples are learning pattern recognition methods such as neural networks, AdaBoost, and support vector machines. In the practical stage, the Hog feature amount X input from the image is subjected to the transformation matrix A1 and A2 calculated in the learning stage, and the angle y1 between the pedestrian 5C and the head 5A, and the pedestrian 5 torso 5C. And the angle y2 of the hand 5B is obtained.
If the absolute value of the difference between y2 and y1 (y2−y1) is 17 degrees or less, it is determined that the pedestrian 5 is viewing the mobile terminal 50. That is, according to the following non-patent literature on ergonomics, the angle between the head and hand using a mobile phone is 7 degrees at the maximum, and the measurement resolution is 5 degrees. Since an error of 10 degrees (= 5 degrees × 2) occurs, 10 degrees + 7 degrees = 17 degrees.
[Non-patent literature]
Prof. Satoru Kubota, Professor, Seikei University at the Ergonomics Symposium 2014 (Friday, March 7th, 2014) held at the flat panel display ergonomics symposium Slide 24 pages

  Further, a hand region is obtained from the image, and if a linear object can be detected from the region (an edge image is obtained and linear approximation is performed), it is determined that the mobile terminal 50 is likely to be held.

  The image processing is pattern matching having a learning function. Since the vehicle driving support device 9 has such a learning function, the performance of the vehicle driving support device 9 can be improved.

[Details of Step 2 (Part 2)]
Next, FIG. 3 showing details of step 2 will be described.
(Step 201 to Step 204)
For the pedestrian 5 in step 201, the edge of the pedestrian 5 is extracted in step 202, and in step 203, the direction of the head 5A and the direction of both hands 5B of the pedestrian 5 are calculated. In step 204, the direction of the pedestrian 5 relative to the vehicle 100 (own vehicle) is estimated.

(Step 205)
In step 205, it is determined whether or not the pedestrian 5 is laterally or obliquely with respect to the vehicle 100.

(Step 206)
If the pedestrian 5 is oriented sideways and obliquely with respect to the vehicle 100, in step 206, the angle [θ] formed by the head 5A and the hand 5B of the pedestrian 5 is calculated. [Θ] = | y2−y1 |.

(Step 207)
In step 207, 0 ≦ [θ] ≦ 17 is determined.

(Step 208)
If 0 ≦ [θ] ≦ 17, it is determined in step 208 whether or not a straight edge comes out of the hand 5B of the pedestrian 5. If 0 ≦ [θ] ≦ 17, it is determined in step 212 that the pedestrian 5 is not looking at the portable terminal 50.

(Step 209)
If it is determined in step 208 that there is no straight edge from the hand 5B of the pedestrian 5, in step 209, the possibility that the pedestrian 5 is looking at the portable terminal 50 is determined to be medium.

(Step 211)
If it is determined in step 208 that a straight edge has come out of the hand 5B of the pedestrian 5, it is determined in step 211 that the possibility that the pedestrian 5 is looking at the portable terminal 50 is high.

(Step 210)
If it is determined in step 205 that the pedestrian 5 is not laterally or obliquely (front or rearward) with respect to the vehicle 100, the orientation of one hand 5B is fixed or the orientation of both hands 5B is determined in step 210. It is determined whether or not it is fixed. If the direction of one hand 5B is fixed or the direction of both hands 5B is fixed, the process proceeds to step 208. If the direction of one hand 5B is not fixed and the direction of both hands 5B is not fixed, the process proceeds to step 212.

[When the detection target is extended not only to the pedestrian 5 but also to the bicycle occupant 5]
In the present invention, the detection target can be extended not only to the pedestrian 5 but also to the bicycle occupant 5. In this case, as shown in FIG. 5, learning images of the bicycle rider 5 who drives the bicycle 60 while looking at the portable terminal 50 are collected. The flowchart and processing method are the same as those for the pedestrian 5. In this case, the pedestrian detection unit 1 also needs means for separating the pedestrian 5 and the bicycle occupant 5.

For example, when T _h = 5.0 s, t _A and t in FIG. 9A are as shown in Table 1 and Table 2 below. The relative speeds in Tables 1 and 2 are relative speeds in the traveling direction of the vehicle 100 and the pedestrian 5.
Table 1 shows a normal pedestrian warning timing: TTC (s) (t _{B in} FIG. 9B), and Table 2 shows a pedestrian determination time with a possibility of contact (FIGS. 9A and 9B). b) t) is shown.

  Using the determination time t in Table 2, it is determined whether the pedestrian 5 is likely to be in contact with the mobile terminal 50 or is a normal pedestrian 5 who is not looking at the mobile terminal 50. Since the recognition accuracy decreases as the distance from the pedestrian 5 increases (the speed increases), the value decreases as the distance from the pedestrian 5 increases.

  Table 1 shows the timing for issuing an alarm for the vehicle 100. The reason why the relative speed is from −8.33 (m / s) is that the distance from the pedestrian 5 is 10 m or more from the flowchart of FIG. 2, so that an alarm is issued if it is −8.33 (m / s) or more. This is because the distance is 10 m or less.

The results for each lap rate in Table 1 indicate that when pedestrian 5 moves sideways (the speed in the traveling direction of vehicle 100 is 0 km / h (V _{ha in} FIG. 10)) after t _A seconds, It shows how the vehicles 100 overlap. The larger the lap rate, the closer the lateral position of the vehicle 100 and the pedestrian 5 is, and the higher the possibility of collision, so the warning is given early (the pedestrian 5 is in the lateral direction). Result when the vehicle 100 goes straight in the direction of travel). In FIG. 10, h is the distance between the pedestrian 5 and the vehicle 100.

The lap ratios in Tables 1 and 2 are set to a ratio in which the pedestrian 5 and the vehicle 100 overlap each other at a position after the TTC of the pedestrian 5 (100% for the entire pedestrian).
Relative speeds in Tables 1 and 2: A difference between the speed (V _ca ) of the vehicle 100 in the traveling direction of the vehicle 100 and the pedestrian speed (V _ha ).
The reason why the relative speeds in Table 1 and Table 2 are not less than −30.666 (m / s) is that the relative speed is less than −30.566 (m / s) and usually it does not travel on the general road. It is.

Next, the timing for issuing the alarm shown in Table 1 is proved.
(Proof of TTC 1.37 (s) when the relative speed is 13.89 m / s and the lap rate is 10%) (* 1 mark in Table 1)
(Proof of TTC 1.47 (s) when the relative speed is 13.89 m / s and the lap rate is 20%) (* 2 mark in Table 1)

From Table 1, V _ha = 0 (m / s), V _hb = 1.388 (m / s), V _ca = 1.89 (m / s), V _cb = 0 (m / s), 1 _h = 1.475 (m), and the dry asphalt (μ = 0.8) stop distance can be determined as follows. (In this case, it is assumed that the deceleration deceleration is constant 0.8G, and the communication delay from the controller that controls the alarm to the ESP controller that controls the brake is 0.3 (s).)

Braking distance (m) = (square of initial velocity) / (2 × road friction coefficient × gravity acceleration) = 13.89 ² /(2×0.8×9.8)=12.30 (m)
Free running distance (m) = Reaction time (s) x Vehicle speed before braking (m / s) = 0.3 x 13.89 = 4.17 (m)
Stop distance (m) = 12.30 + 4.17 = 16.47 (m)
Since an alarm is issued so that it stops 1m before the stop distance,
(16.47 + 1) /13.89 = 1.26 (s)
When the relative speed with the pedestrian 5 is -13.89 (m / s), an alarm is issued when the TTC is 1.26 (s). This is the timing for issuing an alarm when the lap is 0%. When the lap rate is 10%, 0.1475 (m) from the left end of the vehicle 100 (own vehicle width (1 _h = 1.475 (m)). The pedestrian 5 collides with the pedestrian 5. The pedestrian 5 walks from the left end of the vehicle 100 to 10% at 0.106 (s) (0.1475 (m) ÷ 1.388 (m / s). Further, it is necessary to issue an alarm before 0.11 (s).
1.26 + 0.11 = 1.37 (s)

Next, when the lap rate is 20%, the vehicle 100 collides with a pedestrian at 0.295 (m) from the left end of the vehicle 100. When the lap rate is 20%, the pedestrian 5 reaches 0. Since it walks at 21 (s), it is necessary to issue an alarm before 0.21 (s).
1.26 + 0.21 = 1.47 (s)

  The value of lap rate of 30% or more in Table 1 is calculated in the same way as the above formula in order to keep the distance from the pedestrian 5 at least 1 m when stopped, and an alarm will be issued before the lap rate increases. It has become. The determination time 1.5 (s) when the relative speed in Table 2 is -13.89 (m / s) and the lap rate is 10 (%) was determined in consideration of the pedestrian 5 and the relative speed. The table was determined from TTC 5.0 (s) with pedestrians, and since the distance to pedestrians 5 is far, 1.5 (s) was set so as to keep a large amount.

  According to the present invention, it is possible to avoid contact between the vehicle 100 and the pedestrian 5 or the bicycle occupant 5 who is operating the portable terminal 50. Further, the pedestrian 5 or the bicycle occupant 5 uses the mobile terminal 50 to perform the image processing based on the operation posture of the mobile terminal 50 focusing on the positions of the head 5A and the hand 5B. Since it is determined whether or not the operation is in progress, the accuracy of the determination can be improved.

[Second Embodiment]
In 2nd Embodiment, the said pedestrian detection part 1 image | photographs and detects the pedestrian 5 (refer FIG.7, FIG.10 etc.) of the vehicle front Fr with a stereo camera. As shown in FIG. 11, in step 301 (step S301), the pedestrian determination unit 2 (see FIG. 1) determines whether or not the detected pedestrian 5 exists outside the own vehicle traveling path. When the pedestrian 5 exists outside the own vehicle traveling path, the vehicle control unit 3 (see FIG. 1) uses the TTC (predicted contact time for the pedestrian 5) determined from the distance between the vehicle 100 and the pedestrian 5 and the own vehicle speed. Control the vehicle.

  In the next step 302, the pedestrian determination unit 2 determines whether the pedestrian existing outside the traveling path of the vehicle is the pedestrian 5 with contact possibility watching the portable terminal 50 or the normal pedestrian 5 for t seconds. (For details, see the flowchart of FIG. 12 (the flowchart of the pedestrian determination unit 2)).

  In the final step 303, the vehicle control unit 3 performs vehicle control on the determined pedestrian 5 with a possibility of contact and the normal pedestrian 5.

  The timing of issuing a brake alarm and the timing of operating the automatic brake are not in the state of moving while looking at the portable terminal 50 and the pedestrian 5 with a possibility of contact while moving while looking at the portable terminal 50. It differs from pedestrian 5. The timing is as shown in FIGS. 9A and 9B described in the first embodiment.

[Details of Step 302]
FIG. 12 shows details of step 302. As shown in FIG. 12, the pedestrian determination unit 2 performs determination as follows.
(Step 401 to Step 403)
For the pedestrian 5 in step 401, the orientation of the pedestrian 5 is calculated in step 402. In step 403, the Hog feature amount (the Hog feature amount is a feature amount that can represent the shape of an object present in the image). Then, a histogram-like feature amount obtained from the luminance information of each pixel (pixel) of the image and obtained based on the direction and magnitude of the luminance gradient in the local region (cell) in the image is calculated.

(Step 404)
In step 404, it is determined whether or not the pedestrian 5 is facing directly behind or directly in front of the vehicle 100 (own vehicle).

(Step 408) to (Step 410)
If the pedestrian 5 is lateral or oblique with respect to the vehicle 100, a distance image is created in step 408, and in step 409, the "hand 5B, head 5A" position of the pedestrian 5 is calculated from the Hog feature amount. In step 410, it is determined whether or not the hand 5 </ b> B is oriented in the traveling direction of the pedestrian 5. If the hand 5B is not in the direction of travel of the pedestrian 5, the process returns to step 402.

(Step 411)
If the hand 5B is facing the direction of travel of the pedestrian 5, in step 411, an angle y1 formed by the torso 5C and the head 5A of the pedestrian 5 and an angle y2 formed by the torso 5C of the pedestrian 5 and the hand 5B are obtained. .

(Step 412)
In step 412, 0 ≦ | y1-y2 | ≦ 17 is determined.

(Step 208)
If 0 ≦ | y1−y2 | ≦ 17, it is determined in step 407 that the pedestrian 5 is the pedestrian 5 looking at the mobile terminal 50. The reason is as described in the first embodiment. Unless 0 ≦ | y1−y2 | ≦ 17, the process returns to step 402.

(Step 405)
If it is determined in step 404 that the pedestrian 5 is facing frontward or rearward with respect to the vehicle 100, in step 405, the “hand 5B, head 5A” position of the pedestrian 5 is calculated from the Hog feature amount. In step 406, it is determined whether the direction of one hand 5B is fixed or whether the direction of both hands 5B is fixed. If the direction of one hand 5B is fixed or the direction of both hands 5B is fixed, it is determined in step 407 that the pedestrian 5 is the pedestrian 5 looking at the portable terminal 50. If the orientation of both hands 5B is not fixed, the process returns to step 402.

  As described above, the pedestrian determination unit 2 creates a distance image from the stereo camera, calculates a Hog feature amount from the stereo camera image, and uses the height information of the hand 5B and the head 5A of the distance image to generate the Hog feature amount. The positions of the hand 5B and the head 5A are obtained from the change of the brightness gradient of the head 5A. The Hog feature value normalizes a histogram for obtaining a luminance gradient, and thus has an advantage that it is resistant to changes in brightness such as illumination, and that it is easier to detect the positions of the hand 5B and the head 5A than edge detection.

  According to said structure, the pedestrian 5 with the possibility of contact and the normal pedestrian 5 who are walking in front of the vehicle 100 can be separated, and it walks while seeing the portable terminal 50, and does not notice the vehicle 100. A pedestrian 5 with a possibility of contact can be found preferentially. This is because a distance image is created from a stereo camera, and a three-dimensional position can be obtained. Therefore, an angle is obtained by looking at the positions of the hand 5B and the head 5A, and the hand 5B is directed in the traveling direction of the pedestrian 5. This is because it is possible to detect the pedestrian 5 who is absorbed in the mobile terminal 50 and is not aware of the vehicle 100.

In addition, it is possible to avoid a determination error such that a pedestrian 5 with a contact possibility that does not apply to the learning database described in the first embodiment is not determined as a pedestrian 5 with a contact possibility.
In the invention of the second embodiment, the position of the hand 5B and the head 5A is obtained from the place where the brightness gradient of the Hog feature amount is greatly changed by searching from the ground, and the three-dimensional position of the hand 5B and the head 5A from the distance image. Therefore, it is possible to easily determine the pedestrian 5 with a possibility of contact, and there is no fear of omission of determination.

  In the invention of the second embodiment as well, the detection target can be extended not only to the pedestrian 5 but also to the bicycle occupant 5.

1 detector (pedestrian detector)
5 Pedestrians and bicycle riders 5A Head 5B Hand 11 Car-mounted camera 50 Mobile terminal (cell phone, smartphone, etc.)
100 Vehicle Fr Vehicle front

Claims (4)

In a vehicle driving support device including a detection unit that detects a pedestrian or a bicycle occupant existing in front of a vehicle by an in-vehicle camera mounted on the vehicle,
For the pedestrian or bicycle occupant, it is determined whether or not the mobile terminal is being operated by image processing based on the operation attitude of the mobile terminal that focuses on the position of the head and hand, and an obstacle occurs in driving the vehicle. A vehicle driving support device that issues a brake warning to the driver and activates an automatic brake when the determination is made.   2. The vehicle driving support device according to claim 1, wherein the brake warning is classified into two types for the pedestrian and the bicycle occupant when the portable terminal is operated and when the portable terminal is not operated. .   The vehicle driving support device according to claim 1, wherein the image processing is pattern matching having a learning function.   2. The vehicle according to claim 1, wherein the in-vehicle camera is a stereo camera, and calculates a feature amount from an image to determine whether the pedestrian or a bicycle rider is operating the portable terminal. Driving assistance device.

Images