JP5304292B2 - Vehicle approaching object detection device - Google Patents

Description

  The present invention relates to a vehicle approaching object detection device.

2. Description of the Related Art Conventionally, an approaching object detection device for a vehicle that detects an approaching object to a vehicle during traveling from an image captured by a camera installed in the rearward direction of the vehicle has been proposed. Such devices include those using a gradient method based on the principle of spatiotemporal differentiation (see Patent Documents 1 and 2) and those using a block matching method (see Patent Document 3).

Japanese Patent No. 3011566 Japanese Patent No. 3735468 Japanese Patent No. 3776094

However, in the conventional approaching object detection device for a vehicle, when the camera that captures the rear of the vehicle is a wide-angle camera, the following problems occur. For example, when an overtaking vehicle is imaged by a wide-angle camera, the overtaking vehicle may move by several tens of pixels during one frame. In this case, when the block matching method is used, it is necessary to widen a search area for detecting an approaching object, which greatly increases the amount of calculation. Further, when the gradient method is used, the amount of calculation can be suppressed as compared with the block matching method, but the detection accuracy of the approaching object is lowered.

The present invention has been made to solve such a conventional problem, and the object of the present invention is to detect an approaching object while suppressing an increase in calculation amount even when a wide-angle camera is used. An object of the present invention is to provide a vehicle approaching object detection device capable of suppressing a decrease in accuracy.

An approaching object detection device for a vehicle according to the present invention is an approaching object detection device for a vehicle that detects an approaching object from the rear of the vehicle based on an image from a camera that is provided in the vehicle and images the rear side of the vehicle. An input unit that inputs data of an image captured by the camera, and an average value of adjacent pixel values for each pixel in the horizontal direction of the image input to the input unit is used as the pixel value of the pixel. Pre-processing means for executing the averaging processing to be obtained , and data of a plurality of images taken by the camera at different times, the pixel values of predetermined pixels adjacent in the horizontal direction obtained by the pre-processing means the average value, comprising: a differentiating means for determining the spatial rate of change and the temporal change rate, and a approaching object detection means for detecting an approaching object from the obtained rate of change to the vehicle by the differentiating means, the pretreatment Means is characterized by increasing the number of predetermined pixels according to the pixel away from the vanishing point of the image.

According to the present invention, for each pixel in the horizontal direction of the image input to the input means, an averaging process is performed to obtain an average value of pixel values of adjacent predetermined pixels as the pixel value of the pixel, and the spatial change The approaching object to the vehicle is detected by obtaining the rate and the rate of change over time. Thus, since an approaching object is detected by the gradient method without using the block matching method, an increase in the amount of calculation can be suppressed. In addition, since the averaging process is performed, elements that lead to a decrease in the system when removing the rate of change, that is, a sudden change between images, are removed, and a decrease in the detection system for approaching objects can be suppressed. it can. Therefore, even when a wide-angle camera is used, it is possible to suppress a decrease in the accuracy of detecting an approaching object while suppressing an increase in calculation amount.

It is a block diagram which shows the approaching object detection system containing the approaching object detection apparatus for vehicles of this embodiment. It is a graph which shows the mode of calculation of the one-dimensional flow F (t, x) by the conventional gradient method, (a) shows the pixel value of the column x at the time t, (b) is the pixel of the column x at the time t + 1. (C) shows the spatial change rate I _x (t, x) obtained from the data shown in FIGS. 2 (a) and 2 (b), and (d) shows the values shown in FIGS. FIG. 2B shows a temporal change rate I _t (t, x) obtained from the data shown in FIG. 2B, and FIG. 2E shows a one-dimensional flow F () obtained from FIG. 2C and FIG. t, x). FIG. 3 shows an example of an image taken by the camera, (a) shows an example of an image taken at time t, and (b) shows an example of an image taken at time t + 1. It is a graph for demonstrating operation | movement of the approaching object detection apparatus for vehicles which concerns on this embodiment, (a) shows the data which averaged about the pixel value of the column x in the time t, (b) shows in the time t + 1. (C) shows the spatial rate of change Ia _x (t, x) obtained from the data shown in FIG. 4 (a) and FIG. 4 (b). (d) are shown to FIGS. 4 (a) and 4 (b) to show temporal change rate _Ia t obtained from the data (t, x), (e) is, FIG. 4 (c) and 4 ( A one-dimensional flow Fa (t, x) obtained from d) is shown. It is the figure which showed the result of the one-dimensional flow Fa (t, x) graphically. It is a graph which shows the mode of calculation of the one-dimensional flow F (t, x) by the conventional gradient method, Comprising: The case where a Gaussian filter is used is shown, (a) is a Gaussian about the pixel value of the column x in the time t. FIG. 6B shows data when a filter is used, FIG. 6B shows data when a Gaussian filter is used for the pixel value of column x at time t + 1, and FIG. 6C shows FIGS. 6A and 6B. The spatial change rate Ig _x (t, x) obtained from the data shown in FIG. 6 is shown, and (d) shows the temporal change rate Ig _t (shown from the data shown in FIGS. 6A and 6B). t, x), and (e) shows the one-dimensional flow Fg (t, x) obtained from FIG. 6 (c) and FIG. 6 (d). It is the approaching object detection system 2 containing the approaching object detection apparatus for vehicles which concerns on 2nd Embodiment. It is a figure which shows the image at the time of the vertical blurring of a camera.

DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments of the invention will be described with reference to the drawings. FIG. 1 is a block diagram showing an approaching object detection system including a vehicle approaching object detection device of the present embodiment. As shown in FIG. 1, the approaching object detection system 1 includes a camera 10 and a vehicle approaching object detection device 20. The camera 10 is a wide-angle camera that is provided at the rear of the vehicle and captures the rear side of the host vehicle. Here, the wide-angle camera in this embodiment refers to a camera having a diagonal angle of view of about 60 degrees or more.

The vehicle approaching object detection device 20 detects an approaching object (for example, another vehicle) from the rear of the vehicle based on a plurality of images captured by the camera 10 at different times. Such a vehicle approaching object detection device 20 includes an input unit (input unit) 21, a preprocessing unit (preprocessing unit) 22, a differential operation unit (differentiation unit) 23, and an approaching object detection unit (approaching object detection). Means) 24 and output unit 25
And.

The input unit 21 inputs data of an image captured by the camera 10. The preprocessing unit 22 performs an averaging process for obtaining an average value of pixel values of adjacent pixels as the pixel value of each pixel in the horizontal direction of the image input to the input unit 10. . For example, the preprocessing unit 22 sets an average value of seven pixels including three pixels on the left and right of a certain pixel as the pixel value of the center pixel. Thereby, the pre-processing unit 22 averages the pixel values. Note that the preprocessing unit 22 is not limited to the case where the average value is obtained from seven pixels, but six or less or eight.
You may obtain | require from two or more pixels, and you may obtain | require the average value of only the predetermined number of the left of a certain pixel, and the average value of only the predetermined number of right. The pixel value means a luminance value in the present embodiment,
It means not only the luminance value but also any numerical value relating to an image or pixel such as a color number.

The differential calculation unit 23 is data of a plurality of images captured by the camera 10 at different times, and the horizontal change rate (spatial differential value) of the horizontal data averaged by the preprocessing unit 22 is calculated. The time change rate (temporal differential value) is obtained. Specifically,
Assuming that the luminance value of the column x at time t is I (t, x) in a certain row on the image, the spatial change rate I _x (t, x) is I _x (t, x) = (I (t, x + 1) -I (t,
x) + I (t + 1, x + 1) -I (t + 1, x)) / 2.
Further, the temporal change rate I _t (t, x) is equal to I _t (t, x) is _{equal to} I _t (t, x) = (I (t +
1, x) -I (t, x) + I (t + 1, x + 1) -I (t, x + 1)) / 2.

The approaching object detection unit 24 detects an approaching object to the vehicle from the rate of change obtained by the differential calculation unit 23. The approaching object detection unit 24 can obtain a one-dimensional flow in the horizontal direction from the spatial change rate and the temporal change rate. Specifically, the approaching object detection unit 24 calculates the one-dimensional flow F (t, x) of the column x at time t as F (t, x) = − I _t (t, x) / I _x (t,
x). In the above arithmetic expression, I _x (t, x) ≠ 0
It is. In this way, the approaching object detection unit 24 can obtain the one-dimensional flow F (t, x), and therefore can detect the movement of other vehicles in the horizontal direction on the image, and based on this, An approaching object to the vehicle can be detected.

When the approaching object is detected, the output unit 25 outputs the information to a display, a speaker, or the like provided in the vehicle. Accordingly, driving assistance can be performed by notifying the driver of an approaching object by an image or by notifying an approaching object by voice.

Next, prior to describing the operation of the vehicle approaching object detection device 20 according to the present embodiment, FIG.
Referring to FIG. 4, the calculation of the one-dimensional flow F (t, x) by the conventional gradient method will be described. FIG. 2 is a graph showing how the one-dimensional flow F (t, x) is calculated by the conventional gradient method.
FIG. 2A shows the pixel value of column x at time t, and FIG. 2B shows the pixel value of column x at time t + 1. 2C shows the spatial change rate I _x (t, x) obtained from the data shown in FIGS. 2A and 2B, and FIG. 2D shows the spatial change rate I _x (t, x). ) And the temporal change rate I _t (t, x) obtained from the data shown in FIG. Further, FIG. 2 (e) shows a one-dimensional flow F (t, x) obtained from FIG. 2 (c) and FIG.
).

In the example shown in FIG. 2, it is assumed that the vehicle behind the host vehicle approaches about 10 pixels. In this case, assuming that the luminance value is high at the location where the vehicle exists, as shown in FIGS. 2 (a) and 2 (b), a high luminance value is shown in the vehicle existing portion (reference numerals 2a1, 2a2). , 2
b1, b2).

FIG. 3A is a diagram illustrating an example of an image captured at time t, and FIG. 3B is a diagram illustrating an example of an image captured at time t + 1. As shown in FIGS. 3A and 3B, portions 2a1, 2a2, 2b1, and 2b2 in FIGS. 2A and 2B are located at the positions of the vehicles 100 and 101 in FIG. It corresponds.

Then, from the data as shown in FIG. 2A and FIG. 2B, the spatial change rate I _x (t, x
) And the temporal change rate I _t (t, x) are as shown in FIG. 2C and FIG. 2D, and when the one-dimensional flow F (t, x) is calculated from these, As shown in FIG.

As shown in FIG. 2E, the value of the one-dimensional flow F (t, x) is substantially “0”.
For this reason, even if the vehicle 100 behind the host vehicle approaches 10 pixels, the approach is not detected.

On the other hand, in this embodiment, the averaging process is performed. FIG. 4 is a graph for explaining the operation of the vehicle approaching object detection device 20 according to the present embodiment. Note that FIG.
) Shows the data obtained by averaging the pixel values of the column x at time t, and FIG.
Indicates the data obtained by averaging the pixel values of the column x at time t + 1. 4C shows the spatial change rate Ia _x (t, x) obtained from the data shown in FIGS. 4A and 4B, and FIG. 4D shows the spatial change rate Ia _x (t, x). ) and 4 (temporal change rate obtained from the data shown in b) _Ia t (t, shows a x). In addition, FIG. 4 (e) is similar to FIG.
) And FIG. 4D show a one-dimensional flow Fa (t, x).

In the example shown in FIG. 4, it is assumed that the vehicle 100 behind the host vehicle approaches 10 pixels. In this case, assuming that the luminance value is high at a location where the vehicles 100 and 101 exist, as shown in FIGS. 4A and 4B, a high luminance value is shown in the existing portions of the vehicles 100 and 101. (See reference numerals 4a1, 4a2, 4b1, 4b2). However, in the example shown in FIG. 4, since the averaging process is performed, the graph is not represented as a sharp pulse as shown in FIG. 2A and FIG. It looks like a pulse.

From such data, the spatial change rate Ia _x (t, x) and the temporal change rate Ia _t (
4 (c) and FIG. 4 (d) are calculated, and when the one-dimensional flow Fa (t, x) is calculated from these, the result is as shown in FIG. 4 (e). .

As shown in FIG. 4E, the value of the one-dimensional flow Fa (t, x) indicates “−10” (see reference numerals 4e1 and 4e2). For this reason, it is possible to accurately detect that the vehicle 100 behind the host vehicle approaches 10 pixels. FIG. 5 is a diagram illustrating the result of the one-dimensional flow Fa (t, x) in an image. As shown in FIG. 5, the vehicle 100 exists in FIG. 3A, the image location 51 where the vehicle 100 no longer exists in FIG. 3B, and the vehicle 100 does not exist in FIG. In FIG. 3B, the image location 52 where the vehicle 100 exists is clear.

Here, an example in which a Gaussian filter is used as a comparison target is shown. FIG. 6 is a graph showing a state of calculation of the one-dimensional flow F (t, x) by the conventional gradient method, and shows a case where a Gaussian filter is used. 6A shows data when a Gaussian filter is used for the pixel value of the column x at time t, and FIG. 6B shows the data when the Gaussian filter is used for the pixel value of the column x at time t + 1. Shows the data when. 6C shows the spatial change rate Ig _x (t, x) obtained from the data shown in FIGS. 6A and 6B, and FIG. 6D shows the spatial change rate Ig _x (t, x). ) And the temporal change rate Ig _t (t, x) obtained from the data shown in FIG. 6B. Furthermore, FIG. 6 (e) is similar to FIG.
The one-dimensional flow Fg (t, x) calculated | required from (c) and FIG.6 (d) is shown.

In the example shown in FIG. 6, it is assumed that the vehicle 100 behind the host vehicle approaches 10 pixels. In this case, assuming that the luminance value is high at a location where the vehicles 100 and 101 exist, as shown in FIGS. 6A and 6B, a high luminance value is shown in the existing portions of the vehicles 100 and 101. (See symbols 6a1, 6a2, 6b1, 6b2). However, in the example shown in FIG. 6, since a Gaussian filter is used, the graph is not represented as a sharp pulse as shown in FIGS. 2 (a) and 2 (b). It is the target.

From such data, the spatial change rate Ia _x (t, x) and the temporal change rate Ia _t (
6 (c) and FIG. 6 (d) are calculated, and when the one-dimensional flow Fa (t, x) is calculated from these, the result is as shown in FIG. 6 (e). .

As shown in FIG. 6E, the value of the one-dimensional flow Fg (t, x) indicates “−10” or more (see reference numerals 6e1 and 6e2). For this reason, even though the vehicle 100 behind the host vehicle approaches 10 pixels, it is detected that the vehicle 100 approaches further. In particular, if such a detection is performed, there is a possibility that even an object that is not approaching may be erroneously detected as approaching. Therefore,
As in this embodiment, the averaging process can be performed more accurately than using the Gaussian filter.

Thus, according to the approaching object detection device 20 for a vehicle according to the present embodiment, for each pixel in the horizontal direction of the image input to the input unit 21, the average value of the pixel values for the adjacent predetermined pixels is calculated. An averaging process that is obtained as a pixel value of a pixel is executed, a spatial change rate and a temporal change rate are obtained, and an approaching object to the vehicle is detected. Thus, since an approaching object is detected by the gradient method without using the block matching method, an increase in the amount of calculation can be suppressed. In addition, since the averaging process is performed, elements that lead to a decrease in the system when removing the rate of change, that is, a sudden change between images, are removed, and a decrease in the detection system for approaching objects can be suppressed. it can. Therefore, even when a wide-angle camera is used, it is possible to suppress a decrease in the accuracy of detecting an approaching object while suppressing an increase in calculation amount.

Next, a second embodiment of the present invention will be described. Vehicle approaching object detection device 20 according to the second embodiment
Is the same as that of the first embodiment, but the configuration is partially different. Only differences from the first embodiment will be described below.

FIG. 7 shows an approaching object detection system 2 including the vehicle approaching object detection device 20 according to the second embodiment. As shown in FIG. 7, the approaching object detection device 20 for a vehicle according to the second embodiment newly includes a second differentiation operation unit (second differentiation means) 26. The processing contents of the preprocessing unit 22 and the approaching object detection unit 24 are also different from those of the first embodiment.

In the second embodiment, the preprocessing unit 22 performs an averaging process for obtaining an average value of pixel values for adjacent pixels as the pixel value of each pixel in the vertical direction of the image input to the input unit 21. Run. In addition, the second differential calculation unit 26 is a plurality of image data captured by the camera 10 at different times, and the vertical direction data averaged by the pre-processing unit 22 is subjected to a spatial change rate and a temporal change. Find the rate of change. The approaching object detection unit 24 includes a second differential calculation unit 2.
6 is detected based on the rate of change obtained in step 6.

FIG. 8 is a diagram illustrating an image when the camera 10 is vertically blurred. As shown in FIG. 8A, it is assumed that there is no object such as another vehicle on the road. In such a case, it is assumed that the camera 10 is shaken downward. Thereby, an image as shown in FIG. 8B is acquired. That is, when the camera 10 is shaken downward, the position of the horizon 103 is raised by y on the image. And the approaching object detection apparatus 2 for vehicles which concerns on 1st Embodiment
In the case of 0, there is a possibility that it is erroneously determined that some object has left even though there is no object on the road.

However, since the approaching object detection device 20 for a vehicle according to the second embodiment can determine the vertical blur of the camera 10, it can detect an approaching object to the vehicle in consideration of the detected blur. That is, when the camera 10 is shifted upward, the approaching object detection unit 24 performs processing so as to cancel out the moving away movement output from the differential calculation unit 23. When the camera 10 is shifted downward, the differential calculation unit 2
3 is processed to cancel the approaching movement output. Thereby, accuracy can be further improved.

Thus, according to the approaching object detection device 20 for a vehicle according to the second embodiment, as in the first embodiment, even when a wide-angle camera is used, the approach is performed while suppressing an increase in the amount of calculation. A decrease in detection accuracy of an object can be suppressed.

Further, a second differential operation unit 26 for obtaining a spatial change rate and a temporal change rate for the data in the vertical direction is provided. Based on the change rate obtained by the second differential operation unit 26, the vertical blur of the camera 10 is corrected. Detecting an object approaching the vehicle in consideration of the detected blur. For this reason, when the camera 10 is shaken upward and the image is moved downward as a whole, the possibility of erroneously detecting a white line or the like as an approaching object can be reduced.

As described above, the present invention has been described based on the embodiment, but the present invention is not limited to the above embodiment, and may be modified without departing from the gist of the present invention.

For example, the preprocessing unit 22 may increase the value of the predetermined pixel as the vehicle speed increases. As a result, in a situation where the speed of the vehicle increases and a sudden change between images is likely to occur, averaging is performed so as to suppress a more rapid change.
It is because the fall of the detection accuracy of an approaching object can be suppressed further.

The preprocessing unit 22 may increase the value of the predetermined pixel as the pixel moves away from the vanishing point of the image. Because it is possible to perform averaging that suppresses more rapid changes at locations that are more prone to sudden changes between images, away from the vanishing point, and it is possible to further suppress the decrease in detection accuracy of approaching objects. is there. The vanishing point may be obtained based on the optical flow obtained by obtaining the optical flow in the image, or may be obtained by another method.

DESCRIPTION OF SYMBOLS 1, 2 ... Approaching object detection system 10 ... Camera 20 ... Vehicle approaching object detection apparatus 21 ... Input part (input means)
22: Pre-processing unit (pre-processing means)
23 ... Differential operation section (differentiation means)
24 ... Approaching object detection unit (approaching object detection means)
25... Output unit 26... Second differential operation unit (second differential means)

Claims (3)

A vehicle approaching object detection device that detects an approaching object from the rear of the vehicle based on an image from a camera that is provided in the vehicle and captures the rear side of the vehicle,
Input means for inputting data of an image captured by the camera;
Pre-processing means for executing an averaging process for obtaining an average value of pixel values of adjacent pixels as the pixel value of each pixel in the horizontal direction of the image input to the input means;
Spatial change rate and temporal change of data of a plurality of images captured by the camera at different times, and an average value of pixel values of predetermined pixels adjacent in the horizontal direction obtained by the preprocessing unit A differentiating means for determining the rate;
An approaching object detecting means for detecting an approaching object to the vehicle from the rate of change obtained by the differentiating means ,
The vehicle approaching object detection device, wherein the preprocessing means increases the number of predetermined pixels as the pixels move away from the vanishing point of the image . The preprocessing means performs an averaging process for obtaining an average value of pixel values of adjacent pixels as a pixel value of each pixel in the vertical direction of the image input to the input means,
Spatial change rate and temporal change with respect to an average value of pixel values of predetermined pixels adjacent in the vertical direction obtained by the pre-processing means, which are data of a plurality of images taken by the camera at different times A second differentiating means for obtaining a rate;
The approaching object detecting means detects a camera shake in the vertical direction of the camera from the rate of change obtained by the second differentiating means, and detects an approaching object to the vehicle in consideration of the detected shake. The vehicle approaching object detection device according to claim 1. The approaching object detection device for a vehicle according to claim 1, wherein the preprocessing unit increases the number of predetermined pixels as the speed of the vehicle increases.

Images