JPH08166221A - Vehicle recognizing device for night - Google Patents

Abstract

PURPOSE: To recognize another vehicle by detecting light emitted by the vehicle at night without the erroneous recegnition of a street light or the like. CONSTITUTION: In a headlight switching device according to this invention, when the front image of a vehicle is inputted (Step 601), the image is binarized (Step 602), the outline is extracted (Step 603), the size of a bright part (light) is obtained (Step 604), and a pair having symmetry is extracted on the basis thereof (Step 605). In the image obtained after predetermined time, the symmetricalness of the pair is pursued, and when it can be pursued, the light is adopted as the light source of the other vehicle (Step 606) and an inter-vehicle distance is obtained (Step 607), and when the inter-vehicle distance decreases in succession, it is determined that the other vehicle is an on-coming vehicle, and a headlight is switched over to a low beam (Step 608). Accordingly, the headlight is appropriately switched over without erroneously recognizing a street light or the like for the light source of the other vehicle.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device mounted on a vehicle for detecting another vehicle at night.

[0002]

2. Description of the Related Art There has been proposed a device for automatically switching the angle of a headlight when a vehicle is driving at night and passes by an oncoming vehicle. In such a device, it is necessary to correctly detect the approach of an oncoming vehicle.
No. 5153 discloses a device in which light-receiving elements are arranged in the horizontal direction in front of a vehicle and an oncoming vehicle is detected from variations in the amount of received light detected by each light-receiving element. According to this device, when the oncoming vehicle approaches, the light-receiving element facing the headlight receives a large amount of light, but as the distance from the light-receiving element increases, the light-receiving amount decreases drastically with the irradiation direction. A large variation occurs in the amount of received light. On the other hand, when the headlight of one's own vehicle reflected by the guardrail or the like is received, the amount of received light is substantially uniform, so that the oncoming vehicle can be detected.

[0003]

However, according to the above-mentioned device, even when a street lamp or a decorative lamp is received, the amount of received light varies greatly, and the street lamp or the like is erroneously recognized as an oncoming vehicle. In view of this problem, an object of the present invention is to recognize light emitted by another vehicle at night without erroneously recognizing a street light or the like.

[0004]

The invention according to claim 1 of the present invention, which has been made to solve the above-mentioned problems, is a vehicle recognition device which is mounted on a vehicle and detects another vehicle at night. The image in front of the vehicle is divided into predetermined pixels and the brightness of each pixel is detected to shoot as two-dimensional vertical / horizontal luminance information. A light source extraction unit that generates a binarized image based on the image, an image storage unit that stores the binarized image, an image storage unit that stores the binarized image in the image storage unit, and the image storage By means of the means, the behavior of each light source is checked by comparing the two binarized images stored in the image storage unit at different times during the traveling, and each light source is dependent on the light source of another vehicle. Is a thing Characterized in that it comprises a respective determining vehicle recognition means as to whether or not.

The invention described in claim 2 is the same as claim 1.
In the nighttime vehicle recognition device according to, the measuring means for calculating the vertical length and the horizontal length in the image of each light source extracted by the light source extracting means, and the vertical direction calculated by the measuring means. An image between two light sources forming a light source pair for each of the light source pairs extracted by the light source pair generating means for extracting two light sources having substantially the same length and lateral length as a light source pair. Calculating means for calculating the above distance and / or the inclination on the image of the line segment connecting both light sources, and the vehicle recognizing means calculates the light source pair for each light source pair extracted by the light source pair generating means. Based on the distance and the inclination calculated by the above, it is attempted to make the two binarized images correspond to each other, and whether or not the light source pair is caused by the light source of another vehicle Or each Characterized in that it is intended to constant.

Further, the invention described in claim 3 is the same as claim 2
In the nighttime vehicle recognition device according to, the light source pair generation means, the extraction of the light source pair, first extracts one of the light sources, then, the inclination of the line segment connected to the light source is substantially horizontal, The distance from the light source is within a value according to the length in the vertical direction or the length in the horizontal direction of the light source measured by the measuring unit, and the light source is searched for. To do.

According to a fourth aspect of the present invention, in the nighttime vehicle recognition device according to the second or third aspect, the vehicle recognition means sets the midpoint position of both light sources forming the light source pair to the light source pair. And a rectangular area of vertical and horizontal sizes corresponding to the distance between both light sources forming the light source pair is set with reference to the position of the light source pair in one of the binarized images, and the behavior of the light source pair is investigated. It is characterized in that, when comparing the two binarized images, only the binarized images in the rectangular area are examined.

According to a fifth aspect of the invention, in the nighttime vehicle recognition device according to any one of the second to fourth aspects, the other is based on the distance between the light sources of the light source pair calculated by the calculating means. Vehicle-to-vehicle distance calculating means for calculating the distance between the vehicle and the vehicle on which the nighttime vehicle recognition device is mounted, the vehicle recognizing means for the light source pair determined to be due to the light source of another vehicle. Then, based on the change in the inter-vehicle distance calculated by the inter-vehicle distance calculating means, it is determined whether it is due to the light source of the oncoming vehicle or the light source of the preceding vehicle.

[0009]

In the nighttime vehicle recognition apparatus according to claim 1 of the present invention, the image pickup means photographs an image in front of the vehicle as vertical and horizontal two-dimensional luminance information. This brightness information is
The light source extraction unit binarizes the bright portion and the dark portion to form a binarized image, which is stored in the image storage unit by the image storage unit. Stored in this way 2
By comparing the two binarized images, the presence or absence of the bright portion (that is, the light source) and the behavior of the light source can be obtained. Based on the comparison result, the vehicle recognition means determines whether or not the light source is from another vehicle.

That is, in the nighttime vehicle recognition device according to the first aspect, since the luminance information in front of the vehicle is taken in as vertical and horizontal two-dimensional information, that is, plane information, the position of the light source on the plane is limited to the horizontal direction. Instead, it can be grasped vertically. In other words, in the recognition device in which the light receiving elements are arranged horizontally, a plurality of light sources that are identified as the same in the horizontal direction can be individually recognized by the difference in their vertical positions. .

Further, since the image storing means stores the above-mentioned luminance information which has been binarized by the light source extracting means in the image storage section, by comparing the two binarized images stored at different times. It is possible to investigate changes in the position of the light source. By the vehicle recognition means performing this investigation, it is possible to determine whether or not the light source is due to the vehicle. Examples of this determination method include the following methods.

That is, the light sources existing in front of the vehicle mainly include headlights of oncoming vehicles, tail lamps of preceding vehicles, street lights and the like. The magnitudes of these relative velocities with respect to the vehicle equipped with the nighttime vehicle recognition device are generally in the order of the headlight of the oncoming vehicle, the streetlight, and the taillight of the preceding vehicle during traveling, and therefore these magnitudes on the image are The order of severity of behavior is the same. Obey,
These three parties can be discriminated from each other by evaluating the intensity of the behavior, that is, the magnitude of the position change of the light source. Further, since the image capturing means captures the front image as the two-dimensional luminance information, not only the behavior of the light source but also the size of the light source (for example, the length in the vertical direction and the length in the horizontal direction, or the area).
Can also be evaluated. Based on this size, it is possible to identify each light source on the same screen, or to identify the corresponding light source between the two binary images.

Further, in the nighttime vehicle recognition device according to the second aspect, the measuring means calculates the vertical length and the horizontal length in the image of the light source extracted by the light source extracting means. Then, the light source pair generation means extracts two light sources having substantially the same calculated value as a light source pair, and the calculation means calculates a distance between the two light sources forming each extracted light source pair and a line connecting the two light sources. Calculate the slope of the minute.

The vehicle recognizing means associates each light source pair extracted by the light source pair generating means between the two binarized images based on the distance and the inclination calculated by the calculating means. Attempts are made to determine whether or not the light source pair is due to the light source of another vehicle depending on whether or not it is possible to cope with the situation.

In other words, in the nighttime vehicle recognition device according to the second aspect, since the light source pairs that are pairs of the light sources stored in the image storage unit are extracted as a unit, the street lamps that do not form a pair are provided. By removing the light source due to, for example, from the extraction target, the correspondence between the subsequent binarized screens can be efficiently performed. If the light source pair is from the light source of the vehicle, the distance and the inclination are substantially equal in the two binarized screens whose image capturing times are close to each other. By trying to correspond between the binarized screens of
It can be determined whether the light source pair is due to the light source of the vehicle.

When attempting this measure, if the measure is made only for the distance between the two light sources, it is possible to prevent a light source pair having an unreasonably long distance (or a short distance) from being determined to be due to the light source of the vehicle, and also to incline If it is attempted to deal with only the above, it is possible to prevent the light source pair having an unreasonably large inclination from being determined to be due to the light source of the vehicle. All light source pairs exhibiting an unsuitable value such as can be excluded from the determination target, and the reliability of determination increases.

Further, in the nighttime vehicle recognition device according to the third aspect, the light source pair generating means first extracts one light source, and then searches for the other light source with the light source as a reference, thereby pairing the light sources. To extract. Moreover, in this search, the inclination of the line segment connected to the reference light source is substantially horizontal, and the distance from the light source is the vertical length or the horizontal length of the light source measured by the measuring unit. It is performed under the condition that it is within the corresponding value. Therefore, the inclination of the line segment connecting the two light sources is too large for the light source of the vehicle, and the distance between the two light sources is too large for the light source of the vehicle compared to the size of the two light sources. Is not extracted as a light source pair.

That is, in the nighttime vehicle recognition device according to the third aspect, the light source is not searched from the entire image,
Since only the light source pairs satisfying the above conditions are extracted, inappropriate light source pairs by the light sources of the vehicle can be eliminated, and the number of extracted light source pairs can be reduced. Therefore, the correspondence of the light source pair attempted between the two binarized screens after the light source pair is extracted can be efficiently performed.

Still further, in the nighttime vehicle recognition device according to the present invention, the position of the light source pair in one of the two binarized images is tried when the correspondence of the light source pair is attempted between the two binarized images. A rectangular area of vertical and horizontal sizes corresponding to the distance between the two light sources forming the light source pair is set at the same position on the other binarized image, and the light source pair corresponding to the reference is set in the area. Search for light source pairs. The position of the light source pair is represented by the midpoint of both light source positions forming the light source pair.

That is, in the nighttime vehicle recognition device according to the fourth aspect, the light source pairs whose positions are different from each other as the positions of the light source pairs deviate from the area are excluded from the search target, and the light source pairs are the light sources of the vehicle. The determination as to whether or not there is can be made more efficient and highly reliable.

In the nighttime vehicle recognition device according to claim 5, the inter-vehicle distance calculation means determines the nighttime vehicle recognition device from the other vehicles based on the light source distance between the light source pairs calculated by the calculation means. The distance between the vehicle and the vehicle, that is, the so-called inter-vehicle distance is calculated. And for the light source pair determined by the vehicle recognition means to be due to the light source of another vehicle,
A change in the inter-vehicle distance between the two binarized screens is obtained, and it is determined whether the light source pair is an oncoming vehicle or a preceding vehicle. That is, if the inter-vehicle distance is shortened, it is determined that it is due to the light source of the oncoming vehicle, and if there is almost no change, it is due to the light source of the preceding vehicle.

That is, according to the nighttime vehicle recognition device of the fifth aspect, the vehicle-interval distance calculating means obtains the vehicle-interval distance between each light source pair determined to be due to the vehicle, and how the distance changes. It is possible to distinguish the oncoming vehicle and the preceding vehicle from.

[0023]

Embodiments of the present invention will be described with reference to the drawings. FIG. 2 is a schematic configuration diagram showing an overall configuration of a headlight angle automatic switching device 1 to which the nighttime vehicle recognition device of the present invention is applied.

The headlight angle automatic switching device 1 is
A camera 11 as an image pickup means of the present invention for taking a front image 5 of the vehicle 3, an ECU 13 for recognizing a night vehicle,
Vehicle headlight 15 and headlight 1 not shown
5 and an actuator for switching.

The internal structure of the ECU 13 is shown in FIG. EC
U13 binarizes the CPU 21, the CPU bus 23, the binarization threshold value generation circuit 25, and the video signal 27 coming from the camera 11 based on the threshold value to generate a binarized image. Comparator 29 and image bus 31
And a binarized image memory 33 as an image storage unit of the present invention for storing a binarized image, and a 3 × 3 filter arithmetic LSI 35 for extracting the contour line of the binarized image.
And a 3 × 3 filter data ROM 37 for storing the calculation data of the 3 × 3 filter calculation LSI 35, a digital memory
A RAM 39 for storing program data for a signal processor (DSP) and a digital signal processor (DSP) 41.

The headlight switching signal 43 is
It is transmitted to a headlight switching unit (not shown) via the CPU bus 23, and the traveling beam and the passing beam (see FIG. 2) of the headlight 15 are switched by this signal. The processing performed by this apparatus will be described below with reference to the flowchart of FIG. 4 and the explanatory view of FIG. Note that FIG.
4 is a visual representation of how the binarized image stored in the binarized image memory 33 is processed. It is assumed that this process is activated by a well-known timer interrupt at predetermined time intervals.

First, in step 601, an image is input. This is to input the video signal from the camera 11 in FIG. 2 to the video signal 27 in FIG. Then step 602
At, the input image is binarized. This is because the video signal 27 is input to the non-inversion side input terminal of the comparator 29 of FIG. 3, and the binarization threshold value signal 45 generated by the binarization threshold value generation circuit 25 is input to the inversion side of the comparator 29. A binarized video signal 47 is generated by inputting it to the terminal, and this signal is stored in the binarized image memory 33 as a binarized image through the image bus 31. That is, the binarized threshold value generation circuit 25 and the comparator 29 correspond to the light source extraction means of the present invention, and the portion stored in the binarized image memory 33 corresponds to the processing performed by the image storage means. This binarized image is the image 101 of FIG.

Next, in step 603, the contour line of the binarized image is extracted. This is because the 3 × 3 filter arithmetic LSI 35 shown in FIG. 3 accesses the binarized image stored in the binarized image memory 33 through the image bus 31 to access 3 × 3.
The contour line of the binarized image is extracted based on the filter data stored in the filter data ROM 37. The extracted contour image of the binarized image is displayed on the image bus 31 through the DS.
It is stored in the RAM 39 for P. This is the image 1 in FIG.
02, and the bright part of the image 101 is surrounded by the contour line. The portion surrounded by the contour line (the bright portion in the image 101) is the headlight of the oncoming vehicle, the tail lamp of the preceding vehicle, the street light, and the like, but hereinafter, these are simply called lights. .

Next, at step 604, the processing as the measuring means of the present invention for calculating the size of the light from the contour line image is performed. This is because the DSP 41 of FIG.
It is calculated by processing the contour image stored in M39. The size of the light is the length and width of the light shown in the image 103 of FIG. The vertical length and the horizontal length are stored in the DSP RAM 39.

Next, at step 605, light symmetry extraction is performed. This is calculated by the DSP 41 based on the vertical length and the horizontal length of the light stored in the DSP RAM 39 of FIG. 3, and corresponds to the processing as the light source pair generation means of the present invention. This state is shown in the image 104 of FIG.

Next, in step 606, symmetry tracking over the two binarized images is performed. This corresponds to the process of attempting the correspondence of the light source pair performed by the vehicle recognition means of the present invention. The DSP 41 calculates based on the paired data (which write and which write is a pair) stored in the DSP RAM 39 of FIG. This state is shown in the image 105 of FIG.

Next, in step 607, distance calculation is performed. This distance calculation is performed by performing backward calculation from the apparent width on the image of a symmetrical pair as in the image 106 of FIG. 5 in consideration of the magnification of the lens system used when this image was captured. In this method, an accurate distance cannot be obtained unless the true width of the headlight (or tail lamp) of the vehicle for which the pair is formed on the image is known. Approximate distance can be obtained by performing back calculation using the width of). In the case of this device that switches headlights, an accurate distance is not necessary, and this method is also practical.

Next, in step 608, the headlights are switched. If the approximate distance calculated in step 607 is equal to or less than a certain value (for example, 100 m), it is determined that the on-vehicle distance to the oncoming vehicle or the preceding vehicle is short, and the headlight is switched from the traveling beam to the passing beam, and this processing ends. To do.

Details of the symmetry extraction in step 605 will be described with reference to the flowchart of FIG. 6 and FIG. 7 visually showing the processing contents. In addition,
Here, for the sake of simplicity, the description will be limited to the process in which one reference light is set to the light 49 (see FIG. 7) and a pair of lights is searched.

First, in step 701, the larger of the horizontal length Lx (see FIG. 7) and the vertical length Ly (see FIG. 7) of each light is set to L. In the following step 702, the width of the search range is set in accordance with the value of L, which defines the lateral direction of the search range of the light paired with the light 49. Here, this width is set to a length K · L obtained by multiplying L by a proportional constant K, and is distributed to both sides of the center position of the light 49 as shown in FIG. 7. The center position of each light is the intersection of a straight line that divides the vertical length of each light into two equal parts and a straight line that equally divides the horizontal length of each light into two parts. Shall be represented.

Next, at step 703, an angle area is defined as defining the vertical direction of the search range of the light paired with the light 49. The angle area is such that the line segment connecting the pair of lights and the light 49 is substantially horizontal, and is set by setting an allowable angle vertically from the horizontal direction with the light 49 as the center. Here, as shown in FIG. 7, an angle region of 16 degrees in total is defined on the left and right sides of the light 49 by 8 degrees vertically. The search range 51 composed of two triangles is defined by the angle area thus determined and the width of the search range set in step 702.
a and 51b (see FIG. 7) are formed.

Next, at step 704, the search range 51a,
The light whose center position is located within 51b is extracted. Here, in FIG. 7, the light 53 whose center position is within the search range of the light 49 is extracted. When the lights 53 are extracted, in the next step 705, the horizontal size of each light, that is, the horizontal length obtained in step 604 is checked. If the light 49 and the light 53 are a pair of light sources of the same vehicle, the lateral length Lx of the light 49 and the lateral length Lx ′ of the light 53 (see FIG. 7) are almost equal. Is due to the light source of the vehicle. Here, the proportional constants KLrxmax and KLrxmin are set, and KLrxmax · Lx
It is checked whether ≧ Lx ′ ≧ KLrxmin · Lx is satisfied.

Next, in step 706, the size of each light in the vertical direction, that is, the vertical length is similarly checked. That is, regarding the vertical length Ly of the light 49 and the vertical length Ly ′ of the light 53 (see FIG. 7), KLrymax · Ly
It is checked whether ≧ Ly ′ ≧ KLrymin · Ly is satisfied (KLrymax and KLrymin are proportional constants).

Subsequently, at step 707, the existence prohibition area of other lights is set. If the light 49 and the light 53 are paired by the light source of the vehicle, the third light is not located between these two lights. Therefore, another light existence prohibition area 55 (see FIG. 7) is set, and when another light exists in this area, the light 49 and the light 53 are not regarded as a light source pair (step 708).

When all the above check items are satisfied, the lights 49 and 53 are extracted as candidates (hereinafter, simply referred to as a pair of lights) of the light sources of the same vehicle (step 709), and the symmetry of the symmetry is extracted. The extraction process ends. Next, the details of the symmetry tracking in step 606 will be described with reference to the flowchart of FIG. 8 and FIG. 9 visually showing the processing contents.
The symmetry tracking is to check where the pair of lights extracted in step 605 has moved in the next image 59 (see FIG. 9). The next image 59 is a binarized image obtained by performing the same processing as that in steps 601 to 603, and is different from the current image 57 (see FIG. 9) from which the symmetry is extracted. The image becomes the image after the time required for the listed processes.

First, as a quantity that characterizes the lights that form the extracted pair, the distance W on the screen between the center positions of the two lights (see FIG. 9) and the line segment that also connects the center positions are horizontal lines on the screen. An angle θ formed by and (see FIG. 9) is calculated (step 801). Further, the vertical length and the horizontal length of the two lights, which have already been measured in the above step 604, are also set as the quantities that characterize this pair.
In the following, among the lights forming the pair, the right length of the light is Lrx, the vertical length is Lry, the left length of the light is Llx, and the vertical length is Lly. Figure 9
reference). That is, the process performed in step 801 corresponds to the process performed by the calculation means of the present invention.

Next, in step 802, horizontal search range calculation is performed. This is because the light corresponding to the extracted pair of lights in the next image 59 is displayed in the next image 59.
The size of the search range in the horizontal direction is defined when the search is performed. If the pair is a light source pair, it is considered that a pair having a larger width W will move more in the next image 59. Therefore, the proportional constant is set to Kx ′, and 2 in the next image 59.
Kx '· W is distributed around the midpoint (x, y) of the two lights as the horizontal search range.

In the following step 803, step 802
In the same manner as above, the vertical search range is calculated. That is, the proportional constant is set to Ky ′, and Ky ′ · W is distributed around the middle point (x, y) on the next image 59 to be the vertical search range. Next, in step 804, the search range of the next image 59 is corrected in consideration of vertical movement and horizontal movement. When the camera 11 is mounted on a vehicle, it is considered that the position of the light source pair in the screen is shifted for each image due to the pitch and roll of the vehicle 3. Therefore, let the vertical movement of the pair in the image be Δy and the horizontal movement be Δx, and newly set Kx ′ · W + Δx as the lateral search range, Ky ′ ·
Let WΔy be the vertical search range. Above, Step 802
When the horizontal and vertical search ranges are set in step 804, a rectangular search range 61 as shown in FIG. 9 is generated.

Next, at step 805, the next screen pair in which the midpoint positions of the two lights are located in the search range 61 in the next image 59 is extracted. For example, in FIG. 9, a pair in which the midpoint is located at the coordinates (x ', y') within the search range 61 is extracted. With respect to the pair thus extracted, an amount representing a characteristic is extracted (step 806). That is, the distance W ′ between the center positions and the inclination θ ′ (see FIG. 9) are calculated in the same manner as the processing performed in step 801, and the horizontal length Lrx ′ and the vertical length of the right light are calculated. Lry ', the horizontal length Llx' of the left light, and the vertical length Lly 'are also calculated at the same time (see FIG. 9 for both).

Next, at step 807, the size of the right light is checked. If the pair having the midpoint position (x ', y') extracted in the next image 59 is the light source pair having the midpoint position (x, y) moved in the current image 57, the right light It is considered that the size of does not change rapidly. For example, in the horizontal direction, the proportional constant is KLrxma
x, KLrxmin as KLrxmax · Lrx ≧ Lrx ′ ≧ KLrxmi
It is checked whether or not n · Lrx is satisfied, and similarly in the vertical direction as well, it is checked whether or not it is within the allowable range.

Then, the size of the left light is checked at step 808. This is performed in the same way as the right light size check. In the vertical direction, the proportional constants are KLrymax and KLrymin, and KLrymax and Lry are used.
Check if ≧ Lry '≧ KLrymin · Lry is satisfied,
Similarly, in the lateral direction, it is checked whether it is within the allowable range.

Further, in step 809, the inclination of the pair is checked. The midpoint position (x ',
y ') pair is the midpoint position (x,
If the pair in y) is moved, it is considered that the slope of the pair does not change rapidly. Therefore, if the proportionality constants are Kθmax and Kθmin, Kθmax · θ ≧ θ ′ ≧ Kθ
Check if min · θ is satisfied.

A pair satisfying all the checks in steps 807 to 809 is selected as a pair by the light source of the same vehicle (step 810), and this processing for tracking the symmetry of the light is terminated. FIG. 10 shows an example in which the headlight 15 is switched in accordance with the processing described above with reference to FIGS. 4 to 9. When the oncoming vehicle is approaching, the headlight that was switched to the passing beam is switched to the traveling beam when the oncoming vehicle is no longer detected, and when the oncoming vehicle is detected again and approaches, the recognition device The distance to the vehicle where is installed is about 170
It is shown that the beam is switched to the low beam again when m is reached.

According to the above method, since information about the brightness in front of the vehicle is obtained as an image which is two-dimensional data in step 601, the size of the light showing the bright portion in the image is determined in step 604. It can be calculated. Since it is checked in step 705 and step 706 so that the sizes of the lights calculated in this way are approximately the same, street lights with different sizes of bright parts,
Decorative lights are excluded.

Further, when forming this pair, a lateral search range is determined based on the size of the reference light 49 (step 702), and the other light is searched for. With the two lights, too wide intervals are excluded from the search target, and the search can be performed efficiently.

Further, since the angle region is determined in step 703 so that the inclination of the line segment connecting the two lights is substantially horizontal, the inclination of the above line segment may be too large depending on the light source of another vehicle. Excluded (steps 705, 70)
6) The search can be performed more efficiently. Obey,
Even if the size of the bright portion formed in the image by a street light or the like and the size of the light of another vehicle happen to be the same, there is a high possibility that they will be excluded by the setting of this search range.

Moreover, since the existence prohibition area 55 is set (step 707) and the light sources having other light sources in this area are excluded (step 708), the two lights are already caused by the light sources of other vehicles at this stage. Very likely. This pair of very likely lights from another vehicle's light source is tracked in the next image 59,
To examine its behavior (step 606), step 6
After the processing up to 05, even if a bright portion due to a street light or the like still survives the above check, it can be eliminated by examining its behavior.

For example, the size of the bright portion created by the street light in the image is substantially equal to the size of the bright portion of the light of another vehicle, and the distance from the light is also appropriate as the distance of the headlight or tail lamp of the vehicle. Moreover, it is assumed that the street light is extracted as one side of the pair of lights due to the coincidence that the angle formed by the line segment when connected to the light is substantially horizontal. Also in this case, the street light in the next image 59 behaves completely differently from the light of the other vehicle, and thus is not regarded as the light of the other vehicle. The completely different behavior means that, for example, a pair of light sources of another vehicle moves in the next image 59 while keeping the inclinations of both light sources substantially horizontal, whereas a street light is different from the vehicle. Since the vehicle is moving at a speed, the inclination gradually increases. Therefore, the check in step 809 can be caught and eliminated.

The tracking of the light source pair in the next image 59 is performed by searching the midpoint position of both lights forming the light source pair. The search range 61 is determined also in this search, so that the tracking can be performed efficiently. Can be done. That is,
In steps 802 and 803, the search range 61 in the horizontal direction and the search range 61 in the vertical direction are determined according to the distance W between the two light sources forming the light source pair. Therefore, the allowable range of movement in the next image 59 is set to the size of the light source pair. It can be set according to the above, and it is possible to exclude those having an excessive movement amount from the search target.

Moreover, since the search range 61 includes the lateral movement Δx and the vertical movement Δy in consideration of the pitch and roll of the vehicle 3, the retrieval can be performed even when the vehicle 3 shakes. That is, according to the present headlight angle automatic switching device 1, the size of the light in the image in front of the vehicle 3, the distance between the two lights, the horizontality of the line segment connecting the two lights, and the two lights. From the behavior of, it is judged whether the light is due to the light of another car, so
The headlight 15 will not be excessively switched without erroneously recognizing street lights, decorative lights, and the like.

Although the headlight angle automatic switching device 1 to which the nighttime vehicle recognition device of the present invention is applied has been described above, the present invention is not limited to the above application example and can be implemented in various modes. For example, although the present invention is applied to the headlight angle automatic switching device 1 in the above-described embodiment, it may be applied to a device that notifies the approach of an oncoming vehicle. That is, when the approach of an oncoming vehicle is sensed, it is possible to configure a device that notifies the driver by sound or the like and calls attention. With such a device, night driving can be performed more safely.

Further, it is possible to configure a device that detects the distance to the preceding vehicle and informs the driver with a sound or the like to warn the driver when the vehicle approaches too much. With such a device, night driving can be performed more safely. Furthermore, the apparent width of the symmetrical pair on the screen (Llx / 2)
When + W + (Lrx / 2) is Wc, the rate of change of Wc per unit time dWc / dt is proportional to the relative speed of the pair from the shooting position of the camera. Therefore, if the proportional constant determined by the magnification of the camera system is Kw, then Kw · dWc / dt
Is the speed of the vehicle measured from the installation position of the camera. It can be used for nighttime vehicle speed measurement devices.

In the above embodiment, it was explained that the next image 59 for tracking the pair of lights is input again, but without such a method, the image input is only the process corresponding to step 601. You can do it. In this case, for example, symmetry tracking may not be performed for the first image taken, and symmetry tracking may be performed for subsequent images by comparing with the previously input image.

Further, the pair extracted in the processing of steps 807 to 809 is checked, but in addition, the distance between the light sources forming the pair may be checked. For example, this distance is W ', the proportional constant is KWmax, KW
It is checked whether min satisfies KWmax · W ≧ W ′ ≧ KWmin · W. In this way, step 807
Although the check in step 809 is satisfied, erroneous detection of a light source having a wide W'for a vehicle will not be erroneously detected.

[Brief description of drawings]

FIG. 1 is a block diagram illustrating a nighttime vehicle recognition device of the present invention.

FIG. 2 is a schematic configuration diagram showing an overall configuration of a headlight angle automatic switching device 1 that is an embodiment of the present invention.

FIG. 3 is an explanatory diagram showing an internal configuration of an ECU 13 of the embodiment.

FIG. 4 is a flowchart showing a process performed by the ECU 13 of the embodiment.

FIG. 5 is a flowchart showing a process performed by the ECU 13 according to the embodiment.
It is explanatory drawing which showed visually the mode that the binary image stored in the binary image memory 33 changes.

FIG. 6 is a flowchart showing a light symmetry extraction process in the embodiment.

FIG. 7 is an explanatory diagram of symmetry extraction processing according to the embodiment.

FIG. 8 is a flowchart showing symmetry tracking processing in the embodiment.

FIG. 9 is an explanatory diagram related to symmetry tracking processing according to the embodiment.

FIG. 10 is a graph exemplifying how an oncoming vehicle is detected and how the headlight 15 is switched by the automatic headlight angle switching device 1 of the embodiment.

[Explanation of symbols]

1 ... Headlight angle automatic switching device 3 ... Vehicle 5 ... Front image 11 ... Camera 13 ... ECU 15 ... Headlight 21 ... CPU 33 ... Binary image memory 49 ... Reference lights 51a, 51b ... Search range 53 ... Search Light 55 ... Prevention area 57 ... Current image 59 ... Next image 61 ... Search range 101-106 ... Image

Claims (5)

[Claims] 1. A vehicle recognition device mounted on a vehicle for detecting another vehicle at night, wherein an image in front of the traveling vehicle is divided into predetermined pixels,
An image pickup device for taking two-dimensional vertical and horizontal luminance information by detecting the brightness of each pixel; a light source extracting device for generating a binarized image of the luminance information taken by the image pickup device based on the brightness; An image storage unit for storing the binarized image, an image storage unit for storing the binarized image in the image storage unit, and the image storage unit at different times during traveling by the image storage unit. Vehicle recognition means for examining the behavior of each light source by comparing the two binarized images stored in, and for determining whether or not each light source is due to the light source of another vehicle. A nighttime vehicle recognition device, comprising: 2. The nighttime vehicle recognition device according to claim 1, further comprising a measuring unit for calculating a vertical length and a horizontal length in an image of each light source extracted by the light source extracting unit, A light source pair generating means for extracting two light sources having substantially the same vertical and horizontal lengths calculated by the measuring means as a light source pair; and each light source pair extracted by the light source pair generating means,
A calculation means for calculating the distance between the light sources forming the light source pair on the image and / or the inclination of the line segment connecting the both light sources on the image is further provided, and the vehicle recognition means extracts the light source pair generation means. Based on the distance and the inclination calculated by the calculating means, the two light source pairs are made to correspond to each other between the two binarized images. A vehicle recognition device for nighttime, characterized in that it determines whether or not it is due to a light source of the vehicle. 3. The nighttime vehicle recognition device according to claim 2, wherein the light source pair generating means extracts the light source pair by first extracting one of the light sources, and then connecting the light source with a line. By searching for a light source in which the inclination of the minute is substantially horizontal and the distance from the light source is within a value according to the vertical length or the horizontal length of the light source measured by the measuring unit. A night vehicle recognition device characterized by being performed. 4. The nighttime vehicle recognition device according to claim 2, wherein the vehicle recognition means sets a midpoint position of both light sources forming a light source pair as a position of the light source pair, Two binary images are set in order to set a rectangular area having vertical and horizontal sizes corresponding to the distance between both light sources forming the light source pair with reference to the position of the light source pair in the binary image, and to investigate the behavior of the light source pair. The vehicle recognition device for nighttime, characterized in that it is examined only from the binarized image in the rectangular area when comparing the above. 5. The nighttime vehicle recognition device according to claim 2, wherein the nighttime vehicle is distinguished from the other vehicle based on the distance between the light sources of the light source pair calculated by the calculation means. An inter-vehicle distance calculating means for calculating the distance to the vehicle equipped with the recognition device is further provided, and the vehicle recognizing means uses the inter-vehicle distance calculating means for the light source pair determined to be due to the light source of another vehicle. A night-time vehicle recognition device, which determines whether the light source is from a light source of an oncoming vehicle or a light source of a preceding vehicle based on the calculated change in inter-vehicle distance.