JP2004125636A - On-vehicle laser radar device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To prevent deterioration in detection accuracy of a laser radar device caused by splashes. <P>SOLUTION: This laser radar device 2 acquires a two-dimensional scan image by using a laser beam to two-dimensionally scan the backside of a vehicle or the front side thereof, calculates a pixel average value on the scan image, subtracts the average value from the scan image, and determines whether or not a two-dimensional image includes a true target object after the subtraction. Since the average value on the scan image has a value close to pixel values of target objects such as splashes scattered in a wide area while the pixel value of a target object such as a following vehicle has a value exceeding the average value, the target object buried in splashes and the like can be correctly detected by performing the subtraction process. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an in-vehicle laser radar device. More specifically, a vehicle-mounted laser used to detect a following vehicle that may collide with the own vehicle (backward monitoring of the own vehicle) or a preceding vehicle that may possibly collide with the own vehicle (monitoring the front of the own vehicle). The present invention relates to improvement of a radar device.
[0002]
[Prior art]
laser
[LASER] [Light Amplification by Stimulated Emission of Radiation] means an electromagnetic wave whose frequency is in a light region. Since the light has a single wavelength and a uniform phase (referred to as coherent light), it is possible to obtain strong light with low attenuation and which is not easily diffused. The Japanese translation is "amplification of light by stimulated emission of radiation." Also,
Radar
[RADAR] [Radio Detection And Ranging] is a distance measuring device that detects an object using an electromagnetic wave and measures its position (distance, direction, etc.). Therefore, the laser radar device particularly refers to a distance measuring device using “an electromagnetic wave whose frequency is in a light region”. Since radio waves are not used, there is an advantage in that they are not restricted by law (the Radio Law).
[0003]
Today, vehicles such as automobiles are equipped with various safety support devices. For example, there is a known rearward monitoring device that uses a laser radar device to detect a following vehicle that may collide with the own vehicle. (For example, see Patent Document 1).
[0004]
This rear monitoring device emits a laser beam toward the rear of the own vehicle, measures the actual inter-vehicle distance between the own vehicle and the following vehicle from the reflected light, and calculates the actual inter-vehicle distance and the speed of the own vehicle ( The following vehicle speed (subsequent vehicle speed) is calculated from the own vehicle speed), and the safe inter-vehicle distance between the own vehicle and the following vehicle is calculated based on the own vehicle speed and the following vehicle speed. When the vehicle is shorter than the distance, a stop lamp (brake lamp) of the own vehicle is turned on to call attention to a driver of a following vehicle.
[0005]
By the way, the traveling of the vehicle is not limited to the fine weather, and may naturally be the traveling in the rain or after the rain. In such a case, particularly when the amount of rainfall is large, as a problem peculiar to the vehicle, the detection accuracy of the laser radar device may decrease due to a water film (splash) jumped up by a tire of the own vehicle or a preceding vehicle. . Therefore, a technique that focuses on such a splash (hereinafter, referred to as “prior art”) is known (for example, see Patent Literature 2).
[0006]
[Patent Document 1]
Utility Model Registration No. 2565672
[Patent Document 2]
JP 2001-357497 A
[0007]
[Problems to be solved by the invention]
However, the prior art described above
"Because the following vehicle that may collide with the own vehicle is the one following the own vehicle lane and approaching the own vehicle, such a following vehicle should be detected and monitored by the laser radar device." After pointing out,
"Recognition of the own vehicle lane (white line on the road surface) has conventionally been performed using an optical image recognition means such as a CCD camera. The white line recognition by the dynamic image recognition means becomes impossible. "
Although the measures are disclosed,
"The vehicle speed detecting means for detecting the vehicle speed of the own vehicle, the yaw rate detecting means for detecting the yaw rate of the own vehicle, the running locus estimating means for estimating the running locus of the own vehicle from the vehicle speed and the yaw rate, and Of the following vehicles, a following vehicle detecting means for detecting a following vehicle traveling near the traveling locus is only provided as a main element,
In short, the above prior art certainly focuses on "splash", but does not solve any problem caused by the splash (a problem of lowering the detection accuracy of the laser radar device). There is a technical problem intended by the invention.
[0008]
Therefore, an ultimate object of the present invention is to solve the problem of a decrease in detection accuracy of a laser radar device caused by splash.
[0009]
[Means for Solving the Problems]
The vehicle-mounted laser radar device according to the present invention includes: an image acquisition unit configured to two-dimensionally scan a vehicle rear or a vehicle front using a laser beam to acquire a two-dimensional scan image; and calculate a pixel average value of the two-dimensional scan image. Average value calculating means, subtracting means for subtracting the pixel average value from the two-dimensional scanned image, and determining whether or not a true target is included in the two-dimensional image after the subtraction by the subtracting means. Means.
Here, the “true target” refers to a “follower vehicle” that may collide with the host vehicle when the scanning direction of the laser beam is behind the vehicle. Refers to a "preceding vehicle" where a car may collide.
According to the present invention, the average pixel value of the two-dimensional scan image has a value close to the pixel value of the target scattered over a wide range such as a splash, while the pixel value of the target such as a following vehicle has the above average value. Since the value has a value larger than the target value, by performing the above-described subtraction processing, a target object buried in splash or the like can be correctly detected.
Further, an on-vehicle laser radar device according to the present invention includes an image acquisition unit that two-dimensionally scans a vehicle rear or a vehicle front using a laser beam to acquire a two-dimensional scan image, and corresponds to a road surface from the two-dimensional scan image. Extracting means for extracting an image portion to be extracted, specifying means for specifying a road surface angle from an image corresponding to the road surface extracted by the extracting means, and calculating a pixel average value of a portion of the two-dimensional scan image at or above the road surface angle Average value calculating means, subtracting means for subtracting the pixel average value from the two-dimensional scanned image, and determining whether or not a true target is included in the two-dimensional image after the subtraction by the subtracting means. Means.
Here, the “road surface angle” refers to a scan angle (vertical scan angle) of a laser beam that irradiates a road surface that is sufficiently far away from the vehicle position. When the road surface in the monitoring direction is flat, the road surface angle becomes 0 degree or a slight negative elevation angle. Alternatively, the road surface angle becomes a positive elevation angle on a slope that increases in altitude as the road surface in the monitoring direction becomes more distant, or the road surface angle becomes negative in a slope that decreases as the road surface in the monitoring direction becomes more distant. Elevate. “Road angle” may be rephrased as “road height”. This is because the “road angle” indicates a relative relationship between the height of the own vehicle position and the height on a road surface that is sufficiently far from the own vehicle position.
According to the present invention, since the pixel average value of a portion equal to or larger than the road surface angle is adopted as the pixel average value of the two-dimensional scan image, the target object buried in splash or the like can be correctly detected without being affected by the road surface image. .
Further, in a preferred aspect of the present invention, the extracting unit is configured such that, from among the images included in the two-dimensional scan image, the width of the laser beam in the distance direction on the low elevation angle side is narrow and the elevation angle in the distance direction is increased as the elevation angle increases. An image having a tendency to increase in width is extracted as an image portion corresponding to the road surface.
Here, two laser beams emitted from the vehicle (laser beam A and laser beam B for convenience) are considered. If the elevation angles of these two laser beams A and B are φa (elevation angle of laser beam A) and φb (elevation angle of laser beam B), respectively, and if φa <φb <0 degrees, both laser beams A and B are 0 degree. Since the following low-elevation angle beam is satisfied and φa <φb, the laser beam A irradiates a position closer to the vehicle than the laser beam B. Further, these two laser beams A and B have a predetermined vertical beam width. Therefore, the detection length of the road surface at the irradiation point of the laser beam A (D1 in the description of the embodiment) is shorter than the detection length of the road surface at the irradiation point of the laser beam B (D2 in the description of the embodiment). That is, the shape of the road surface detected by the on-vehicle laser radar device does not correspond to the actual road surface shape at a position close to the own vehicle, and is detected shorter as the vehicle is closer to the own vehicle. The extracting means in the preferred embodiment extracts an image portion corresponding to a road surface from a two-dimensional scan image by applying this principle.
[0010]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, an embodiment of the present invention will be described with reference to an example of a “rear monitoring device” that detects a following vehicle that may collide with the own vehicle.
[0011]
In the following description, the specification or examples of various details and examples of numerical values, character strings, and other symbols are merely reference for clarifying the idea of the present invention, and all or some of them Obviously, the idea of the invention is not limited. In addition, well-known techniques, well-known procedures, well-known architectures, and well-known circuit configurations (hereinafter, “known matters”) will not be described in detail, but this is also for the purpose of simplifying the description, It does not intentionally exclude all or some of the matters. Such a well-known matter can be known to those skilled in the art at the time of filing the present invention, and is naturally included in the following description.
[0012]
FIG. 1 is an overall block diagram of the backward monitoring device. The rear monitoring device 1 includes a laser radar device (vehicle laser radar device) 2, a vehicle speed detecting unit 3, a yaw rate detecting unit 4, an ECU (electronic control unit) 5, a stop lamp 6, a warning lamp 7, a buzzer 8, and a hazard lamp 9. And the seat belt device 10 and the like.
[0013]
The laser radar device 2 includes a control circuit (image acquisition means) 2a that controls the overall operation of the device and a pair of optical lenses (light projecting lens 2b and light receiving lens 2c) in XY directions (X direction is the height direction of the vehicle). , Y direction is the width direction of the vehicle), a scanning circuit 2e for generating and outputting a driving signal for driving the lens driving device 2d according to a control signal from the control circuit 2a, and a control circuit. A light emitting circuit (LD) 2f for lighting and driving a light emitting element (LD: laser diode) 2f in accordance with a control signal from 2a, and a control circuit 2a that converts light received by a light receiving element (PD: photodiode) 2h into an electrical signal. And a two-dimensional memory 2j for image processing (to be described in detail later).
[0014]
FIG. 2 is a front view of the laser radar device 2 (particularly, a part that receives and emits laser light). The light projecting lens 2b and the light receiving lens 2c located in front of the light emitting element 2f and the light receiving element 2h are integrally connected to a lens driving device 2d via shafts 2k and 2m. It is driven up, down, left, and right by the device 2d (the internal linear motor). When the shaft 2k is driven right and left within a predetermined distance range α by the lens driving device 2d, the optical axes of the light projecting lens 2b and the light receiving lens 2c are scanned in the same distance range α (scan in the vehicle width direction). Has become. Similarly, when the shaft 2m is vertically driven by the lens driving device 2d in a predetermined distance range β, the optical axes of the light projecting lens 2b and the light receiving lens 2c are scanned in the same distance range β (scan in the height direction of the vehicle). It is supposed to.
[0015]
FIG. 3 is a conceptual view of the scan of the laser radar device 2. In this figure, a laser beam 11 emitted to the rear of the vehicle through a light projecting lens 2b is directed to a predetermined angle range (an angle range corresponding to the above-mentioned distance range α in the vehicle width direction (Y direction) of the vehicle); For convenience of explanation, this angle range is also assumed to be α.), And a predetermined angle range (an angle range corresponding to the distance range β described above) in the height direction (X direction) of the vehicle; For this reason, the angle range is also set to β.), So that an object behind the host vehicle (such as the following vehicle 12 or the road surface 13) can be detected by these XY scans. The cross-sectional shape of the illustrated laser beam 11 is a vertically long elliptical shape. This is to reduce the beam width in the vehicle width direction (Y direction) of the own vehicle to increase the resolution (horizontal resolution) in that direction. However, if the required resolution can be obtained, other cross sections are used. The shape may be, for example, a circle or a horizontally long ellipse.
[0016]
FIG. 4A is a conceptual scanning diagram of the laser radar device 2 viewed from the lateral direction. In this figure, of the laser beams 11 scanned in a predetermined angular range β in the height direction (X direction) of the own vehicle, the laser beams 11L_1 and 11L_2 with low elevation angles are on the road surface 13 immediately behind the own vehicle 14. , And the laser beam 11M having a medium elevation angle irradiates the following vehicle 12, and the laser beam 11H having a high elevation angle is radiated into the air.
[0017]
Here, assuming that the distances from the vehicle 14 to the irradiation points of the low elevation angle laser beams 11L_1 and 11L_2 are R1 and R2, respectively, R1 <R2 and the laser beam 11 has a predetermined vertical beam width. As shown in FIG. 4B, the detection length D1 of the road surface 13 at the distance R1 is shorter than the detection length D2 of the road surface 13 at the distance R2. In other words, the shape of the road surface 13 detected by the laser radar device 2 does not correspond to the actual shape of the road surface 13 at a position close to the vehicle 14, and is detected shorter as the vehicle is closer to the vehicle 14. This principle is used in “road surface angle detection” (step S13b in FIG. 7) described later.
[0018]
Next, the operation of the laser radar device 2 will be described.
FIG. 5 is a diagram showing a flowchart of a control program repeatedly executed in the control circuit 2a of the laser radar device 2 at every predetermined calculation cycle.
[0019]
According to this flowchart, the control circuit 2a first drives the lens driving device 2d to move the light projecting lens 2b and the light receiving lens 2c one drive step in the vehicle width direction (Y direction) (step S11). The “drive step” is defined as the maximum drive step when the scan angle (α) in the vehicle width direction (Y direction) is divided by the angle θ corresponding to the horizontal beam half width of the laser beam 11. This refers to the driving steps from “first” to “max”, which are numbers (max). For example, assuming that α = 5 degrees and θ = 0.1 degrees for convenience, the division result is “5 ÷ 0.5 = 10”, so the first to tenth drive steps are performed.
[0020]
Next, while maintaining the driving step position in the Y direction, the lens driving device 2d is driven to scan the light projecting lens 2b and the light receiving lens 2c once in the vehicle height direction (X direction) within a predetermined angle range β. While the light emitting element 2f is turned on, the signal received by the light receiving element 2h during the scan is fetched and stored in the two-dimensional memory 2j (step S12), and then the above processing (steps S11 and S12) is performed. Is repeated for a predetermined angle range α in the vehicle width direction.
[0021]
When the above processing is completed for all the driving steps within the same angle range α (first to tenth driving steps according to the above-described convenience example) (“YES” in step S13), the processing exits the loop and will be described later. Is executed (step S14), and the result of the subsequent vehicle detection is output to the ECU 5 (step S15).
[0022]
The ECU 5 measures the actual inter-vehicle distance between the own vehicle and the following vehicle based on the following vehicle detection result output from the laser radar device 2, and detects the actual inter-vehicle distance and the own vehicle detected by the vehicle speed detection unit 3. The vehicle speed (subsequent vehicle speed) of the following vehicle is calculated from the vehicle speed (self vehicle speed), and the safe inter-vehicle distance between the own vehicle and the following vehicle is calculated based on the own vehicle speed and the following vehicle speed. When the vehicle distance is shorter than the safe inter-vehicle distance, the stop lamp 6 is turned on (or blinks), the hazard lamp 9 is blinked to call attention to the driver of the following vehicle, and the warning light 7 is turned on ( Or blinks the buzzer 8 to warn the driver of the own vehicle of the approach of the following vehicle, and activates the pretension of the seat belt device 10 to prepare for an emergency (back-end collision). Take safety measures.
[0023]
Next, an example of information stored in the two-dimensional memory 2j in step S12 will be described.
FIG. 6 is a conceptual diagram of information storage in the two-dimensional memory 2j. In this figure, a two-dimensional memory 2j is a two-dimensional map having a vertical axis as an angle (scan angle of the laser beam 11 in the X direction) and a horizontal axis as a distance (a distance from the vehicle 14 to a reflection point of the laser beam 11). It has a storage space. The two-dimensional memory 2j stores information corresponding to the intensity of the reflected light of the laser beam 11 reflected at an arbitrary angle and at an arbitrary distance at the intersection of those angles and distances. This corresponds to an image memory for storing a two-dimensional image of a distance.
[0024]
Now, in the illustrated two-dimensional memory 2j, images of the following three objects are stored. The first is an image of the road surface 13 (hereinafter, referred to as “road surface image 13a”). As described above, the road surface image 13a has a characteristic that the road surface width D1 is small near the host vehicle 14 and the road surface width D2 is widened as the distance increases. In other words, the road surface width gradually increases as the distance from the host vehicle 14 increases, and the road surface has a characteristic that the laser beam 11 is not visible at a certain distance (when the scan angle of the laser beam 11 in the X direction is substantially horizontal) or more. Therefore, the angle at which the road surface image 13a is interrupted and becomes invisible (or an angle obtained by adding a slight margin angle to the angle) can be regarded as the “road surface angle”.
[0025]
The second is an image of the following vehicle 12 (hereinafter, referred to as “following vehicle image 12a”). The difference between the following vehicle image 12a and the road surface image 13a is that the following vehicle image 12a is located at an angle higher than the "road surface angle", and the following vehicle image 12a is observed as one lump. (It is not an image having a wide distribution that is continuous over a certain distance from the vicinity of the vehicle 14 as in the road surface image 13a.) The following vehicle image 12a corresponds to the “true target” described in the gist of the invention.
[0026]
As described above, since the first image (road image 13a) and the second image (following vehicle image 12a) have the above-mentioned clear difference, the two are distinguished by using this difference, and The succeeding vehicle 12 that may possibly collide with the vehicle 14 can be correctly detected.
[0027]
An object of the present invention is to solve a problem of a decrease in detection accuracy of the laser radar device 2 caused by a splash.
According to the present inventors, the splash image has different characteristics from the above-described first image (road surface image 13a) and second image (following vehicle image 12a). Specifically, FIG. As shown, the image due to the splash (hereinafter referred to as “splash image 16”) has a wide angle and is distributed over a wide range from a short distance to a medium-to-long distance. It shows features distinctly different from the second image (the following vehicle image 12a), and furthermore, the pixel value of the image (the splash image 16) averages the first image (the road surface image 13a) and the second image Paying attention to the fact that the pixel value is smaller than the pixel value of the (following vehicle image 12a), and detecting the second image (following vehicle image 12a) buried in the splash image 16 using these characteristics. It was.
[0028]
Incidentally, the reason that “the pixel value of the splash image 16 is on average smaller than the pixel value of the road surface image 13a and the following vehicle image 12a” is that the intensity of the reflected laser light from the road surface 13 and the following vehicle 12 is sufficient. This is because the intensity of the reflected laser light scattered on the surface of the water droplet (splash is a collection of water droplets) is considerably weaker than that of the laser beam.
[0029]
FIG. 7 is a flowchart showing a subroutine of the following vehicle detection process (step S14 in FIG. 5).
In this flowchart, first, the vehicle speed of the host vehicle 14 is fetched from the vehicle speed detection unit 3, and the acceleration / deceleration of the host vehicle 14 is calculated from the degree of change (for example, a differential value) of the vehicle speed. Then, it is determined whether the acceleration / deceleration is within a range between a predetermined deceleration (for example, about -0.1 G) and a predetermined acceleration (for example, about +0.1 G), and whether the acceleration / deceleration is other than 0. A determination is made (step S13a). In short, these acceleration / deceleration determination conditions are to determine whether or not the host vehicle 14 is in a non-stop state (acceleration / deceleration ≠ 0) and is running at a constant speed without sudden acceleration or sudden deceleration. become.
[0030]
Then, when the acceleration / deceleration determination condition is satisfied, the angle θ at which the road surface becomes invisible is obtained (step S13b; extracting means), and the angle θ is learned (for example, the angle θ from the ignition ON to the present time). An average value is obtained, and this is set as a road surface angle.), And the learning result is set to “road surface angle” (see FIG. 6) (step S13c; specifying means), and the subsequent processing (steps S13d to S13f; determination) On the other hand, if the acceleration / deceleration determination condition is not satisfied, the process passes steps S13b and S13c, and proceeds to the subsequent processes (steps S13d to S13f).
[0031]
Here, among the above-mentioned acceleration / deceleration determination conditions, a condition for determining whether or not the vehicle is traveling at a constant speed is that, when rapid acceleration or rapid deceleration is performed, the rear portion of the vehicle body sinks (in the case of rapid acceleration) or rises (suddenly). For example, when the vehicle is subject to a deceleration of about -0.1 G or less or an acceleration of about +0.1 G or more, depending on the hardness of the suspension of the vehicle, This is because the optical axis of the laser beam 11 is inclined by a considerable amount (1 degree or more) in the height direction (X direction) of the vehicle body with the lifting, and the road surface angle is learned by the laser beam 11 having such an inclination. Even so, the road surface angle cannot obtain sufficient reliability.
[0032]
For this reason, in the present embodiment, in order to learn the road surface angle with sufficient reliability, the above-mentioned "condition for determining whether or not the vehicle is traveling at a constant speed" is added. The other determination condition (non-stop state; acceleration / deceleration ≠ 0) is that the road surface angle is learned only during traveling without sudden acceleration / deceleration. The reason is that the vehicle is not in the non-stop state. At that time (that is, when the vehicle 14 is in a stopped state), attention should be paid to the danger of rear-end collision, and maximum attention should be paid to approaching the following vehicle 12 rather than learning the road surface angle. It is.
[0033]
In step S13a, the acceleration / deceleration of the host vehicle 14 is calculated from the vehicle speed detected by the vehicle speed detection unit 3. For example, acceleration / deceleration detection means such as an acceleration sensor is provided, and the output signal is output. Of course, you can use it.
[0034]
When the learning of the road surface angle is completed, a process of detecting the following vehicle 12 that may collide with the own vehicle 14 from the stored information of the two-dimensional memory 2j (steps S13d and S13e; average value calculating means, subtracting means) ).
[0035]
FIGS. 8A to 8C are conceptual diagrams of the above-described process, that is, the process of detecting, from the information stored in the two-dimensional memory 2j, the following vehicle 12 that may collide with the host vehicle 14. In FIG. 8A, A and B are two representative image extraction target distances. A dashed line A 'intersecting the distance A and a dashed line B' intersecting the distance B are extraction lines of the extracted image corresponding to the distances (A, B). Hereinafter, for convenience, the extracted image extracted along the dashed line A 'is referred to as an A-th extracted image 17, and the extracted image extracted along the dashed line B' is referred to as a B-th extracted image 18. The A-th extracted image 17 is formed only of the splash image 16, and the B-th extracted image 18 includes the following vehicle image 12 buried in the splash image 16.
[0036]
As described above, the pixel value of the subsequent vehicle image 12a has a larger value than the pixel value of the splash image 16. This is because the splash image 16 is formed by weakly reflected laser light scattered by water droplets.
[0037]
As shown in FIG. 8B, the distribution of pixel values of the A-th extraction image 17 is substantially constant regardless of the angle, but as shown in FIG. Since the distribution of the pixel values protrudes in the existing angle range of the following vehicle image 12a, the average values Aave and Bave of each of these two extracted images (the A-th extracted image 17 and the B-th extracted image 18) are obtained. By subtracting the average value Aave of the A-th extracted image 17 from the pixel value of the A-th extracted image 17 and subtracting the average value Bave of the B-th extracted image 18 from the pixel value of the B-th extracted image 18, FIG. The results shown in the right diagrams of (b) and (c) can be obtained.
[0038]
In other words, the A-th extracted image 17 'after the subtraction has only useless information about the noise level, whereas the B-th extracted image 18' after the subtraction has useful information corresponding to the following vehicle image 12a. .
[0039]
Therefore, by rewriting the storage information (the storage information at the road surface angle or more) of the two-dimensional memory 2j with the above subtraction results (the A-th extracted image 17 'and the B-th extracted image 18'), unnecessary image information (splash image 16), a high-quality image of only the following vehicle image 12a, such as an image behind the own vehicle in fine weather, can be stored in the two-dimensional memory 2j.
[0040]
Lastly, by extracting a set of pixel values having a specific “lump” that is equal to or larger than “road surface angle” from the information stored in the two-dimensional memory 2j, the set portion is detected as the following vehicle 12. (Step S13f), and the detection result can be output to the ECU 5 (Step S15 in FIG. 5). The “lump” is a set portion of pixel values having a size larger than the projected area of the vehicle, and specifically has a size of 1 m or more.
[0041]
In order to obtain information as shown in the left diagram of FIGS. 8B and 8C, the dynamic range of the light receiving circuit 2i needs to be as wide as possible. If the dynamic range is narrow, the A-th extraction image 17 and the B-th extraction image 18 become saturated, and the subsequent vehicle image 12a cannot be taken out even if the average value is subtracted. For example, a logarithmic amplifier has a high dynamic range and is therefore preferably used for the light receiving circuit 2i. Further, an amplifier having a high dynamic range may be used for the light receiving circuit 2i, and the amplification gain of the amplifier may be made variable. In this case, when the gain is adjusted in accordance with the average values Aave and Bave, specifically, when the gain is adjusted so as to decrease the amplification gain as the average values Aave and Bave are larger, a substantial dynamic range is obtained. This is particularly preferable in terms of measures against splash during a heavy rain, since the effect of expanding can be obtained.
[0042]
As described above, according to the present embodiment, not only during the fine weather, but also as shown in FIG. 9, the vehicle 14 is affected by the splash (or fog or snow) jumped up in the rainy weather or the like. Without the prior art, it is possible to detect a following vehicle 12 that may collide with the host vehicle 14 without the need for a special effect.
[0043]
Further, in the present embodiment, since the “road angle” (see FIG. 6) can be accurately grasped, for example, a sports type vehicle having a low vehicle height as shown in FIG. 10 (most of these vehicles have an air resistance). The hood and windshield are greatly inclined to reduce the amount of light, and the reflected light of the laser beam is considerably weakened.) Can be detected.
[0044]
Alternatively, in a vehicle having a high vehicle height (a large truck or the like) as shown in FIG. 11, only a weak reflected light can be obtained because a low elevation angle portion of the laser beam passes under the vehicle body. The vehicle can be detected from the “lumps” of pixel values existing at or above the “road angle”.
[0045]
Note that the technical concept of the present invention is not limited to the above-described embodiment, and various modifications may be made.
In the above embodiment, the influence of the splash is eliminated by “calculating the average value of the received light amount at an angle equal to or larger than the road surface angle and subtracting the average value from the received light amount” (step S13e). Not limited. For example, since the height of a vehicle that can travel on a road is restricted to “3.8 m or less” (road law) except when permitted, as shown in FIG. Normally, the height of the included following vehicle image 12a does not exceed the angle G corresponding to the above-described limit value. Paying attention to this point, the range from the angle G to the road surface angle is set as the following vehicle existing region (hereinafter, referred to as “vehicle existing region”), and the range larger than the angle G is set as the splash determination region. It may be. The illustrated splash area includes the “signboard image 19”, which will be described later.
[0046]
FIG. 13 is a modification of the above-mentioned flowchart (FIG. 7), and corresponds to the storage information of the two-dimensional memory 2j shown in FIG. The difference from the above flowchart lies in the processing performed for each distance (step S13e '). That is, in this modified example, the true target object (the following vehicle 12) is obtained by calculating the average value of the received light amount in the angle region (splash determination region) equal to or larger than the vehicle existing region and subtracting the average value from the received light amount. I try to detect. According to this, since the average value of the received light amount in the splash area does not include the following vehicle image 12a, the detection accuracy of the true target (the following vehicle 12) can be improved.
[0047]
By the way, in the angle area larger than the vehicle existence area, not only the splash, but also the image of the signboard such as a bridge crossing the road or a road sign installed at a high position on the road (hereinafter collectively referred to as these). "Signboard image 19"). In such a case, the sign 19 such as a signboard is included in the average value of the amount of received light, which may reduce the detection accuracy of the true target (the following vehicle 12). It is a non-moving object) and moves away backward at the same speed as the host vehicle 14. Therefore, the presence or absence of the signboard image 19 is determined by performing speed discrimination (step S21). If the signboard image 19 is not detected, An angle range that is equal to or greater than the vehicle presence area is set as a splash determination area (step S22). On the other hand, when the signage image 19 is detected, an angle range that is equal to or greater than the vehicle presence area and other than the signage image 19 is set as a splash determination area ( In step S22), by using the average value of these splash areas in step S13e, the influence of the signboard such as a road sign is eliminated, and the detection accuracy of the true object (the following vehicle 12) is reduced. It can be avoided.
[0048]
In the above-described embodiment, a vehicle located behind the own vehicle is assumed as a vehicle (follower vehicle 12) that may collide with the own vehicle 14. However, for example, as shown in FIG. Considering the case where the vehicle 20 and the vehicle 21 are approaching from behind when the vehicle 14 stops after passing through the curve, the vehicle 20 is traveling on the adjacent lane 22 and the vehicle 21 , The vehicle that should pay attention to the own vehicle 14 (the following vehicle that may collide with the own vehicle 14) is not along the vehicle 20 located behind the own vehicle 14 but along the own lane 21. The vehicle 21 is approaching. In order to detect such a vehicle 21 as a dangerous following vehicle, the shape (curvature or the like) of the own lane 23 is calculated from the yaw rate value before the stop detected by the yaw rate detection unit 4, or the stop of the own vehicle 14 is performed. The previous traveling locus may be estimated, and the vehicle 21 approaching along the lane shape or the traveling locus may be regarded as a dangerous following vehicle.
[0049]
In the above embodiment, the laser beam 11 is scanned in the X direction and the Y direction. For example, as shown in FIG. 15, the laser beam 11 having a horizontally long elliptical cross section is one-dimensionally drawn in the vertical direction (X direction). You may scan. The scanning mechanism and the scanning operation in the Y direction can be omitted.
[0050]
Further, in the above-described embodiment, the rear monitoring device has been described as an example. However, the present invention is not limited to this, and may be applied to a device that performs front monitoring of the vehicle 14. That is, as shown in FIG. 16, the laser radar device 2 may be mounted in front of the own vehicle 14 to detect a road surface or a preceding vehicle 24 located in front of the own vehicle 14. Even in this case, similarly to the above-described embodiment, it is possible to eliminate the influence of splash, rain, fog, and the like that the preceding vehicle 24 jumps up, and detect the preceding vehicle 24 with stable accuracy. When the distance to the preceding vehicle 24 (inter-vehicle distance) falls below a safe distance according to the speed of the own vehicle 14, an alarm is issued to the driver of the own vehicle 14 or braking is automatically applied. Can be.
[0051]
Also, in the above embodiment, the laser beam 11 is scanned in the X direction and is moved by one drive step in the Y direction. Conversely, the laser beam 11 is scanned in the Y direction and is moved by one drive step in the X direction. You may move each. In this case, a three-dimensional memory is prepared, the received light amounts in the Y direction, the X direction (angular direction) and the distance direction are stored. After the movement in the X direction is completed, the three-dimensional memory is scanned in the Y direction. 7 and FIG. 13 may be performed.
[0052]
Further, in the above-described embodiment, as an algorithm for eliminating the splash, a method of calculating the average value of the received light amount for each distance is exemplified, but the present invention is not limited to this. Furthermore, advanced image processing techniques and pattern recognition techniques may be used.For example, as a two-dimensional image of distance and angle, a detected object is determined by a technique such as edge detection, or a road surface has a similar shape. This may be used to determine the road surface angle.
[0053]
Further, in the above embodiment, the road surface angle is set, and all the angles smaller than the road surface are set as the road surface. Alternatively, the processing may be performed at a road surface angle or more at that distance.
[0054]
Further, in the above embodiment, when the acceleration is not in the range of -0.1G to 0.1G in step S13a of FIG. 7, the road surface angle is not calculated. For example, as shown in FIG. Alternatively, it may be determined whether or not the door has been opened and closed (step S31), and when the door has been opened and closed, the learning result of the road surface angle θ may be reset (step S32). Further, the road surface angle may be learned for each acceleration. This is because a situation where the own vehicle is actually dangerous is a situation where the sudden braking must be applied, and it is necessary to recognize a following vehicle at that time even when the own vehicle is tilted forward and backward. Therefore, a road surface angle corresponding to the acceleration of the own vehicle is required.
[0055]
In FIG. 17, first, it is determined whether or not the door is opened and closed in a step S31. This is because the angle at which the road surface can be seen may change due to the inclination of the vehicle body depending on the number of passengers, the amount of luggage, and the like. Therefore, it is determined whether or not the boarding door or the rear door has been opened and closed. If the door has been opened and closed, the learning result of the road surface angle θ is reset in step S32. On the other hand, if the condition in step S31 is not satisfied, an angle θ at which the road surface becomes invisible is obtained in step S13b. In step S13c, the learning is performed for each acceleration of the own vehicle, and the learning result is set as the road surface angle. Regarding the road surface angle, since the vehicle body is tilted during acceleration and deceleration, a table is provided for each acceleration of the vehicle and learning of the road surface angle is performed for each acceleration. As an example of the learning method, an average value of the road surface angle θ is obtained for each acceleration in a table in which the acceleration is divided every 0.01 G. Subsequent processing is the same as in FIG.
[0056]
Although FIG. 14 shows an example of the traveling locus using the yaw rate, it is highly probable that the own vehicle 14 is stopped or traveling at a rapid deceleration when a subsequent vehicle is assumed to collide. This means that if there is traffic congestion immediately after a sharp curve, the risk of collision with the following vehicle is the highest. However, the yaw rate value has very poor accuracy when the speed of the vehicle is low, and cannot be used at all when the vehicle stops. To compensate for this, as shown in FIG. 18, for example, a two-dimensional memory having 50 × 50 storage cells is provided. The trajectory of the host vehicle is plotted in the two-dimensional memory from the yaw rate and the host vehicle speed, the lane shape is calculated from the host vehicle trajectory, and the following vehicle is estimated. Even when the vehicle is traveling at a considerably low speed, the danger of a collision can be accurately determined.
[0057]
【The invention's effect】
According to the present invention, the pixel average value of the two-dimensional scan image has a value close to the pixel value of the target scattered over a wide range such as a splash, while the pixel value of the target such as a following vehicle has the average value. Since the value has a value larger than the value, the target object buried in splash or the like can be correctly detected by performing the above-described subtraction processing.
Further, according to the present invention, since the pixel average value of a portion equal to or larger than the road surface angle is adopted as the pixel average value of the two-dimensional scan image, the target object buried in the splash or the like without being affected by the road surface image. Can be detected correctly.
According to a preferred aspect of the present invention, "the road surface shape detected by the on-vehicle laser radar device does not correspond to the actual road surface shape at a position close to the own vehicle, and is detected as being shorter as the vehicle is closer to the own vehicle. By applying the principle of `` being done '', the image part corresponding to the road surface is extracted from the two-dimensional scan image, so that the detection accuracy of the road surface can be improved, the road surface angle is correctly recognized, and regardless of the type of vehicle It is possible to accurately detect a following vehicle or the like existing on the road surface.
[Brief description of the drawings]
FIG. 1 is an overall block diagram of a rear monitoring device.
FIG. 2 is a front view of the laser radar device 2 (particularly, a portion that receives and emits laser light).
FIG. 3 is a conceptual diagram of scanning of the laser radar device 2.
FIG. 4 is a conceptual view of scanning of the laser radar device 2 as viewed from a lateral direction.
FIG. 5 is a diagram showing a flowchart of a control program repeatedly executed in a control circuit 2a of the laser radar device 2 at a predetermined calculation cycle.
FIG. 6 is a conceptual diagram of information storage in a two-dimensional memory 2j.
FIG. 7 is a flowchart showing a subroutine of a following vehicle detection process (step S14 in FIG. 5).
FIG. 8 is a conceptual diagram of a process of detecting a following vehicle 12 that may collide with the host vehicle 14 from information stored in a two-dimensional memory 2j.
FIG. 9 is a view for explaining the effect of the present invention (elimination of the influence of splash).
FIG. 10 is a diagram (sports type vehicle detection) illustrating the effect of the present invention.
FIG. 11 is a diagram (large truck detection) illustrating the effect of the present invention.
FIG. 12 is a diagram illustrating another example of information stored in the two-dimensional memory 2j.
FIG. 13 is a diagram showing a modification of the above-described flowchart (FIG. 7).
14 is an explanatory diagram for detecting a vehicle 21 that poses a danger to the host vehicle 14 in consideration of the shape of the traveling lane of the host vehicle 14. FIG.
FIG. 15 is a conceptual diagram of one-dimensional scanning of a laser beam 11 having a horizontally long cross section in a vertical direction (X direction).
FIG. 16 is an explanatory diagram of mounting the laser radar device 2 in front of the host vehicle 14 and detecting a road surface and a preceding vehicle 24 located in front of the host vehicle 14;
FIG. 17 is a diagram showing a modification of the flowchart in FIG. 7;
FIG. 18 is a conceptual diagram of a two-dimensional memory for plotting a traveling locus of the own vehicle.
[Explanation of symbols]
S13b Step (extraction means)
S13c Step (identifying means)
S13e step (mean value calculating means, subtracting means)
S13f Step (determination means)
2 Laser radar device (vehicle laser radar device)
2a control circuit (image acquisition means)
11 Laser beam
12a Subsequent car image (true target)

Claims (3)

Image acquisition means for acquiring a two-dimensional scan image by two-dimensionally scanning the rear of the vehicle or the front of the vehicle using a laser beam,
Average value calculating means for calculating a pixel average value of the two-dimensional scan image,
Subtraction means for subtracting the pixel average value from the two-dimensional scan image,
A determining unit for determining whether or not a true target is included in the two-dimensional image after the subtraction by the subtracting unit; Image acquisition means for acquiring a two-dimensional scan image by two-dimensionally scanning the rear of the vehicle or the front of the vehicle using a laser beam,
Extracting means for extracting an image portion corresponding to a road surface from the two-dimensional scan image,
Specifying means for specifying a road surface angle from an image corresponding to the road surface extracted by the extracting means,
Average value calculation means for calculating a pixel average value of a portion of the two-dimensional scan image of the road surface angle or more,
Subtraction means for subtracting the pixel average value from the two-dimensional scan image,
A determining unit for determining whether or not a true target is included in the two-dimensional image after the subtraction by the subtracting unit; The extracting means may select, from among the images included in the two-dimensional scan image, an image in which the width in the distance direction is small on the low elevation angle side of the laser beam and the width in the distance direction is increased as the elevation angle is increased. 3. The on-vehicle laser radar device according to claim 2, wherein the laser beam is extracted as an image portion corresponding to the road surface.

Images