JP2013161434A - Periphery monitoring device for vehicle - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a periphery monitoring device for a vehicle, capable of removing a stereoscopic object that cannot be an obstacle.SOLUTION: Cameras 11, 12 photograph the front of a vehicle with a stereo image from different viewpoints. A target detection unit 21 detects a target included in a pair of images photographed by the cameras 11 and 12. A target removal unit 22 removes, out of the detected targets, a target separated from other targets through a space with a reference width or more and having a reference height or more, from a detection result. A stereoscopic object determination unit 23 determines a stereoscopic object captured in the image, on the basis of the target in which unnecessary things are removed by the target removal unit 22.

Description

  The present invention relates to a vehicle periphery monitoring device, and more particularly to a device that is mounted on a vehicle and detects a three-dimensional object by monitoring the surroundings of the vehicle.

  In recent years, in order to realize vehicle driving support such as collision avoidance with a three-dimensional object and automatic traveling, various apparatuses for detecting a three-dimensional object that obstructs traveling by monitoring the periphery of the vehicle have been developed. In this three-dimensional object detection device, for example, various targets existing around the vehicle are detected, and according to a predetermined method, it is determined whether or not the detected target is a three-dimensional object that becomes an obstacle. Do. Such a technique is proposed in Patent Document 1 and the like.

  In the technique described in Patent Document 1, a three-dimensional real space is divided into a plurality of partitioned spaces extending up and down, and a detection point of a target detected around the vehicle is one of the plurality of partitioned spaces. Project to. And in the technique described in this patent document 1, the histogram regarding the distance of a detection point is produced for every division space, Based on the result of this histogram, the identification of the solid object in each division space, and from a vehicle to a solid object The distance is calculated.

JP 2009-176091 A

  However, since the technique described in Patent Document 1 uses a partitioned space that extends vertically, a target (for example, a pedestrian or a vehicle) located near the road surface and the height of the vehicle. If a target at a higher position (for example, a signboard or a tree) exists at the same distance and direction from the vehicle, the detection points detected for each target are the same partition space. Will be projected.

  For this reason, depending on the result of the histogram, a target at a position higher than the height of the vehicle that cannot become an obstacle to vehicle travel is grouped with a target that becomes an obstacle at a position close to the road surface. There exists a problem that a possibility that it may be misdetected as a three-dimensional object which can become a collective obstacle arises. This erroneous detection of a three-dimensional object causes a decrease in accuracy in driving support control such as collision avoidance.

  Note that the above Patent Document 1 also discloses a technique for preventing erroneous detection by not reflecting a detection point whose simple height from the road surface is a predetermined value or more in the histogram. However, in order to accurately determine the height from the road surface, estimation of the road surface is essential, and there is a problem that the processing load of the apparatus increases. In addition, if an error occurs in the estimation of the road surface, detection of a three-dimensional object will be affected. In addition, it is necessary to accurately consider the slope of the road surface while traveling uphill or downhill, and it is very difficult to determine whether a solid object is subject to processing based on the height from the road surface. It can be said that it is difficult.

  Therefore, an object of the present invention is to avoid the identification of a target that cannot be an obstacle and a target that can be an obstacle as one solid object, and to improve accuracy in driving support control such as collision avoidance. It is an object of the present invention to provide a vehicle periphery monitoring device capable of preventing a decrease.

  The present invention is directed to an apparatus for monitoring the periphery of a vehicle. In order to achieve the above object, a vehicle periphery monitoring apparatus according to the present invention includes a target information acquisition unit that acquires information about a target around the vehicle, and a target acquired by the target information acquisition unit. Among them, a plurality of targets whose distance from the vehicle is within a predetermined range are extracted, and a space having a predetermined width or more in the height direction from the first target (for example, a target on the road surface) is extracted. A target that excludes the second target that is separated from the detection result, and a target that is output from the target exclusion unit, and identifies a three-dimensional object that exists in the vicinity of the vehicle And a three-dimensional object specifying unit.

  Typically, the target detection unit detects the characteristic locations of the target as a plurality of detection points. The target exclusion unit is configured such that the distribution area of the plurality of detection points of the first target and the distribution area of the plurality of detection points of the second target detected by the target detection unit are high. It is determined whether or not they are separated via a space having a predetermined width or more in the vertical direction.

  Preferably, the target excluding unit excludes a second target having a predetermined reference height or more from the detection result. The predetermined reference height may be determined based on the height distribution state of detection points of a plurality of targets at a distance within a predetermined range from the vehicle.

  Also, typically, the three-dimensional object specifying unit creates a histogram based on the detected target for each of a plurality of partitioned spaces extending in the vertical direction formed by dividing the three-dimensional real space. And identifying a three-dimensional object based on the histogram

  According to the vehicle periphery monitoring device described above, it is possible to avoid specifying a target that cannot be an obstacle and a target that can be an obstacle as a single three-dimensional object. Driving support control such as high collision avoidance can be performed.

The figure which shows schematic structure of the periphery monitoring apparatus 1 of the vehicle which concerns on one Embodiment of this invention. A diagram showing a scene in the surrounding environment where a traveling vehicle encounters The figure which shows an example of the detection point of the target detected in one scene shown to FIG. 2A The figure which shows an example of the solid object which the solid object determination part 23 specified as a result of a grouping process The flowchart which shows the procedure of the process which excludes the target which is not the target which the target exclusion part 22 performs. The figure explaining the judgment method of the target exclusion which the target exclusion part 22 performs The figure explaining the judgment method of the target exclusion which the target exclusion part 22 performs The figure for demonstrating an example of the two-dimensional plane area | region which the solid-object determination part 23 defines The figure for demonstrating the case where the detection point of the target detected by FIG. 2B is projected on the two-dimensional plane area | region shown to FIG. 6A. The figure for demonstrating the grouping process with respect to the detection point projected on the two-dimensional plane area | region shown to FIG. 6B The figure for demonstrating the grouping process with respect to the detection point projected on the two-dimensional plane area | region after the target was excluded by the target exclusion part 22. FIG. The figure which shows an example of the solid object which the solid object determination part 23 specified as a result of the grouping process of FIG. 6C. The figure which shows an example of the solid object which the solid object determination part 23 specified as a result of the grouping process of FIG. 6D.

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a diagram showing a schematic configuration of a vehicle periphery monitoring device 1 according to an embodiment of the present invention. In FIG. 1, the vehicle periphery monitoring device 1 according to the present embodiment includes a target information acquisition unit 10 and a three-dimensional object identification unit 20.

First, each configuration included in the vehicle periphery monitoring device 1 of the present embodiment will be described.
The target information acquisition unit 10 is used to acquire information related to a target such as an obstacle existing around the vehicle. The target information acquisition unit 10 is typically realized by a configuration including a stereo camera device capable of acquiring images obtained by capturing the same target from two different viewpoints, or radar waves such as electromagnetic waves. It can be realized by a configuration including a radar device that can transmit toward a target and receive a reflected wave, or can be realized by a configuration including a stereo camera device and a radar device.
In the following embodiment, the target information acquisition unit 10 having a configuration including a stereo camera device will be described as an example.

  The target information acquisition unit 10 includes two cameras 11 and a camera 12. The cameras 11 and 12 are image sensors such as CMOS (Complementary Metal Oxide Semiconductor) or CCD (Charge Coupled Device). Preferably, the camera 11 and the camera 12 have the same angle of view and the number of pixels, and have a predetermined interval at the same height position from the road surface in a state in which a front visual field region can be imaged inside the vehicle windshield. It is attached. Then, the camera 11 and the camera 12 capture the front of the vehicle in synchronization with each other at a predetermined period, and output the captured pair of images to the three-dimensional object specifying unit 20.

  In the present embodiment in which the target information acquisition unit 10 is a stereo camera device, the three-dimensional object identification unit 20 corresponds to an image processing device. In this case, the three-dimensional object specifying unit 20 includes a target detection unit 21, a target exclusion unit 22, and a three-dimensional object determination unit 23.

The target detection unit 21 inputs a pair of images taken from different viewpoints taken by the camera 11 and the camera 12 of the target information acquisition unit 10, and detects a target included in the pair of images.
Typically, the target detection unit 21 uses a so-called stereo technique, performs a stereo matching process on the image captured by the camera 11 and the image captured by the camera 12, and the like. Find the correlation of images. And the target detection part 21 is based on the principle of the triangulation using the parallax with respect to the same target, The target position in a three-dimensional real space, ie, the distance from a vehicle (its reference point) to a target , And the direction from the vehicle (reference point) to the target.

The process performed by the target detection unit 21 will be described using a specific example.
For example, consider a scene where a vehicle is traveling on a road as shown in FIG. 2A. In this scene, the target detection unit 21 detects a target detection point (see FIG. 2B) as a target included in the image area 55 captured from the traveling vehicle as shown in FIG. 2B with respect to the pedestrian 51 and the street tree 52. Multiple black circles) will be detected. The location detected as the detection point is a characteristic location of the target such as an edge portion or a portion where a contrast difference is generated. And the target detection part 21 extracts the positional information regarding the distance from a vehicle and the direction from a vehicle regarding this detected several detection point, respectively.

  In the technique executed by the three-dimensional object determination unit 23 to be described later, a target that cannot be an obstacle (in this example, the roadside tree 52 in the same distance from the vehicle shown in FIG. 2B (a distance within a predetermined range from the vehicle)). ) And a plurality of detection points of a target that can become an obstacle (in this example, pedestrian 51) are detected simultaneously, as shown in FIG. It is recognized as a three-dimensional object 62a (grouping process). Such recognition processing significantly reduces the accuracy of driving support control such as collision avoidance.

Therefore, the present invention newly includes a configuration of the target excluding unit 22 and analyzes the target (the pedestrian 51 and the roadside tree 52 in the example of FIG. 2A) detected by the target detecting unit 21. However, the feature is that it is possible to avoid “identifying a target that cannot be an obstacle and a target that can be an obstacle as one solid object” that could not be implemented.
Below, the target exclusion part 22 which is the characteristic structure of this invention is demonstrated.

FIG. 4 is a flowchart showing a procedure of processing for excluding a target performed by the target excluding unit 22 shown in FIG.
First, the target exclusion unit 22 selects one detection point group at a distance within a predetermined range from the vehicle for a plurality of detection points detected by the target detection unit 21 (step S401). Then, the target excluding unit 22 confirms the distribution state of the detection points for the selected detection point group (step S402). In accordance with the distribution situation of the detection points, the target exclusion unit 22 determines whether there is an object with a space width w between the targets equal to or larger than a predetermined reference width for any two targets (step). S403).

  If it demonstrates with a specific example, as shown to FIG. 5A, the detection point 51a of the pedestrian 51 (1st target) and the detection point 52a of the street tree 52 (2nd target) will detect one. It becomes a point group. Then, it is determined whether or not the space width w from the detection point 51a of the pedestrian 51 to the detection point 52a of the roadside tree 52 is greater than or equal to a predetermined reference width. Here, the fact that the two targets are separated via a space having a predetermined reference width or more means that the two targets do not form one solid object.

If it can be confirmed that the second target is present from the first target through the space width w greater than or equal to the predetermined reference width (step S403: Yes), the target excluding unit 22 next It is determined whether or not the second target exceeds a predetermined reference height h (step S404, FIG. 5B). This determination is performed in order to determine that the second target is a target at a high position that cannot form one solid object together with the first target. The predetermined reference height h is set based on, for example, the following equation [1] using the average height Ave of a plurality of detection points included in the selected detection point group and the standard deviation σ of the detection point distribution. The
h = Ave + 2σ [1]

When it is determined that the height of the second target exceeds the predetermined reference height h (step S404: Yes), the target excluding unit 22 detects the second target (street tree 52). The point 52a is excluded from the detection result (step S405).
On the other hand, when the 2nd target which exists via space width w beyond a predetermined standard width cannot be checked from the 1st target (Step S403: No), or the height of the 2nd target is If it is determined that the predetermined reference height h is not exceeded (step S404: No), the target excluding unit 22 does not exclude the detection point 52a of the second target (street tree 52) from the detection result.

  Note that the processing described in steps S402 to S405 is executed for all detection point groups at a distance within a predetermined range from the vehicle (steps S401 and S406).

The three-dimensional object determination unit 23 determines a three-dimensional object based on the detection result output from the target exclusion unit 22.
Typically, the three-dimensional object determination unit 23 defines a two-dimensional plane region that is parallel to the road surface divided into a plurality of regions in a lattice shape, and is detected by the target detection unit 21 and the target exclusion unit The target detection point in the three-dimensional real space from which the target that is not the target in 22 is excluded is projected onto any divided region on the two-dimensional plane region. Then, the three-dimensional object determination unit 23 determines a three-dimensional object existing on the road surface according to the projection result of the detection point on the two-dimensional plane region.

The process performed by the three-dimensional object determination unit 23 will be specifically described using a scene where the vehicle is traveling on the road as illustrated in FIG. 2A described above.
Assume that the three-dimensional object determination unit 23 defines a two-dimensional planar region divided into a plurality of partitioned regions 60 in a lattice shape as shown in FIG. 6A. In this case, the three-dimensional real space is divided into a plurality of partitioned spaces extending in the up-down direction, which is a prismatic region perpendicular to the road surface with the partitioned region 60 as a cross section. The three-dimensional object determination unit 23 projects (votes) the plurality of detection points in the real space detected as shown in FIG. 2B to the corresponding divided areas 60 (also called ballot boxes) as shown in FIG. 6B. ). That is, regardless of the height position in the three-dimensional space, the detection points existing in the same partition space are projected onto one partition region 60. Next, the three-dimensional object determination unit 23 determines the number of detection points projected in each of the divided areas 60, and obtains a divided area 60 on which a number of detection points equal to or greater than a predetermined threshold is projected. Then, the three-dimensional object determination unit 23 groups (groups) the adjacent areas into one group and determines the three-dimensional object as an area where the obtained detection points are equal to or larger than the threshold value.

For example, in the target exclusion unit 22, when the processing in the three-dimensional object determination unit 23 is performed without excluding the detection point 52a of the roadside tree 52 that is the second target, it is shaded with a diagonal line in FIG. 6C. It is determined that the divided area 62 including the detection point of the street tree 52 is an area where a three-dimensional object exists.
The three-dimensional object determination unit 23 collects a single region from the detection point closest to the road surface to the detection point farthest from the road surface, based on the position information of the detection points for each of the divided regions 62 determined to have a three-dimensional object. Interpreted as a three-dimensional object.
The three-dimensional object interpreted as one by the three-dimensional object determination unit 23 is specified as one solid object 62a represented by a large rectangular frame including both the pedestrian 51 and the roadside tree 52, as shown in FIG. 7A. The

On the other hand, in the target excluding unit 22, when the processing in the three-dimensional object determining unit 23 is performed by excluding the detection point 52a of the road tree 52 that is the second target, it is shaded in FIG. 6D. It is determined that the segment area 61 that does not include the detection point of the roadside tree 52 is an area where a three-dimensional object exists.
The three-dimensional object determination unit 23 collects a single region from the detection point closest to the road surface to the detection point farthest from the road surface based on the position information of the detection points for each of the divided regions 61 determined to have a three-dimensional object. Interpreted as a three-dimensional object.
Therefore, the three-dimensional object interpreted as one by the three-dimensional object determination unit 23 in this case is specified as one solid object 61a represented by a small rectangular frame including only the pedestrian 51 as shown in FIG. 7B. .

  In addition, the process which the target information acquisition part 10 mentioned above, the target detection part 21, and the three-dimensional object determination part 23 perform is the existing technique regarding specification of a three-dimensional object. Accordingly, the configuration and processing method described in the present embodiment are merely examples, and the technique described in Patent Document 1 (a three-dimensional real space is divided into a plurality of partitioned spaces extending vertically, and a plurality of A three-dimensional object may be specified using a technique for creating a histogram for each of the divided spaces and specifying a three-dimensional object based on the histogram) or a conventional technique known to those skilled in the art.

As described above, according to the vehicle periphery monitoring device 1 according to the embodiment of the present invention, a target that cannot be an obstacle and a target that can be an obstacle are simultaneously at a distance within a predetermined range from the vehicle. Even if it is detected, if there is a target from a specific target through a space width w that is greater than or equal to a predetermined reference width and exceeds a predetermined reference height h, the target Is excluded from the detection result.
As a result, it is possible to avoid the identification of a target that cannot be an obstacle and a target that can be an obstacle as a single three-dimensional object. It becomes possible to perform driving support control.

  In the above-described embodiment, the case where the camera 11 and the camera 12 are provided inside the windshield of the vehicle and the image in the vehicle traveling direction is taken has been described. However, in addition to this, the camera 11 and the camera 12 may be provided inside the rear glass or side glass of the vehicle, and images of the rear and side of the vehicle may be taken.

Note that some or all of the functional blocks constituting the vehicle periphery monitoring device in the embodiment of the present invention are a central processing unit (CPU), a storage device (memory (ROM, RAM, etc.), hard disk, etc.), and input / output. It is realized by using hardware resources such as an apparatus, and is typically embodied as an IC (LSI, system LSI, super LSI, ultra LSI, etc.) that is an integrated circuit. These functional blocks may be individually made into one chip, or may be made into one chip so as to include a part or all of them.
Further, the method of circuit integration is not limited to ICs, and implementation using dedicated circuitry or general purpose processors is also possible. Also, an FPGA (Field Programmable Gate Array) that can be programmed after manufacturing the IC or a reconfigurable processor that can reconfigure the connection and setting of the circuit cells inside the IC may be used.
Further, if integrated circuit technology (such as biotechnology) that replaces ICs appears as a result of the advancement of semiconductor technology or another technology derived therefrom, functional blocks may naturally be integrated using this technology.

  The three-dimensional object specifying method executed by the vehicle periphery monitoring device according to the embodiment of the present invention described above includes predetermined program data that can execute the procedure of the three-dimensional object specifying method stored in the storage device. It may be realized by being interpreted and executed by the CPU. In this case, the program data may be introduced into the storage device via a recording medium such as a CD-ROM or a flexible disk, or may be directly executed from the recording medium. The recording medium refers to a semiconductor memory such as a ROM, a RAM, or a flash memory, a magnetic disk memory such as a flexible disk or a hard disk, an optical disk memory such as a CD-ROM, a DVD, or a BD, and a memory card. The recording medium is a concept including a communication medium such as a telephone line or a conveyance path.

  The vehicle periphery monitoring apparatus according to the present invention can be used for a vehicle equipped with a vehicle driving support system such as collision avoidance with an obstacle or automatic driving, and particularly for detecting a target that cannot be an obstacle. This is useful when, for example, it is desired to prevent a decrease in accuracy in driving support control by excluding the result.

DESCRIPTION OF SYMBOLS 1 Vehicle periphery monitoring apparatus 10 Target information acquisition part 11, 12 Camera 20 Three-dimensional object specific | specification part 21 Target detection part 22 Target exclusion part 23 Three-dimensional object determination part 51 Pedestrian 52 Street tree 55 Image area 51a, 52a Detection point 60, 61, 62 Division area 61a, 62a Three-dimensional object

Claims (6)

A device for monitoring the periphery of a vehicle,
A target information acquisition unit that acquires information about targets in the vicinity of the vehicle;
Among the targets acquired by the target information acquisition unit, a plurality of targets whose distance from the vehicle is within a predetermined range are extracted, and a space having a predetermined width or more in the height direction from the first target. A target excluding unit for excluding the second target that is separated from the detection result,
A vehicle periphery monitoring device comprising: a three-dimensional object specifying unit that specifies a three-dimensional object existing around the vehicle based on a target detection result output from the target excluding unit. The target detection unit detects a characteristic location of the target as a plurality of detection points,
The target exclusion unit includes a plurality of detection point distribution areas of the first target and a plurality of detection point distribution areas of the second target, which are detected by the target detection unit. The vehicle periphery monitoring device according to claim 1, wherein it is determined whether or not the vehicle is separated through a space having a predetermined width or more in the height direction.   The vehicle periphery monitoring device according to claim 2, wherein the target exclusion unit excludes the second target having a predetermined reference height or more from a detection result.   The said predetermined reference height is determined based on the height distribution situation of the detection point which the several target in the distance within the predetermined range from the said vehicle has, The characterized by the above-mentioned. Vehicle periphery monitoring device.   The vehicle periphery monitoring device according to claim 1, wherein the first target is a target on a road surface.   The three-dimensional object specifying unit creates a histogram based on the detected target for each of a plurality of partitioned spaces extending in the vertical direction formed by dividing a three-dimensional real space, and based on the histogram The vehicle periphery monitoring device according to claim 1, wherein a three-dimensional object is specified by the method.

Images