JPH1097699A - Obstacle detecting device for vehicle - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an obstacle detecting device for vehicle which can extract horizontal and vertical edges in a shorter time and is improved in processing efficiency by limiting the area for extracting the horizontal edge based on left and right guide lines and the area for extracting the vertical edge based on the extracted horizontal edge. SOLUTION: An obstacle detecting device for vehicle is provided with a horizontal edge extracting means M2 which extracts a horizontal edge in an area between recognized left and right guide lines, an area estimating means M3 which estimates an area having a possibility of the existence of an obstacle based on the horizontal edge, a vertical edge extracting means M4 which extracts a vertical edge in the possible area, and a discriminating means M5 which discriminates the obstacle based on the extracted vertical edge. Therefore, the time required by the obstacle detecting device for extracting the horizontal and vertical edges can be shortened and the processing efficiency and processing speed of the device can be improved, because the areas for extracting the edges can be contained in a taken picture.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vehicle obstacle detecting device, and more particularly to a device for detecting an obstacle from an image in front of a vehicle.

[0002]

2. Description of the Related Art Conventionally, there has been proposed an obstacle detecting apparatus which performs image processing on an image captured in front of a vehicle to detect an obstacle such as a preceding vehicle. For example, Japanese Patent Application No. 4-15
No. 1343 discloses an image in front of a vehicle, detects a horizontal edge and a vertical edge of the image, and detects an area in which both a horizontal edge continuous area and a vertical edge continuous area exist. Judge as an obstacle.

[0003]

In the conventional apparatus, horizontal edges are detected over the entire area of a captured image, and
Since an obstacle is determined by detecting a vertical edge, the horizontal edge and the vertical edge are also detected in an area where no obstacle can exist, and there is a problem that processing efficiency is low and processing takes time. Was.

[0004] The present invention has been made in view of the above points,
By limiting the area where horizontal edges are extracted based on the left and right guidelines and the area where vertical edges are extracted based on the extracted horizontal edges, the time required for extracting horizontal edges and vertical edges can be reduced, and processing can be reduced. It is an object of the present invention to provide a vehicle obstacle detection device with improved efficiency.

[0005]

According to a first aspect of the present invention, there is provided a vehicle obstacle detecting apparatus for detecting an obstacle from an image M0 taken in front of a vehicle, as shown in FIG. A guide line recognizing means M1 for recognizing the left and right guidelines, and a horizontal edge extracting means M2 for extracting a horizontal edge having a large difference in luminance between vertically adjacent pixels in a region sandwiched between the recognized left and right guidelines.
And an area estimating means M3 for estimating a possibility area where an obstacle is present based on the extracted horizontal edge, and a large difference in luminance between horizontally adjacent pixels in the estimated possibility area. A vertical edge extracting unit M4 for extracting a vertical edge, and a determining unit M5 for determining an obstacle based on the extracted vertical edge.

As described above, the region from which the horizontal edge is extracted is limited based on the recognized left and right guidelines, and the existence possibility region is estimated based on the extraction result of the horizontal edge. Because only vertical edges are extracted, the area where horizontal and vertical edges are extracted can be part of the captured image, reducing the time required for extraction, improving processing efficiency, and enabling high-speed processing. Becomes

[0007]

FIG. 2 is a block diagram showing an embodiment of the apparatus according to the present invention. In FIG. 1, a video camera 10 captures an image of a road ahead of the host vehicle. The image signal is supplied to an image input circuit 11, A / D converted, and supplied to an interface circuit 12. The image data read from the interface circuit 12 under the control of the CPU 20 is transmitted to the bus 14.
Is supplied to and stored in the video RAM 15.

A correlation processor 21 performs a correlation operation between the template image data read and supplied from the RAM 16 under the control of the CPU 20 and the search area image data read and supplied from the video RAM 15. The resulting correlation value is supplied to the CPU 20. CPU
20 executes a processing program stored in the ROM 18 to determine an obstacle such as a preceding vehicle in the image. At this time, the RAM 16 is used as a work area. Also, C
The PU 20 transmits the result of the obstacle determination to the interface circuit 17.
To an external system such as an automatic operation control system.

FIG. 3 shows a flowchart of the obstacle judging process executed by the CPU 20. This process is executed every predetermined time. In the figure, in step S10, image data of one screen image captured by the video camera is input from the interface circuit 12 to the video RAM 15. Next, white line recognition is performed in step S12 corresponding to the guideline recognition means M1. Note that the guide line is not limited to the white line, and may be a yellow overtaking prohibited line.

Here, an image 30 of a road shown in FIG.
Search area 33 ₁ ~ 33 _2n against the white line 31 is a guideline of the inner (n is, for example, 9) is provided, each search region 33 ₁ ~ 33 _2n is the vertical position and the horizontal position in the image 30 within Has been fixed.

When one screen has 512 × 512 pixels,
Each search area 33₁~ 33_2nIs, for example, 32 x vertical
It is composed of 23 pixels in the direction. In addition, the search area 33₁~ 3
3_2n2n templates corresponding to each are stored in the RAM 16
Each template is stored in the horizontal direction 16 × vertical direction
FIG. 4B shows a search area 33. ₁
Shows a template corresponding to. However, in contrast to FIG.
35 is white line, 30 is asphalt
G (background part).

In the correlation operation processor 21, the search area 33
A _first image data, when this is supplied with image data of a corresponding template, by performing a correlation operation of both the white line portion in the search area 33 ₁ position and the correlation value CPU2
Transfer to 0. A search area 33 ₂ ~ 33 _2n, respectively in the same manner, this is carried out correlation calculation between corresponding template, the position and the correlation value of the white line portion in the search area each are transferred to the CPU 20. The CPU 20 selects, from these search results, only those for which the correlation value is recognized as a white line when the correlation value is equal to or less than the predetermined reference value, and based on the position of the white line portion in each of the selected search areas, for example, by the least square method, the white line
The position and inclination of each of the first and the second 32 are obtained, the position of the vanishing point 34 is further obtained, and white line recognition is performed.

When the white line recognition is completed, the process proceeds to step S14 corresponding to the horizontal edge extracting means M2, and the white line 3
A horizontal edge is detected in an area sandwiched between 1 and 32, that is, an area within a white line. The horizontal edge is a position where the difference in luminance between vertically adjacent pixels is equal to or greater than a predetermined threshold.

Next, the horizontal edge detected in step S16 is projected on the Y axis. Here, the projection means to count horizontal edges having the same Y coordinate and different X coordinates on the screen for each Y coordinate. Next, proceeding to step S18, the Y-axis position at which the Y-axis projection value is equal to or larger than the threshold value TH (y) is extracted,
The region candidate positions Cy (k) are set in order from the one with the largest Y coordinate. At the same time, the number of region candidate positions Cy (k) is obtained. For example, in the image shown in FIG. 5, it is assumed that the Y-axis projection value of the horizontal edge is as shown by the solid line in FIG. Threshold T
H (y) is a function of the Y coordinate, and the value increases as the Y coordinate increases, as indicated by a dashed line. This is Y
This is because the larger the axial position, the closer the obstacle is to the host vehicle, and the greater the Y-axis projection value of the horizontal edge. Thus, the region candidate positions Cy (0), Cy (1), Cy
(2) is extracted.

Next, at step S20, the candidate counter k is set to 0.
And the process proceeds to step S22. Area estimation means M
In step S22 corresponding to No. 3, an area where an obstacle may exist is estimated. This existence possibility area is a rectangle, the area candidate position Cy (k) is set as a lower limit, and the X coordinate of a point where the Y coordinate of the left white line 31 matches Cy (k) is set as Ll (k).
The X coordinate of the point where the Y coordinate of the right white line 32 coincides with Cy (k) is defined as Lr (k), and a value Wy (k) obtained by multiplying the distance between Lr (k) and Ll (k) by a predetermined coefficient α. ), Wy (k) = α · (Lr (k) −L1 (k)) An area represented by upper left coordinates and lower right coordinates as follows.

(Ll (k), Cy (k) -Wy
(K)), (Lr (k), Cy (k)) where the coefficient α is, for example, α = １ ／, which is Cy
This is because, when (k) is set as the lower limit of the preceding vehicle, the height of the vehicle is almost within 1/2 of the lane width (the distance between the white lines 31 and 32). In the image shown in FIG.
The existence possibility area corresponding to (0) is points β1, β2, β
3, inside the rectangle of β4.

Next, the process proceeds to step S24 corresponding to the vertical edge extracting means M4, and a vertical edge in the existence possibility area is detected. A vertical edge is a position where the difference in luminance between horizontally adjacent pixels is equal to or greater than a predetermined threshold. Next, the vertical edge detected in step S26 is projected on the X axis. That is, horizontal edges having the same X coordinate and different Y coordinates on the screen are counted for each X coordinate. Then, in step S28, X-axis positions at which the X-axis projection value is equal to or greater than a predetermined threshold value THx (y) are extracted, and are set as Vx (i) in ascending order of the X-axis position. For example, when the X-axis projection value of the vertical edge for the points β1, β2, β3, and β4 is as shown by the solid line in FIG.
The position V is obtained by comparison with a threshold value THx (y) indicated by a dashed line.
x (0), Vx (1), and Vx (2) are extracted, and together with this, the number Nx of the extraction points Vx (i) is obtained. Threshold value THx
(Y) is a function of the Y coordinate (the value of Cy (k)),
The value increases as the value of (k) increases.

Subsequently, in step S30, it is determined whether or not the number of extracted points Nx is greater than 1. If Nx> 1, step S30 is executed.
Go to 32. In step S32, all extraction positions Vx
From (i), all combinations of a pair of positions are performed, and the X-direction width W (j) of each pair is obtained. As shown in FIG.
When (0), Vx (1), and Vx (2) are extracted, the width W (0) = | Vx (0) −Vx (1) |, W (1) =
| Vx (0) −Vx (2) |, W (2) = | Vx (1)
−Vx (2) | is obtained.

Thereafter, in step S34, it is determined whether or not WTH1 (y) <W (j) <WTH2 (y) is satisfied for each W (j). The above threshold value WTH1 (y), W
TH2 (y) is a function of the Y coordinate Cy (k),
The value increases as y (k) increases. Threshold value WTH1
(Y) corresponds to the minimum width of the vehicle at the Y coordinate Cy (k), and the threshold value WTH2 (y) corresponds to the maximum width of the vehicle at the Y coordinate Cy (k). Here, WTH1 (y) <
If W (j) <WTH2 (y) is equal to (W)
(0) or W (1) or W (2)), if satisfied, the process proceeds to step S36. If not all W (j) are satisfied, the width of the detected object does not match the vehicle. And the process proceeds to step S38.

In step S36, for a width W (j) satisfying the condition of step S34, a pair of extraction positions Vx (i) from which the width W (j) is obtained, and a candidate region position Cy (k) at that time are obtained. Are output as the left and right ends and the lower end of the obstacle, and the process ends. The above steps S30 to S40 correspond to the determination means M5.

On the other hand, if Nx> 1 in step S30, since there is only one edge extending in the vertical direction or there is no edge, it is assumed that there is no obstacle in this existence possibility area.
Proceeding to step S38, the value of the subscript k is incremented by one. Thereafter, the process proceeds to step S40, where it is determined whether or not k <Nc. If k <Nc, the region candidate position Cy (k) still remains, so the process proceeds to step S22 to perform the processes of steps S22 to S40. repeat. If k ≧ Nc, the processing for detecting the left and right ends of the obstacle from the existence possibility area has been completed for all of the area candidate positions Cy (k), and the processing ends.

As described above, the region where the horizontal edge is extracted is limited based on the recognized left and right white lines, and the existence possibility region is estimated based on the extraction result of the horizontal edge. Just to extract the vertical edges,
The region for extracting each of the horizontal edge and the vertical edge can be a part of the captured image, so that the time required for extraction can be reduced, the processing efficiency can be improved, and high-speed processing can be performed.

In step S14, white lines 31, 32
Since the horizontal edge is extracted only in the region sandwiched between the white lines 31 and 32, the horizontal edge of the building or the like existing in the region outside the white lines 31 and 32 is not detected, and the horizontal edge of the building or the like is mistaken as the horizontal edge of the obstacle. There is no fear. The same is true for vertical edges, which improves the accuracy of obstacle detection.

[0024]

As described above, the first aspect of the present invention provides
In an obstacle detection device for a vehicle that detects an obstacle from an image captured in front of a vehicle, a guideline recognition unit that recognizes left and right guidelines in the image, and a vertical direction within an area sandwiched between the recognized left and right guidelines. Horizontal edge extracting means for extracting a horizontal edge having a large difference in luminance between pixels adjacent to the pixel, and area estimating means for estimating a possibility area where an obstacle is present based on the extracted horizontal edge. Vertical edge extracting means for extracting a vertical edge having a large luminance difference between horizontally adjacent pixels in the existence possibility area, and determining means for determining an obstacle based on the extracted vertical edge. .

As described above, the region from which the horizontal edge is extracted is limited based on the recognized left and right guidelines, and the existence possibility region is estimated based on the result of extracting the horizontal edge. Because only vertical edges are extracted, the area where horizontal and vertical edges are extracted can be part of the captured image, reducing the time required for extraction, improving processing efficiency, and enabling high-speed processing. Becomes

[Brief description of the drawings]

FIG. 1 is a principle diagram of the present invention.

FIG. 2 is a block diagram of the present invention.

FIG. 3 is a flowchart of an obstacle detection process according to the present invention.

FIG. 4 is a diagram for explaining a captured image and a template.

FIG. 5 is a diagram for explaining a captured image.

FIG. 6 is a diagram showing a Y-axis projection value of a horizontal edge.

FIG. 7 is a diagram showing an X-axis projection value of a vertical edge.

[Explanation of symbols]

 Reference Signs List 10 Video camera 11 Image input circuit 12, 17 Interface circuit 15 Video RAM 16 RAM 18 ROM 20 CPU 21 Correlation processor M0 Image M1 Guideline recognition means M2 Horizontal edge extraction means M3 Area estimation means M4 Vertical edge extraction means M5 Judgment means

Claims (1)

[Claims] 1. An obstacle detecting device for a vehicle, which detects an obstacle from an image captured in front of a vehicle, comprising: a guideline recognizing means for recognizing left and right guidelines in the image; Horizontal edge extraction means for extracting a horizontal edge having a large difference in luminance between vertically adjacent pixels in an area, and area estimation for estimating a possible area where an obstacle is present based on the extracted horizontal edge Means, vertical edge extracting means for extracting a vertical edge having a large luminance difference between horizontally adjacent pixels in the estimated existence possibility area, and determining an obstacle based on the extracted vertical edge. An obstacle detection device for a vehicle, comprising: a determination unit.