JP3134749B2 - White line detection method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a white line detecting method for detecting a white line drawn on a road surface along a road.

[0002]

2. Description of the Related Art There has been proposed a vehicle that autonomously runs on a road surface by detecting a white line drawn along the road surface and preventing the vehicle from deviating from the road. . In such an apparatus, it is necessary to accurately recognize or capture a target white line.
As a method of detecting a white line, there is a method of detecting an edge of a target white line in an image obtained by capturing an image of the front of the vehicle viewed from the driver's seat with a camera, a method of extracting a feature amount, and the like. .

[0005] There are two methods for detecting an edge: a method for extracting a contour line and a method for extracting an original image. The edge is
The brightness of the road surface changes abruptly, and the white line indicates that the road surface changes from black to white. The method of extracting the contour line is to take a contour image of the input image, set a horizontal search area, search from the center of the road, set the first edge as a white line candidate point, and check the neighborhood of the white line detected in the previous search area. , A new search area is set, and the same search is performed to extract a contour line. The method of extracting the original image is to set a plurality of detection lines in the vertical direction, detect edges on the lines, and detect the edges of the detected valleys having a common attribute (such as direction). It is a white line candidate.

A method of extracting a feature value is to take a contour image (binary image) of an input image, perform labeling of the contour image, and apply a validity as a white line to a labeled area, for example, Evaluate that the slender course does not change suddenly,
The white line candidates are extracted.

[0005]

However, in the extraction of the contour line in the edge detection method, since the search is performed in the horizontal direction from the center of the road, the detection accuracy is remarkably high due to shadows on the road, characters drawn on the road surface, and the like. Getting worse. Also, the method of extracting the original image is difficult to cope with a road that changes from a straight line to a curve or a road that has a long curve, for example, in order to recognize the original image with a common attribute.

In the method of extracting a feature value, it is difficult to detect a long-distance white line (particularly, a broken line) because it is detected based on the validity as a white line. There is a problem that it takes a very long process to evaluate the validity one by one. Furthermore, when a white line is hidden by a vehicle traveling ahead of the host vehicle, the white line cannot be detected, and may deviate from a regular white line position or may be erroneously detected. There are also problems such as difficulty in recognizing.

The present invention has been made in view of the above points, and can detect a white line drawn on a road surface on a road that curves in either direction, and the white line is hidden by a preceding vehicle. An object of the present invention is to provide a white line detection method capable of preventing erroneous detection in a case where the white line is detected at a long distance with high accuracy.

[0008]

According to the present invention, in order to achieve the above object, according to the present invention, a search area having a predetermined length from a predetermined position on a screen on a road surface in front of a captured vehicle is defined as the search area. The road surface is set along the white line drawn along the road and inclining in the direction in which the white line curves , and the luminance value is measured in each search area to determine the road surface and the white line. A white line is detected, the presence or absence of a preceding vehicle is recognized from the captured image, and when the search area is within the area of the preceding vehicle, the white line is processed as not detected.

[0009] In claim 2, when recognizing the preceding vehicle, when the luminance value of the search area is lower than a predetermined value, or brightness values of <br/> search area is equal to or higher than a predetermined value, and the service < The search area is canceled as white line not detected when any of the white line candidate points in the area is in the area of the preceding vehicle. According to the third aspect, when a preceding vehicle is recognized, a white line search area is set, and when the search area overlaps the area of the preceding vehicle, the search area is canceled as white line not detected.

In the first or second aspect, the road surface is distinguished from the white line by using a threshold value that is higher than the luminance value of the road surface by a predetermined value, and a portion having a luminance value higher than the threshold value is determined as a white line. You may do so. In claim 1,
The brightness value of the road surface may be an average value of brightness values at a plurality of locations on the road surface immediately before the vehicle. An image of the road surface in front of the vehicle seen from the driver's seat is captured and displayed on a screen, and a search area of a predetermined length is set at a predetermined position on the screen. And
The luminance value in each search area is measured, the road surface and the white line are discriminated, and the white line is detected. This search area is set along the white line drawn along the road from the predetermined position on the screen to the road surface.
Then, the white line is tracked and set sequentially while being inclined in the direction in which the white line curves . On the other hand, the vehicle recognizes the preceding vehicle traveling ahead of the own vehicle and determines that the brightness value of the search area is predetermined.
Is lower than the value, or brightness values of the search area is higher than a predetermined value, when and white line candidate points in the search area is within the area of the front <br/> next vehicle, or the front vehicle area When the search area is overlaid, the white line is not detected and the search area is canceled. As a result, erroneous detection by the preceding vehicle can be prevented, and the white line can be accurately recognized up to a long distance.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The embodiments of the present invention will be described below by way of examples. (Embodiment) The white line detection method of the present invention focuses on the fact that a white line drawn on a road surface along a road has a higher luminance value than the surrounding road surface, and detects the white line with high accuracy up to a long distance. Things. FIG. 1 is a block diagram of a system configuration of a detection device for carrying out the method of the present invention. In FIG. 1, a white line detecting device is installed at a predetermined location in the vehicle, for example, a position near a rear-view mirror, and is a monitor monochrome camera 1 that captures an image of a road surface ahead as viewed from a driver's seat. It is composed of an image processing device 2 and a computer 3 which take in monochrome image information, perform necessary image processing, and display it on a display device (not shown).

FIG. 2 shows an image screen in front of the driver's seat taken by the camera 1, and white lines 6 and 7 are displayed on the road surface 5 a of the road 5.
Are drawn along the road 5, for example, a white line 6 on the left side of the screen indicates a boundary line between the road shoulder and the traveling lane, and a white line 7 on the right side indicates a boundary line between the traveling lane and the overtaking lane. ing. Note that the vehicle is traveling in the traveling lane, and the road 5 is slightly curved to the right.

Next, a method for detecting a white line will be described.
As is clear from FIG. 2, the position of the left white line 6 when viewed from the driver's seat when traveling in the driving lane can be predicted to some extent on the screen. Is found once, and thereafter, the found white line 6 is sequentially traced. The white line 6 is tracked in the horizontal direction (X direction) from the left to the right of the screen, and in the vertical direction (Y direction) from the bottom to the top of the screen, from left to right, and from top to bottom in the positive direction. .

Since the white line 6 is brighter and has a higher luminance value than the road surface 5a, a threshold value of the luminance value for discriminating between the road surface 5a and the white line 6 is set. First, a plurality of, for example, three points (i, j, k) in the horizontal direction are taken as reference points of the luminance value of the road surface 5a immediately before the vehicle, and the luminance value of each of these points is detected. To determine. VTH = MAX (i, j, k) + α (1) or VTH = (i + j + k) / 3 + α (2) where α is, for example, a luminance value of about 10 to 30. I do. Since the brightness value of the white line 6 is clearly higher than the average value compared to the other portions, a value higher than the average value by the value α is set as the threshold value VTH to ensure the detection of the white line 6.

Next, at the approximate location where the white line 6 should exist on the road surface 5a, that is, at the lower left corner of the screen,
Is set as the first search area A1 for detecting. This search area A1 is represented by a straight line of a predetermined length in the vertical direction of the screen. Then, the luminance is measured in the range of the search area A1. Taking the distribution of luminance in the search area A1, if there is a white line 6, there is a part where the luminance value is extremely high. Therefore, a point K (1) in which the luminance value is extremely high is left as a white line 6 for the time being.

Similarly, a second search area A2 is set at a predetermined interval d in the horizontal direction from the point K (1), and a luminance distribution in the search area A2 is obtained. The point K (2) is defined as a white line 6. Second point K (2)
Is very close to the first point K (1) and is always higher than the point K (1), so that it can be easily detected. The positions to set the first search area A1 and the second search area A2 are determined.

After the two white line detection points K (1) and K (2) are determined from the first search area A1 and the second search area A2 as described above, the third and subsequent search areas are determined. Are sequentially set in the rightward and vertical directions of the screen at the predetermined interval d, the search area gradually deviates from the white line 6 because the road 5 is curved. If the road 5 has a sharp curve, the white line 6 may not be detected.

Further, when the road 5 is curved to the left, the curve of the white line 6 on the left side is very tight compared to the case where the road 5 is curved to the right as shown in FIG. The direction changes from the positive direction to the negative direction on the way. Therefore, if the search area is simply moved rightward from the left side of the screen, for example, the fourth search area A4 indicated by a dotted line in the figure completely deviates from the white line 6, and the white line 6 cannot be detected. Therefore, the vehicle cannot respond to the left curve of the road 5. Similarly, when the search area is taken from the right side of the screen, it becomes impossible to deal with a right curve.

Therefore, for the third and subsequent search areas, the next white line detection expected position is estimated from the previous white line detection point and the current white line detection point. That is, as shown in FIG. 4, the search area is inclined in the direction in which the white line 6 curves, and the search area is set so as to be always perpendicular to the white line 6. As a result, the search area approaches the white line 6, and even when the search area of the white line 6 is taken from the left side of the screen, it is possible to deal with the rightward and leftward curves of the road 5.

Next, a method for setting the third and subsequent search areas will be described. For the third and subsequent search areas, a center point of the next search area is set on an extension of a straight line connecting the previous white line detection point and the current white line detection point, and the center point of the current search area and the next search area are set. The distance d between the center point of the search area and the center point of the previous search area is set to the same length as the distance d between the center point of the previous search area and the center point of the current search area. That is,
The distance d between the center points of the search areas is made constant. Also,
The length of each search area is a fixed length except for a special case described later (when not detected). Then, a search area having a fixed length is set so as to pass through the set center point and be orthogonal to the extension line.

FIG. 5 shows a method of setting the center position of the third and subsequent search areas. The center points of the search areas A1 and A2 are M1 and M2, and the positions of the upper and lower ends are P11 and P12.
These are indicated by P21 and P22. Hereinafter, similarly, the center point of each search area and the positions of the upper and lower ends are represented. First search area A1
And each white line detection point K (1) and K (2) in the second search area A2
A point M3 is taken on an extension line (indicated by a dotted line) connecting the center point M2 and a distance d from the center point M2, passes through the point M3, and is orthogonal to the extension line to a predetermined length (radius r). (Diameter of a circle) and set this area as the third search area A
Assume 3.

First, a case where the next detection expected position Mi is estimated from the previous detection point K (c-1) and the current detection point K (c) when the number of times of detection of the white line 6 is 2 or more will be described. At this time,
Mi is represented by a straight line passing through the detection points K (c-1) and K (c).
1, a distance between Mi) = d (constant). As a result, the length between the white line detection points becomes substantially the same, and the detection conditions can be made substantially the same.

Therefore, each point Mi, Mi-1, K (c), K (c-1)
Are defined as Mi (xmi, ymi) and Mi-1 (xmi-1, y
mi-1), K (c) (x (c), y (c)), K (c-1) (x (c-1),
y (c-1)), the points xmi and ymi are expressed as follows. (1) ang ≧ 0 when xmi = xmi-1 - {d 2 / (1 + ang 2)} 1/2 ··· (3) ymi = ymi-1 - {d 2 / (1 + ang 2)} 1/2 · ang ··· (4) (2 ) ang <xmi = xmi-1 + {d 2 / (1 + ang 2)} when ^{0 1/2 ··· (5) ymi =} ymi-1 + {d 2 / (1 + ang ² )｝ ^1/2 · ang (6) where ang = (y (c) −y (c−1)) / (x (c) −x (c−
1)). This ang is positive when the inclination of the line connecting the previous detection point and the current detection point is downward to the right on the screen, and negative when it is upward to the right. D is, for example, a value of about 25 pixels. Next, the point M is orthogonal to a straight line passing through the detection points K (c-1) and K (c),
Intersections of a straight line passing through i and a circle having a radius r centered on the point Mi are Pi1 and Pi2, respectively, and Pi1 (xi1, yi2), Pi2
When (xi2, yi2), the coordinates of the point Pi1, xi1 = xmi - {r 2 / (1 + ang_1 2)} 1/2 ··· (7) yi1 = ymi - {r 2 / (1 + ang_1 2)} 1 ^{/ 2} · ^ang_1 coordinates (8) point Pi2 ^{is, xi2 = xmi + {r 2} / (1 + ang_1 2)} 1/2 ··· (9) yi2 = ymi + {r 2 / (1 + ang_1 2) ｝ ^1/2 · ang_1 (10) Here, ang_1 = −1 / ang, r is a value of about 15 pixels which is half the length of the search area having a predetermined length.

A line segment connecting these points Pi1 and Pi2 is set as the next search area. Thus, FIG.
A search area is set as shown in FIG. This makes it possible to detect the white line 6 for the left curve as shown in FIG. Of course, the right curve shown in FIG. 2 can be dealt with. When the white line 6 could not be detected in the search area A3 (not detected), as shown in FIG. 6, on the extension line (shown by a dotted line) connecting the white line detection points K (1) and K (2). And a distance 2d from the center point M2 of the search area A2
A fourth search area A4 is set at a point M4, passing through the point M4 and orthogonal to the extension line. Then, the length of the fourth search area A4 is set to be longer than the predetermined length, that is, the length of the search areas A1 and A2.

That is, when the number of times of detection of the white line 6 is two or more, instead of the radius r of the circle centered on the point Mi, the radius r 'is changed according to the number of undetections (T), and r' = ( r + α
The search area is enlarged as T) (see FIG. 6). Here, the value α is about 5 pixels. In FIG. 6, the length of the search area A4 is exaggerated. When the white line 6 can be detected in the search area A4, the detected point is set to K (4). In the fifth search area, since the white line 6 could not be detected in the third search area A3, the second search area A2
A point M5 is located at a distance d from the center point M4 of the search area A4 on an extension of a straight line connecting the white line detection point K (2) of the search area A4 and the white line detection point K (4) of the fourth search area A4. Then, a fifth search area A5 is set centering on the point M5 and orthogonal to the extension line.

If the white line 6 cannot be detected even in the fourth search area A4 (not detected), the white line detection points K (1) and K (2) of the search areas A1 and A2 are obtained. Distance 3 from the center point M2 of the search area A2 on the extension line connecting
A point M5 is set at the position d, and a fifth search area A5 is set passing through the point M5 and orthogonal to the extension line.
Then, the length of the search area A5 is set longer than the search area A4. In this way, after the white line was detected last time,
If undetected before the next white line is detected,
The length of the next search area is set to be sequentially longer according to the number of undetections (T) occurring during that time, so that the possibility of detecting the white line 6 is increased.

However, when the number of undetected white lines 6 is increased and the detection point is missed, the estimated position of the next search area is deviated too much from the previous detection position, and the next search area is set. However, there is a possibility that the white line 6 cannot be detected. This is especially true in the case where the curve of the white line 6 is a sharp left curve as shown in FIG. Therefore, the number of undetected times is counted, and a predetermined number (for example, 5
When this occurs, the search for the white line 6 on the screen is stopped, and the next screen is displayed, and the detection of the white line 6 is started.

When the number of detections is less than two, that is, when the second search area A2 is set, it is not possible to estimate the expected detection position by the above-described calculation method. Therefore, the center position M2 of the search area A2 at this time
Are the coordinates (Xp, Yp) of the detection point of the first search area A1.
(1)), the Y coordinates (Yp1 ',
Yp2 ') alone is determined by the following equation. Note that the coordinates Xp ′
Is obtained by adding a predetermined distance d to Xp.

Xp ′ = Xp + d (11) Yp1 ′ = Y (1) + h1 (12) Yp2 ′ = Y (1) + h2 (13) where h1 is −30 pixels The value of about h2 is a value of about 10 pixels. The second search area is always at a position higher than the detection position in the first search area, and sequentially moves to a higher position.

However, the center point Mi (xmi, ymi) is determined by the following equation. xmi = xP '(14) ymi = (yP1' + yP2 ') / 2 (15) By the way, when the left curve of the road 5 becomes tighter as shown in FIG. In the setting, for example, as shown in the search area A5 in FIG.
Detection may not be possible. In addition, on a normal road, the minimum radius of the curve is about 50 m. Therefore, it is preferable that the white line 6 can be detected even in the left curve having the minimum radius.

Therefore, the search area obtained by the above equations (7) to (10) faces the curve of the white line 6 about the center point Mi (counterclockwise in FIG. 7) by a predetermined angle θ. Rotate. As a result, the search area can be brought closer to the white line 6, and the white line 6 can be detected more. In general, coordinates (x ′, y ′) after a certain point (x, y) is rotated counterclockwise by θ ° around the point (α, β)
Is

[0032]

(Equation 1)

## EQU2 ## Therefore, the coordinates Pi after the points Pi1 and Pi2 are rotated counterclockwise by θ ° around the point Mi.
1 '(xi1', yi1 ') and Pi2' (xi2 ', yi2') are the points Pi
At 1 ', xi1' = xmi + cosθ.xi1 + sinθ.yi1 (17) yi1 '= ymi-sinθ.xi1 + cos.theta.yi1 (18) At point Pi2', xi2 '= xmi + cos.theta. xi2 + sinθ · yi2 (19) yi2 ′ = ymi−sinθ · xi2 + cosθ · yi1 (20) Here, θ is an angle of about 30 ° to 45 °.

The line connecting these two points Pi1 'and Pi2' is used as the final next search area. FIG. 8 shows a case where the white line 6 is detected by the search area set in this way. This makes it possible to detect the left white line 6 even in the case of a left curve having a small radius. Also,
When there are two or more detection points in one search area, for example, when two white lines, a white line and a dotted line, or a white line and a curb are close to each other, the threshold V is set in the same search area.
Two points exceeding the TH (white line candidates) may be detected. In such a case, a point closer to the expected position is defined as a white line. That is, assuming that the Y coordinates of the two white line candidates are Y1 and Y2, the white line detection position Y (c) is the smaller of the absolute values of Y1 and Y2 and the expected detection position Yc.

When | Yc−Y1 | <| Yc−Y2 |, Y (c) = Y1 (21) When | Yc−Y1 |> | Yc−Y2 |, Y (c) = Y2.・ ・ (22) If the absolute values are equal, the point with the higher luminance value is set as the white line detection position. Next, the procedure of the white line detection method will be described with reference to the flowchart of FIG. First, as shown in FIG. 2, a threshold value VTH for detecting the brightness value of the road surface 5a immediately before the vehicle and distinguishing the road surface 5a from the white line 6 is determined by the above equation (1) or (2) (step S2).
1). Next, the search area for detecting the white line 6 is expressed by the following equation.
Set by (3) to (15) and (17) to (20) (Step S
2) A luminance value is measured in the search area (step S3). It is determined whether or not the luminance value of the detected point is equal to or higher than the threshold value VTH (step S4). If the luminance value is equal to or higher than the threshold value VTH, an undetected counter for counting the number of undetected times T is reset (step S5). Is 2 times
It is determined whether the number is equal to or more than the number of times (step S6). When the number of times of detection of the white line 6 is two or more, it is determined whether or not there are two or more points having a luminance value equal to or more than the threshold value VTH (step S7). As a white line candidate, the previous equation (2
Check whiteness likeness by 1) and (22) (Step S8)
Then, the white line and the curb are distinguished from each other, and the next search area is set near the detection position (step S9).

In step S6, when the number of detections of the luminance value equal to or higher than the threshold value VTH in the search area is two or less, that is, once, and when there are no two or more points equal to or higher than the threshold value VTH in step S7, A point having a luminance value is set as a white line candidate (step S10), and the process proceeds to step S9. Also,
If there is no point having a luminance value equal to or greater than the threshold value VTH in step S4, the count value of the counter for non-detection is advanced by 1 (step S11), and the next search area is set near the previous search area (step S12). When the number of undetected times exceeds 5, the detection is stopped and the screen shifts to the next screen.

Further, the threshold value VTH of the luminance value may be changed for each search area. Road surface 5a nearby in FIG.
Although the threshold value is set based on the brightness value of the road surface 5, since the brightness of the entire road surface decreases as the road surface 5a increases, the brightness of the white line 6 also decreases. For this reason, the luminance of the near road surface and the luminance of the distant white line do not change much, and it becomes difficult to detect the distant white line 6. Therefore, an average luminance value SAV in the search area is obtained, and a certain value α is added to this average luminance value to obtain a threshold VST
Set H. That is, VSTH = SAV + α (α is a value of about 10 to 30) (23). Since the white line part has a clearly higher luminance value than the other parts,
A line higher than the threshold value VSTH is defined as a white line. As a result, it is possible to detect a distant white line satisfactorily.

By the way, many vehicles are traveling on the road, and other vehicles are usually traveling in front of the own vehicle with an appropriate inter-vehicle distance. Therefore, the white line is often hidden by the preceding vehicle, particularly when the road is curved. For this reason, there is a possibility that a place other than the regular position may be detected as a white line, and the subsequent search area itself may deviate from the regular white line position.

Therefore, it is recognized whether or not another vehicle is running ahead of the own vehicle. If the white line candidate detection point in the white line search area is within the recognition area of the preceding vehicle, Cancel the detection point and make it undetected,
It is preferable to continue the subsequent processing. Therefore, a processing method after recognizing whether there is another vehicle (or an obstacle) ahead of the own vehicle and recognizing that there is another vehicle (or an obstacle) will be described.

When recognizing a target object ahead of the host vehicle, in particular, the front vehicle, the front vehicle is recognized by paying attention to the fact that the lower part of the vehicle is dark due to shadows and the tires are black. is there. In FIG. 1, an image processing device 2 and a computer 3 process a monochrome image input from a monochrome camera 1 by a recognition logic as shown in FIG. That is, first, a black portion in the captured monochrome image is expanded. (Step S15). Next, the monochrome image is binarized to be divided into a black region and a white region (step S16), and noise such as small points is removed (step S17). Then, the vehicle-likeness of the black area determined after the noise is removed is checked, and the parts other than the vehicle are removed. That is, it is checked whether or not the black area is a vehicle (step S1).
8).

Next, a method for recognizing a vehicle by the recognition logic of FIG. 10 will be described. For example, suppose that the scene in front of the vehicle viewed from the driver's seat when the vehicle is running looks as shown in FIG. The monochrome camera 1 captures the same scenery as in FIG. The monochrome camera 1 outputs an image signal corresponding to the captured monochrome image and adds it to the signal processing device 2. In the figure, dark portions are hatched, and the pitches of the hatched portions are finer in darker portions. In addition, the lower part of the front vehicle is darkened by shadows, and the tires are black, so that they are darker than the other parts and are completely black.

The image processing apparatus 2 expands a black portion (including a gray portion) in the image of FIG. 11 to convert a hole (white portion) inside the black portion and a nearby black region into one region. Summarize (FIG. 12). The expansion of the black region is expressed by E = MIN (A, B, C, and E) in the region (matrix) formed by 9 pixels of 3 × 3 A to I in the original image as shown in FIG. D, E, F, G, H, I) The processing of (24) is sequentially performed to shrink a bright portion (area).

That is, focusing on the pixel E at the center (middle) of the nine pixels A to I, the pixels A to D and F to I around the pixel E If there is a pixel having a luminance value smaller (lower) than the luminance value, the pixel of E is also adjusted to the luminance value of the smallest pixel. Since the above equation includes the pixel E itself, when the luminance value of the pixel of the E is the smallest, the luminance value does not change. Then, the pixel brought to the center of the matrix is moved one by one (scanned), and the luminance value of the center pixel is set sequentially. For example, FIG.
When the monochrome image shown in (1) is composed of 512 × 448 pixels, the scanning is performed for all these pixels. As a result, as shown in FIG. 12, the black area becomes larger as a whole. That is, the black area expands. FIG.
After expanding the black area as shown in, binarization is performed,
An image of a white area and a black area as shown in FIG. 14 is obtained.

Next, noise removal of the binarized image is performed. After the binarized black area is expanded with a circle having a predetermined radius r1 (for example, a radius having a size of 5 pixels), the area is further reduced with the circle and returned to the original size. This process is referred to as “process 1”. In the expansion of the black region, a circle having a radius r1 is rolled on a contour (outer periphery) forming the black region, and a locus drawn by the circle is set as a new black region.
As a result, as described above, the holes (bright portions) inside the black region and the nearby black regions can be combined into one. Next, the expanded black area is reduced by a circle having the same radius r1. In this reduction, the circle having the radius r1 is rolled inside the outline forming the expanded black region in terms of image, and the inside of the drawn locus is set as a new black region. Thus, the black area has the original size.

By the way, when the holes inside the black area and the adjacent black areas are combined into one by the above-described processing 1, even objects that are not desired to be combined into one rarely are combined into one. Need to be restored (divided). Therefore, the black area after the processing 1 is reduced by a circle having a predetermined radius r2 (for example, a radius having a size of three pixels).
Inflate again with the circle of the radius and return to the original size. still,
This processing is referred to as “processing 2”. This processing 2 is similar in image to the processing 1 described above.

Next, the processing 2 is described with reference to FIGS.
8 will be described more specifically. The binarized image shown in FIG. 15 is for explaining the noise removal processing and is different from the binarized image shown in FIG. In the image shown in FIG. 15, four black areas 20 to 23
Are different objects that you do not want to be put together. However, when the processing 1 is performed on the black areas 20 to 23, as shown in FIG.
1, 22, and 23 are connected by the connecting portions 24, 25. Therefore, when the black regions 20 to 23 are reduced by the circle having the radius r2 as described above, the connecting portions 2 to 23 shown in FIG.
4, 25 are removed and the black areas 20 and 21, 22 and 2
3 is cut off. Next, when the separated black regions 20 to 23 are expanded again with the circle having the radius r2,
As shown in FIG. 18, the size returns to the original size with the black regions 20 to 23 separated.

The image of FIG. 19 shows a black region after the above-described noise removal processing has been performed on the binarized images 26 to 30 shown in FIG. Then, the likeness of the vehicle in the black area of the image shown in FIG. 19 is checked, and other than the vehicle is removed.
This vehicle-likeness check is performed according to the procedure shown in FIG. The vehicle-likeness is checked by the most efficient method. First, it is determined whether the position of the center of gravity of the black area remaining after the removal is within the set range (step S2).
1) When it is not within the set range, it is determined that the black area is not a vehicle, and the subsequent determination work is terminated. The front vehicle seen from the normal driver's seat can be seen substantially at the center of the screen or slightly below the center as shown in FIG. Therefore, the center of gravity of the black area representing the lower part of the preceding vehicle is located at the center or in a range slightly below the center. Therefore,
Black area 2 in black areas 26 to 30 in FIG.
6 and 27 are determined not to be a vehicle, and the black areas 28
30 are determined to be vehicle-like. The setting range is
In consideration of the slope, pitching of the own vehicle, and the like, the portion may be slightly lower than the center of the screen and lower.

In step S21, black areas 28 to 3
When it is determined that the center of gravity of 0 is within the set range, it is determined whether or not the area of the black region is within the set range (step S22). That is, when the black areas (areas) 28 to 30 are smaller than the area set assuming the normal inter-vehicle distance between the preceding vehicle and the host vehicle, it is determined that the vehicle is obviously not a car. In step S22, the black areas 28 to
If it is determined that 30 is likely to be a vehicle, these black areas 2
It is determined whether or not the uppermost coordinates of 8 to 30 are equal to or smaller than a set value (step S23). For example, as shown in FIG. 17, the coordinates of the uppermost portions 20a and 20b of the black region 20, which is substantially V-shaped upward from the approximate center of the screen, are located at the upper center and upper right of the screen. is not.

If it is determined in step S23 that the vehicle is likely to be a vehicle, it is next determined whether the ratio of the length and width of the black area is within the set range (step S24). The size of the vehicle is such that the aspect ratio is within a certain range. In the case of a normal vehicle as shown in FIG. 11, the size of the vehicle is binarized, and the black area after noise removal includes the lower part of the vehicle. It is the part of the tire, which is horizontally long. Of course, depending on the vehicle, the vertical and horizontal directions may be the same or vertical. For example, in the case of a vehicle equipped with a black hood or a cover, the vehicle may be vertically long.
Therefore, the aspect ratio is set to a range of about 1: 1. When the ratio is within this range, the vehicle is recognized as a vehicle. FIG. 1
It is determined that the black areas 28 and 29 on the screen 9 are likely to be vehicles (step S25). In this way, the computer 3 checks the uniqueness of the vehicle from the four items from step S21 to step S24 (step S25).
As shown at 0, the preceding vehicles 31, 32 are displayed on the screen.
The same applies to the case where an obstacle itself on the road is detected.

If the white line candidate detection point is within the recognition area of the preceding vehicle, the detection point is canceled, and the subsequent processing is performed as undetected. At this time, there are two methods for canceling: (1) when the white line candidate detection point is within the front vehicle area, and (2) when the white line search area is over the front vehicle area. First, the processing method (1) when the white line candidate detection point is in the front vehicle area will be described with reference to the flowchart in FIG.

When the computer 3 recognizes the preceding vehicle as described above (step S30), the computer 3 measures the luminance value of the search area An of the white line 6 in the area of the recognized preceding vehicle (step S31). It is determined whether or not the luminance value is equal to or greater than a threshold value (formula (1) or (2)) (step S32). The computer 3 searches the search area A
If the luminance value of n is equal to or greater than the threshold value, it is determined whether or not the white line candidate point in the search area An is within the area of the preceding vehicle (step S33). Is determined as a white line candidate point, and processing at the time of detection is performed (step S34). Further, the computer 3 determines each of the detection points when the luminance value of the search area An does not reach the threshold value in step S32 or when the white line candidate point in the search area An is in the area of the preceding vehicle in step S33. Processing (cancellation) is performed as undetected (step S35).

Next, the processing method when the white line search area of (2) overlaps the front vehicle area will be described with reference to the flowchart of FIG. After recognizing that another vehicle is running ahead of the host vehicle (step S40), the computer 3 sets a white line search area An (step S41), and sets the search area An to the area of the preceding vehicle. It is determined whether or not the vehicle is hanging (step S42). If the vehicle is not in the area of the preceding vehicle, the above-described normal processing is performed (step S43), and it is determined whether or not there is a white line detection candidate point (step S44). ). And
If there is a white line candidate point, the search area An is set as a search area when a white line is detected (step S45). When the search area An overlaps the front vehicle area in step S42, or when the white line candidate point is not detected in step S44,
It is determined that the white line has not been detected, and the search area An is processed (canceled) as the search area at the time of not being detected (step S).
46).

In this way, when the white line is hidden by the preceding vehicle, the detection points in the search area are canceled, thereby preventing erroneous detection of the white line due to the shift of the search area from the regular white line position. This makes it possible to accurately recognize a white line over a long distance.

[0054]

As described above, according to the present invention, according to the first aspect, a search area having a predetermined length from a predetermined position on a screen of a road surface in front of the imaged vehicle is formed along the road on the road surface. Along the white line being drawn and the car of the white line
In the search direction, measuring the brightness value in each search area, discriminating the road surface and the white line, detecting the white line, and recognizing the presence or absence of a preceding vehicle from the captured image. By processing the white line as undetected when the search area is within the area of the preceding vehicle, erroneous detection of the white line by the preceding vehicle can be prevented, and the white line can be accurately recognized up to a long distance. Become.

[0055] In claim 2, when recognizing the preceding vehicle, when the luminance value of the search area is lower than a predetermined value, or brightness values of <br/> search area is equal to or higher than a predetermined value, and the service < The search area is canceled as a white line not detected at any time when a white line candidate point in the area is within the area of the preceding vehicle, thereby preventing erroneous detection of a white line by the preceding vehicle. Can be. In claim 3, when a preceding vehicle is recognized, a white line search area is set, and when the search area overlaps the area of the preceding vehicle, the search area is canceled as white line undetected,
It is possible to prevent erroneous detection of a white line by a preceding vehicle.

[0056] Incidentally, in claim 1 or 2, discrimination between the road surface and white lines, a threshold value higher by a predetermined value than the luminance value of the road surface, to determine the portion of the high brightness value than the threshold value and the white line it leads to an effect such as that that Do is possible to reliably determine the white lines and the road surface. Further, the claims
In 1, the road surface brightness value is an average value of the brightness values at a plurality of locations on the road surface immediately before the vehicle, thereby providing an effect that a more accurate road surface brightness value can be obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram of a system configuration of a white line detection device for implementing a white line detection method according to the present invention.

FIG. 2 is a screen showing a right-curve road and a left-side white line detection search area obtained by imaging the front of a driver's seat by the camera of FIG. 1;

FIG. 3 is a screen showing a left curve road and a left white line detection search area obtained by imaging the front of a driver's seat by the camera of FIG. 1;

FIG. 4 is an explanatory diagram of a case where a white line is detected by inclining a search area in accordance with a curve of the white line on the screen of FIG. 3;

FIG. 5 is an explanatory diagram showing a method for setting a search area for white line detection on the screen of FIG. 3;

FIG. 6 is an explanatory diagram showing a method for setting a search area when a white line has not been detected in FIG. 5;

FIG. 7 is a screen showing a relationship between a left white line and a search area when a left curve of a road is tighter.

8 is an explanatory diagram showing a method for setting a search area for white line detection on the screen of FIG. 7;

FIG. 9 is a flowchart illustrating a procedure of a white line detection method according to the present invention.

FIG. 10 is a flowchart showing a logic for recognizing a vehicle traveling ahead of the host vehicle.

FIG. 11 is a screen showing a forward image viewed from the driver's seat by the camera of FIG. 1;

FIG. 12 is a screen showing a state where a black area of the screen of FIG. 11 is expanded.

FIG. 13 is an explanatory diagram showing a means for expanding a black area on the screen of FIG. 11;

14 is a screen showing a state where the screen of FIG. 11 is binarized.

FIG. 15 is a screen for performing noise removal processing on a binarized screen.

16 is a screen showing a state where a black area of the screen of FIG. 15 is expanded.

17 is a screen showing a state where a black area of the screen of FIG. 16 is reduced.

18 is a screen showing a state where a black area of the screen of FIG. 17 is expanded to its original size.

FIG. 19 is a screen after noise removal processing has been performed on the binarized image of FIG. 14;

FIG. 20 is a screen displaying a preceding vehicle from the image of FIG. 19;

FIG. 21 is a flowchart showing a procedure for checking vehicle-likeness from the image shown in FIG. 19 after noise removal processing is performed.

FIG. 22 is a flowchart illustrating a procedure of a white line detection processing method when a white line is hidden by a vehicle ahead.

FIG. 23 is a flowchart illustrating a procedure of another white line detection processing method when a white line is hidden by a preceding vehicle.

[Explanation of symbols]

 Reference Signs List 1 camera 2 image processing device 3 computer

Continuation of front page (58) Fields investigated (Int.Cl. ⁷ , DB name) G06T ⁷ /00-7/60 B60R 21/00 G01C 15/00 G05D 1/02 G06T 1/00

Claims (3)

(57) [Claims] 1. A search area having a predetermined length from a predetermined position on a screen of a road surface in front of a vehicle, which is imaged, is searched for along a white line drawn along the road on the road surface and the power of the white line.
In the search direction, measure the brightness value in each search area, determine the road surface and the white line to detect the white line, and recognize the presence or absence of the preceding vehicle from the captured image. When the search area is within the area of the vehicle in front, a white line is detected as not detected and processed. 2. A search area when a preceding vehicle is recognized.
When the luminance value is lower than the predetermined value, or brightness values of the search area is equal to or higher than a predetermined value, and the white line candidate points in the search area is either when in the area of the front vehicle
2. The white line detection method according to claim 1, wherein the search area is canceled when white lines are not detected. 3. A white line search area is set when the preceding vehicle is recognized, and the search area is canceled as a white line not detected when the search area overlaps the area of the preceding vehicle. The white line detection method according to claim 1.