JP3345995B2 - Road white line detector - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a road white line detecting device, and more particularly to a road white line detecting device for detecting a white line drawn on a road from an image captured by an image sensor.

[0002]

2. Description of the Related Art Conventionally, a road white line detecting device using an image sensor has been proposed as an elemental technology for realizing automatic driving of a vehicle. For example, the applicant has filed Japanese Patent Application No.
In -265655, a captured image obtained by capturing an image of a road ahead of a vehicle by a camera is scanned with a search window of a predetermined size, and an area having the highest correlation with a template set in advance so as to conform to a white line shape. There is disclosed a road white line detection device that recognizes as a white line.

In this apparatus, when a white line is picked up by a camera arranged in the center of the vehicle, the white line on the left side of the vehicle is normally raised to the right, and the white line on the left side of the vehicle is picked up to the left, and each image is picked up with higher brightness than the asphalt portion. In view of the above, a template in which a white line is drawn in advance to the right, to the left, etc. is facilitated, and a correlation between an image cut out as a search window from a captured image and the template is calculated, and a scanning position where the highest correlation is obtained. Is grasped as a white line area.

[0004] By the way, when the traveling road approaches a curve, the white line at a position relatively far from the vehicle may be such that the white line on the left side of the vehicle rises to the left or the white line on the right side of the vehicle rises to the right. On the other hand, in order to use the result of white line detection in vehicle control such as automatic control, it is necessary to detect a relatively distant white line position. For this reason, practical white line detection cannot be realized if the template referred to in white line detection is a fixed template.

Therefore, in the above apparatus, a white line is detected by a fixed template immediately before a vehicle where the white line shape is relatively stable. Thus, a method of sequentially detecting distant white lines is adopted. In this case, the shape of the template used for white line detection does not largely differ from the actual white line shape depending on the curve of the road, and the white line position can be detected at a position separated from the vehicle.

[0006]

However, in the technique of realizing white line detection at a position separated by a predetermined distance while sequentially selecting a template from immediately before a vehicle as in the above-described conventional apparatus, a white line at a position to be monitored is detected. It is necessary to perform a large amount of arithmetic processing before it can be detected, and it is not possible to meet the demand for rapid processing.

In addition, the above-described conventional apparatus performs white line detection by obtaining a correlation between a preset template and an actual captured image. Therefore, the white line position may be erroneously detected depending on the luminance of the captured image. Since the correlation value between the template and the captured image is obtained by accumulating the absolute value of the luminance difference between the corresponding pixels of the two images, for example, when the average luminance of the captured image changes between daytime and nighttime, the luminance change is considered. This is because an appropriate matching is not always obtained in the relationship with the fixed template set beforehand.

In this case, it is possible to prepare a fixed template to be referred to for several types of average luminance and to perform processing for selecting an optimal template in relation to the luminance of the captured image, but the storage capacity of the template is increased. However, the processing time required for template selection is prolonged, which is not preferable.

[0009] It is also possible to binarize the obtained picked-up image by luminance and perform gradation conversion so that the asphalt portion and the white line portion have a reference luminance, but such processing is generally complicated. The calculation involves complicated processing contents and a prolonged processing time, which is not preferable.

[0010] The present invention has been made in view of the above points, and ascertains an asphalt area immediately before a vehicle as a low-luminance area, and scans an area to be monitored horizontally and has the highest correlation with the low-luminance area. An object of the present invention is to provide a road white line detection device that can solve the above-described problem by recognizing a low region as a white line region.

[0011]

FIG. 1 is a block diagram showing the principle of a road white line detecting apparatus for achieving the above object. That is, as shown in FIG. 1, the above object is to capture an image of the road ahead of the moving object by the image sensor M1, set a search window in the obtained captured image, scan the captured image, and scan the image in the search window. In a road white line detection device that detects a white line on a road based on a correlation between the white line on a road and a predetermined template, a part of a low-luminance region that is immediately in front of a moving object and is probably predicted to be not a white line is used as the template. A low-luminance template setting means M2, a correlation calculating means M3 for obtaining a correlation between the template set by the low-luminance template setting means M2 and an image in a search window for scanning the captured image, This is achieved by a road white line detection device having white line recognition means M4 for recognizing a region where a low correlation is detected as a white line.

Further, in the road white line detecting device having the above-mentioned structure, the correlation calculating means M3 determines a region farther than the immediately preceding portion of the moving object, which is adopted as the low-luminance region by the low-luminance template setting means M2, as a scanning region by a search window. This configuration is effective for detecting a distant white line.

Further, when a white line is detected on both sides of the road by the white line recognizing means M4, a low luminance area setting means M5 for setting the vicinity of the center of these two white lines as a low luminance area is provided. Template setting means M2
However, at the beginning of the execution of the road white line detection process, a part of the low luminance area immediately before the moving body is partially replaced by the low luminance area setting means M.
When a low-luminance area is newly set according to 5, a part of the low-luminance area is adopted as a template, and the correlation calculating means M3 is used until the scanning area by the search window reaches the monitoring area in front of the moving object. The configuration in which the scanning area of the search window is updated by a predetermined distance farther than the area adopted as the low-luminance area by the low-luminance template setting means M2 is a white line when the luminance of the monitoring area set far and the luminance immediately before the vehicle are different. Effective for detection.

[0014]

In the road white line detecting device according to the present invention, the low-luminance template setting means M2 includes the image sensor M
A part of the region immediately before the vehicle in the captured image of No. 1 and which is probably not a white line is adopted as a template.

In this case, immediately before the moving body, the white line on the road is always located on both sides of the moving body regardless of whether the traveling path is a straight road or a curve. If it is highly probable that such a region is a low-brightness region, the set template will surely capture the brightness of the asphalt surface.

Accordingly, the correlation calculation means M3 performs a correlation calculation between the brightness of the asphalt surface and the comparison image in the search window, and when the comparison image is cut out from the asphalt surface in the search window. A high correlation is obtained when the comparison image is cut out from a portion occupied by the white line in the search window, and a relatively low correlation is obtained.

For this reason, the white line recognizing means M4 recognizes an area having the lowest correlation as a white line area,
The white line position in the scanning area of the search window is accurately detected without being affected by the average luminance of the road surface.

As described above, in the configuration in which the state of the asphalt surface is copied as a template, white lines can be detected simply by searching for an area having a low correlation, and the white line shape in the comparison image cut out from the search window is problematic. No.

Therefore, when the correlation calculation means M3 sets the scanning area of the search window to a predetermined monitoring area far from the moving object, the correlation calculation with the template set by the low-luminance template setting means M2 allows the correlation area to be set in the monitoring area. Are easily detected.

When the white line recognizing means M4 detects white lines on both sides of the road within the scanning area of the search window, it is highly likely that the middle of the white lines is asphalt.
At this time, if the low-luminance template setting means M2 captures the area as a low-luminance area and copies this area as a template, the template is constantly updated in the scanning area of the search window.

Therefore, in addition to the above configuration, when the correlation calculating means M3 sets a scanning area that is farther from the area copied as a template until the scanning area of the search window reaches a predetermined monitoring area, In addition, the template does not deviate from the asphalt area during the movement of the scanning area, and the template is finally copied from the monitoring area.

In this case, since the template is copied from the low-luminance area in the monitoring area, even if the brightness of the part immediately before the moving object and the brightness of the distant part set as the monitoring area are temporarily different. , The white line detection processing is executed without being affected by this.

[0023]

FIG. 2 is a block diagram showing an entire configuration of a road white line detecting apparatus according to an embodiment of the present invention. In the figure, a camera 10 corresponds to the above-described image sensor M1,
It is connected to the image frame memory 14 via the / D converter 12. Thereby, the image for one screen obtained by the camera 10 is stored in the image frame memory 14 as digital data.

The image frame memory 14 is connected to a CPU 16 that implements the low-luminance template setting means M2, the white line recognition means M4, and the low-luminance area setting means M5. The CPU 16 reads out image data of a predetermined portion of the image frame memory 14 and performs various arithmetic processes to realize these functions.

The CPU 16 is connected to a correlation operation device 18 corresponding to the above-mentioned correlation operation means M3. The correlation operation device 18 is supplied with the data relating to the template set by the CPU 16 based on the image data in the image frame memory 14 and the image data in the search window set by the CPU 16. And the correlation operation device 1
8 performs a correlation operation by template matching based on these data.

Hereinafter, a method of correlation calculation by template matching will be briefly described with reference to FIG.

FIG. 3A shows an m × n pixel template used for template matching, and FIG. 3B shows a captured image in which an image in the template is partially captured. When performing template matching, first, FIG.
A search window (S) is roughly set in the captured image shown in (B). Then, an image having the same size as the template is extracted from the search window as a comparison image (C), and the sum of absolute values of the luminance differences of the corresponding pixels is obtained as a correlation value D.

In this case, as shown in FIGS. 3A and 3B, when the image in the template and the comparison image do not match, the correlation value D is large, and the image in the template and the comparison image are different. If they completely match, the correlation value D = "0"
Becomes

Therefore, when the comparison image is set by shifting one pixel at a time in the set search window and the correlation value D with each template is obtained, the position where the correlation value D becomes the minimum is the most correlated. It is a high position, and the position can be specified as a position where the image in the template exists.

As described above, the correlation operation based on the template matching is an effective method for detecting an image similar to the image in the template in the search window.
The road white line detection device of the present embodiment is a device that detects the position of a white line drawn on a road by using such a principle.

That is, the CPU 16 is provided with a correlation operation device 18.
The white line position is detected on the basis of the correlation calculation result, and is output to a predetermined vehicle control device (not shown) via the white line position output device 20. In other words, the vehicle has steering, accelerator,
A vehicle control device that controls brakes etc. is installed,
Automatic control is performed based on the information on the white line position thus supplied.

A CRT 24 is connected to the output side of the image frame memory 14 via a D / A converter 22, and the images stored in the image frame memory 14 are sequentially displayed on the CRT 24.

When a white line position is detected by template matching, generally, a template in which the asphalt portion has low luminance and the white line portion has high luminance as shown in FIG. 4A, and as shown in FIG. A method of comparing a comparison image cut out from the obtained captured image is employed.

In this case, the result of the white line detection is usually used for control of automatic driving of the vehicle and the like, and in order to execute practical white line detection, a time for reflecting the detected white line position in the control is secured. In this sense, it is necessary to detect a white line at a position separated from the vehicle to some extent. On the other hand, FIG.
As shown in (2), the tangential direction of the white line at a position separated from the vehicle changes due to the influence of the curve of the road, and when a comparative image to be compared with the template is cut out, the shape of the white line greatly changes in the comparative image.

For this reason, conventionally, a white line is detected using a fixed template at a vehicle near point where the white line shape is relatively stable, a comparative image recognized as a white line is used as a next template, and the search window is scanned. As described above, a method is employed in which the area is gradually separated away from the vehicle toward a desired monitoring area, and as a result, it takes a long time to detect a white line.

Further, as shown in FIG. 4 (C), the average luminance of the captured image differs depending on the environment where the vehicle is placed.
In the correlation calculation with the fixed template in which the change in luminance is not considered, a smaller correlation value D is obtained for the comparison image cut out from the position not including the white line than the comparison image cut out at the position including the white line. That is, a case where a high correlation is obtained may occur.

Therefore, the road white line detecting device of this embodiment is
From the captured image stored in the image frame memory 14 as shown in FIG. 5, the road surface immediately before the center of the vehicle, which is likely to be asphalt (in FIG. 5A, the coordinates (X
t, Yt)) as a template, and then a comparative image is cut out from a scanning area defined as a monitoring area at a position of Y coordinate = Ya, and a correlation value D calculation with the template is performed.

In this case, a portion having a low correlation with the state of asphalt captured as a template, that is, a portion where a white line occupies a large proportion in the comparative image has a higher correlation value D.
Is calculated, and the CPU 16 executes the processing shown in FIG.
As a result, a correlation value D curve having a peak at a portion corresponding to the position of the white line is obtained.

Therefore, if the X coordinate of the peak position is detected, the coordinate values XL and XR can be grasped as white line positions. Then, the XL and XR are not affected by the change in the average luminance of the captured image, and the possibility of erroneously detecting the position of the white line is suppressed to be extremely small.

In this case, the vicinity of the center between XL and XR detected as the position of the white line has a high degree of asphalt. For this reason, in the present embodiment, FIG.
As shown in (c), coordinates ((XR-XL) / 2, Ya)
It is decided to copy the template from a certain point. As a result, even when the road is curved, it is possible to obtain the template from within the monitoring area set far from the vehicle while ensuring that the area copied to the template is always the asphalt area. The detection of the white line along is realized.

FIG. 6 shows a CPU 16 for realizing the above functions.
2 shows a flowchart of a main routine executed by the user. As shown in the figure, when this routine is started, first, in step 100, initial processing is executed. This initial process is a process realized as described later by an initial routine whose flowchart is shown in FIG. 7, and is a position separated by a predetermined distance from the vehicle (a position of Y = Ya).
The low-luminance area in which the template is to be copied in the monitoring area set in is set.

Next, in step 200, white line detection processing is performed based on a template in which a part of the monitoring area is copied. This white line detection processing is specifically realized by a white line detection routine shown in the flowchart of FIG. 8, and a process of recognizing a position having the lowest correlation with the template in the monitoring area as a white line position is performed.

After completing these processes, step 300
Then, it is determined whether the area recognized as the white line in the above step 200 is truly a white line. This white line determination processing is realized by a white line determination routine whose flowchart is shown in FIG. 9, and as described later, determines whether the two positions recognized as white lines are in an appropriate positional relationship with the lane width. That determination is made.

If the white line has been properly detected, it is determined that the template copied from the monitoring area is functioning properly.
00 is repeatedly executed to continuously execute white line detection.
On the other hand, if it is determined that the white line has been missed, the initial process shown in step 100 is executed again in order to surely copy the low luminance area to the template.

The specific contents of steps 100 to 300 will be described below.

When the initial processing routine (FIG. 7) corresponding to step 100 is started, first, at step 10
In step 2, a template is set. In this case, since there is no other effective means other than assuming an asphalt highly dumb region as a low-brightness region and using it as a template, in the image stored in the image frame memory 14, the coordinates ( Xt, Yt) (see FIG. 5A)
An image of m × n pixels is cut out and stored in a memory. still,
In this routine, this step 102 corresponds to the low-luminance template setting means M2 described above.

After the completion of this processing, the routine proceeds to step 104, where initial values are set to detect the position of the white line drawn on the left side of the vehicle. Here, a configuration is adopted in which the scanning area of the search window is directly set to Y = Ya corresponding to a desired monitoring area, and the upper left coordinate (x, y) of the comparison image composed of m × n pixels is represented by (Xf / Y). 2, Ya). Here, in this embodiment, the captured image is composed of Xf × Yf pixels, and Xf / 2 corresponds to the center position of the captured image.

In this step 104, along with the above processing, Dbuf and Dbuf storing the maximum value of the correlation values Dx and y between the template to be calculated later and the comparison image having the coordinates (x, y) at the upper left are detected. The Xbuf storing the X coordinate x of the set position is cleared to “0”, and the counter i incremented under a predetermined condition is set to “1”.
When a white line is detected in the comparison image, a process of setting “1” to a flag “flag” and setting a flag dir to “0” to indicate that a white line on the left side of the vehicle is being detected is executed. I do.

Step 106 is a step for realizing the above-described correlation operation means M3 together with the correlation operation device 18, and performs processing for calculating and storing the correlation values Dx, y by template matching.

Step 108 is a step for determining whether or not flag is "0". In accordance with the principle described later, it is determined that a white line has been detected in the comparison image cut out from the search window, and flag = "1". Until "", it is determined that the condition is satisfied.

In this case, step 110 is executed. Step 11 is a step of determining whether or not the correlation value Dx, y is equal to or greater than a predetermined determination value Kth.
That is, the asphalt area immediately before the center of the vehicle is used as the template, and the correlation values Dx, y increase as the proportion of the white line in the comparison image increases.

Therefore, by setting Kth to an appropriate value, it can be determined whether or not a white line exists in the comparison screen from the establishment of Dx, y ≧ Kth. From this point of view, a process of determining whether a white line has been detected as a result of scanning the comparison image in the search window is executed.

Steps 112 and 114 are skipped while the condition is not satisfied. On the other hand, if the condition is determined to be satisfied, the correlation value Dx, y at that time is set to Dbuf at step 112, and Store the upper left coordinates (x, y) of the comparison image of Xbuf in step 1
At step 14, a process of setting flag = "1" is executed. Therefore, when this step is subsequently started, it is determined that the condition of step 108 is not satisfied.

Step 116 is a step for determining which side of the vehicle is executing white line detection. The determination is made based on whether or not the flag dir is "0". This time, since the value of dir has not been changed after the setting in step 104, it is determined that dir = "0", and the process of step 118 is executed thereafter.

Step 118 is an X coordinate value x of the comparison image.
Is a step of setting In this routine, the comparison coordinates are scanned one pixel at a time when performing template matching. In this step 118, x-1 obtained by subtracting one pixel from the current X coordinate value x is set as a new x.

When the above processing is completed, step 10
6, the calculation and storage of the correlation values Dx, y are performed again. As described above, since "1" is set in the flag this time, the process proceeds to step 120, and it is determined whether or not the counter i has reached 10.

This counter sets the first correlation value D calculation when it is determined in step 110 that Dx, y ≧ Kth is satisfied for the first time, that is, when it is recognized that a white line is detected for the first time in the comparative image. This is a counter that counts how many times the correlation value D has been calculated.

In this routine, the white line drawn on the road is 5 in the captured image due to the relationship with the pixel size of the camera 10.
Scanning of the comparative image image is performed until i = 10 because the size of the comparison image falls within 6 pixels.
Is a step of substantially determining the continuity of scanning.

By performing such processing, for example, even when a high-luminance building or the like is imaged further outside the white line in the captured image, the building or the like is not erroneously recognized as a white line. As a result, the white line detection accuracy is improved.

Then, in step 120, i>
If it is determined that 10 is not established, step 12
After incrementing i in step 2, the process proceeds to step 124,
It is determined whether the current correlation value Dx, y is larger than the value of Dbuf storing the maximum value up to the previous time.

If Dx, y> Dbuf is satisfied, Dbuf should be updated to a new correlation value Dx, y. In step 126, Dbuf and Xbuf are updated. On the other hand, if Dx, y> Dbuf is not satisfied, it is not necessary to update Dbuf. In this case, step 126 is jumped to step 116. As a result, Dbuf always has the calculated correlation value D.
The maximum value of x and y is stored, and the X coordinate value x of the comparative image indicating the maximum correlation value Dx, y is stored in Xbuf.

On the other hand, in step 120, i> 1
If 0 is established, it is determined that the detection of the left white line has been completed, and the process proceeds to step 128 to detect the right white line.
In step 128, it is determined whether or not "1" is set in dir. If this time is immediately after the end of the detection of the left white line, it is determined that the condition is not satisfied because “0” is still set in dir, and step 130 is executed thereafter.

In step 130, the white line position output device 20
, The value of Xbuf is set as the left white line position XL, and
In this step, the value of buf is output as the correlation value DL of the left white line.

After the processing in step 130 is completed, the flow advances to step 132 to set an initial value for detecting the right white line. In this step, the coordinates (x, y) of the comparison image at the start of detection are set to (Xf / 2,
Ya), Dbuf = "0", Xbuf = "0", i
= "1" and flag = "0", while dir = "1" to indicate that the right white line is detected.

Thereafter, steps 106 to 116 are executed in the same manner as the left white line detection processing described above. However, since dir = "1" this time, dir is set to dir in step 116.
= “0” is determined not to be established. Therefore, when the right white line is detected, step 134 is executed following step 116, and the X-coordinate value x is incremented by one pixel to realize scanning of the comparison image.

Then, on the right side of the vehicle, Dx, y ≧ K
The correlation value Dx, y that satisfies th is calculated, and then i
The processes of steps 106 to 126 are repeatedly executed until> 10 is satisfied, and step 128 is executed on the condition that i> 10 is satisfied.

This time, since dir = “1”, the condition of step 128 is satisfied. Thereafter, in step 136, the value of Xbuf is output as the left white line position XL, and the value of Dbuf is output as the correlation value DL of the left white line. To end this routine.

As described above, the initial processing routine shown in FIG. 7 is a routine for detecting the left and right white line positions in the monitoring area (Y = Ya) based on the template copied from the area immediately before the center of the vehicle. In this sense,
In FIG. 7, steps 124, 126, 130, and 136 for determining the values of XL and XR representing the position of the white line are steps for implementing the white line recognition means M4 described above in this embodiment.

When the white line is detected by the above-described initial processing routine, the position of the white line is grasped as the position having the worst correlation, and is not affected by the change in the average luminance of the captured image. In this sense, the CPU 16
However, even in the configuration in which the initial processing routine is repeatedly executed, the white line detection processing that does not cause erroneous detection and has a higher processing speed than the conventional method can be realized.

However, for example, in the situation immediately before the vehicle enters the tunnel, Y = Y set as the monitoring area
The brightness of the road surface may be different between the region a and the portion immediately before the center of the vehicle set to be cut out as a template. In this case, the average luminance of the template and the average luminance of the asphalt in the monitoring area are clearly different, and may cause erroneous detection of the position of the white line.

The white line detection routine shown in FIG. 8 is executed by the CPU 16 in order to eliminate such an adverse effect and improve the white line detection accuracy.
Is a routine to be executed. Note that FIG.
Steps that execute the same processes as those described in (1) are given the same reference numerals, and descriptions thereof are omitted.

That is, in this routine, which is executed immediately after startup, in step 202, the coordinates (XL + X) between XL and XR detected in the initial processing routine are described.
R) / 2 and Ya corresponding to the Y coordinate of the monitoring area are stored in the memory as a template with an image of m × n pixels having the X coordinate and the Y coordinate at the upper right, respectively.
No. 4,206, the feature is that the initial X coordinate of the comparison image for the detection of the left and right white lines is set to (XL + XR) / 2.

In this sense, in the present embodiment, the above-mentioned step 202 which has the function of changing the set coordinates of the template is performed by the low-luminance template setting means M described above.
2 and the low basic area setting means M5.

The processing executed to detect the position of the white line with respect to these initial settings will be described below with reference to FIG.
Is exactly the same as the initial routine shown in FIG. In this case, the image cut out as the comparison image and the image cut out as the template are both present in the monitoring area of Y = Ya, and it can be assumed that both environments are practically the same. There is no room for the occurrence of white line detection, and white line detection can be performed with high accuracy.

In this embodiment, in the initial processing routine, a template is cut out from the area immediately before the center of the vehicle, and a white line is detected at a position (Y = Ya) directly separated from the vehicle by a predetermined distance using the template. However, a configuration may be adopted in which processing is performed to gradually separate the scanning area of the search window from the position where the template is cut out to the monitoring area.

For example, the area immediately before the vehicle is set to Y = Yc, and Y
Assuming that b is an intermediate point between Ya and Yc, a template is initially cut out at a position of Y = Yc, a white line is detected at a position of Y = Yb based on the template, and XL and XR obtained as a result are represented by ( A new template with (XL + XR) / 2, Yb) as the upper left coordinate may be cut out, and an initial processing routine may be used to execute white line detection in the area of Y = Ya based on this template.

In this case, the white line detection in the Y = Ya region performed in the initial processing routine sets the initial coordinates to ((XL
+ XR) / 2, Ya), so that even if (Xf / 2, Ya) falls outside the two white lines due to the effect of the curve, appropriate White line detection can be realized.

FIG. 9 is a flowchart of a white line determination routine for determining whether the position detected as a white line by the white line detection routine shown in FIG. 8 is a white line drawn on a road. When a high-luminance object other than a white line is picked up in a captured image, the object may be recognized as a white line, and therefore, the processing is executed to improve the reliability of detection accuracy.

That is, when the routine shown in FIG. 9 is started, first, a calculation of the distance Wx = | XR-XL | Then, in step 304, the or Wx is within a predetermined range, i.e., in relation to the upper limit value W _H and the lower limit value W _L which defines by assuming a typical lane width of W _H> Wx> W _L See if the relationship holds.

If the above condition is satisfied, it is determined that the white line has been successfully detected, and the flow advances to step 306 to return to the white line detection process (step 200). On the other hand, if the above condition is not satisfied, it is determined that the white line has been missed, and the process proceeds to step 308, where the process returns to the initial process (step 100) to execute the white line detection from the beginning.

As described above, in the road white line detection device of the present embodiment, the template is cut out from immediately before the center of the vehicle at the beginning of the execution of the white line position detection processing, and finally, the template is cut from the vicinity of the center of the desired monitoring area. Since the cutout is performed to detect the white line, the detection accuracy of the white line position does not deteriorate due to the change in the average luminance of the captured image, and a high processing speed is realized even though the monitoring area is set to a distant place. It has the effect that it can be done.

In the above embodiment, the size of m × n pixels is set as the template and the comparison image. However, in view of the fact that it is sufficient to detect the position with the worst correlation as a white line, it is not always necessary. The height direction is not important, and a size of m × 1 pixel can be used as long as m pixels having substantially the same width as the white line width are secured. In this case, the processing time required for the correlation operation by the template matching is reduced, and further higher speed can be realized as compared with the above embodiment.

[0083]

As described above, according to the first aspect of the present invention, the template referred to by the correlation calculation means for correlating with the image in the search window is a copy of the low-luminance area immediately before the moving object. Yes, since the area having the lowest correlation with the template is configured as a white line area, the detection accuracy of the white line is not affected by the average luminance of the captured image. White lines drawn on the road can be detected with appropriate accuracy.

According to the second aspect of the present invention, while the low-luminance template setting means copies the portion immediately before the moving body as a template, the correlation calculating means scans the monitoring area search window which is separated from the moving body by a predetermined distance. Since the area is set as an area, it is possible to quickly and accurately detect a white line far away from the moving object while eliminating the influence of the average luminance of the captured image.

According to the third aspect of the present invention, the area set as the scan area of the search window by the correlation calculation means and the low-luminance area where the template is copied are gradually changed from the immediately preceding portion of the moving body to the desired monitoring area. Move away from the moving object. At this time, the low-luminance area is set near the center of the two detected white lines, and is always kept within the asphalt area and separated from the monitoring area.

Therefore, regardless of whether the road is a straight road or a curve, the template always captures a low-luminance area on the asphalt surface, and the road white line detection device of the present invention uses the low-luminance area on both sides of the low-luminance area. By performing a correlation operation with the template, white line detection can be performed.

For this reason, even when the average brightness of the road differs between the monitoring area and the area immediately before the moving body, for example, when the monitoring area enters a tunnel, the difference in the average brightness affects the detection accuracy of the white line. There is a feature that accurate and quick white line detection can be realized under a wide traveling condition without giving.

[Brief description of the drawings]

FIG. 1 is a principle configuration diagram of a road white line detection device according to the present invention.

FIG. 2 is a block diagram illustrating an overall configuration of a road white line detection device according to an embodiment of the present invention.

FIG. 3 is a diagram for explaining the principle of correlation calculation by template matching.

FIG. 4 is a diagram for explaining a road white line detection method using a conventional device.

FIG. 5 is a diagram for explaining a road white line detection method according to the present embodiment.

FIG. 6 is a flowchart of an example of a main routine executed in the embodiment.

FIG. 7 is a flowchart of an example of an initial processing routine executed in the embodiment.

FIG. 8 is a flowchart of an example of a white line detection processing routine executed in the present embodiment.

FIG. 9 is a flowchart of an example of a white line determination routine executed in the embodiment.

[Explanation of symbols]

 M1 Image sensor M2 Low brightness template setting means M3 Correlation calculation means M4 White line recognition means M5 Low brightness area setting means 10 Camera 14 Image frame memory 16 CPU 18 Correlation calculation device

──────────────────────────────────────────────────続 き Continued on the front page (58) Fields surveyed (Int. Cl. ⁷ , DB name) G08G 1/09 G01B 11/24 G01C 3/06 G05D 1/02 G06T 1/00

Claims (3)

(57) [Claims] An image of a road ahead of a moving object is captured by an image sensor, a search window is set in the obtained captured image, the captured image is scanned, and a search is performed based on a correlation between the image in the search window and a predetermined template. A road white line detecting apparatus for detecting a white line on a road by using a low-luminance template setting unit that employs, as the template, a part of a low-luminance area that is immediately in front of a moving object and is probably not a white line; Correlation calculating means for obtaining a correlation between the template set by the luminance template setting means and an image in a search window for scanning the captured image; and a white line for recognizing, as a white line, an area where the lowest correlation is detected by the correlation calculating means. A road white line detection device, comprising: a recognition unit. 2. The method according to claim 1, wherein the correlation calculation unit sets a region farther than a portion immediately before the moving object, which is adopted as the low luminance region by the low luminance template setting unit, as a scanning region by a search window. Road white line detector. 3. A low-luminance template, comprising: a low-luminance area setting unit configured to set a vicinity of the center of these two white lines as a low-luminance area when a white line is detected on both sides of a road by the white-line recognition unit; The setting means sets a part of the low-brightness area immediately before the moving body at the beginning of the execution of the road white line detection process, and thereafter, when a new low-brightness area is set by the low-brightness area setting means, A part is adopted as a template, and the correlation calculation means adopts a scan area by a search window as a low luminance area until the scan area by the search window reaches a monitoring area in front of the moving object. 2. The road white line detection device according to claim 1, wherein the road white line is updated a predetermined distance from the area.