JP2004240636A - White line detection device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a white line detection device detecting a white line with high accuracy by use of only an image of a road photographed from a vehicle while suppressing increase of a processing amount even if the image includes the white line and an object except the white line. <P>SOLUTION: In this white line detection device, a differential value of a luminance is calculated in a prescribed area of the image photographed from the vehicle to detect an edge part. Next, by referring to a distribution or the number of the detected edge parts, it is decided whether the edge parts are noise parts or not, and desired white line candidate point is extracted. Lastly, the white line drawn in a boundary of a lane on which the vehicle travels is detected from positions of a plurality of white line candidate points. Thus, the high-accuracy white line detection device with low mounting cost can be realized. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a white line detection device that suppresses erroneous recognition of a white line due to the presence of an object other than a white line and detects the white line with high accuracy in a vehicle traveling on a road on which a white line is drawn.
[0002]
[Prior art]
Conventional white line detection devices detect objects other than white lines, such as preceding vehicles, using a distance measuring device such as radar or stereo vision, and perform white line detection by limiting the range to an area where no objects other than the white lines exist. (Patent Document 1). This is because, when an object other than the white line exists, it is difficult to distinguish the image from the white line, and a white line is erroneously recognized.
[0003]
[Patent Document 1]
JP-A-11-14376 (FIG. 1)
[0004]
[Problems to be solved by the invention]
However, the white line detection device described in Patent Literature 1 requires both an imaging unit for acquiring an image and a distance measuring unit for detecting the position of an object other than a white line. Was a major problem in terms of
[0005]
In order to reduce the mounting cost, it is necessary to have a means for detecting white lines by distinguishing white lines from objects other than white lines using only the acquired images. The realization of the device is required. The present invention provides a white line detection device that detects a white line with high accuracy while suppressing an increase in the processing amount using only the image, even if the acquired image includes a white line and an object other than the white line. With the goal. Another object is to provide a white line detection device capable of distinguishing a white line from another vehicle using only an acquired image and accurately determining whether or not the other vehicle exists.
[0006]
[Means for Solving the Problems]
The present invention for solving at least one of the above-mentioned problems is as follows.
The invention according to claim 1 is a white line detection device that detects a white line drawn on a boundary of a lane on which a vehicle travels, wherein an imaging unit that acquires an image in front of or behind the vehicle, and a pixel from the image. By detecting an edge portion where the amount of change in information exceeds a predetermined value, extracting white line candidate points, and removing noise by referring to the number or distribution of the white line candidate points extracted in a predetermined region of the image, White line candidate point extracting means for obtaining a desired white line candidate point; and white line detecting means for detecting a white line drawn on the boundary of the lane on which the vehicle runs from the positions of the plurality of white line candidate points. It is a white line detection device.
[0007]
In the invention according to claim 2, the white line candidate point extracting means accumulates the number of the white line candidate points extracted in a predetermined range of the image, and determines that the accumulated number of the white line candidate points is a predetermined number. 2. The white line detection device according to claim 1, wherein when the value is larger than the value, the white line candidate points detected in the predetermined range are regarded as noise and removed.
[0008]
The invention according to claim 3 is characterized in that the white line candidate point extracting means removes white line candidate points continuously extracted in a substantially vertical direction of the image as noise. Is a white line detection device.
[0009]
According to a fourth aspect of the present invention, the white line candidate point extracting means has a function of recognizing the presence of another vehicle based on an edge portion not attributable to the white line, and the recognition is made at each time. The erroneous recognition in the presence / absence determination of the other vehicle is suppressed by averaging the presence / absence of the other vehicle over a predetermined time longer than the detection cycle, thereby suppressing the erroneous recognition. It is a white line detection device of the description.
[0010]
[Action]
According to the first aspect of the present invention, the noise is removed by referring to the number or distribution of the edge portions detected in the predetermined area of the image indicating the front or rear of the host vehicle. Therefore, the detected edge is caused by the white line. It is possible to effectively determine whether or not the white line is present, and it is possible to detect a white line with high accuracy. Further, the processing described in claim 1 does not require a sensor or the like other than an image, so that it is possible to reduce the cost of mounting on a vehicle.
[0011]
According to the invention according to claim 2, the number of edge portions in a predetermined region of the photographed image is accumulated, and it is determined that the number is larger than a predetermined value, and the edge portion detected in the region is determined. Since the object is regarded as noise, it is possible to clearly distinguish and recognize an object other than the white line which is the cause of the noise.
[0012]
According to the third aspect of the present invention, since the edge portion of the captured image that is continuously detected in the substantially vertical direction is regarded as noise, the white line and the object other than the white line that causes the noise are clearly distinguished. It can be recognized separately.
[0013]
According to the invention according to claim 4, when it is determined whether or not another vehicle is present based on the edge portion regarded as noise, the presence or absence of the other vehicle determined at each time is determined based on the detection cycle. By averaging over a long predetermined time, it is possible to effectively suppress erroneous recognition in the determination of the presence or absence of another vehicle.
[0014]
BEST MODE FOR CARRYING OUT THE INVENTION
<First embodiment>
FIG. 1 shows a processing flow of the white line detection device according to the first embodiment of the present invention. FIG. 1A is a flowchart showing the principle of white line detection, and FIG. 2 shows an image corresponding to the white line detection processing when there is no object other than the white line.
[0015]
When the white line detection process is started (S11), an image in front of or behind the own vehicle is photographed (S12). For photographing, an arbitrary camera such as visible light or infrared light can be used. Next, as shown in FIG. 2A, a plurality of horizontal lines 11 (here, five lines) are sequentially scanned to calculate a change amount of luminance as pixel information. In practice, it is most preferable to calculate the luminance differential value e. The pixel information for calculating the amount of change is not limited to luminance, and information such as color (wavelength) may be used. If the calculated differential value e is an edge portion larger than a predetermined value (threshold value), it is estimated that there is a boundary between the road surface and the white line, and the white line candidate point 12 is determined (FIG. 2B). ). In FIG. 2, there is no object other than the white line 10, and thus no noise exists. Therefore, the white line candidate point extracting means considers that all the detected white line candidate points are caused by the white line (S13), and detects the position 13 of the white line from these white line candidate points (S14, FIG. 2 (c)). ).
[0016]
Next, a detailed process of the white line candidate point extracting means S13 when an object other than the white line (here, the preceding vehicle) exists in the captured image will be described with reference to FIG. FIG. 3 shows images corresponding to these processes.
[0017]
First, assuming a plurality of horizontal lines (five in this case) in a captured image, the lowermost row is i _max , and the uppermost row is i _min . First, in the i _max line (S22), the horizontal line is set (S23), a differential operation of luminance is executed, and an edge portion is detected to obtain a white line candidate point (S24). The extraction of the white line candidate points at this time will be described later with reference to FIG. Next, the number of white line candidate points in the line on which the differential operation has been performed is accumulated. It can be seen that the total number of white line candidate points (hereinafter referred to as an accumulated value) in the i _max line shown in FIG.
[0018]
In the situation shown in FIG. 3, when only a normal white line is detected, the accumulated value of the white line candidate points of the horizontal line is about 4 to 6 or less. Therefore, based on the accumulated value of the white line candidate points detected in the line i _max , it is determined that there is no noise (an object other than the white line) on this horizontal line (S25), and these edge portions are defined as white line candidate points. It can be.
[0019]
Next, in the horizontal line of i = i _max −1 (S27), the differential value is calculated in the same manner as described above, and the accumulated value of the white line candidate points is calculated. In the example of FIG. 3, the accumulated value of the white line candidate points on the horizontal line of i = 4 is 4, and it can be determined that there is no noise on this line. Therefore, it is recognized that these white line candidate points are not noise. Is done. Hereinafter, the same processing is repeated until i = i _min (S26).
[0020]
In the horizontal line at i = 2 in FIG. 3, the accumulated value of the white line candidate points is 11. When only a normal white line appears in the image, the accumulated value is about 4 to 6 or less, so it can be estimated that the white line candidate point of the horizontal line with i = 2 contains noise. Therefore, the white line candidate points detected on this line are regarded as noise (S25) and are removed (not considered in the detection of the white line).
[0021]
On a general road as shown in FIG. 3, when noise caused by a preceding vehicle is detected on a horizontal line of i = 2, a horizontal line of i = 1 above (distant) is always ahead. Since noise due to the vehicle is included, it is not necessary to extract white line candidate points for horizontal lines above the horizontal line from which the noise was detected. Then, in the case of FIG. 3, a white line is detected from white line candidate points that do not contain noise and are detected on horizontal lines from i = 3 to i = i _max (S14).
[0022]
The means for calculating the accumulated value (indicated by ctr) of the white line candidate points in each horizontal line will be described below with reference to FIG. Here, the position of a pixel in each horizontal line is represented by j, and the left end of the horizontal line to be processed is j _start and the right end is j _end .
[0023]
First, the processing is started from j = j _{start on} the horizontal line to be processed. At this time, ctr is 0 (S32). Next, a differential value e of the luminance of each pixel is calculated (S33), and it is determined whether or not the differential value exceeds a certain threshold (S34). When the differential value of the luminance is larger than the threshold, it can be estimated that the edge portion is a white line candidate point. At this time, 1 is added to ctr (S35), and at the same time, the edge portion is stored as a white line candidate point.
[0024]
The white line candidate point extraction process described in S36 analyzes the edge portion estimated as the white line candidate point in S34 in detail. More specifically, referring to the absolute value of the differential value calculated in S33 and the distance between the edge portion having a positive differential value and the edge portion having a negative differential value, detection of a more accurate white line is performed. To contribute. However, it is not possible to grasp all noises by this processing.
[0025]
The above processing is repeated until j = j _end (S37, S38), and finally the accumulated value ctr of the white line candidate points of the horizontal line is calculated. Then, as described above, it is determined whether or not noise is included by referring to the value of the cumulative value ctr of the white line candidate point (S25). If the noise is included, the white line candidate point is removed. A white line is detected.
[0026]
<Second embodiment>
A second embodiment for determining whether or not the extracted white line candidate point is noise will be described below. The image shown in FIG. 4 is the same as that described in the first embodiment, and the extraction of the white line candidate points along the horizontal line is the same as described above.
[0027]
In the present embodiment, as shown in FIG. 4, it is determined whether or not noise is present in consideration of the distribution of the extracted white line candidate points. In general, when an image in front of or behind a vehicle is taken, the edge portion caused by the white line does not continue in the vertical direction of the image, but is always detected in the oblique direction (FIG. 4). On the other hand, an edge portion caused by a preceding vehicle or a following vehicle tends to be continuously detected in a substantially vertical direction of the image (FIG. 4). That is, when the extracted white line candidate points are substantially continuous in the vertical direction, it can be estimated that those white line candidate points are noise caused by another vehicle. In this way, considering the distribution of the detected white line candidate points (difference in continuity), the white line candidate points can be distinguished from those caused by the white line and noises caused by other vehicles.
[0028]
Here, “substantially in the vertical direction” means that the edge portion due to the white line is continuously detected in the oblique direction, and is continuous in the vertical direction to the extent that it can be sufficiently distinguished from the oblique direction. (See FIG. 4).
[0029]
When the white line candidate points are successively extracted in the substantially vertical direction in this way, it can be estimated that those white line candidate points are caused by other vehicles, and such white line candidate points are removed as noise. Is done. By detecting a white line based on the white line candidate points from which such noise has been removed, a more accurate white line detection device can be realized.
[0030]
<Third embodiment>
In the above description, the edge portion regarded as noise is considered to be caused by the preceding vehicle or the following vehicle. That is, it is estimated that when there is noise, another vehicle exists, and when there is no noise, there is no other vehicle. Therefore, it is possible to determine whether or not another vehicle exists in the vicinity by processing the edge portion regarded as the noise as described below.
[0031]
In the first and second embodiments, FIG. 5 shows a case where it is determined that another vehicle exists and a case where it is determined that another vehicle does not exist as time elapses. FIG. 5 (a) shows the presence / absence information (existence or non-existence) of another vehicle at each time. However, in reality, the other vehicle is not present for a very short time (about the detection cycle). X in FIG. 5A is considered to be erroneous. In such a case, if a predetermined time longer than the detection period is averaged and observed, and the presence or absence of another vehicle is determined in consideration of temporal continuity, the presence or absence of another vehicle can be accurately determined. In the case as shown in FIG. 5, it is possible to ignore a point like X and determine that another vehicle is present.
[0032]
As an example, the detection of the presence or absence of another vehicle is detected at a detection cycle of about 0.1 second, but the predetermined time to be averagely observed is preferably several to ten times the detection cycle. By arranging each detection result for a predetermined time in this manner, it is possible to accurately detect whether or not another vehicle exists.
[0033]
Note that an object other than the other vehicle is detected as being present only in a very short time since the own vehicle is moving, and is regarded as non-existent when observed over a predetermined period of time. That is, it does not affect the detection of the white line or the presence / absence of another vehicle.
[0034]
<Modification>
In the above embodiment, the process of extracting the white line candidate points along the horizontal line in the captured image was performed. However, the present invention described in the claims is not limited to such an embodiment, and various modifications can be considered.
[0035]
For example, in the process of detecting another vehicle ahead or behind, instead of considering all the edge portions regarded as noise in the captured image, the edge is limited to a predetermined area as shown in FIG. It is possible to analyze the section and determine the presence or absence of another vehicle. Such a predetermined area can be arbitrarily determined according to the situation.
[0036]
In the embodiment, the detection of the edge portion is performed along the horizontal line. However, the detection is not necessarily performed on the horizontal line, but may be performed in a vertical or oblique direction.
[0037]
【The invention's effect】
According to the white line detection device of the present invention, it is possible to accurately distinguish a white line from an object other than the white line using only the captured image. In addition, since processing can be performed only with an image, it can be easily mounted on an automobile and cost reduction can be achieved. Further, since the white line detection process is simple, the white line detection process can be sufficiently performed using in-vehicle devices having limited performance.
[Brief description of the drawings]
FIG. 1 is a diagram showing a processing flow of a white line detection device according to a first embodiment.
FIG. 2 is a diagram illustrating white line detection processing when no object other than a white line exists.
FIG. 3 is a diagram illustrating white line detection processing when an object other than a white line exists.
FIG. 4 is a diagram illustrating processing of a white line detection device according to a second embodiment.
FIG. 5 is a diagram showing a process for detecting another vehicle according to the third embodiment.
FIG. 6 is a diagram showing a modification in which an area for detecting an edge portion is limited.
[Explanation of symbols]
Reference Signs List 10 White line 11 on image 11 Horizontal line 12 for detecting edge 12 Extracted white line candidate point 13 White line 21 detected Other vehicle

Claims (4)

A white line detection device that detects a white line drawn on the boundary of the lane on which the vehicle travels,
Imaging means for acquiring an image in front of or behind the vehicle,
An edge portion where the amount of change in pixel information exceeds a predetermined value is detected from the image to extract white line candidate points, and noise is removed by referring to the number or distribution of the white line candidate points extracted in a predetermined region of the image. By doing, white line candidate point extraction means to obtain a desired white line candidate point,
A white line detecting device for detecting a white line drawn on the boundary of the lane in which the vehicle travels from the positions of the plurality of white line candidate points. The white line candidate point extracting means accumulates the number of the white line candidate points extracted in a predetermined range of the image, and when the accumulated number of the white line candidate points is larger than a predetermined value, 2. The white line detection device according to claim 1, wherein the white line candidate points detected in the range are regarded as noise and removed. 2. The white line detection device according to claim 1, wherein the white line candidate point extraction unit removes white line candidate points continuously extracted in a substantially vertical direction of the image by regarding the white line candidate points as noise. 3. The white line candidate point extraction means has a function of recognizing the presence of another vehicle based on an edge portion not caused by the white line,
2. The recognition according to claim 1, wherein the presence / absence of another vehicle determined at each time is averaged over a predetermined time longer than a detection cycle to suppress erroneous recognition in the presence / absence determination of the other vehicle. The white line detection device according to claim 3.

Images