JP2000242899A - White line recognizing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a white line recognizing device that can obtain horizontal center of an optical axis and the angle of depression with respect to a road image which depends on an on-vehicle camera attachment condition with excellent precision. SOLUTION: The device is provided with an approximate straight line calculating part 6 for detecting the candidate points of white lines in the road image picked-up by the on-vehicle camera 1 and respectively obtaining the approximate straight lines of the left and right white lines, a virtual loss point calculating part 7 for obtaining the intersecting point of the approximate straight lines as a virtual loss point, a histogram generating part 8 for respectively generating calculation frequency histograms concerning a horizontal coordinate and the vertical one of the virtual loss point and a learning part 9 for respectively obtaining the horizontal center of the optical axis and the angle of depression with respect to the road image as the distribution center of the respective histograms of the horizontal and vertical coordinates. It is favorable to obtain the optical axis horizontal center and the angle of depression as the distribution center of the moving average output of the histograms.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for recognizing a white line from a road image in front of a vehicle taken by a vehicle-mounted camera.
In particular, the present invention relates to a white line recognition device that can accurately determine the horizontal center of the optical axis and the depression angle with respect to the road image depending on the mounting conditions of the vehicle-mounted camera.

[0002]

2. Related Background Art A pavement road is usually provided with a white line defining a driving lane or the like. Recognizing such a white line from a road image ahead of the vehicle captured by an on-board camera,
Various attempts have been made to use the vehicle for safe driving.

[0003]

When estimating the shape of a white line on a road from white line information recognized from a road image, only a screen-view coordinate system (camera coordinates) and a top-view coordinate system (ground coordinate system) are used. Is performed. This coordinate conversion is performed using parameters including the mounting height of the vehicle-mounted camera, its optical axis center, and the depression angle. However, depending on the mounting condition of the vehicle-mounted camera, the values of the above parameters may differ from the actual mounting height, the optical axis center, and the depression angle of the vehicle-mounted camera, and an error occurs in coordinate conversion. An error occurs in the white line shape.

In order to eliminate such an error, for example, Japanese Patent Application Laid-Open No. 10-40499 discloses a technique for correcting a depression angle or the like by checking whether or not two white lines obtained by coordinate transformation are parallel. Is done. However, there is a problem that a great deal of processing is required each time the coordinate transformation is performed to check the parallelism of the white line. The present invention has been made in view of such circumstances, and an object thereof is to easily and accurately obtain an optical axis horizontal center and a depression angle with respect to a road image depending on a mounting condition of an on-vehicle camera, and obtain an accuracy of white line recognition. An object of the present invention is to provide a white line recognition device that can be enhanced.

[0005]

In order to achieve the above object, a white line recognition device according to the present invention detects a candidate point of a white line in a road image picked up by a vehicle-mounted camera, and respectively approximates right and left white lines. An approximate straight line calculating means for obtaining, a virtual vanishing point calculating means for obtaining an intersection of these approximate straight lines as a virtual vanishing point, and a histogram creating means for respectively generating a calculation frequency histogram for the horizontal coordinate and the vertical coordinate of the virtual vanishing point; Learning means for calculating the horizontal center and the depression angle of the optical axis with respect to the road image from the histograms of the horizontal coordinate and the vertical coordinate, respectively.

That is, a virtual vanishing point is obtained as an intersection of approximate straight lines of two right and left white lines obtained on a road image, and a calculation frequency histogram for the coordinates of the virtual vanishing point is obtained from
The horizontal coordinate having the highest frequency is obtained as the horizontal center of the optical axis with respect to the road image, and the vertical coordinate having the highest frequency is obtained as the depression angle with respect to the road image.

Preferably, the learning means determines the horizontal center of the optical axis and the depression angle for the road image using the moving average output of each of the histograms of the horizontal coordinate and the vertical coordinate. Features.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A white line recognition device according to one embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a schematic diagram of a main part of a white line recognition device according to this embodiment.
Reference numeral 1 denotes a vehicle-mounted camera. An image signal indicating a road image ahead of the vehicle captured by the on-vehicle camera 1 is amplified to a predetermined level via an amplifier 2, converted into a digital signal via an A / D converter 3, and stored in an image memory 4. Written.
The writing of the road image into the image memory 4 is repeatedly performed at a predetermined frame cycle.

The white line recognition process for the road image obtained in the image memory 4 is performed by first obtaining a white line candidate point in the road image by the white line candidate point detection unit 5.
For the detection of the white line candidate points, the luminance signal is discriminated (binarized) at a predetermined level, for example, by utilizing the fact that the luminance level of the white line portion in the road image is higher than the main surface of the road. And so on. Alternatively, a change in the luminance signal level may be checked along a predetermined white line search line, and a white line candidate point may be obtained while detecting a position where the change in the luminance level is steep and the amount of change is large as an edge of the white line. By such white line candidate point detection processing, for example, a plurality of white line candidate points P on the road image G captured including the white line W in FIG. It should be noted that the white line W usually exists at least one on each of the left and right sides to define the traveling lane in which the vehicle is traveling, so that the white line candidate point P is obtained along each of the white lines W.

Then, the approximate straight line calculation unit 6 calculates the white line candidate point P
The white line candidate points are individually recognized as one piece of white line information for each group, and a straight line approximation formula indicating straight lines LL and LR approximating the left and right white lines W respectively is calculated. Specifically, according to the coordinates of a highly reliable white line candidate point P detected in the lower area of the screen in the road image G, the approximate straight line LL,
A linear approximation formula indicating LR is calculated.

However, the feature of this device is that
The intersection of the approximate straight lines LL and LR of the two left and right white lines W obtained as described above in the virtual vanishing point calculation unit 7 is obtained as the virtual vanishing point E in the road image G.
The coordinates (horizontal coordinates and vertical coordinates) of the virtual vanishing point E can be obtained by calculating the intersections of the linear approximation formulas indicating the approximate straight lines LL and LR, respectively. Then, the calculation frequency of the coordinates of the virtual vanishing point E obtained for each road image G in this manner is calculated by the histogram creation unit 8.
The learning unit 9 obtains the coordinates of the mode, for example, as information on the horizontal center of the optical axis with respect to the road image G and the depression angle according to the histogram. That is, each time the virtual vanishing point E is obtained from the road image G as described above, the number of times of detection as the virtual vanishing point is incremented for the horizontal coordinate and the vertical coordinate, and thus the detection shown as Hh and Vh in FIG. Create a frequency histogram. And these detection frequency histograms Hh, Vh
For example, the horizontal coordinate Hp and the vertical coordinate Vp forming the frequency peak are calculated by using the optical axis horizontal center with respect to the road image G,
And depression angle information.

When creating the detection frequency histograms Hh and Vh for the horizontal and vertical coordinates of the virtual vanishing point E, for example, an approximate straight line L whose vehicle speed is equal to or higher than a predetermined value is used.
Various conditions such as an error between L, LR and the white line being within a predetermined value and an absolute value of a gradient of the left and right approximate straight lines LL, LR being within a predetermined value are determined. Only when these conditions are satisfied, It is desirable to add the information of the virtual vanishing point E to the histograms Hh and Vh. By applying such a condition, for example, the information of the inappropriate virtual vanishing point E in the case where the road (white line W) is largely bent or the case where the vehicle is turning is removed while the above-described information is excluded. The histograms Hh and Vh can be created with high accuracy.

Further, instead of obtaining the peak points Hp, Vp of the detection frequency histograms Hh, Vh, moving averages Have, Vave of the detection frequency histograms Hh, Vh are obtained, and the distribution center of the moving averages Have, Vave is obtained. It is also possible to obtain Hc and Vc as information on the optical axis horizontal center and the depression angle, respectively. FIG. 3 shows a flow of a series of processing for calculating the optical axis horizontal center and the depression angle for the road image G described above. The processing procedure will be briefly described. First, a road image G captured by the on-vehicle camera 1 is input [Step S1], and left and right white line approximate straight lines LL, LL are determined based on white line candidate points P obtained from the road image G. LR is calculated [Step S2]. Next, these white line approximate straight lines LL,
The coordinates of the intersection of the LR and the virtual vanishing point E in the road image G
[Step S3].

Thereafter, the white line approximate straight lines LL, LR
The deviation from the above-mentioned white line candidate point P is obtained as a deviation,
It is determined whether the deviation is within a predetermined threshold [Step S4]. That is, the white line approximation straight lines LL, LR
Is verified whether or not is approximated within a predetermined allowable error range. When the white line approximation straight lines LL and LR are not found within the predetermined allowable error range, the white line W is largely bent, and the virtual vanishing point E obtained as described above is shifted to the optical axis horizontal of the road image G. It is excluded as inappropriate information for obtaining the center and the depression angle. In this case, the process returns to the above-described step S1 to start the process for the next road image G.

On the other hand, if the deviation is within the predetermined threshold [Step S4], then the gradients (gradients) of the white line approximate straight lines LL and LR are obtained, and the difference between the absolute values of the gradients is determined by the predetermined value. It is determined whether or not it is within the threshold of [Step S5]. That is, it is verified whether the white line approximate straight lines LL and LR are obtained in the road image G substantially symmetrically, in other words, whether the vehicle is positioned substantially parallel to the road. When the difference between the absolute values of the gradients of the white line approximation straight lines LL and LR is large, the vehicle direction is inclined with respect to the road, and the information of the white line W is used to determine the optical axis horizontal center and the depression angle of the road image G. The information of the virtual vanishing point E obtained as described above as being inappropriate to use is excluded. Also in this case, the aforementioned step S1
Then, the process for the next road image G is started.

Further, when the conditions for the gradients of the white line approximation straight lines LL and LR are satisfied, it is determined whether or not the running speed (vehicle speed) of the vehicle is equal to or higher than a predetermined threshold [Step S6]. In other words, it is verified whether or not the road image G is obtained while the vehicle is traveling at a certain speed or higher, and the information of the virtual vanishing point E obtained from the substantially similar road image G and regarded as the same is obtained. Exclude. In other words, in order to use the information of the virtual vanishing point E obtained from the road images G of different scenes, information repeatedly obtained from the same type of road image G is excluded. Also in this case, the process returns to the above-described step S1 to start the process for the next road image G.

When the usefulness of the information of the virtual vanishing point E is confirmed through the above-described determination processing, the information of the horizontal coordinates and the vertical coordinates of the virtual vanishing point E is detected. It is added to the histograms Hh and Vh representing the frequencies [Step S7]. Specifically, the detection frequency count value indicated by the corresponding coordinates of the histograms Hh and Vh is incremented (+1).

Thereafter, the number of information of the virtual vanishing point E added to the histograms Hh and Vh (total number of histograms)
Is greater than or equal to a predetermined threshold, for example, 100 samples [Step S8]. That is, the histogram H
h, Vh, a sufficient amount of information for obtaining the horizontal center and the depression angle of the optical axis of the road image G is obtained from the histogram Hh,
It is determined whether or not Vh has been obtained. Then, the histograms Hh and V with sufficient information amount to analyze the contents
If it is confirmed that h has been obtained, the center of the optical axis is obtained by obtaining the distribution center of the histograms Hh and Vh [Step S9]. The horizontal center and the vertical coordinates are obtained as information of the depression angle with respect to the road image G.

On the other hand, when the information amount of the histograms Hh and Vh is small, for example, the histograms Hh and Vh
Then, it is determined whether or not the mode value in step S1 exceeds a predetermined threshold value (step S10). That is, it is determined whether or not the distribution characteristics of the histograms Hh and Vh are locally sharp.
If it has a sharp distribution characteristic and its peak value (mode value) exceeds a predetermined threshold, [Step S1
0], the coordinates indicating the peak value are detected as the distribution center of the histogram, and the information is obtained as the optical axis center coordinates [Step S9].

If the information amounts of the histograms Hh and Vh are small and their peak values (mode values) are less than a predetermined threshold value, the histograms are used to calculate the horizontal center of the optical axis and the depression angle with respect to the road image G. It is determined that the information amounts of Hh and Vh are insufficient [Steps S8 and S10], and the processing returns to the processing procedure from Step S1 described above to repeatedly execute the processing for the next road image G in the same manner.

Thus, the histograms Hh and Vh of the virtual vanishing point E obtained from the road image G are created as described above,
According to the present apparatus for calculating the horizontal center of the optical axis and the depression angle with respect to the road image G as the distribution center, the information (white line shape) of the white line W obtained from the road image G can be converted into the ground coordinate system without performing coordinate conversion. The horizontal center of the optical axis and the depression angle can be easily obtained. Therefore, even when there is an attachment error with respect to the vehicle-mounted camera 1, the actual horizontal center of the optical axis and the depression angle of the road image G captured by the vehicle-mounted camera 1 can be accurately obtained.
When the information of the white line W (shape of the white line) obtained from the above is converted into a coordinate on the ground surface coordinate system and used as information for safe driving, the information accuracy can be made sufficiently high.

Incidentally, regarding the optical axis deviation of the vehicle-mounted camera 1, the deviation amount w from the center of the screen of the road image G to the horizontal center of the optical axis is obtained, and the focal length of the vehicle-mounted camera 1 is f. Based on w and the depression angle θ, the correction amount ψ
May be obtained as tanψ = (w / f) cos θ and corrected.

According to the present apparatus, a histogram of the virtual vanishing point E obtained from the road image G is created, and the horizontal center of the optical axis and the depression angle are obtained as the distribution center of the histogram. The factors can be normalized, and the horizontal center of the optical axis and the depression angle with respect to the road image G can be obtained with high accuracy regardless of changes in the road environment. Further, as described above, when creating the histogram, the virtual vanishing point E obtained from the inappropriate road image G is excluded, and the virtual vanishing point E obtained from the road image G that can be regarded as the same information is excluded. So
A histogram of the virtual vanishing point E can be created with high accuracy and high reliability. As a result, effects such as the detection accuracy of the horizontal center of the optical axis and the depression angle can be sufficiently improved.

The present invention is not limited to the above embodiment. For example, when analyzing the histogram, it is possible not only to detect the peak value and to obtain the distribution center of the moving average output, but also to perform an appropriate filtering process. Also, the condition determination at the time of creating the histogram can be appropriately changed according to the road environment and the like. In addition, the present invention can be variously modified and implemented without departing from the gist thereof.

[0025]

As described above, according to the present invention, the horizontal axis of the optical axis and the depression angle are obtained according to the histogram of the virtual vanishing point obtained from the road image. Accuracy can be sufficiently improved. Therefore, when utilizing the white line information obtained from the road image for safe driving, effects such as stably assuring the information accuracy can be obtained.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram of a main part of a white line recognition device according to an embodiment of the present invention.

FIG. 2 is a diagram showing a road image for explaining a processing concept according to the present invention, a virtual vanishing point thereof, and a relationship between a histogram of the virtual vanishing point, an optical axis horizontal center, and a depression angle.

FIG. 3 is a flowchart illustrating an example of a series of processing procedures for calculating an optical axis horizontal center and a depression angle with respect to a road image;

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 vehicle-mounted camera 4 image memory 5 white line candidate point detection unit 6 approximate straight line calculation unit 7 virtual vanishing point calculation unit 8 histogram creation unit 9 optical axis horizontal center and depression angle learning unit

Continued on the front page (72) Inventor Hiroshi Fujii 5-33-8 Shiba, Minato-ku, Tokyo Inside Mitsubishi Motors Corporation (72) Inventor Hisashi Ishikura 2-3-2 Marunouchi, Chiyoda-ku, Tokyo Mitsubishi Electric Co., Ltd. Company F term (reference) 5B057 AA16 CF05 DA07 DB02 DC07 DC08 DC19 5H180 AA01 CC04 CC24 LL01 LL08 5L096 BA04 CA04 EA31 FA10 FA32 FA35 FA62 FA67 FA69 GA59

Claims (2)

[Claims] 1. A white line recognition device for recognizing a white line from a road image ahead of a vehicle captured by an on-vehicle camera, comprising: detecting candidate white line points in the road image to obtain approximate straight lines of left and right white lines. Straight line calculating means, virtual vanishing point calculating means for finding an intersection of these approximate straight lines as a virtual vanishing point, histogram creating means for creating detection frequency histograms for the horizontal and vertical coordinates of the virtual vanishing point, respectively, A white line recognizing device comprising: learning means for obtaining an optical axis horizontal center and a depression angle with respect to the road image from respective histograms of coordinates and vertical coordinates. 2. The white line recognition device according to claim 1, wherein said learning means obtains an optical axis horizontal center and a depression angle with respect to said road image using a moving average output of each of said horizontal coordinate and vertical coordinate histograms.