JPH05164569A - Run road detector of traveling vehicle - Google Patents

Abstract

PURPOSE:To obtain the detector capable of performing highly precise run road detection without increasing the operation quantity of image processing. CONSTITUTION:An outside image is input from an image input means 1 and a particular search area for performing the search of a run road end in the image is set in a search area setting means 2. Within a candidate point detection means 3 only the inside of the search area is searched and a candidate point presumed to compose a part of a run road end is detected. In a run road end approximation means 4 an approximation straight line passing each detected candidate point is found as the run road end. In a search area position correction means 5 the set position of the search area is corrected on the basis of the approximated straight line. As above-mentioned the search area is set to always pursue the position of the run road end.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a traveling road detecting device for a moving vehicle, and more particularly to a device for inputting an image of the outside of the moving vehicle and detecting the traveling road based on this image.

[0002]

2. Description of the Related Art It is important to detect a traveling path correctly when controlling traveling of an autonomous vehicle. As described above, in order to detect the traveling road of a moving vehicle, an image input device such as a camera is used to input an image in front of the traveling road, and for this image, for example, image processing for detecting a white line indicating the end of the traveling road. It is general to detect the traveling path by applying the above. For example, Japanese Patent Application Laid-Open No. 64-26913 discloses a method of detecting such a traveling road.

[0003]

However, in the conventional traveling road detecting apparatus, if the detection accuracy of the traveling road is increased, the calculation load of the image processing is increased, and if the calculation load of the image processing is reduced, the traveling road is detected. However, there is a problem that the detection accuracy of is reduced. As described above, in order to detect the road, it is necessary to recognize the white line indicating the road edge from the input image. However, the higher the accuracy of recognizing such white lines, the greater the amount of calculation for image processing.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a traveling road detecting device for a moving vehicle, which can perform traveling road detection with high accuracy without increasing the calculation amount of image processing.

[0005]

[Means for Solving the Problems]

(1) A first invention of the present application is, in a traveling path detection device for a moving vehicle, image input means for inputting an image of an outside world of the moving vehicle,
Search area setting means for setting a specific search area to search for a road edge in the input image, and a candidate presumed to form a part of the road edge by searching the set search area. A candidate point detecting means for detecting a point, a traveling road edge approximating means for obtaining an approximate straight line passing through each of the detected candidate points as a traveling road edge, and a search for correcting the set position of the search area based on the obtained approximate straight line The area position correcting means is provided.

(2) A second invention of the present application is, in the traveling path detecting apparatus for a moving vehicle according to the first invention, detecting an edge portion where a density value change of a pixel is abrupt in an image in a search area. The means and the means for detecting a white area portion composed of pixels having the density value of the white line indicating the road edge are provided in the candidate point detecting means.

(3) The third invention of the present application is, in the traveling path detecting apparatus for a moving vehicle according to the above-mentioned first invention, the distance C of the approximate straight line with respect to a predetermined origin is changed beyond a predetermined reference value. Sometimes, the position correction for a specific search area is stopped.

[0008]

[Work]

(1) According to the first aspect of the present invention, the candidate point for the road edge is searched only for the inside of the search area set in the input image. Then, an approximate straight line is obtained for the searched candidate points, and this approximate straight line is recognized as the road edge. Therefore, the calculation load of image processing is significantly reduced as compared with the conventional apparatus that detects the road edge of all input images. Moreover, since the position of the search area is corrected based on the currently recognized road edge, the search area is always set at the position where the road edge is most likely to exist. Become. Therefore, it becomes possible to accurately detect the traveling path.

(2) According to the second invention of the present application, two pieces of information, that is, the information indicating the area and the information indicating the edge position of the white line forming the road edge are detected and combined. Since the white line is detected in this manner, the white line can be detected with high accuracy.

(3) According to the third aspect of the present invention, when the position change amount of the approximate straight line exceeds the reference value, the position correction for the specific search area is stopped. That is, the setting position of the specific search area remains the same as the previous time. By stopping the position correction in this way, it is possible to continue to secure a correct set position even in a search area in which the road edge is accidentally detected incorrectly.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below based on illustrated embodiments. FIG. 1 is a block diagram showing the basic configuration of a traveling road detection device for a mobile vehicle according to the present invention. Image input means 1
Has a function of inputting an image of the external environment of the moving vehicle, and in this embodiment, is constituted by a video camera. The search area setting means 2 has a function of setting a specific search area in which the search for the road edge is to be performed in the input image. The candidate point detection means 3 has a function of searching the set search area and detecting a candidate point that is estimated to form a part of the road edge. Is composed of an edge detecting section 3a for detecting a steep edge section and a white area detecting section 3b for detecting a white area section consisting of pixels having a density value of a white line indicating a road edge. The traveling road edge approximating means 4 has a function of obtaining an approximate straight line passing through each of the detected candidate points, and this approximate straight line is recognized as the traveling road edge. Further, based on this approximate straight line, the search area position correction means 5 corrects the set position of the search area in the search area setting means 2.

Next, the operation of this apparatus will be described with reference to a concrete example. Now, consider a case where an image as shown in FIG. 2A is input from the image input means 1. This image is
It is an image of a traveling road ahead of the moving vehicle, and in the example of the figure, a white line L indicating a left traveling road end and a white line R indicating a right traveling road end.
And are included in the image. If the image processing for white line search is performed over the entire area of this image in order to detect such white lines L and R, the calculation load becomes very large. Therefore, in the present invention, the image processing for the white line search is performed only within the specific search area set by the search area setting means 2. For example, in the search area setting means 2, four search areas W1 to W1 as shown in FIG. 2B are displayed on the input image shown in FIG.
Consider the case where W4 is set. The candidate point detection means 3 may perform the search process for detecting the candidate points estimated to form a part of the white line only within the search areas W1 to W4.

In the apparatus of this embodiment, this candidate point search processing is performed as follows. Now, for example, consider a case where an image as shown in FIG. 3A is obtained in the search area W. Here, the white line L is shown as a line having a thickness, the white area portion corresponding to the white line L is composed of pixels having a density value corresponding to white, and the other area portions are , A pixel having a density value corresponding to the traveling road surface. The edge detection unit 3a performs edge detection by passing such image information in the search area W through a filter that emphasizes edges. On the other hand, the white area detection unit 3b
Detects the white area by passing the image information in the search area W through a filter for detecting white pixels. These processes will be described below with reference to one-dimensional image processing. For example, when the scanning line S is defined for the image in the search area W shown in FIG. 3A, and one-dimensional scanning is performed along the scanning line S, the scanning line S corresponds to the density value of each pixel. When the resulting signal waveform is passed through a filter that emphasizes edges, the edge waveform shown in Fig. 3 (b) is obtained, and when passed through a filter that detects white regions, the white region waveform shown in Fig. 3 (c) is obtained. Is obtained. The edge waveform includes accurate information on the edge position, and the white area waveform includes information indicating which area among the areas bounded by the edge is the white area. Therefore, the original white line L can be accurately detected by combining these two waveforms. The above is the one-dimensional processing performed only for the pixels on the scanning line S, but in reality, such processing is performed for the two-dimensional image. As a filter for emphasizing edges, for example, a 3 × 3 two-dimensional filter to a 3 × n two-dimensional filter may be used. In the case of general road images, sufficient processing is possible with n <5. As a filter for detecting a white area, a filter that selects pixels having a density value equal to or higher than a preset white level may be used. The white level can be set by a method such as an optimal binarization method, but in this embodiment, the density value of 1.5 times the average of the density values of pixels is simply used as the boundary value of the white level. There is.

By performing the image processing as described above,
Candidate points estimated to form part of the white line are detected in each search area. In this embodiment, to detect candidate points,
The so-called token tracker method is used. This method is a method of estimating the current candidate point position using the Kalman filter based on the past candidate point position.
Detailed description of this method is omitted in this specification. Here, as shown in FIG. 4A, each search region W1
The following description will be continued on the assumption that candidate points P1 to P4 have been detected within W4. Of course, actually, a plurality of candidate points may be detected within one search area. The traveling road edge approximating means 4 obtains an approximate straight line passing through the detected candidate points P1 to P4 as a white line L indicating the traveling road edge. That is, the white line L is obtained as shown in FIG. The white line L is shown in FIG.
The two-dimensional coordinate system on the image as shown in and the three-dimensional coordinate system in the real world including the road are obtained. The white line L obtained on the two-dimensional coordinate system is used for position correction in the search area position correcting means 5, and the white line L obtained on the three-dimensional coordinate system is used as information for performing traveling control of this moving vehicle. Will be

Now, when the white line L indicating the road edge is obtained in this way, the search area position correcting means 5 corrects the position of the search area. The principle of position correction in this embodiment is very simple. That is, the correction is performed so that the center point M of the search area is moved in the horizontal direction so that it is located on the obtained white line L. For example, as shown in FIG. 5 (a), when the white line L is found at the position shown in the figure with respect to the current search area W, the position of the search area W is horizontally moved to the left in the figure. Position correction may be performed so as to obtain the state shown in FIG. By performing such position correction, the search area is always set at a position where the probability that a white line exists is highest in a new input image.

As described above, when the white line L is recognized by the image processing of one screen, the position of the search area for performing the image processing of the next screen is corrected based on the recognized white line L. It is a feature. In other words, this search area W
1 to W4 are in a state in which the white line L is traced while slightly moving in the horizontal direction on the input image. By the way, the weakness of tracking due to such position correction is that once a line other than the white line, for example, a shadow or a scratch on the road surface is mistakenly recognized as a white line, the shadow on the road surface is The point is that you will continue to track the wound.
Therefore, in this embodiment, the following processing is performed in order to eliminate such erroneous tracking.

Now, as shown in FIG. 6, it is assumed that the white line L is recognized on the input screen, and the inclination angle θ of this white line.
And the distance C from the origin O of this white line (defined in the upper left of the screen in this example) are defined as shown in the figure. Here, when the white line L obtained for each input screen is tracked in time, it is considered that the position of the white line L will change abruptly if the moving vehicle is traveling on a straight road in a normal state. Hateful. In particular,
It is considered that the variation of the inclination angle θ indicates the variation of the traveling road and the variation of the distance C indicates the variation of the vehicle. In view of this, in the present embodiment, when the distance C changes abruptly while the vehicle is traveling in a normal state, it is recognized that a shadow or scratch on the road surface may have been erroneously detected as a white line. I am trying to make such a check. That is, the four search areas W1 to W4
, And if one or two of the four position correction amounts are significantly large and exceed a predetermined reference value, erroneous detection is detected for the one or two search areas. Position correction is stopped and position correction is stopped (that is, for the search area determined to have been erroneously detected, this time without position correction,
Leave the original position). Generally, it is considered that the shadows and scratches on the road surface are partially present. Therefore, when the position correction amount is large only for one or two of the four search areas, the road surface is It is highly possible that the shadows and scratches of the are detected as white lines. If the position is corrected in such a case, the shadows and scratches on the road surface will be traced as white lines thereafter. Therefore,
In such a case, the position correction is stopped and the search region is fixed at the original position so that the white line can be traced. In contrast, all four or three of four
If the one position correction amount exceeds the predetermined reference value,
It is determined that there is no erroneous detection, and the position is corrected as it is.

Next, another method for avoiding the above-mentioned erroneous detection will be disclosed. in this way,
A stereo camera is used as the image input means 1. That is, the left image and the right image are input from two video cameras installed at a predetermined distance on the left and right sides, and a correction process for erroneous detection is performed based on the both images. For example, consider a case where a left image and a right image as shown in FIG. 7 are obtained. Here, WL1 to WL indicated by solid lines
Reference numeral 4 denotes a newly set search area on the left image, and solid lines WR1 to WR4 denote newly set search areas on the right image. The area indicated by the broken line is the past position of each search area indicated by the solid line. Here, it can be seen that, out of the eight search areas indicated by the solid line, only the area WL3 has changed extremely from the past position. The corresponding change in the region WR3 is small as in the other search regions. In this way, the basic concept of this method is to judge the false detection by comparing the left and right images.

Specifically, the left and right corresponding regions, that is, WL1 and WR1, WL2 and WR2, WL3 and WR, respectively.
3, the position change amounts for each pair of WL4 and WR4 are compared, and when only one change amount is extremely large,
It is determined that the false detection is performed. Generally, it is considered that shadows and scratches on the road surface are rarely erroneously detected as white lines in both left and right images by the stereo camera. Therefore, for the paired areas,
When the amount of positional change in one area is extremely large, it can be determined that a shadow or scratch on the road surface has been erroneously detected.

The comparison of the amount of positional change can also be performed using virtual images. If the conditions such as mutual distance between stereo cameras, installation angle, zoom magnification, etc. are known, the corresponding virtual right image can be calculated based on the left image, and conversely, the virtual right image can be calculated based on the right image. The corresponding virtual left image can be calculated. Therefore, for example, it is assumed that the virtual left image is calculated based on the right image shown in FIG. On this virtual left image, the regions WR1 to WR4 are
The regions WL1 ′ to WL4 ′ are obtained as the projections of
Then, theoretically, the regions WL1 to WL4 on the left image
And the areas WL1 ′ to WL4 ′ on the virtual left image should match. So, for example, as shown in FIG.
When the region WL3 and the region WL3 'are largely deviated from each other, it can be determined that the erroneous detection is performed in either the region WL3 or the region WR3. Normally, when the erroneous detection is performed, the amount of position change becomes large, so that it can be recognized that the erroneous detection is performed for the one with the larger amount of position change.

As a measure against erroneous detection, one method is to leave the same position as the previous time without correcting the position of the search area determined to have been erroneously detected. In the example of FIG. 7, the search region WL3 determined to be erroneously detected is left at the previous position shown by the broken line instead of being corrected to the new position shown by the solid line. The other method is a method of correcting the position by utilizing the position of the area on the virtual image. In the example of FIG. 7, the search area WL3 determined to be erroneously detected is corrected to the position of the area WL3 ′ obtained on the virtual left image, instead of being corrected to the new position indicated by the solid line. ..
In any of the methods, it is possible to prevent the shadows and scratches on the road surface from being tracked as they are based on erroneous detection.

Although the present invention has been described above based on the illustrated embodiment, the present invention is not limited to this embodiment and can be implemented in various modes other than this. For example, in the above-described embodiment, only the process for the white line L indicating the left end of the road is described, but it goes without saying that the same process is performed for the white line R indicating the right end of the road. Further, in the above-described embodiment, an example in which four search areas are set has been described, but the number of search areas is not particularly limited to four. Further, as the position correction direction of this search region, the example in which the position is moved in the horizontal direction has been shown in the above-described embodiment, but the position may be corrected by moving it in the vertical direction, or may be moved in the diagonal direction to perform the position correction. It doesn't matter.
Further, in the above-mentioned embodiment, although the initial setting position of the search area is not particularly mentioned, for example, only when the apparatus is started, the search processing is performed on the entire input image to recognize the white line, and the white line is recognized. Each search area may be initialized on the line.

[0023]

As described above, according to the present invention, in a traveling road detecting apparatus for a moving vehicle, a search area is set in a part of the input image, and the traveling road edge search processing is performed only in this search area. Since the position of the search area is corrected based on the result, it is possible to detect the traveling path with high accuracy without increasing the calculation amount of the image processing.

[Brief description of drawings]

FIG. 1 is a block diagram showing a basic configuration of a traveling road detection device for a mobile vehicle according to the present invention.

FIG. 2 is a diagram showing a state in which four search areas are set in a road image input in the device shown in FIG.

FIG. 3 is a diagram showing a candidate point detection process within a set search area.

FIG. 4 is a diagram showing a process of obtaining an approximate straight line for a detected candidate point.

FIG. 5 is a diagram showing a process of correcting the position of a search region based on the obtained approximate straight line.

FIG. 6 is a diagram showing a principle of checking erroneous detection in the detection device shown in FIG.

FIG. 7 is a diagram illustrating a method of performing processing for erroneous detection using a stereo camera.

[Explanation of symbols]

1 ... Image input means (camera) 2 ... Search area setting means 3 ... Candidate point detection means 3a ... Edge detection part 3b ... White area detection part 4 ... Road edge approximation means 5 ... Search area position correction means L, R ... White line M ... Center point O ... Origin point P1 to P4 ... Candidate point S ... Scan lines W, W1 to W4, WL1 to WL4, WR1 to WR4, W
L1 'to WL4' ... Search area

Claims (3)

[Claims] 1. An image input unit for inputting an image of the outside world of a moving vehicle, a search region setting unit for setting a specific search region in the input image to search for a road edge, and the inside of the search region. And a candidate point detecting means for detecting a candidate point estimated to form a part of the road edge, and a road edge approximating means for obtaining an approximate straight line passing through each detected candidate point as a road edge, A search region position correction unit that corrects the set position of the search region based on the approximate straight line, and a traveling path detection device for a mobile vehicle. 2. The traveling road detecting apparatus according to claim 1, wherein the candidate point detecting means detects an edge portion of the image in the search area in which a change in pixel density value is abrupt and a traveling road end. And a means for detecting a white area portion consisting of pixels having the density value of the white line shown, and a traveling path detection device for a moving vehicle. 3. The traveling road detecting apparatus according to claim 1, wherein when the distance C of the approximate straight line with respect to a predetermined origin changes by exceeding a predetermined reference value, the position correction for the specific search area is stopped. A traveling path detection device for a moving vehicle, characterized in that: