JPH03158976A - Traffic lane discriminating method - Google Patents

Abstract

PURPOSE:To discriminate the traffic lane of a traveling road by extracting a straight line corresponding to a traffic lane sectioning strip plotted on the traveling road from among a straight line group approximate to the align of feature points, and identifying a short dashed line from the length of an interrupted part, and judging the positional relation of each traffic lane sectioning strip on the basis of the short dashed line. CONSTITUTION:An original picture in a figure (a) is differentiated and converted into an edge picture, and this edge picture is Huff-transformed, and the straight line group approximate to the align of the feature point is obtained. The straight line corresponding to the traffic lane sectioning strip is selected from among this straight line group on the basis of the road width of the traveling road. As the result, these straight lines shown in the figure (b) are obtained. Then, the straight line 6 corresponds to the short dashed line 2, and the straight line 7 to the traffic lane sectioning stripe 3, and the straight line 8 to the traffic lane sectioning stripe 4, and the straight line 6 corresponding to the short dashed line 2 is discriminated from among these straight lines 6 to 8, and the positional relation of every detected straight line 6 to 8 is discriminated. Thus, as shown in the figure (d), the straight line 6 is judged to be a center line, and the straight line 7 is decided to be the traffic lane sectioning stripe 3 to show the end part of a left side traffic lane, and the straight line 8 is decided to be the traffic lane sectioning stripe 4 to show the end part of a right side traffic lane.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a driving lane discrimination method for recognizing lane markings drawn on a driving road by image processing and determining the lanes of the driving road.

[Conventional technology]

In order to recognize the lanes on a driving road, it is necessary to identify lane markings. For this purpose, it is necessary to perform image processing on data obtained by a TV left camera, etc., and extract line segments corresponding to lane markings. As one method for this, Japanese Patent Application Laid-Open No. 6224310 and
There is a method using the Hough transform disclosed in Japanese Patent No. 61. According to this, by performing Hough transform on each feature point on an image obtained as image data, it is possible to obtain a group of straight lines corresponding to the distribution of feature points.

[Problem to be solved by the invention]

Broken lines are sometimes used as lane markings, and by identifying the broken lines, it is possible to understand the symbolic meaning of the broken lines as road signs.

In other words, if the lane marking is a dashed line, it means that the lane marking is the center line of the road, or that it is possible to veer out of the lane you are driving in order to overtake the vehicle in front of you. It also means that it is possible to change lanes. however,
The conventional driving lane determination method described above can determine the presence or absence of lane markings, but it is difficult to determine the type of lane markings, and the image information captured by the camera is not effectively used. Ta.

On the other hand, in order to determine whether a lane marking is a broken line, it is conceivable to rule that if the lane marking is not continuous, it is a broken line, or if the lane markings are regularly interrupted, it is a broken line. However, on actual driving roads, lane markings that should be solid lines are blurred or broken, and the length captured by the camera changes depending on the driving conditions, so the above rules cannot be applied in absolute terms. This is not true, and it is difficult to determine the type of lane marking.

[Means to solve the problem]

The present invention has been made to solve these problems, and focuses on at least two or more straight lines detected by the above method, and calculates the maximum length of the discontinuous part of the feature point sequence for each dividing line. The method comprises the steps of determining the lane markings having the longest discontinuous portion among these maximum lengths as a broken line, and determining the positional relationship of each lane marking based on the broken lines. .

In addition, the sum of the lengths of the discontinuous parts of the feature point sequence is calculated for each of at least two or more straight lines detected by the above method, and the lane marking line with the largest sum of these lengths is identified as a broken line. and a step of determining the positional relationship of each lane marking based on this broken line.

[Effect]

The lane marking with the maximum length of the broken part or the lane marking with the largest sum of the lengths of the broken parts is identified as a broken line,
The meaning of each driving lane is determined based on this broken line.

〔Example〕

FIG. 1 is a diagram illustrating the basic concept of an embodiment in which the present invention is applied to a method for determining the east line on which an automobile is traveling.

The scene captured by the camera mounted on the moving vehicle is
For example, suppose that it is as shown in FIG. 1(a). That is, the driving path 1 is divided into two lanes, and a broken line 2 is drawn in the center of each lane. Further, it is assumed that the end of the left lane is indicated by a lane marking 3 and the end of the right lane is indicated by a lane marking 4, and each marking continues to the horizon 5.

The original image of such a scene is differentiated and converted into an edge image, which is represented by a dot display (not shown). Then, a Hough transform is performed on this edge image to obtain an array of feature points and a group of approximate straight lines.

From this group of straight lines, a straight line corresponding to the lane marking is selected based on the width of the road. As a result, three straight lines shown in FIG. 3(b) are obtained. Straight line 6 is broken line 2, straight line 7
corresponds to the lane marking 3, and the straight line 8 corresponds to the lane marking 4, and among these straight lines 6 to 8, the straight line 6 corresponding to the broken line 2 is identified as described later. The concept in this case is shown in the same diagram (C).
is shown. Then, the positional relationship between the detected straight lines 6 to 8 is determined as will be described later, and as shown in FIG. The straight line 8 is determined to be the lane marking 4 indicating the end of the right lane. This determination result is used to determine the direction of travel of the vehicle traveling on the road 1.

The main calculation methods based on the above-mentioned basic concept will be explained in detail below.

Among the detected straight lines 6 to 8 shown in FIG. 1(b), identification of the straight line 6 corresponding to the broken line 2 as shown in FIG. 1(C) is performed as follows. . Here, assuming the detection straight line shown in FIG. 2 as the object of identification, the x+Y coordinates of the starting point are (XS, YS), and the x, y coordinates of the ending point are (XE).

YE). Further, this detection straight line has an inclination of 45'' or more and less than 135° with respect to the X axis.

I will give an overview of identification. First, the edge image is scanned along a detection straight line by a fixed width W dot as shown by the arrow in the figure. Then, by comparing the intensity of each scanned edge point with a predetermined threshold and detecting the discontinuous part of the straight line, that is, the discontinuous part of the edge point sequence, the length of the maximum discontinuous part of the detected straight line is calculated. demand. This process is performed for each detected straight line, and the maximum length of the interrupted portion is determined for each detected straight line. Furthermore, the longest broken part is determined from among the maximum lengths of broken parts, and the lane marking line having this longest broken part is set as a broken line.

FIG. 3 is a flowchart showing details of this identification process.

First, set the starting point of the dot series along the straight line. The X and Y coordinates of each dot are (x.

y), the coordinates (x, y) of the starting point are x
-XS -W/2. y””YS. In addition, the memory MAX is used to store the maximum length of the interrupted part of a straight line.
is cleared (step 101). Next, each coordinate (x,
y) is read from the edge image along the detection straight line (step 102). Then, the read intensity data is set to a predetermined threshold T
It is determined whether it is larger than HL (step 103).

This threshold value THL is a value determined by principal component analysis when performing the Hough transform.

If the read edge strength is smaller than the threshold value, the edge point is assumed to correspond to a discontinuous part of the straight line, and a counter Break is incremented (step 104). This counter Break counts the number of edge points where the intensity data is less than a predetermined threshold value, and thereby measures the length of the broken part of the straight line. Then, it is determined whether the X coordinate of the read edge point is Xn+W/2 (step 105). Here, X
n is the X coordinate of the starting point of each scan when scanning with width W along the detection straight line, and corresponds to the X coordinate of the starting point of each arrow along the detection straight line in FIG. This X coordinate is Xn+W
/2 is determined by determining whether the end point of the scan along the arrow has been reached.

If the end point of the arrow has not been reached, 1 is added to the X coordinate value of the edge point to be scanned, and scanning is continued along the arrow (step 106). Also, when the end point of the arrow is reached, 1 from the y coordinate of the edge point to be scanned.
and prepare for the next scan along the arrow (step 1
07). Next, it is determined whether the X coordinate value of this subtraction result is smaller than the y coordinate value YE corresponding to the end point of the detection straight line, and it is determined whether scanning along the detection straight line has ended ( Step 108). If the X coordinate value of the subtraction result is equal to or larger than YE, it is assumed that the checking of the detected straight line has been completed, and the next other detected straight line is checked (step 109). If the X coordinate value of the subtraction result is smaller than YE, the X coordinate value of the edge point to be scanned is
W/2, and the coordinates of the next edge point to be scanned are adjusted to the starting point of the next arrow (step 110).

After the processing of step 1.06 or step 110 is completed, it is determined whether or not the read coordinate value of the edge point exists within the Hough transform area H (see FIG. 2) (step 111). If it is outside the Hough transform area H, the process returns to step 105 and the once set coordinate values of the edge point to be scanned are reset. If it is inside the Hough transform area H, return to step 02, proceed with scanning along the arrow about the detected straight line, and measure the length of the interrupted part by counter Brea.
It is measured by the count value of k.

On the other hand, if the intensity data of the edge point read in step 103 is larger than the threshold value THL, it is determined that the edge point corresponding to the read data is a continuous part of the straight line and that the interrupted part has ended. Then, it is determined whether or not the count number of the counter Break is larger than a value corresponding to the length of the broken part of the straight line stored in the memory MAX (step 112). Counter Br
If the count number of eak is larger than the number of memories stored in the memory MAX, the number of memories stored in the memory MAX is added to the counter Br.
It is rewritten to the count number of eak and used as new straight line discontinuation portion length data (step 113). Since no data is stored in memory MAX at the beginning of processing,
The count number of the counter Break is larger than the number stored in the memory MAX. In addition, memory MAX
1 2 - data is stored, the maximum discontinuous portion length data up to the processing time is stored in the memory MAX.

If the judgment result in step 112 indicates that the count number of the counter B r e a k is smaller than the number stored in the memory MAX, the length of the detected discontinuous portion is determined by the length of the detected discontinuous portion up to the processing point. This means that it is shorter than the length of the largest broken part. Therefore, the count number of the counter Break is set to O, and the length data of the currently detected broken part is cleared in preparation for the next detection of the broken part length (step 114).

In the explanation of the basic concept in FIG. 1(d), the positional relationship of each detected straight line is determined as follows, and the meaning of each straight line is assigned.

For example, the three detection straight lines g1°β2. shown in FIG.
Assume that ρ3 is obtained. In this case, the straight line Ωn]
Suppose that (m-1 to 3) are identified as broken lane markings, and the meaning of the other detected straight line ρ1 (i-1 to 3) at this time is determined, that is, the other straight line Ωl with respect to the straight line pm. The positional relationship is determined as follows. Here, each of the detection straight lines g1 to g3 is expressed using ρ and θ shown in FIG. ρ is the length of a perpendicular line drawn from the origin of the X, Y coordinates, which are Hough transform coordinates, to the detection straight line, and θ is the rotation angle from the X axis.

FIG. 6 is a flowchart showing the process of assigning meaning to each detected straight line.

First, the straight line Dm identified as a broken line is assumed to be the center dividing line of two lanes. According to the above expression method, the straight line Ωm is expressed as ρ−ρm
, θ−θm (step 201). Fourth
Among the detection straight lines Ω1 to Ω3 illustrated in the figure, the straight line fI3 corresponds to the straight line Ωm.

Next, the magnitude of θm is 0″ or more and less than 90° (0°50m
90°) (step 202). When this condition is met, θi roughly exceeds θm and becomes 90°.
It is determined whether the angle is less than (θm × θi × 90°) (step 203).

If this condition is satisfied, the straight line Ωi is the center dividing line 1
It is determined that the boundary line indicates the edge of the right lane of rn (step 204).

The relative relationship between the positions of the straight line (1 m, straight line 11i) at this time is shown in Figure 7 (a). That is, θm is within the range of 0° to 90°, and gi is larger than θm and 90° The dotted line in FIG. 7 indicates the horizon. Also, in step 203, gi is
If m x gi >90'' is not satisfied, the straight line gi is determined to be the boundary line indicating the end of the left lane of the center dividing line j?m (step 205).The straight line ρm at this time.

The relative relationship of the positions of gi is shown in FIG. 7(b).

That is, θm is within the range of 06 to 90°, but gi is greater than 90″.

Also, in step 202, θm is the conditional expression 0@≦θm
If gi does not satisfy 90°, then it is determined whether gi exceeds 90° and is less than 6 m (90'<gi x θm) (step 206). If this condition is met,
The straight line 11i is determined to be the center dividing line, a boundary line indicating the end of the left lane of Qm (step 207). The relative relationship between the positions of the straight lines Nm and jli at this time is shown in FIG. 7(C). In other words, θm exceeds 90° and gi is 9
It exceeds 0° and is smaller than θm. Further, in step 206, if gi does not satisfy the conditional expression 90° and θm, it is determined that the straight line ρi is a boundary line indicating the end of the right lane of the center marking line IIm (step 208). .

The relative relationship between the positions of the straight lines ρm, , 17i at this time is shown in FIG. 7(d). In other words, θm exceeds 90°, but gi is smaller than 90°.

Note that although the above description has been made regarding the case where there are three detection straight lines, it is not necessary to be limited to this number, and the number may be two, for example.

According to this embodiment, it is not only possible to determine the lanes of a traveling road as shown in FIG. 1(a), but also has the following advantages. That is, even if the entire traveling route 1 shown in FIG. 8 is not shown in the image, by identifying the broken line 2, it is possible to determine where the own army is located on the traveling route 1.

In addition, in the figure, the same part as in Figure 1 (a) is 5. 6 The same symbols are used for all cases.

Many technologies have been provided that allow vehicles or robots to run independently by recognizing white lines drawn on the road under limited conditions (such as inside a factory). There has been little introduction of technology to accurately recognize and run autonomously. Therefore, according to this embodiment, it is an effective technique for self-driving on a realistic driving road where there are broken lines and continuous lines are blurred or interrupted.

In addition, in the explanation of the above embodiment, broken lines were identified by finding the length of the maximum broken part of each detected straight line, and then finding the longest broken part of this maximum length. By performing such processing in accordance with the above embodiment, broken lines can be identified in the same manner as in the above embodiment. In other words, the sum of the lengths of the discontinuous portions of the straight lines is calculated for each detected straight line, and the detected straight line having the maximum sum among these sums can be set as a broken line. Even in this case, the sum of the lengths of the discontinuous portions can be determined by scanning the edge image along each detected straight line and comparing the intensity data of each edge point with a predetermined threshold.

Also, instead of finding the length of the broken part of the straight line, find the length of the continuous part of the straight line, and use the same method as in the above example to determine whether the lane marking is a broken line or not based on the length of the continuous part. can do. In this case, if the intensity data of each edge point along the detected straight line is larger than a predetermined threshold value, it is determined that it is a continuous portion.

〔Effect of the invention〕

As explained above, according to the present invention, the lane marking with the maximum length of the broken part or the lane marking with the largest sum of the lengths of the broken parts is identified as a broken line, and each lane marking is determined based on this broken line. A meaning is given to the driving lane.

Therefore, according to the present invention, it is not only possible to determine the presence or absence of lane markings, but also the type of lane markings, and the image information captured by the camera can be effectively used. .

7 8

[Brief explanation of the drawing]

Fig. 1 is a diagram for explaining the basic concept of an embodiment of the present invention, Fig. 2 is a graph for explaining the comparison between a detected straight line and an edged data image in this embodiment, and Fig. 3 is a graph for explaining the basic concept of an embodiment of the present invention. FIG. 4 is an image diagram for explaining an example of a straight line detected in this embodiment; FIG. 5 is a graph for explaining a method of expressing detected straight lines; FIG. FIG. 6 is a flowchart showing the meaning of each detected straight line in this embodiment (=I), FIG. 7 is a graph showing the relative relationship of the positions of each detected straight line obtained from the processing results of FIG. 6, and FIG. are other original images taken by the camera. 1... Driving path, 2... Broken line, 3... Boundary line indicating the edge of the left lane, 4... Indicating the edge of the right lane. border line,
5...Horizon, 6,7.8...Detection straight line.

Claims (1)

[Claims] 1. A first step of capturing an image of a driving route to obtain original image data, a second step of obtaining feature points included in the original image data, and a straight line that approximates the arrangement of the feature points. a third step of determining a group of straight lines; a fourth step of extracting a straight line corresponding to a lane marking drawn on the driving route from the group of straight lines; a fifth step of determining the length and identifying a broken line based on the length; and a sixth step of determining the positional relationship of the respective lane markings based on the broken line, A driving lane determination method characterized by determining a driving lane. 2. A claim characterized in that in the fifth step, the maximum length of the interrupted portion of the straight line is determined for each straight line extracted in the fourth step, and broken lines are identified based on the maximum length. The driving lane determination method described in 1. 3. The driving lane determining method according to claim 2, wherein in the fifth step, a straight line having the longest interrupted portion among the determined maximum lengths of interrupted portions is identified as a broken line. 4. In the fifth step, the sum of the lengths of the broken parts of each straight line extracted in the fourth step is calculated, and the straight line with the maximum sum of lengths is identified as a broken line. The method for determining a driving lane according to claim 1. 5. In the fifth step, each straight line extracted in the fourth step is compared with the feature point found in the second step, and if the intensity of the feature point is below a predetermined level, it is determined as an interrupted part. The driving lane discrimination method according to claim 1, claim 2, claim 3, or claim 4, characterized in that: 6. In the sixth step, the positional relationship of each lane marking is determined from the magnitude relationship between the angle of the broken line and the angle of each other lane marking. 3. The driving lane determining method according to claim 4 or claim 5. 7. A first step of imaging the driving route to obtain original image data, a second step of finding feature points included in the original image data, and a third step of finding a group of straight lines that approximate the arrangement of the feature points. a fourth step of extracting a straight line corresponding to a lane marking drawn on the driving route from the group of straight lines; and determining the length of a continuous portion of the straight line for each of the extracted straight lines. A fifth step of identifying a broken line based on the length, and a sixth step of determining a positional relationship between the respective lane markings based on the broken line, and determining the lane of the traveling road. A driving lane determination method characterized by: