JPH0656619B2 - White line detector - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a white line detection device for detecting a white line drawn on a traveling path such as an automated guided vehicle.

(Prior Art) In order to control the traveling of a traveling vehicle such as an automated guided vehicle, an external environment is captured by an image capturing unit such as a television camera attached to the traveling vehicle, and the external environment is recognized from the captured two-dimensional image. The technology to do is known. Conventionally, as means for recognizing the external environment for such a purpose, line-shaped or spot-shaped marks (markers) that can be sensed have been installed on a road surface or a wall surface, and those marks are detected by a light-receiving element provided on a traveling vehicle. It is known how to do it. (For example, JP-A-59-224505).

However, in such a conventional method, it is necessary to install a special marker on a road surface or a wall surface, so that it is difficult to change the layout of the external environment or there is a problem that the marker is dirty. It is desired to develop a method for stably recognizing the external environment without modification.

(Object and Structure of the Invention) The present invention has been made in view of the above points, and an object of the present invention is to recognize the external environment without changing the external environment such as re-installing markers when changing the layout or adding the markers. Therefore, pay attention to the fact that a white line is drawn in the factory to distinguish the passages of people and traveling vehicles from the workplace, and in order to achieve the above purpose, the overall configuration is shown in Fig. 1. In addition, from the two-dimensional observation image obtained by imaging the external environment including the white line, the edge detecting unit 100 detects the point where the gray value changes from dark to bright and the point where the gray value changes from bright to dark, and the set of detected points is detected. The edge at the left end of the area formed by this edge is the point where the gray value changes from dark to bright, and the edge at the right end is the point where the gray value changes from light to dark. That width should be detected An area within a predetermined width range corresponding to the thickness of the white line is detected by the white line candidate portion detection means 200 as a white line candidate portion, and the horizontal center point of the white line candidate portion is obtained by the pattern classification means 300 for the entire white line candidate portion. The skeletal line of the white line candidate portion is obtained by connecting the central points, the skeletal line is classified into an upper right line and a lower right line, and the white line detecting means 400 performs HOUGH conversion on the classified pattern to obtain a white line. Is configured to detect.

(Example) The present invention will be described below with reference to the drawings.

FIG. 2 shows a block diagram of an embodiment of the white line detecting apparatus according to the present invention.

This apparatus includes an image pickup unit 101 and an image memory 102 for storing image data picked up by the image pickup unit 101.
A differential calculation unit 103 that calculates edge information from the information stored in the image memory 102, a binarization unit 104 that binarizes the calculated edge information with a threshold value α, and the calculated edge information is inverted with a threshold value β. A binarizing unit 105 that binarizes, and thinning units 106 and 107 that thin the edge information extracted by these binarizing units 104 and 105 into a width of 1 respectively.
And an addition unit 108 for adding the two thinned images
A vertical edge removing unit 109 that removes vertical edges from the addition result, a white line candidate portion extracting unit 110 that detects a region that is considered to be a white line, and a center line (skeleton line) from the white line candidate region. A skeletal line extraction unit 111 for extracting,
From the extracted skeleton line, a pattern classification unit 112 that divides the skeleton line into upper right and left upper patterns, and an upper right upper skeleton line storage image memory 11 that stores the obtained pattern
3, the same image memory 114 for storing the right down skeleton line,
It is composed of a white line detection unit 115 which detects a white line from the data in these memories 113 and 114.

Next, the operation will be described with reference to FIG.

An image signal A (i, j) such as that shown in FIG. 9 (A) taken from the image pickup unit 101 is stored in the image memory 102 and sent to the differential operation unit 103, and is sent to the 3 × 3 differential filter P. By the action of, the edge image A ′ (i, j) is obtained.
For example, when the image to be handled can have 8-bit grayscale information (grayscale value 0 to 255) per pixel A '(i, j) = INT [{A (i + 1, j-1) + 2A (i + 1, j) + A (i + 1, j + 1)-
A (i-1, j-1) -2A (i-1, j) -A (i-1, j + 1) +1024} / 2 ³ ]. Where INT (X) is a function that gives the largest integer that does not exceed X, 1024 and 2 ³ is a section for both normalized.

The edge image A ′ (i, j) thus obtained is sent to the binarization units 104 and 105, and the binarization unit 104 binarizes A ′ (i, j) with a threshold value exceeding the gray value 128. By doing so (for the pixel whose gray value in A '(i, j) has finished the threshold value, the gray value is replaced with 255, and the gray value 0 is replaced for the pixels that do not exceed the threshold value). , A gray-scale change point from bright to dark that is equal to or larger than a certain value when raster scanning is performed is detected as a pixel having a gray-scale value 255, and in the binarization unit 105,
By inversion binarizing A ′ (i, j) with a certain threshold value which does not exceed the gray value 128, the gray value is set to 0 for the pixel whose gray value exceeds the threshold value in A ′ (i, j). If the pixel is replaced and the threshold value is not exceeded, the gray value 128
The scanning is performed by replacing the above) with a raster scanning), and a gray-scale change point from bright to dark that is equal to or larger than a certain value when raster scanning is performed is detected as a pixel having a gray-scale value 255.

Next, the images obtained by the binarization units 104 and 105 are thinned to a thickness of 1 by the thinning units 106 and 107, and a grayscale value of 128 at the grayscale change point obtained from the thinning unit 107 from light to dark. Then, the addition unit 108 adds the gray values of the same pixels as the results of the thinning unit 106 and the thinning unit 107. By the above processing, for example, the original image A (i, j) shown in FIG. 3 (a) to the edge component point image B (i, j) shown in FIG. 3 (b) are obtained.

Next, in the vertical edge removing unit 109, the vertical run with the gray value of 255 and the vertical run with the gray value of 128 in the edge composing point image B (i, j) are removed, respectively. Such an image C (i, j) is obtained.

Next, in the white line candidate portion extraction unit 110, the gray value of the vertical edge removed image C (i, j) is checked, and the gray value at the left end is 25.
5. Find a horizontal area where the gray value at the right end is 128 and the width x at both ends satisfies W / 2xW with respect to W given in advance in FIG. , A gray value 255 is given as a white line candidate portion, and an image D (i, j) as shown in FIG.

FIG. 4 is a flowchart of white line candidate portion extraction. Vertical M
Vertical edge-removed image C (i,
Regarding j), raster scanning is performed in the horizontal direction from the upper left of the screen (F-1) with i = 1 and j = 1. During this time, image C
The gray value of (i, j) is checked, and the point where the gray value becomes 255 is checked (F-2). When a pixel having a gray value of 255 is found in the image, the horizontal search range W determined by the following equation
Is set (F-3).

W = INT (j / 20) Here, 20 in the formula is an example of a value predetermined by the user depending on the layout of the camera, the thickness of the white line, and the like. ,
Next, it is confirmed that W is not 0 (F-4), and when it is confirmed, K is calculated from K = i + W (F-5), and until K ≧ N (F-6), the image C (i, j) is obtained. ) The gray value is 128
(F-7). When the gray value becomes 128, L = i is set (F-8), and D (L, j) = 255.
Then, the pixel to be obtained is obtained (F-9). After this processing, the process returns to step (F-2) and the above-described processing is repeated for i = 1 to N and j = 1 to M. In this way, the screen is searched from left to right to extract the maximum area that does not exceed W from the image C (i, j), and the white line candidate portion image D (i, j) as shown in FIG. j) is obtained.

That is, also in this processing, it is assumed that the pixel at the upper left of the screen is first scanned in the horizontal direction, and i = 1 and j = 1 (P-1).

First, it is checked whether or not the gray value of the white line candidate portion image D (i, j) is 0 (P-2), and if it is 0, it is determined that there is no white line candidate portion and the pixel is moved horizontally by one pixel. On the other hand, image D
If the gray value of (i, j) is not 0, it is assumed that there is a white line candidate portion and A = 1 is set (P-3), and then i is set to i + 1 (P-
4) Until i = N (P-5), it is continuously checked whether or not the density value of the image D (i, j) is 0 (P-6). 1 scanning point
Every time the pixel is moved in the horizontal direction, A is incremented by 1 (F-7). When D (i, j) = 0 in step (P-6), Is calculated to obtain K (P-8), and a pixel E (k, j) = 255 having the K as the x coordinate is calculated (P-9).

Returning to step (P-2), the above-described processing is repeated for i = 1 to N and j = 1 to M to obtain the skeleton line for the entire screen. As a result, a skeletal line image E (i, j) as shown in FIG.

Next, in the pattern separation unit 112, the skeleton line image E
An upper right skeleton line F (i, j) and a lower right skeleton line G (i, j) are extracted from (i, j). Ie E
In (i, j), the skeleton line constituent pixel (shading value 255
When a 3 × 3 mask such that the (pixels of the pixel) comes in, the pattern in the mask becomes one of the patterns shown in FIG. 6, and therefore the upper right pattern and the lower right pattern can be easily classified. . The upper right skeleton line F obtained in this way
The images of (i, j) and the right down skeleton line G (i, j) are as shown in FIGS. 3G and 3H and are stored in the image memories 113 and 114, respectively.

Next, in the white line detection unit 115, these skeleton lines F
White lines can be detected by applying HOUGH conversion to (i, j) and G (i, j), respectively. The white line detection result is as shown in FIG. Figures l _L and l _R are the detected white lines.

The HOUGH transform is a method for detecting a straight line from a point sequence such as F (i, j) and G (i, j), and is widely used in the field of image processing. The HOUGH conversion used this time will be described below with reference to FIG. 7 by taking F (i, j) as an example.

FIG. 7B shows the reference line y = α provided on the same FIG.
When the point P is changed from P ₀ to P _n , the straight line connecting the point P and the skeleton-constituting point A represents the locus of the point Q intersecting the image end. Points on the parameter plane (P _i , Q
_{In i} ), the gray value is incremented each time the trajectory passes in order to represent how many skeleton line constituent points are on the line segment connecting P _i and Q _i . For example, in Figure 7 (b),
Given gray value 2 at the intersection of the projection pattern l _B projection pattern l _A and point B of the point A, other l _A, a point on l _B gives the gray value 1.

FIG. 8 shows a flowchart of the HOUGH conversion of F (i, j), in which the grayscale value of the parameter plane is R (P,
Q).

First, the point P is set to P ₀ (W-1), the pixel at the left end on the reference line (i = 1, j = α) is started (W-2), and the shade of the upper right skeleton line F (i, j) is changed. Increase i from 1 to P and j from α to M until the value reaches 255. Q when the gray value of the upward skeletal line F (i, j) reaches 255
₁ is calculated (W-4), and the gray value R of the parameter plane is calculated.
Increment (P, Q) by 1 (W-5). Here, the maximum gray value of the parameter plane V = 0, X = P ₀ , Y =
Starting from Q ₀ (W-6), X from P ₀ to P _n , Y
Is changed from Q ₀ to Q _n, and the gray value R of the point (X, Y) is changed.
A point (X, Y) at which (X, Y) becomes larger than the maximum gray value V (= 0) is obtained.

Let P _i (X) giving the maximum gray value V be S, and let Q _i (Y)
Let T be the maximum gray value V (W-8). When the maximum gray value V is larger than the threshold value V _TH , there is a white line, and when it is small, there is no white line (W-9).

In this way, all the given skeleton constituent points are processed, and the point having the maximum gray value on the parameter plane (P, Q) can be regarded as the optimum straight line.

Here, for example, considering that only one white line is visible in the original image, for example, when only the white line in the upper right corner is visible, there is no white line data in G (i, j) extracted as the skeletal line in the lower right direction. It is not included. Therefore G
It is expected that the maximum gray value of the parameter plane when HOUGH conversion of (i, j) becomes smaller than that when the white line is visible. Therefore, by setting a threshold value to the maximum gray value V on the parameter plane when the HOUGH conversion is performed, it becomes possible to deal with even when only one white line is visible. In the above example, the skeletal line G (i, j) is descending to the right.
Since the maximum gray value of the parameter plane obtained from the above is less than V, the white line on the right is not detected.

The white line position detected in this way can be regarded as a passage end position, and can be used for, for example, traveling control of an automated guided vehicle.

Incidentally, it can be regarded as vanishing point of FIG. 3 the two was determined by HOUGH transform as shown in (i) white line l _L and l _R fulcrum V _P of the observation image.

(Effects of the Invention) As described above, in the present invention, in the observation screen, the point where the density changes from dark to bright is the left end, and the point where the density changes from bright to dark is the right end is close to the white line width. It is assumed that the center point of the area is detected as a white line candidate part, and the white line is detected by setting a threshold value to the maximum gray value of the parameter plane obtained by HOUGH conversion. Since it is possible to recognize the external environment without using a marker installed on the wall or a wall surface, there is no need to re-install the marker or newly install the marker even if the layout of the factory is changed or added, so it is highly versatile. The facility cost can be significantly reduced, and there is no troublesome maintenance, which is advantageous in running cost. In addition, the detection is not affected by some dirt on the white line,
The white line position can be detected stably. Further, there is an advantage that the HOUGH conversion for obtaining the white line can be stably and quickly performed.

[Brief description of drawings]

FIG. 1 is an overall configuration diagram of a white line detecting device according to the present invention, and FIG.
FIG. 3 is a block diagram of an embodiment of a white line detecting apparatus according to the present invention, FIG. 3 shows images obtained at each step of image processing of white line detection according to the present invention, (a) is an observation image,
(B) is an edge point image, (C) is a vertical edge removed image, (D) is a horizontal search range, (E) is a white line candidate image, (H) is a skeleton line image, and (G) is at the upper right. Skeletal image,
(H) is a right-down skeleton image, (L) is an image showing white line detection results, FIG. 4 is a flowchart of white line candidate portion extraction, FIG. 5 is a flowchart of skeleton line image refining, FIG. 6 is a skeleton line pattern, FIG. 7 is an explanatory diagram of HOUGH conversion, and FIG. 8 is H
It is a flowchart of OUGH conversion. 100 ... Edge detecting means, 200 ... White line candidate portion detecting means, 300 ... Pattern classifying means, 400 ... White line detecting means

Claims (1)

[Claims] 1. Edge detection for detecting an edge formed by a set of detected points by detecting points at which gray values in an observed image including a white line change from dark to bright and points at which light and dark change throughout the image. Means, and the left edge of the area formed by the edges is the point where the gray value changes from dark to light, and the right edge is the point where the gray value changes from light to dark, and its width is the width of the white line. White line candidate portion detection means for detecting a region within a predetermined width range corresponding to the thickness as a white line candidate portion, and a horizontal center point of the white line candidate portion is obtained for the entire white line candidate portion and the white line is connected by connecting the center points. A pattern classifying unit that obtains the skeleton line of the candidate portion and classifies the skeleton line into an upper right line or a lower right line, and a white line detecting unit that detects a white line by performing HOUGH conversion on the classified pattern. White line detection apparatus characterized by having.