JPS63103381A - Erase point detector - Google Patents

Abstract

PURPOSE:To quickly and accurately detect an erase point by forming a cumulative line image in accordance with the frequency in inclination of lines connecting the constituting points of an object and erase point candidate points and determining the erase point by this cumulative line image. CONSTITUTION:The image of the object is picked up by an image pickup part 201, and the edge constituting point of the object is extracted from the picked-up image by an edge extracting means 100. The cumulative line image is formed in accordance with the frequency in inclination of line connecting edge constituting points extracted by the edge extracting means 100 and erase point candidate points to be on a prescribed line. The erase point is determined in accordance with this cumulative line image by an erase point discriminating means 102.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a vanishing point detection device that is attached to an automatic guided vehicle or the like and used to identify three-dimensional objects or the like.

(Prior Art) In order to control the running of an automatic guided vehicle, it is necessary to identify, for example, a three-dimensional object from a two-dimensional image captured by an imaging unit such as a television camera attached to the automatic guided vehicle. Vanishing point detection technology is conventionally known for detecting and identifying the orientation, position, etc. of a three-dimensional object from a two-dimensional image.

Generally, when an object is imaged as a two-dimensional image using a fluoroscope,
As shown in FIG. 12(a), rectangular parallelepiped objects 401, 4
02 parallel ridge lines (hereinafter referred to as parallel lines >403,404
．． Or, the apparent distance between 405 and 406 decreases as the parallel lines move away from the camera, eventually reaching O''.

This point AQ at "O" is called a "vanishing point". That is,
Vanishing point detection technology detects the vanishing point from the apparent distortion that occurs as a result of imaging a three-dimensional object with depth as a two-dimensional image, and uses this to detect information about the three-dimensional object, such as the inclination of the object's surface. This is a technology that detects things such as direction.

To detect this vanishing point, KENDER first expressed the parameters of the linear equation in the two-dimensional drawing into a pseudo-Hough plane using HOUGH transformation. Since one straight line in the real image space is represented as one point on the pseudo-Hough plane, the vanishing point in the real image space can be determined by searching for a straight line on the pseudo-Hough plane. KENDER
In this way, the pseudo-Hough plane was roughly Hough-transformed, a straight line on the pseudo-Hough plane was determined, and the vanishing point AQ was detected.

In addition, as shown in Figure 12 (b), 0HTA et al. derive the distance ratio from the area ratio when affine transformation is applied to the two texture components 407 and 408 in the screen, and from this, the vanishing point BQ is calculated. For example, 1981, when determining the location.
February issue of “Information Processing” Vol. 24, No. 12, pp. 1421-
Page 1427, Katsushi Ikeuchi).

When the vanishing point is determined in this manner, the inclination of the plane of the three-dimensional object is determined from the vanishing point, and thereby information regarding the three-dimensional object can be obtained from the two-dimensional image.

However, in such conventional vanishing point detection devices, in order to detect the vanishing point, it is necessary to apply Hough transform twice according to KENDER's algorithm, and the algorithm of 0HTA et al. Since information had to be detected, if there are many edge constituent points, a huge amount of calculation is required and it takes a considerable amount of time to detect the vanishing point, making it difficult to quickly control the movement of automatic guided vehicles etc. The problem is that it can't be done.

(Object and Structure of the Invention) The present invention has been made in view of the above-mentioned points, and an object of the present invention is to quickly detect a vanishing point. To achieve this object, the overall structure is shown in FIG. As shown, the imaging unit 2
01, the edge constituent points of the subject are extracted from the captured image by the edge extracting means 100, and the edge constituent points extracted by the edge extracting means 100 and the vanishing point candidate points that are supposed to be on a predetermined line are The cumulative straight line image forming means 101 forms a cumulative straight line image based on the degree of inclination and chainness of straight lines connecting the straight lines, and the vanishing point is determined by the vanishing point determining means 102 from the cumulative straight line image thus formed. .

(Example) The present invention will be explained below based on the drawings.

FIG. 2 is a block diagram of an embodiment of the vanishing point detection device according to the present invention.

In FIG. 2, the vanishing point detection device includes an imaging unit 201, an image memory 202 that stores images captured by the imaging unit 201, and a series of vanishing point detection units for the images stored in the image memory 202. An image processor 203 that processes
, image memories 207 and 208 in which results of processing by the image processor 203 are stored, and a coordinate storage area 209.

The imaging unit 201 is a camera 31 that captures an image from an object.
0, and an image input unit 311 that converts an analog image from a camera 310 into a digital image and sends it to the image memory 202.

The image processor 203 performs edge detection processing and binarization processing on the image stored in the image memory 202 and stores the resulting edge image in the image memory 207.

As is well known in the edge detection process, for example, a 5OBEL operator that performs the negative differentiation of +A+28+C-G-2H-I l +lA+2D+10G-C-2F-1l using filtering coefficients as shown in FIG. etc. are used.

Furthermore, the image processor 203 extracts edge constituent points (XH,
yH) and store it in the coordinate storage area 20.
9, and then a vanishing point detection process is performed.

In the vanishing point detection process of this embodiment, assuming that the camera 310 is oriented horizontally with respect to an object (not shown), the vanishing point candidate point on the flat road is as shown in FIG. The vanishing point candidate point is on the center line (y□=N/2> of an image with NXM pixels because it is not actually Hiratanji). ).

That is, when detecting the vanishing point, the image processor 203
As shown in Figure 5, first set a point (x□, y□) on the center line (y□=N/2>, then set this point (X
, V□) and each edge constituent point (xH, y haze (i=1~K
) is determined as j=(x・-y.)/(XHV□) ■, and the angle θ is calculated from the slope j of this straight line as t an-
1j, and as a result, the linear slope frequency αj for each angle tan-1j is determined as shown in FIGS. 6 and 7. Points (x□, y□)
is a vanishing point candidate point. In the example of FIGS. 6 and 7, a certain point (x□) on the center line.

The angle of inclination of the straight line t an-1j with yo) is 70”~
This means that there are 10 edge constituent points (×·2y haze) that are 80', that is, the frequency αj of 70'' to 80' is ``10''.

The straight line y-yo=j (x-x□) is accumulated in the image memory 208 as a straight line image passing through the current point (x□, y□) as shown in FIG. 6th
In the example shown in the figure and FIG. 7, the inclination angle jan-1j=70
'~80'' (for example, adopt the median value of 75°)
The cumulative straight line image H(x,
The pixel value of y) is increased by "'1".

The image processor 203 calculates all points on the center line (x□
, y□ ) (x□ = 1~M, y□ = N/2
)J (X Xo)' and store it in the image memory 20.
8 is repeated to complete the cumulative straight line image H(X, V) as shown in FIG. 9, and based on this cumulative straight line image H(X, y), the maximum cumulative value MAX (H(
The coordinates (X, y) of the image memory 208 giving the coordinates (X, y)
) as the vanishing point AP or the cumulative straight line image H(
The vanishing point AP is determined as the position of the center of gravity of X, V).

The operation of the vanishing point detection device having such a configuration will be explained using the flowcharts shown in FIGS. 10 and 11.

Figure 10 is a flowchart showing the overall processing flow of the vanishing point detection device, and Figure 11 is step (F-8) in Figure 10.
3 is a flowchart showing a detailed flow of vanishing point detection processing in FIG.

In FIG. 10, in step (F-1>) the imitation image unit 20
In step 1, an image input process of inputting an image of a target object (not shown) is performed, and in step (F-2>, the input image is stored in the image memory 202. Next, in step (F-3>), the image processor 203 is the image memory 20, for example, by the 5OBEL operator as described above.
Edge detection processing is performed on the image stored in 2.
After that, binarization processing is performed on the image subjected to edge detection processing in step (F-4>), and step (F-5
The edge image binarized in > is stored in image memory 2.
Store in 07.

Next, in step (F-6>), the image processor 203 extracts the coordinates (×·, 'y'H) (i=1 to K) of the edge constituent points from the edge image stored in the image memory 207. Then, in step ( At F-7> step (coordinates □J, yH of the edge constituent points extracted at F-6>
) is stored in the coordinate storage area 209.

The coordinates (×・, yi
) (i=1 to K), vanishing point detection processing is performed in the stamp (F-8>).

As mentioned above, the vanishing point detection process of this embodiment is performed at the center line of an image having NXM pixels (VO=N
/2> Assuming that the vanishing point is above, this center line (y□
= N/2) on the point (xo.

The vanishing point is detected by creating a straight line cumulative image from the slope frequency of the straight line connecting V□) and the coordinates (×·, y haze) of the edge constituent points.

Referring to FIG. 11, in step (P-1), the image memory 208 in which the linear cumulative image is created is cleared and initialized. Next, in step (P-2), a point (X□
, y□), and X□ = 0. ¥0 =N/
Set it to 2.

Next, in step (P-3>, the center line (y□=\/2
) to increase the X coordinate X□ of the point (x□, y□) on
Add "1" to the coordinate xo. In this case, since the X coordinate Xo was cleared in step (P-2>), the X coordinate WXo becomes "1" in step (P-3>).

Next, in step (P-4), the cheaples of frequency α· for each angle tan-1j as shown in FIG. 6 are cleared. That is, in the table shown in FIG. 6, the frequencies αj for all angles are cleared. Next, in step (P-5), the coordinates (X·, y of the edge constituent point identifier i of the haze are cleared).

Now, the processing from steps (P-6) to (P-10) consists of one point (X□, y□) on the center line (y□=N/2>) and the edge component point (xH).

Vi) The inclination frequency α· of the straight line connecting (i='l to K) is determined, and when the frequency αj is greater than the threshold value, the cumulative straight line image @H(X, V) is incremented.

First, in step (P-6), the coordinates (×
, yH) is incremented by "1". In this case, since the identifier i is cleared in step (P-5>), the identifier i becomes at 1 to in step (P-6).In other words, initially, the coordinates (×, yH) of the edge constituent points are Next, step (P
-7), the coordinates of this edge constituent point (×・, 'y'i
) and the step (Pl -3) is the point (xo,
V□ ') with a straight line as shown in Figure 5, and the slope j of this straight line is j = (yl 'io)/ (xl -x□ )-1
- Calculate the angle tanJ-θ1 from this slope j, and add 111 ft to the frequency αj corresponding to this angle θ1. For example, if this angle θ1 is in the range of 70' to 80°, then in FIG.
"1 pa is added to the frequency αj corresponding to the range of 70' to 80°. In this case, since the processing is related to the coordinates (X, yl) of the first edge constituent point, the angle j an"
The frequency αj corresponding to j of 70' to 80 degrees is "1 Pa," and the frequencies corresponding to other angles are "OP1."

Next, in step (P-8>), it is determined whether the @degree αj added in step (P-7) has become larger than the threshold value T. For example, if 7'' is set as the threshold value T, in this case, the frequency αj Since IJ 1 pt is smaller than the breakage threshold T, the process proceeds to step (P-9>).

In step (P-9>, the coordinates of the edge constituent points (×・,
The identifier i of Vi) is the identifier of the last edge constituent point ゛K
It is determined whether the value becomes equal to 11 or not.

In this case, the identifier i is "1P" and not KI?, so focus on the coordinates (X2, V2) of the next edge constituent point and perform the processing of steps (P-1> and (P-8>). Return to step (P-6) again to repeat.

In step (P-6), the identifier i is incremented by "1" to 44299. This allows step (P-7>
Now, in the same way, the coordinates of the next edge constituent point (x2.\2
> and the center line (r'(
>=N/2) on the point (Xo.

yo) with a straight line as shown in Figure 5, and the slope j of this straight line is j = (V2-V□) / (x2-x□)-1
・ Calculate the angle tan J = θ2 from this inclination j, and add 1" to the difficulty level αj corresponding to this angle θ2. For example, if this angle θ2 is 70' like the angle θ1,
If it is within the range of ~80°, the angle tan is shown in Figure 6.
-1j ranges from 70° to 80°, the frequency αj is 742 If
Therefore, the frequency αj corresponding to the other angles is "Off". Unlike this, the frequency αj corresponding to the angle ~80' is "1", and the frequency αj corresponding to the angle 160° to 170° is "Off".
becomes ゛1pa, and the frequency αj corresponding to other angles is Ote
It becomes.

In this way, the coordinates (xi) of the edge constituent points.

yH) is increased from "1pa" to "Kn," and the coordinates (x・, yl) of each edge constituent point and the center line (V(>=
In the process of adding up the frequency αj for the slope j of the straight line connecting a certain point (Xg, 'i10) on N/2>,
When the frequency αj becomes larger than the threshold T in step (P-8>, the angle t corresponding to this frequency α・
Proceed to step (P-10>) to generate a straight line of an' j and accumulate it in the image memory 208. For example, in the example shown in FIGS.
The frequency αj where j corresponds to 70' to 80' is when the identifier i is "
Since it becomes "10'' when K 11, the frequency αj reaches the threshold T by the time the identifier i becomes at K te,
The process will proceed to step (P-10>).

In step (P-10>, from the slope j when the angle is 70' to 80', the straight line y-y□=j(x
x□) and accumulates it in the image memory 208. As a result, the cumulative straight line image H(x,
y), the pixel value that the straight line y-y□ = j (x-x□) passes through increases by "1", and one point (x□, y(> ) for the cumulative straight line image H(
X.

End the accumulation of y). And the center line (y□ = N/2
) returns to step (P-3>) again to perform the accumulation to the cumulative straight line image H(X, V) for the next point (X□, V□) on ).

On the other hand, unlike the examples shown in FIGS. 6 and 7, the identifier i
becomes ``K tt, then all angles t an-1j
When the frequency α・does not reach below the threshold for
Therefore, the cumulative straight line image H(x, y
＞ without participating in the formation of the central line (yO=N/2＞
Next point above (x□.

yo) to the cumulative straight line image H(X,V), the process returns to step (P-3>) again.

In step (P-3>), the X coordinate X□ of the point (x, y) on the center line (V□ = N/2) is increased by "1".In this case, the X coordinate X□ is "1". ′°, so it becomes 2″. By this, the point (X□, ’1
0 ) X coordinate X. In the case where is "2", steps (P-4> to (P-10)) are repeated in the same way to obtain the cumulative straight line image H(X).

y) is accumulated.

In this way, in step (P-11>) the center line (
'110=N/2>X of point (X□, '!□) above
When the coordinate Xo becomes “M”, the center line (V□=N/
2> All points on (x(>,y□) (xo=1~M)
Cumulative straight line image H(X.

The accumulation of y) is completed. As a result, the cumulative straight line image H(X) is stored in the image memory 208 as shown in FIG.

y) is completed.

Next, the process proceeds to step (P-12>, where a vanishing point determination process is performed based on the completed cumulative straight line image H(X, V). That is, the processor 203 processes the cumulative straight line image H(X, V) on the image memory 208. ), the maximum cumulative value MAX
The coordinates (x, y) that give (H(x, y)) are determined as the vanishing point AP.

Alternatively, the processor 203 may generate a cumulative straight line image H(X).

y) center of gravity position can be determined as the vanishing point AP.

In this way, the vanishing point detection device of this embodiment connects the point (X□, y□) on the center line where the vanishing point is supposed to exist and each edge constituent point, and calculates the slope frequency αj of the S-line from the point ( x.
Generate one straight line y V□=j(X-Xo>) passing through y□), generate such a straight line for all points (X□, y□) on the center line, and use them as a cumulative straight line image. Since the vanishing point is detected by simple arithmetic processing from the cumulative straight line image, it is possible to detect the vanishing point at high speed and with high accuracy.

(Effects of the Invention) As explained above, in the present invention, a cumulative straight line image is formed from the slope frequency of the straight line connecting the edge constituent points of the object and the vanishing point candidate points that are supposed to be on a predetermined line. Since the vanishing point is determined from the cumulative linear image, the vanishing point can be detected quickly and accurately, making it possible to improve the travel control speed of an automatic guided vehicle or the like.

[Brief explanation of the drawing]

FIG. 1 is an overall configuration diagram of a vanishing point detection device according to the present invention, and FIG. 2 is a block diagram of an embodiment of a vanishing point detection device according to the present invention.
Figure 3 is a diagram showing the filtering coefficients of the 5OBEL operator, and Figure 4 is a diagram showing the center line (y
□=N/2) Figure 5 is a diagram for explaining the vanishing point candidate point (X□, 'J□) on the center line shown in Figure 4 (x□. y□) Figure 6 is a diagram explaining the calculation of the slope frequency αj of the straight line connecting the edge component point (X・, yH). Figure 7 is a diagram showing the frequency αj with respect to angle as a histogram based on the table shown in Figure 6, and Figure 8 is a diagram showing the frequency αj table and histogram shown in Figures 6 and 7. 9 is a diagram illustrating a cumulative straight line image obtained by accumulating straight lines as shown in FIG. 8, and FIG. 10 is a vanishing point detection device of the present invention shown in FIG. 2. 11 is a flowchart showing the overall process flow of the process shown in FIG. 10, and FIG.
A) is a diagram explaining the concept of a vanishing point, and FIG. 12(B) is a diagram showing a conventional vanishing point detection process. 201... Imaging unit, 202, 207, 208... Image memory, 203... Image processor, 209...
Coordinate storage area, αj...frequency, H(x,y)...cumulative straight line image

Claims (1)

[Claims] An imaging unit that captures an image of a target object; an edge extraction unit that extracts edge constituent points of the target object from an image captured by the imaging unit; A cumulative straight line image forming means for forming a cumulative straight line image from the slope frequency of straight lines connecting the vanishing point candidate points, and a vanishing point determining means for determining a vanishing point from the cumulative straight line image formed by the cumulative straight line image forming means. A vanishing point detection device comprising: