JPH0896131A - Method and device for detecting figure - Google Patents

Abstract

PURPOSE: To easily detect a straight line or arc which passes a known pixel position by extending and applying Hough transformation, used to detect the straight line in an image, to the detection of the arc. CONSTITUTION: A start point search means 2 scans an image memory in one direction to search for the pixel at the start point of the figure, and decides the pixel as a start point 8. When this position is regarded as an origin and other pixel positions constituting the figure are denoted as (x1 , y1 ) (i=1, 2...n, n: the number of pixels having pixel values other than 0), a characteristic curve accumulating means 3 accumulates the pixel values in a cell of a cumulative memory 4 satisfying ϕ=(xi cosω+yi sinω)/(x<2> i +y<2> i ) in a (ω, ϕ) space dispersed into plural cells. A peak detecting means 5 finds the peak of the accumulated value as an integration point (ω<0> , ϕ0 ) and detects the figure being an arc which has its center r0 =1/2ϕ0 away from the origin in the direction of an angle ω0 and has a radius r0 when ϕ0 ≠0 and detects the figure being a straight line which passes the origin and is in the direction of an angle ω0 +π/2.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a figure detecting method and apparatus for reading figure data on a drawing and digitizing the drawing.

[0002]

2. Description of the Related Art Conventionally, a straight line and a curve drawn on a drawing are automatically detected and read by image processing technology and are digitized. Among them, the Hough transform is known as a method resistant to noise. (See, for example, Pattern C by Richard O. Duda and Peter E. Hart.
lassification and Scene Analysis ”, John Wiley & S
ons1973, page 336) A straight line detection method by Hough transform will be described with reference to FIG. At an arbitrary point (x, y) on the straight line 11,

[0003]

Is established. Therefore, obtaining θ and ρ from the image including the obtained straight line is a problem.

In digital image processing, when a straight edge in an image is obtained, an edge extraction process is first performed on the image, and then a straight line is detected from the edge image.
In an actual image, the plurality of edge pixels 12 constituting the straight line 11 completely lie on the straight line 11 due to the characteristics of the edge extraction processing and noise due to image input, and the pixel 12 forms a straight line without interruption. I can't expect that.

The Hough transform solves this in the form of accumulation in the parameter space. First, the (θ, ρ) space is divided into the accuracy and range necessary for obtaining a straight line as shown in FIG. Each of the divided areas is called a cell 15. Each cell has a memory for counting. Next, for each pixel 12, its position coordinates (x, y) are substituted into equation (2),

[0007]

According to the above, ρ is calculated for 0 ≦ θ <2π, and a graph of a characteristic curve as shown in FIG. 4 is obtained. That is, the pixel 2 on one straight line limits the parameters (θ, ρ) of the equation of the straight line to the values on the characteristic curve 13 in FIG. The value in the memory of the cell of FIG.
increase. Characteristic curves 13a, 13b, 13c obtained from a plurality of pixels 12a, 12b, 12c (the diagram conceptually shows the principle, and the correspondence between the pixel positions in FIG. 2 and the curve positions in FIG. 3 is not accurate) , 14 on the (θ, ρ) plane
When intersecting at, this point is called the accumulation point, and the coordinates of this point (θ
₀ , ρ ₀ ) are the parameters of the straight line to be found,

[0009]

Is the equation of the straight line.

That is, in order to obtain the parameters (θ ₀ , ρ ₀ ) of the straight line, the characteristic curve is drawn for all "1" pixels in the region, and the cells passing through are counted. Then, in order to detect the accumulation point 14, the count value of each cell in FIG. 3 is checked, and the linear parameter (θ ₀ , ρ ₀ ) is obtained from the position of the cell having the maximum value. By doing this, even if the pixels of the straight line are not correctly obtained during the image processing, the straight line can be stably detected.

The above is the Hough transform that is usually used to detect a straight line in an image. As described above, the detection of a certain figure can be performed by converting it into the accumulation in the parameter space with the parameters defining the figure as coordinate axes.
Attempts have been made to detect other figures. In particular, when a drawing is automatically read from an image, there are many demands for detecting a circle or an arc frequently drawn in the drawing. The above Hough transform is extended to the detection of circles / arcs as follows.
The extension method will be described with reference to FIG.

The circle 16 is represented by the coordinates (x ₀ , y ₀ ) of the center and the radius r ₀ , and is written as in equation (4).

[0014]

As described above, since the circle is represented by the three parameters x ₀ , y ₀ , r ₀ , it can be compared to the above Hough transform, and as shown in FIG. 6, (x ₀ , y ₀ , r ₀ ) Consider a three-dimensional parameter space. Then, this is divided into cells 17, and each cell has a memory for counting. "1" in the image
Equation (4) is modified for each pixel of the position (x _i , y _i ),

[0016]

Is calculated for each (x ₀ , y ₀ ),
(X ₀ , y ₀ , r ₀ ) The existence range of the parameter is represented as the curved surface 6 in the three-dimensional space. The plurality of pixels 18a, 18b, 18c. ． ． , A curved surface 19a, 1 indicating the existence range of the parameter as shown in FIG.
9b, 19c. ． ． (The correspondence between the pixel position in FIG. 5 and the curved surface in FIG. 7 is not accurate), and the accumulation point 2 in the three-dimensional space 2
0, the parameters of the arc to be obtained (x ₀ , y ₀ ,
r ₀ ) is required.

[0018]

However, in order to execute this similarly to the case of the straight line detection, the parameter space (x ₀ , y ₀ , r ₀ ) is discretized and divided into cells 17,
It is necessary to accumulate the number of times the curved surface 6 passes through the cell 19. In this case, the parameter space is a three-dimensional space,
Since the range of parameters is two-dimensional on a curved surface, the number of discretized cells becomes enormous, and the amount of processing for accumulating one pixel 18 becomes enormous. Therefore, this method was not very practical.

It is an object of the present invention to easily detect straight lines and arcs, which are important elements of a drawing, in a practical situation where a pixel through which a part of a figure passes is known in reading the drawing. To provide the equipment.

[0020]

According to a first aspect of the present invention, when a position of a part of a figure forming a straight line or an arc on a drawing is known, this position is set as an origin and another pixel position forming the figure is set. (X _i , y _i ) (i = 1, 2, ..., N; n is 0
The number of pixels with non-pixel values) in the (ω, φ) space discretized into multiple cells

[0021]

Pixel values are accumulated in cells satisfying the above condition, and the peak of the accumulated value is obtained as an integration point (ω ₀ , φ ₀ ), and φ ₀ ≠ 0
In the case of, r figure from the origin in the direction of the angle of ω _{0 0} =
It is detected that it is an arc having a radius r ₀ centered at a position ½φ ₀ away, and when φ ₀ = 0, the figure passes through the origin,
This is a figure detection method for detecting a straight line in the direction of an angle of ω ₀ + π / 2.

According to a second aspect of the present invention, there is provided an image memory for holding a drawing as an image, starting point searching means for scanning the image and detecting a part of a figure as a starting point, and discretely dividing the (ω, φ) space into a plurality of cells. a cumulative memory turned into the a position of the start point as the origin, the positions of pixels having a pixel value non-zero that is regarded as forming the figure _{(x i, y i) (} i = 1,2, ... ,,
n; n is the number of pixels with non-zero pixel values),
For the cell position (ω, φ) of the cumulative memory,

[0024]

A characteristic curve accumulating means for accumulating pixel values in cells satisfying the following conditions; peak detecting means for obtaining a peak of the accumulated value of the accumulating memory as an accumulation point (ω ₀ , φ ₀ ); From the position (ω ₀ , φ ₀ ), if φ ₀ ≠ 0, the figure is r ₀ = １ ／ in the direction of the angle ω ₀ from the origin.
An output is an arc with a radius r ₀ centered at a position φ ₀ away. If φ ₀ = 0, the figure passes through the origin and ω ₀ + π
And a circle / straight line calculating means for outputting a straight line in the direction of the angle of / 2.

[0026]

The present invention is intended for the case where the starting point 22 of the figure 21 is found on the image by a known method as shown in FIG. For example, when a straight edge portion is obtained as a binary image having “1” and the background is “0” by a process such as edge extraction, the entire image is scanned in the scanning direction 2 from the upper left.
3 is scanned, and the found “1” pixel is the starting point 22.
Is.

With this point as the origin and the coordinates of the center 24 of the figure 21 which is the arc to be detected are (x ₀ , y ₀ ) and the radius is r ₀ ,

[0028]

Holds, and any point (x, y) on this circle
By substituting into equation (4),

[0030]

When it is transformed,

[0032]

[0033]

Like the equation (1), the equation (8) is composed of the coordinates (x, y) of each point on the figure and two parameters ω and φ representing the figure. Therefore, the pixels 25a, 25b ,. ． ． Is calculated for the position (x _i , y _i ) of equation (8) with respect to 0 ≦ ω <2π, a graph 26 on the ω-φ plane as shown in FIG.
a, 26b,. ． ． Is obtained.

[0035]

Therefore, the ω-φ plane is divided into cells, and FIG.
Curves 26a, 26b ,. ． ． The coordinates (ω ₀ , φ ₀ ) of the accumulation point 27 can be obtained by accumulating the number of times of passing through and detecting the peak. The angle ω ₀ from the origin to the center 24 of the arc in FIG.
From the obtained φ ₀ , the distance (radius of the arc) r ₀ is

[0037]

[0038]

In the present invention, a straight line can be obtained similarly. Since the straight line corresponds to r ₀ = ∞, φ ₀ = 0. Therefore, as shown in FIG. 10, a straight line 29 passing through the starting point 28 has a direction 30 perpendicular to the straight line 29 as ω ₀ of the accumulation point 27.

[0040]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described with reference to FIG.
The image memory 1 stores a binary image including a figure, and the starting point searching means 2 scans the image memory 1 in a certain direction to search for a "1" pixel which is a starting point of the figure. For example, when the starting point searching means 2 scans the image memory 1 from the upper left in the scanning direction 7 like a television, a pixel which is a part of a circle is found. This pixel is set to start point 8 (X _s ,
Y _s ).

As shown in FIG. 11, the accumulating memory 4 is a counting memory in which the ω-φ space is divided into cells with required accuracy, and each cell has the accuracy of counting up to the number of pixels occupied by the detection range 9 below. . All counting memories constituting the accumulating memory 4 are all cleared to 0 when the starting point 8 is searched. In FIG. 11, the accumulative memory includes only the area of φ ≧ 0. According to the equation (9), a calculation result of φ <0 also appears, which can also be accumulated, but (ω,
If an accumulation point appears at (φ) = (ω ₁ , φ ₁ ), an accumulation point always appears at (ω, φ) = (ω ₁ + π, −φ ₁ ) in principle. Configuration.

In this embodiment, since a circle and a straight line having a radius equal to or less than a predetermined radius are detected, the characteristic curve accumulating means 3 sets a detection range 9 in the image memory 1,
The following processing is performed for each "1" pixel within this range.

(A) Each pixel coordinate (X _i , Y _i ) is translated to the coordinate system having the starting point coordinate (X _s , Y _s ) as the origin, and the normalized coordinate (x _i , y _i ) is converted. obtain.

[0044]

(B) For each (x _i , y _i ), φ is calculated for each ω (0 ≦ ω <2π) by the equation (9),
1 is added to the value of the cell (ω, φ) of the accumulation memory.

The above processing is performed for all "1" s in the detection range 9.
Is executed for the pixel of.

When this is completed, the peak search means 5 is started. The peak searching means 5 scans all the cells of the cumulative memory 4 and obtains (ω, φ) of the cell having the maximum count value, which is (ω ₀ , φ ₀ ). The circle / straight line calculating means 6 first checks whether or not φ ₀ is 0, and in this case, outputs ω ₀ + π / 2 as the direction of the straight line. Straight start (X _s, Y _s)
An external device (not drawn) from this result,

[0048]

Can be obtained as follows.

If φ ₀ is not ₀ , r ₀ = １ ／ φ
The distance from the starting point to the center of the circle is calculated by the formula (6)
, The center coordinate becomes (x ₀ , y ₀ ). Also in this case, the external device also reads the coordinates (X _s , Y _s ) of the starting point,
The coordinates of the center of the circle

[0051]

Can be obtained as follows.

In the embodiment, straight lines and circular arcs are obtained as binarized figures as shown in FIG. 1. However, when the figure is an image with shading, each pixel is either "1" or "1". An intermediate value other than “0” indicating the likeness of a figure can be taken.
Even in such a case, the present invention is applicable, and it is possible to detect a straight line / arc similarly by accumulating the pixel value in the accumulating memory 4 and finding the integration point.

[0054]

As described above, by adopting the configuration of the present invention, a straight line or an arc on the drawing can be stably detected from an image under the condition that some of its pixels have been obtained. You can do it. As a feature of the present invention, it is not necessary that the direct and circular arcs are completely connected, and even if there are a few noise pixels, they can be obtained by peak detection. Another great effect is that a straight line and a circular arc can be detected simultaneously with one configuration.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an exemplary embodiment of the present invention.

FIG. 2 is an explanatory diagram of a Hough transform for detecting a straight line according to a conventional technique.

FIG. 3 is an explanatory diagram of a Hough transform for detecting a straight line according to the related art.

FIG. 4 is an explanatory diagram of Hough transform for detecting a straight line according to a conventional technique.

FIG. 5 is an explanatory diagram of a method for detecting a circular arc according to the related art.

FIG. 6 is an explanatory diagram of a method of detecting an arc according to a conventional technique.

FIG. 7 is an explanatory diagram of a method of detecting an arc according to a conventional technique.

FIG. 8 is an explanatory diagram of the principle of the present invention.

FIG. 9 is an explanatory diagram of the principle of the present invention.

FIG. 10 is an explanatory diagram of the principle of the present invention.

FIG. 11 is an explanatory diagram showing a configuration of a cumulative memory of the present invention.

[Explanation of symbols]

 1 image memory 2 starting point searching means 3 characteristic curve accumulating means 4 accumulating memory 5 peak detecting means 6 circle / straight line calculating means

Claims (2)

[Claims] 1. When the position of a part of a figure forming a straight line or an arc on a drawing is known, this position is used as an origin, and other pixel positions forming the figure are defined as (x _i , y _i ) (i = 1
2 ,. ． ． , N; n is the number of pixels having a pixel value other than 0)
In the (ω, φ) space discretized into a plurality of cells, Is accumulated in a cell satisfying the following equation, and the peak of the accumulated value is obtained as an accumulation point (ω ₀ , φ ₀ ). When φ ₀ ≠ 0,
R ₀ = 1 / (2 figure is from the origin in the direction of the angle of ω ₀
φ ₀ ) An arc with a radius r ₀ having a center at a distant position is detected. When φ ₀ = 0, the figure passes through the origin and ω ₀ + π
A figure detecting method for detecting a straight line in the direction of the angle of / 2. 2. An image memory for holding a drawing as an image, a starting point searching means for scanning the image and detecting a part of a figure as a starting point, and a cumulative memory for discretizing (ω, φ) space into a plurality of cells. And the position of a pixel having a non-zero pixel value regarded as constituting a figure as (x _i , y _i )
(I = 1, 2,..., N; n is the number of pixels having a pixel value other than 0), and the cell position (ω, φ) of the accumulation memory is Characteristic curve accumulating means for accumulating pixel values in cells satisfying the following, and a peak of the accumulated value of the accumulating memory at an accumulation point (ω ₀ ,
a peak detection means for determining as a phi _0), the integrated point (ω _{0, φ} from the position of _0), .phi.0 the case of ≠ 0 is, r from figure origin in the direction of the angle ω _{0 0} = 1 /
An arc having a radius r ₀ centered at a position (2φ ₀ ) is output, and in the case of φ ₀ = 0, the figure passes through the origin, and ω ₀ +
and a circle / straight line calculating means for outputting a straight line in the direction of the angle of π / 2.