JPS6292085A - Recognizing system for symmetrical characteristics of graphic form - Google Patents

Abstract

PURPOSE:To recognize accurately the contents of the symmetrical characteristics of a graphic form even though this symmetrical graphic form has some distortions, by using the detailed contents of the pattern symmetrical characteristics as the feature parameter of the graphic form. CONSTITUTION:A symmetrical picture data producing means 110 produces the symmetrical picture data on the symmetrical graphic form obtained by giving a symmetrical shift to the relevant graphic form based on the input picture data. A symmetrical characteristics recognizing part 120 recognizes the symmetrical characteristics of the graphic form to be recognized based on the input character data or the picture data obtained by processing said input picture data and the symmetrical picture data supplied from the means 110. Thus the symmetrical characteristics of the graphic form can be recognized in detail from the input picture data even though the graphic form to be recognized has some distortions. This detailed symmetrical characteristics is used as a feature parameter for graphic recognition. As a result, the recognizing errors of the graphic form can be decreased.

Description

[Detailed Description of the Invention] [Summary] Symmetrical image data regarding a symmetrical figure of the figure is generated from input image data of the figure, and the input image data or image data obtained by processing it (for example, thickening process) and the symmetrical image data are generated. (for example, based on the degree of agreement between the two). As a result, even if the figure has some distortion, its symmetrical characteristics can be recognized in detail.

[Industrial application field]

The present invention is a figure symmetry characteristic recognition method that recognizes the symmetry characteristics of a figure, for example, whether the figure is axially symmetrical or rotationally symmetrical, from input image data obtained by reading figures such as symbols and characters. Regarding.

[Conventional technology]

Conventionally, as a figure recognition method that recognizes the original figure from input image data obtained by reading figures such as symbols and characters, there is a method that recognizes the original figure based on various characteristic parameters extracted from the input image data. FIG. 9 is a block diagram showing a conventional figure recognition method based on extracted feature parameters.

The observation unit 210 reads the figures drawn on the drawing and sends the obtained input image data to the preprocessing unit 220.

The preprocessing unit 220 performs noise removal and normalization processing on the input image data and sends the data to the feature extraction unit 230 . By performing this preprocessing, the recognition rate improves.

The feature extraction unit 230 extracts various feature parameters regarding the figure based on the input image data. Extracted features include, for example, straightness, end points, branch points, intersections,
There are concavities, convexities, loops, etc., and their positional relationships.

The recognition dictionary 240 stores data on characteristic parameters that figures belonging to each figure category should have.

The recognition unit 250 compares the feature parameters input from the feature extraction unit 230 with the feature parameters of each figure category read from the recognition dictionary 240, and detects a figure category that matches the feature parameters extracted from the input image data. Then, the shape of the input image data is recognized.

[Problem to be solved by the invention 3 There are many figures drawn in drawings that have symmetry, such as axially symmetric figures and rotationally symmetric figures, and especially figures written as design information in drawings have symmetry. There are many shapes.

If such a symmetrical figure is drawn by hand, it is difficult to draw the correct symmetrical figure, so the symmetrical figure actually drawn will be slightly deviated from the correct symmetrical figure.

However, as mentioned above, the feature parameters extracted in conventional figure recognition methods include linearity, end points, branch points, concavity,
These are convexities, loops, intersections, etc., and their positional relationships, and the detailed contents of the symmetry characteristics of the figure are not extracted feature parameters. For this reason, in conventional figure recognition methods,
When a figure entered as a symmetrical figure has some distortion, there is a problem in that it is often mistakenly recognized as another figure without symmetry.

The present invention makes it possible to recognize the figure correctly even if there is some distortion in the figure drawn as a symmetric figure in a drawing by using the detailed content of the symmetry characteristic of the figure as the feature parameter of the figure. Considering this, even if there is some distortion in the symmetrical figure drawn in the drawing as part of the figure feature extraction process,
The object of the present invention is to provide a method for recognizing symmetry properties of figures that can reliably recognize the contents of the symmetry properties.

[Means for solving problems]

The means taken by the present invention to solve the above-mentioned problems in the conventional figure recognition system will be explained with reference to FIG.

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

In FIG. 1, 110 is symmetrical image data generation means;
Based on input image data obtained from the recognized figure, symmetrical image data regarding a symmetrical figure obtained by symmetrically moving the figure is generated.

Symmetry characteristic recognition means 120 recognizes the symmetry characteristic of a figure based on input image data or image data obtained by processing the input image data and symmetric image data.

[For production]

Input image data obtained from a figure that is a recognition symmetry or image data obtained by processing the same is applied to a symmetric image data generation means 110 and a symmetric characteristic recognition means 120.

The symmetrical image data generating means 110 generates symmetrical image data regarding a symmetrical figure obtained by symmetrically moving the figure based on the input image data. Symmetrical movement processing of a figure includes a reversal movement process in which the drawing is reversed to a symmetrical position with respect to a predetermined axis including the X-axis and the Y-axis, and a rotational movement process in which the figure is rotated by a predetermined angle with respect to the origin.

The symmetrical characteristic recognition means 120 uses input image data or image data obtained by processing the input image data and the symmetrical image data generation means 11.
Based on the symmetrical image data input from 0, recognize the symmetrical characteristics of the figure that is symmetrical, such as which axis it is symmetrical about, how many rotation angles it is symmetrical about, etc. .

The number of symmetrical properties obtained from one figure is not limited to one,
For example, there may be a plurality of squares.

As described above, there is some distortion in the recognized figure,
The symmetrical characteristics of a figure can be recognized in detail from human image data. By using this detailed symmetry characteristic as a feature parameter during figure recognition, it is possible to reduce figure recognition errors.

〔Example〕

An embodiment of the present invention will be described with reference to FIGS. 2 to 8.

2 is an explanatory block diagram of the configuration of an embodiment of the present invention, FIG. 3 is an explanatory diagram of a figure to be recognized in the embodiment, and FIG. 4 is an explanatory diagram of a smoothing filter pattern used in the embodiment.
FIG. 5 is an explanatory diagram of the thick processing pattern and thick processing used in the same embodiment, FIG. 6 is an explanatory diagram of the figure center selection process in the same embodiment, and FIG. 7 is a symmetrical image data generation process in the same embodiment. FIG. 8 is an explanatory diagram of the lyricism recognition process in the same embodiment.

(Description of Configuration) In FIG. 2, the symmetric image data generation means 110 and the symmetric characteristic recognition means 120 are the same as shown in FIG.

In the symmetrical image data generation means 110, a figure center selection section 111 selects the center position of the figure to be recognized.

Reference numeral 112 denotes a symmetrical image data generation unit that generates various symmetrical image Y-data of figures to be recognized from input image data.

130 is a smoothing processing unit that performs processing for smoothing human image data.

Reference numeral 140 denotes a thickening processing unit that performs thickening processing on smoothed input image data as a processing process for the input image data.

(Description of Operation) The operation shown in FIG. 2 will be explained by dividing the operation of each component with reference to FIGS. 3 to 8.

(A) Description of operation of smoothing section 130 The operation of the smoothing section 130 will be explained with reference to FIGS. 3 and 4.

In FIG. 3, the white dots represent a figure PS consisting of a trapezoidal symbol, and the black dots represent noise PN (PN, .about.PNS) that is one pixel beyond the figure PS. In addition,
Although PSo and PN are at the same level (for example, H), they are shown with white dots and black dots in order to easily distinguish between them.

Noise consisting of isolated points existing in the vicinity of the figure ps, and whisker-like noise that protrudes by one or more pixels from the figure ps can be easily removed by various known smoothing processes, but as shown in FIG. The noise PN that jumps out by one pixel is often not erased and remains.

If such noise PN remains, the center of the figure PS will not be found correctly, and the degree of coincidence between the figure PS and the symmetric figure subjected to the symmetric movement process will be inaccurate (about this point,
(Refer to (E) Operation description of symmetrical characteristic recognition means described later), the smoothing processing unit 130 performs processing to remove such noise PN as well as isolated point and whisker-like noise. The noise PN removal process will be described below.

In Figure 4, there are 20 types of 3x3 dot-shaped smoothing filter patterns (PATH) (PATs+ to PARsz).
It is shown. White circles indicate pixels of the figure, and black circles indicate black pixels. Note that although both are at the same level (for example, H), in order to make the distinction between the two easier to understand, they are shown as white circles and black circles.

When the smoothing processing unit 130 detects any of the 20 types of patterns shown in FIG. 5 in the input image data,
Processing is performed to rewrite the central black pixel (H level) to a white pixel (L level).

This makes it possible to remove all the noise PN remaining in the figure ps. For example, PN is pattern PAT,
, PN is based on the pattern PA T s lz, PN is based on the pattern PATs+a, PN4 is based on the pattern FATS+6, and PN5 is based on the pattern PATS2.
Each is removed by ゜.

(B) Description of operation of thick processing section 146 The operation of thick processing section 140 will be explained with reference to FIG.

When figures drawn on a drawing are drawn by hand, the lines that make up the figures often have distortions such as small irregularities and curvatures. Such distortion around the figure can be removed by smoothing the periphery, but this process is complicated.

However, by processing the figure to make it thicker, the above-mentioned figure distortion can be easily removed and the symmetrical characteristics of the figure can be better recognized. The present invention is based on this new knowledge to perform thickening processing on the Kugata image data.

FIG. 5(B) shows a thick processing pattern (PAT). Thick processing pattern FAT.

is a 3x3 dot pattern 2, and when a black pixel (H level) is detected in the center, all adjacent white pixels (1 to 8, L level) in the 8 neighborhood are changed to black pixels. .

FIG. 5(A) shows a thick processing pattern PAT on the figure Pso.
The extended figure P obtained by performing one round of thick processing by ,
This shows Se.

The white dots indicate the original figure PSo, and the black dots indicate the figure parts expanded by the thickening process.

Note that, as already explained, the two are actually at the same level.

The thickening process is performed once or multiple times, and the number of times is as follows:
It is appropriately selected depending on the amount of distortion existing in the figure PSo.

The greater the amount of distortion, the greater the number of thickening processes. If the figure PSo is filled in accurately, this thickening process is not particularly necessary.

(C) Description of Operation of Graphic Center Selection Unit 111 The operation of the graphic center selection unit 111 will be explained with reference to FIG.

The center of the figure is located at the smoothing processing unit 130, as shown in FIG.
The figure ps is smoothed by ps.

The minimum rectangle RECT surrounding the figure is determined, and the center Cp of this rectangle RECT is set as the center of the figure. Obtained figure center C
The coordinate data of p is applied to the symmetric image data generation section 112 and the thick processing section 140.

(D) Description of Operation of Symmetrical Image Data Generation Unit 112 The operation of the symmetrical image data generation unit 112 will be explained with reference to FIG.

As shown in FIG. 7, the symmetrical image data generation unit 112 generates an X-axis symmetrical figure PSK, a Y-axis symmetrical figure PS7.90'' rotated figure ps, ., a 180° rotated figure P of the original figure ps-o.
S+a. , 2706 rotating figure PSZ? Each symmetrical image data for O is generated.

Now, the coordinates of the figure ps before the symmetrical movement process are (x,
y) and the corresponding coordinates of each symmetrical figure after the symmetrical movement process are (X, Y), then the symmetrical image data of each symmetrical figure can be obtained by the following calculations.

(aJ Symmetrical image data of X-axis symmetrical figure psX
t Y=-y (b) Symmetrical image data X=- of Y-axis symmetrical figure PSv
x, Y=y (symmetrical image data of C190° rotated figure PS,.) C
=-y, Y=x (symmetrical image data X of dl180° rotated figure PS+ao
=-x, Y=-y (e1270° rotated figure PS,... Symmetrical image data X
=y, Y=-x Note that the coordinates of the figure center Cp do not change due to each of these coordinate transformations.

(Explanation of operation of symmetric characteristic recognition means 120) The operation of the symmetry characteristic recognition means 120 will be explained with reference to FIG.

The enlarged figure ps obtained by the thickening process and the symmetrical figure (for example, 1800 rotation figure PS+ao) are overlapped with their figure centers Cp aligned as shown in Fig. 9, and the degree of coincidence of the overlap between the two is measured. .

- The degree is the total number of pixels of the symmetric figure (this is the original figure PSo
), and the symmetrical figure is an extended figure PS,
It is defined as the ratio of the number of pixels in the area that does not overlap with the .

In FIG. 8, the 180' rotation figure Psis.

The total number of pixels is 67 (this is equal to the number of pixels in the original figure PS), and the number of pixels in the part that overlaps with the extended figure PS (indicated by ■) is 45, and in the part that does not overlap The number of pixels (indicated by black circles) is 22. Therefore, its -significance is 22/67. Similarly, if we measure the degree of matching for other symmetric figures, we get the following results.

(a) X-axis symmetric figure PSX: 0/67=O%Cb
l Y-axis symmetric figure psV: 23/67=34% (C
190' rotated figure PS9°: 14/67=21% (d1
180° rotated figure P S+ao: 22/ 67=
33% te1270° rotated figure P 5zvo: l 2/
67=18% As is clear from this result, when the -growth is large,
If the figure is not symmetrical and the - degree is small, the figure is determined to be symmetrical, and the symmetrical characteristic of the original figure ps can be recognized from the symmetrical figure corresponding to the degree of coincidence. .

- The standard percentage value below which a figure is judged to be symmetrical is statistically selected based on data obtained by actually performing recognition processing for symmetry characteristics on many figures.
In this embodiment, 10% is used as a reference value, and if the value is less than 10%, it is determined that it is symmetrical.

If the reference value is 10%, the above-mentioned figure PSO is recognized to be symmetrical only about the X axis.

The symmetrical characteristic recognition means 120 uses the image data of the expanded figure manually generated from the thick processing section I40 and the symmetrical image data generation section 1.
Based on each piece of symmetrical image data input from 12, the degree of coincidence of each piece of symmetrical image data is determined, and its symmetrical characteristics are recognized.

As a result, the figure ps has X-axis symmetry, Y-axis symmetry, and 90'
It is possible to recognize rotational symmetry, 180° rotational symmetry, and 270″ rotational symmetry.

Although one embodiment of the present invention has been described above, each structure of the present invention is not limited to the structure of the above embodiment.

For example, in addition to the X-axis and the Y-axis, any desired angular axis may be employed as the symmetry axis. The rotation angle can also be selected to be any angle other than 9 (16, 180', 270').

A ratio value between the number of overlapping pixels and the total number of pixels may be used as the degree of matching for determining the presence or absence of symmetry.

〔Effect of the invention〕

As described above, according to the present invention, even if the figure to be recognized has some distortion, the symmetrical characteristics of the figure can be recognized in detail from the input image data. Then, by using this detailed symmetry characteristic as a feature parameter during figure recognition, it is possible to reduce misrecognition errors of figures.

[Brief explanation of drawings]

Fig. 1: A detailed explanatory diagram of the present invention; Fig. 2: an explanatory diagram of the configuration of an embodiment of the present invention; Fig. 3: an explanatory diagram of a figure to be recognized in the embodiment; Figure 4・
...Smoothing filter pattern (PA
Explanatory diagram of Ts), Fig. 5... Thick processing pattern used in the same example (
PAT, ) and thickening processing. Figure 6: An explanatory diagram of the figure center selection process in the same embodiment. Figure 7: An explanatory diagram of the symmetrical image data generation process in the same embodiment. Figure 8: Symmetry characteristic recognition in the same embodiment. Explanatory diagram of processing, FIG. 9: An explanatory diagram of a conventional figure recognition method. In FIG. 1 and FIG. 2, 110... Symmetrical image data generation means, 120... Symmetrical characteristic recognition means, 130... Smoothing processing unit, 140...
・Thick processing part.

Claims (4)

[Claims] (1) A figure symmetry characteristic recognition method that recognizes the symmetry characteristic of a figure from input image data obtained from the figure to be recognized, which (a) relates to a symmetric figure in which the figure is symmetrically moved based on the input image data; (b) symmetrical image data generation means (110) for generating symmetrical image data; (b) symmetrical characteristic recognition means (120) for recognizing symmetrical characteristics of a figure based on the input image data or the processed image data and the symmetrical image data; ), a method for recognizing symmetry characteristics of shapes. (2) The symmetrical characteristic recognition means (120) recognizes the symmetrical characteristic of the figure based on the degree of coincidence between the input image data or the image data processed therefrom and the symmetrical image data. A method for recognizing symmetry characteristics of figures as described in scope 1. (3) The graphic symmetry characteristic recognition method according to claim 1, wherein the processing for the input image data is a thickening process. (4) A graphic symmetry characteristic recognition method according to claim 1, wherein the input image data has been subjected to smoothing processing.