JPS6253869B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides a figure shape recognition processing method, particularly for a figure to be recognized that is vertically and/or horizontally symmetrical. Alternatively, addresses for read access to the image memory storing the image of the symmetrical figure are generated in the direction toward the central axis parallel to the Y axis and in the direction from the central axis to the figure frame. This invention relates to a figure shape recognition processing method for recognizing figures.

The shapes of symbols used in pattern diagrams used on printed wiring boards, maps, weather maps, etc. are often symmetrical vertically and horizontally. If general shape recognition is used to recognize such characteristic figures, complicated hardware will be used for feature extraction, and shape recognition will take a long time. For example, when recognizing the shapes of symbols with symmetry such as black circles, white circles, black triangles, and white triangles shown in Figures A to D in Figure 1, the distance from the cutting frame of these symbol shapes to the "black" ground is calculated. When performing shape recognition,
In the case of the black circle in A in the same figure, the distance from the cutting frame to the "black" ground in the horizontal direction, the distance to the "black" ground, and the "black"
For feature extraction that measures the distance from the background to the cropping frame, in the case of the white circle in Figure B, the distance from the cropping frame to the "black" background in the horizontal direction, the distance to the "black" background,
White distance, again "black" distance and "black"
Feature extraction is performed to measure the distance from the ground to the cutting frame, resulting in a different hardware configuration for each figure to be recognized. Alternatively, when performing feature extraction by measuring the distance from the center line passing through the center point of the figure to the point where it changes from "white" to "black" or from "black" to "white" toward the cutting frame, different Therefore, multiple hardware configurations are required for feature extraction of figure recognition.

The present invention aims to solve the above-mentioned problems, and focuses on the fact that the figure to be recognized has symmetry. The purpose of this invention is to create a single address on the image memory in the direction of the figure frame, and perform read access to the image memory based on the address information, thereby unifying the hardware configuration for feature extraction. Therefore, the figure shape recognition processing method of the present invention is
A figure shape recognition process in which a figure to be recognized that is symmetrical about an axis and/or a Y-axis is stored in an image memory, and the contents of the image memory are read out in the X-axis or Y-axis direction to extract features of the figure to be recognized. The method is equipped with an XY projection extraction unit that finds the center point of the figure stored in the image memory, and also extracts a central axis parallel to the X or Y axis passing through the center point from the figure frame in which the figure exists. A first address generation section that generates an address on the image memory in the direction toward the graphic frame, and a second address generation section that generates an address on the image memory in the direction from the central axis toward the graphic frame. The image memory is characterized in that read access to the image memory is performed based on address information generated by the first address generation section and the second address generation section, respectively. This will be explained below with reference to the drawings.

FIG. 2 shows an embodiment of the configuration for generating an address for read access to an image memory according to the present invention;
The figure is an explanatory diagram explaining how to count the number of pixels,
FIGS. 4A, B, C, and D are explanatory diagrams illustrating addresses for performing read access to the image memory.

In FIG. 2, 1 is an XY projection extractor,
This is a circuit unit that projects a symmetrical symbol figure onto the X-axis and the Y-axis, respectively, and finds the minimum and maximum points of each to obtain a circumscribing frame of the figure to be recognized. 2,
3 is an address generation counter section, 4 to 7 are address generation sections, and 8 is an image memory in which an image of a figure to be recognized is stored. 9 represents a feature extraction section.

The XY projection extraction unit 1 extracts the existence area of the symbol having symmetry stored on the image memory 8,
Determine its center position. That is, the video information of the figure of the symbol to be recognized is sequentially read from, for example, the image memory 8 and input to the XY projection extraction unit 1, and the
In the projection extraction unit 1, the video information is projected onto the X-axis and the Y-axis, respectively. As a result, the coordinates of the existence area of the recognition target symbol A (Xs, Ys), B (X _E , Ys), C
(Xs, _YE ) and D( _XE , _YE ) are obtained. Graphical frames A, B, C, and D in FIG. 3 represent the area where the recognition target symbol exists. Now that the coordinates of figure frames A, B, C, and D have been determined, the center point (X ₀ , Y ₀ ) of the figure can be calculated as {(X _E +Xs)/2, (Y _E +
Ys)/2}, its coordinates can be found.

The address generation counter sections 2 and 3 are counters that count the number of pixels in the graphic frames A, B, C, and D, of which the address generation counter section 2
counts the number of pixels from the upper left to the right along the X axis as follows. That is, half of the X-axis width AB = AE =
Shift one row down in the Y-axis direction every (X _E - Xs)/2, and 2AC = 2, which is twice the Y-axis width AC.
Count up to (Y _E - Ys) and press the above figure frame ABCD.
Instead of counting the total number of pixels in , the number of pixels in half of the figure frame AECF is counted twice. Here, E and F represent the midpoints of AB and CD, respectively.

Similarly, the address generation counter section 3 counts the number of pixels from the top left to the bottom along the Y axis as follows. In other words, half of the Y-axis width AC = (Y _E -
Ys)/2, shift one row to the right in the X-axis direction, and 2AB=2(X _E −
Ys), and instead of counting the total number of pixels in the above graphic frame ABCD, that graphic frame
Count the number of pixels in ABGH twice. Here, G and H represent the midpoints of AC and BD, respectively.

_The address generating unit 4 in FIG. 2 moves from the left and right graphics frames _AC and BD in FIG. An address for accessing video information stored on memory 8 is generated. This is used, for example, when extracting the external left and right contours of a cut out figure as shown in FIG. 4A. Above left and right figure frame AC
The address P (X, Y) on the image memory 8 in the horizontal direction from BD to the central axis EF is determined by the address generation counter unit 2 that counts the pixel G (Xc, Yc) in the figure AEFC in FIG. , the address generator 4 generates the address based on the following equation.

(1) Yc in the Y-axis direction of address generation counter section 2
is counting 0≦Yc≦Y _E −Ys Here, Xs and Ys are the coordinates of point A in the graphic frame ABCD. (Similarly below) Equation (i) shows the generation of an address horizontally from the left graphic frame AC in the direction of the central axis EF.

(2) Yc in the Y-axis direction of address generation counter section 2
is counting Y _E −Ys<Yc≦2(Y _E −Ys) Here, X _E and Y _E are the coordinates of point D in the graphic frame ABCD. (Similarly below) Equation (ii) shows the generation of an address horizontally from the formwork BD in the right figure toward the central axis EF.

_The address generation unit 5 in FIG. 2 is connected to the left and _right graphic frames AC and
Addresses for accessing video information stored on the image memory 8 are generated for each BD. This is used, for example, when extracting the internal left and right contours of a cut out figure as shown in FIG. 4B. Above center axis
Address P (X, Y) on image memory 8 in the horizontal direction from EF to graphic frames AC and BD
Similarly to the above, the address generation counter section 2 is the third one.
When counting the pixels G (Xc, Yc) in the figure AEFC in the figure, the address generator 5 generates an address based on the following equation.

(1) Yc in the Y-axis direction of address generation counter section 2
is counting 0≦Yc≦Y _E −Ys Here, X ₀ =(X _E +Xs/2 (the same applies hereinafter) Equation (iii) indicates the generation of an address horizontally from the center axis EF in the direction of the graphic frame AC.

(2) Yc in the Y-axis direction of address generation counter section 2
is counting Y _E −Ys<Yc≦2(Y _E −Ys) Equation (iv) indicates the generation of an address horizontally in the direction of the graphic frame BD from the center axis EF.

The address generation unit 6 in FIG. 2 moves from the upper and lower graphic frames AB and _CD in FIG _{. 3} to the central axis GF parallel to the Generates an address for accessing the video information stored on the 8. This is used, for example, when extracting the external upper and lower contours of a cut out figure as shown in FIG. 4C. The address P (X, Y) on the image memory 8 in the vertical direction from the upper and lower graphic frames AB and CD toward the center axis GH is determined by the address generation section counter section 3 in the figure shown in FIG.
When counting pixels G (Xc, Yc) in ABGF, the address generator 5 generates an address based on the following equation.

(1) Xc in the X-axis direction of address generation counter section 3
is counting 0≦Xc≦E _x −Es Equation (v) shows the generation of an address perpendicularly toward the central axis GH from the upper figure frame AB.

(2) Xc in the X-axis direction of address generation counter section 3
is counting X _E −Xs<Xc≦2(X _E −Xs) Equation (vi) indicates the generation of an address perpendicularly to the central axis EF from the lower figure frame CD.

The address generation unit 7 in FIG. 2 passes through the center point (X ₀ , Y ₀ ) of the graphic frame ABCD in FIG.
For each CD, an address for accessing the video information stored on the image memory 8 is generated. This is used, for example, when extracting the upper and lower internal contours of a cut out figure as shown in FIG. 4C. Above center axis
Address P (X, Y) on image memory 8 in the vertical direction from GH to graphic frames AB and CD
Similarly to the above, the address generation counter section 3 is
When counting the pixels (Xc, Yc) in the figure ABGH in the figure, the address generator 6 generates an address based on the following equation.

(1) Xc in the X-axis direction of address generation counter section 3
is counting 0≦Xc≦X _E −Xs Here, Y ₀ =(Y _E +Ys)/2 (the same applies hereinafter) Equation (vii) indicates the generation of an address perpendicularly toward the graphic frame AB from the central axis GH.

(2) Xc in the X-axis direction of address generation counter section 3
is counting X _E −X _Y <Ec≦2(X _E −Y _E ) Equation (viii) indicates the generation of an address perpendicularly in the direction of the graphic frame CD from the central axis GH.

Generates an address on the image memory 8 in the direction from the figure frame ABCD in which the figure to be recognized exists to the central axes GH and EF parallel to the X-axis or Y _- axis passing through the center point (X ₀ , Y 0 ) of the figure. The address generators 6 and 4 generate the addresses on the image memory 8 in the direction from the central axes GH and EF toward the graphic frame, and the address generators 7 and 5 generate the addresses, respectively. A read access is executed to the image memory 8 based on the address information. Reading access from each address generation section 4 to 7 to the image memory 8 is performed in a time-division manner, and the black and white binary video signals of the figure to be recognized corresponding to the read address from each address generation section are time-divisionally accessed. The image is read out from the image memory 8. The video signal read out from the image memory 8 in a time-division manner is input to the feature extraction section 9. Feature extraction section 9
Then, the distance from the graphic frame ABCD to the point where the white/black binary video signal changes from "white" to "black" or from "black" to "white" from the center axis EF, GH, and the center axis EF, The distances from "white" to "black" or from "black" to "white" of the above white/black binary video signal heading from GH to graphic frame ABCD are processed and recognized in a time-sharing manner. Extract the features of the target shape. The feature extraction unit 9 observes the distance from “white” to “black” or from “black” to “white”.The hardware is the address generation unit 4 to 7.
Symmetrical shape recognition can be constructed with a single circuit that counts exactly the same number of "white" or "black" for each video signal read out based on the address information generated by the Feature extraction for this purpose can be realized with the above-mentioned single circuit configuration.

As explained above, according to the present invention, the direction from the figure frame in which the figure to be recognized exists toward the central axis;
By generating an address for read access of the image memory in the direction from the central axis toward the figure frame and performing feature extraction processing of the image signal stored in the image memory, it is possible to improve the shape recognition process of symmetric figures. The hardware configuration of the feature extraction section can be made single, and feature extraction can be realized with a single circuit according to the required feature quantity.

[Brief explanation of the drawing]

FIGS. 1A to 1D are explanatory diagrams illustrating the distance of each symmetrical figure from the reference line to the figure, and FIG. 2 is an exemplary configuration for generating four types of addresses for read access to the image memory of the present invention. FIG. 3 is an explanatory diagram for counting the number of pixels within a figure frame of a figure to be recognized, and FIGS. 4A to 4D are explanatory diagrams for explaining addresses for making read access to the image memory. In the figure, 1 is an XY projection extraction section, 2 and 3 are address generation counter sections, 4 to 7 are address generation sections,
Reference numeral 8 represents an image memory, and reference numeral 9 represents a feature extraction unit.

Claims (1)

[Claims] 1 A figure shape in which a figure to be recognized that is symmetrical about the X-axis and/or Y-axis is stored in an image memory, and the contents of the image memory are read out in the X-axis or Y-axis direction to extract features of the figure to be recognized. In the recognition processing method, an XY projection extraction unit is provided to obtain the center point of the figure stored in the image memory, and a center parallel to the X or Y axis passing from the figure frame in which the figure exists to the center point. a first address generation section that generates an address on the image memory in a direction toward the axis; and a second address generation section that generates an address on the image memory in a direction from the central axis toward the graphic frame. , and read access to the image memory is performed based on address information generated by the first address generation section and the second address generation section, respectively. Shape recognition processing method.