JPS5830622B2 - Diagonal intersection detection method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention is a diagonal intersection detection method, and is particularly suitable for original figure information obtained by scanning a figure to be recognized without performing thinning processing on the figure to be recognized. The present invention relates to a diagonal intersection detection method that accurately detects diagonal intersections.

In graphic recognition devices, particularly in recognizing handwritten characters, the intersection of character strokes is extracted as one important feature.

In order to extract such intersections, a so-called thinning process is conventionally performed to extract the skeleton part of the recognition target figure, and the thinned recognition target figure (thinned figure) is converted into, for example, nXm.
It was designed to scan and extract intersections using the mesh observation window.

However, if the stroke width of the recognition target figure (character), that is, the character line width, is thick, the thinned figure obtained by the above-mentioned thinning process may not necessarily be a faithful representation of the skeleton of the original figure. However, undesirable intersections may occur at unexpected locations.

In consideration of the above points, the present invention aims to detect diagonal intersections by using original figure information as it is without performing line thinning processing.

The diagonal intersection detection method of the present invention extracts features based on the scanning information obtained by scanning the recognition target figure, and in a pattern recognition device that performs recognition processing of the recognition target figure, the recognition target figure is moved in one direction. A singular line segment extraction unit is provided for extracting a singular line segment representing a space sandwiched between a first graphic stroke extending diagonally and another second graphic stroke extending diagonally when scanned. Equipped with a closed end extraction function that extracts a closed end where one end of a singular line segment is blocked by a figure, and extracts the actual stroke width on two figure strokes that sandwich the extracted closed end in the scanning direction. The present invention is characterized in that a diagonal intersection candidate point is set at the upper center point, and the diagonal intersection is determined from the coordinate positions of a plurality of set candidate points.

This will be explained below with reference to the drawings.

FIG. 1 is an explanatory diagram illustrating an aspect of diagonal intersection detection according to the present invention, and FIG. figure,
FIG. 3 shows the configuration of one embodiment of the main part of the present invention.

In Fig. 1, 1 is a figure (character) to be recognized, 2 is a figure (character) frame, and 3 and 4 are horizontal scanning singular line segments obtained when the figure to be recognized 1 is scanned in the horizontal direction. , 5 and 6 are vertical scanning singular line segments obtained when the recognition target figure 1 is scanned in the vertical direction, and 3-1゜4-1.5-1.61 are singular line segments, respectively. , 3-2.4-2.5-2.6-2 are closed ends of singular line segments, and 7-1 and 7-2 are diagonal intersection candidate points that sandwich the closed end 3-2. 8-1 and 8-2 are diagonal intersection candidate points, respectively, obtained on the graphic stroke sandwiching the closed end 4-2, 9-1 and 9-2 are diagonal intersection candidate points found on the graphic stroke sandwiching the closed end 5-2, and
Reference numerals 10-1 and 10-2 are diagonal intersection candidate points obtained on the graphic strokes sandwiching the closed end 6-2, and 11 represents the diagonal intersection.

In the case of the present invention, as will be described later, (i) singular line segments 3.4, 5 representing the white area space sandwiched between figure strokes;
, 6 are extracted, and (11) closed end 3-2 of the singular line segment,
4-2, 5-2, and 6-2 are extracted, and diagonal intersection candidate points 7-1, 72.8- are extracted corresponding to each closed end (111).
1.8-2.9-1.9-2.101°10-2 is set, and (iV) if the intersection candidate points on each side are legitimate candidate points, for example at the center of gravity of each candidate point. A diagonal intersection 11 is determined.

Hereinafter, with reference to FIG. 2, the extraction of singular line segments, the extraction of closed ends, and the setting of diagonal intersection candidate points will be briefly explained.

In FIG. 2, 12 to 19 are counters, 2
0 to 22 each represent an adder circuit.

The figure to be recognized 1 shown in FIG.
CN2.5CN3, . . . , for example, during horizontal scanning, a clock signal CL synchronized with the movement of the scanning point position for each counter 12 to 17, or 1 having a frequency half that of the clock signal CL. /2 clock signal 1.
/2CL or 1/4 clock signal 1/4-CL is supplied.

The counter 12 starts counting when the scanning point reaches the x mark position shown in Figure 1, and stops counting at the point where the white mesh changes to the black mesh, that is, the point shown in Figure 1. .

Therefore, when counting is completed, the contents of the counter 12 represent the coordinate information of the dot.

The counter 13 starts counting when the scanning point reaches the x mark position shown in FIG. 1, and stops counting at the point shown in the drawing.

Therefore, when the counting is finished, the contents of the counter 13 are ×
It represents the coordinate information of a point distanced by b/2 from the mark position.

Similarly, the counter 14 shows the value (ed)/2, the counter 15 shows the value d, the counter 16 shows the value d/2, the counter 1γ shows the value (e/d)/4, and the counter 18 shows the value C and the counter 19 shows the value c/2.

As a result, the output of the adder circuit 20 indicates the coordinate value of point a, that is, the intermediate point between points e and d shown in FIG. The output of is the value (
a+c)/2.

For singular line segment 3, point a is determined for each scanning line as described above, and it is checked whether points a and a on vertically adjacent scanning lines are continuous in the vertical direction, and If they are continuous, it can be considered that each point a is connected and extracted as a line segment.

During this time, it is checked whether the starting end and the ending end of the singular line segment are open to the figure frame 2 or the white mesh area, or whether they are closed by the figure black mesh area.

When it is closed by the graphic black mesh area, the starting end and/or the ending end can be considered to be the closed end 3-2.4-2.5-2.6-2 as described above.

Note that the extraction of singular line segments and the extraction of closed ends are determined one after another, for example, as the horizontal scan progresses one after another.

On the other hand, as mentioned above, the coordinates of point C, the coordinates of point C, (a+b)/
A value of 2, a value of (a+c)/2, is obtained for each scan.

Under this condition, when the closed end of the singular line segment, for example 32, is extracted, the point 1-1 indicated by the above value (a+b)/2
, the point 7 to 2 indicated by the value (a+c)/2 is set as the diagonal intersection candidate point according to the present invention.

Diagonal intersection candidate points 8-1 and L-2 are similarly set during horizontal scanning.

Also, diagonal intersection candidate points 9-1.9-2 and 10-1.10-
2 is similarly set during vertical scanning.

FIG. 3 shows an embodiment of the present invention for setting the above diagonal intersection candidate points.

In the figure, 23 to 27 are coordinate tables, 28 and 29, respectively.
．． Reference numeral 30 represents a husband register, 31 a differential circuit, 32 a comparison circuit, 33 a singular line segment length counter, and 34 to 37 a husband AND circuit.

The case of setting the above diagonal intersection candidate points 7-1 and 7-2 will be described below as an example.

As mentioned above, each time the horizontal scan progresses, the values a, b, (a+b)/2, C1 appear on the coordinate table 23 to 27.
(a+c)/2 is successively set.

During this time, each time the value a is set, the line segment length counter 33 is incremented by 1 and the length of 30 singular line segments is measured.

Now assume that singular line segment 3 starts at scan line #2 and ends at scan line #n, and that singular line segment 3 is a closed end at coordinate position an.

At this time, coordinate information an of the closed end 3-2 is set in the register 30.

Also, the coordinate information bn and cn obtained at the #n scanning line
do not match the coordinates of the graphic frame, AND circuits 34 and 35 are both turned on, and the coordinate information (an+b
n)/2 and (an cn)/2 are set in registers 29 and 28, respectively.

The difference circuit 31 outputs the distance between points 7-1 and 3-2 shown in FIG. 1, and the distance between points γ-2 and 3-2.

Then, the comparison circuit 32 outputs logic "1J" when each distance is within the threshold value a.

On the other hand, when the content of the line segment length counter 33 is equal to or greater than a predetermined threshold, the line segment length counter 33 is set to logic "1".
Output.

Therefore, the AND circuit 37 generates a logic "1", and the contents of the registers 29 and 28 are outputted via the AND circuit 36.

That is, the coordinate information is output as the diagonal intersection candidate point.

Note that the reason why the comparison circuit 32 compares with the threshold value α as described above can be considered to be due to the following arrangement.

That is, if the points 7-1 and 7-2 shown in FIG. -2 can be considered to be created and eliminated.

Furthermore, the reason why the line segment length of the singular line segment is checked by the line segment length counter 33 as described above may be considered to be due to the following consideration.

For example, in a handwritten character such as the character "人" where the starting edge of a stroke diagonally downward to the right protrudes slightly beyond a stroke diagonally downward to the left, a diagonal intersection candidate point is set in an undesired manner at the protruding portion. You can think of it as excluding things that are done.

Diagonal intersection candidate point 7-1゜7-2 set as described above,
Based on the coordinate information of 81, 8-2, 9i, 9-2, 101, and 10-2, the coordinates of the center of gravity are determined, and the diagonal intersection 11 is determined as shown in FIG.

As described above, the diagonal intersection 11 is determined, but if necessary, the slope of the figure stroke extending from the diagonal intersection 11 may be checked to confirm that it is a diagonal intersection, or the slope information may be used in subsequent processing. Sometimes I use it.

In this case, the illustrated stroke is scanned using, for example, a 3×3 mesh observation window W shown in FIG. 1, and the inclination of the stroke is investigated.

As described above, according to the present invention, it is possible to directly detect intersections based on original figure information without performing so-called thinning processing.

Note that the extraction of singular line segments and the extraction of closed ends according to the present invention are not only performed for detecting intersections according to the present invention, but are also widely used for extracting other features.

Therefore, the necessity of extracting singular line segments is not disadvantageous to the present invention.

[Brief explanation of the drawing]

FIG. 1 is an explanatory diagram illustrating a mode of diagonal intersection detection according to the present invention, and FIG. 2 is an explanatory diagram illustrating a mode of temporarily determining the coordinates of a singular line segment and the coordinates of a diagonal intersection candidate point according to the present invention. ,
FIG. 3 shows the configuration of one embodiment of the main part of the present invention. In the figure, 1 is the figure to be recognized, 5.6 is a singular line segment, 3 6-2 is a closed end, 7-1 12.9-1, 9-2, 12 is a figure frame, 3, 4° 2.4-2.5-2°, 7-2, 8-1. 0-1 and 10-2 are diagonal intersection candidate points, 11 is diagonal intersection, 12 to 19
are counters, 20 to 22 are adder circuits, and 23 are respectively adder circuits.
28 to 30 are registers, 31 is a difference circuit, 32 is a comparison circuit, and 33 is a line segment length counter.

Claims (1)

[Scope of Claims] 1. In a pattern recognition device that performs recognition processing of the recognition target figure by extracting features based on scanning information obtained by scanning the recognition target figure, the recognition target figure is scanned in one direction. A singular line segment extraction section is provided for extracting a singular line segment representing a space sandwiched between a first graphic stroke extending in an oblique direction and a second graphic stroke extending in another diagonal direction. Equipped with a closed end extraction function that extracts a closed end whose half end is blocked by a figure, the extracted closed end is extracted on two figure strokes that sandwich the extracted closed end in the scanning direction, and the stroke width is substantially equal to that of the stroke width. A diagonal intersection detection method characterized in that a diagonal intersection candidate point is set at a center point, and the diagonal intersection is determined from the coordinate positions of a plurality of set candidate points. 2. Claims characterized in that the closed end used to set the diagonal intersection candidate point is determined based on the singular line segment whose line segment length is greater than or equal to a predetermined value. The diagonal intersection detection method described in Section 1. 3. The diagonal intersection candidate points used to determine the diagonal intersection are limited to diagonal intersection candidate points whose distance from the closed end is within a predetermined value. The diagonal intersection detection method according to item 1 or 2. 4. The diagonal intersection detection method according to any one of claims 1 to 3, wherein the diagonal intersection is determined by examining the inclination angle of a graphic stroke extending from the diagonal intersection.