JPH0122952B2 - - Google Patents

Description

[Detailed description of the invention]

(Technical Field) The present invention relates to a figure recognition device, and particularly to figures that include characters or symbols (hereinafter referred to as "internal characters") inside an outer frame that can be expressed as a geometric figure, such as a seal stamp or a flowchart. The present invention relates to a method of separating a frame portion and an internal character portion. (Background Art) Conventionally, in order to input a program into a computer, the steps are generally to create a flowchart, then code it, punch it on a card or tape, and then input it using an input device. It requires a lot of effort. If there is an input device that can recognize flowcharts,
It is possible to input a program directly into a computer from a flowchart, and the labor-saving effect is enormous. By the way, in order to recognize the flowchart,
First, the symbols used in the flowchart, such as processing (rectangle), judgment (diamond, hexagon), terminal (ellipse), and connector (pentagon, circle, triangle), are identified and classified, and then their internal characters are recognized. There is a need. By the way, when recognizing internal characters, symbols generally only get in the way, so if only the internal character part can be separated and extracted, the recognition performance of internal characters will improve. Conventionally, as a seal matching device, which is an example of a seal recognition device, after aligning the positions of the seal imprint pattern to be compared and the registered seal imprint pattern, the two patterns are matched and the authenticity of the tested seal imprint is determined based on the matching rate. However, in order to improve the accuracy of determining authenticity and reduce the time required for determination, it is possible to extract information only from the outer frame of the seal, such as the external shape and dimensions of the seal. It is said that it is more effective to extract as much as possible features that can be easily extracted from among the so-called external features and the so-called internal features such as the number of characters, character size, and font of the characters inside the stamp and use them for authenticity determination. By the way, when extracting internal features, it is often said that the outer frame of the seal imprint is only an obstacle. However, since there is no simple method for separating and extracting only internal character parts, there is an inconvenience in that internal features must be extracted from the entire seal imprint pattern including the outer frame. (Problems to be solved by the invention) The purpose of the present invention is to solve these drawbacks.
It is designed to easily separate the outer frame and the inner text in a figure that contains internal characters inside an outer frame that can be expressed as a geometric figure. In a figure recognition method that separates and identifies internal figure patterns,
The original pattern is filled in by adding a predetermined number to the white points and black points of the original pattern in at least one direction of the axial direction and the Y-axis direction, and the outermost shell contour of the filled pattern is traced and extracted. Calculate the original white point number ratio, which is the ratio of the number of original white point cells that are filled white points to the total number of cells of the outermost contour line, and when the original white point number ratio is smaller than a predetermined value. repeats the deletion, value replacement, or sign replacement of the outermost cell of the outermost shell contour until the original white point number ratio reaches the predetermined value, and the remaining pattern is identified from the internal figure pattern. It is in the figure recognition method. (Structure and operation of the invention) FIG. 1 shows a first embodiment of the present invention, in which a binarized pattern is filled in a fill-in pattern creation section 1, and the inside of the outer frame is filled in to obtain a fill-in pattern. Next, while tracking the outermost shell of the filled pattern by the outermost shell contour tracking unit 2, points that were white points in the total number of contour points and the binarized pattern but have become black points due to filling (hereinafter referred to as (referred to as the original white point), that is, the original white point number,
Original white point number ratio calculation unit 3 calculates the original white point number ratio = original white point number /
The total number of contour points is calculated, and when the original white point number ratio has not yet exceeded a predetermined threshold in the end determination section 4,
The outermost shell contour eraser 5 erases the outermost shell contour that has just been tracked, returns to the outermost shell contour tracing 2, and
After erasing the outermost contour, the newly outermost contour of the filled pattern is tracked, and the same operation is repeated until the original white point number ratio exceeds a predetermined threshold. Here, the filling pattern creation section 1 will be explained using FIG. 2. The X coordinate is Xi and the Y coordinate is
The point Yj is represented by P(Xi, Yj), and the point P(Xi,
Let the value of Yj) be expressed as |P(Xi, Yj)|.
In addition, in FIG. 2, for convenience of explanation, the main scanning interval and the sub-scanning interval are shown extremely rougher than they actually are. For a binary pattern in which the black point value is 1 and the white point value is 0, raster scanning is performed with the X-axis direction as the main scanning direction and the Y-axis direction as the sub-scanning direction, as shown in Figure 2B. , for each main scan, the X coordinate value X _S of the black point with the minimum X coordinate among the black points existing on the main scanning line, and the X coordinate value X _E of the black point with the maximum X coordinate.
is detected, and an operation is performed in which, for example, 2 is added to the values of all points between X _S and X _E on the main scanning line. For example, in the case of main scanning line l, |P(X _S , Y _l )|=|P(X _S , Y _l )|+
2 to |P(X _E , Y _l )|=|P(X _E , Y _l )|+2, so the values at each point are shown in a diagram as shown in FIG. 2B. Such processing is performed over the entire area in the sub-scanning direction, that is, from l=1 to n, and the filling in the X-axis direction is completed. Next, as shown in FIG. 2C, the graphic pattern that has been filled in the X-axis direction is raster scanned with the Y-axis direction as the main scanning direction and the X-axis direction as the sub-scanning direction, and is scanned for each main scanning. A point on the main scanning line whose value is 3 (the value of a black point in a binarized figure pattern is always rewritten to 3, and although it was a white point in the binarized figure pattern, it is A point that becomes a new black point by filling in will always have a value of 2), the Y coordinate value Y _S of the point with the minimum Y coordinate, and the Y coordinate value Y _E of the point with the maximum Y coordinate. Detect Y _S on the main scanning line
For example, add 4 to the values of all points in ~ _YE . For example, in the case of main scanning line k, |P(X _k , Y _S )|=
Since |P(X _k , Y _S )|+4 to |P(X _k , Y _E )|+4 are performed, the values at each point are shown in a diagram as shown in FIG. 2C. Such processing is performed over the entire area in the sub-scanning direction, that is, from k=1 to m, to complete filling in both the X and Y axes directions. An example of the fill pattern thus obtained is illustrated in FIG. 2D. Furthermore, the second
In Figure D, the intersections (for example, Q, R, and S) of the horizontal line, which is a fill line in the X-axis direction, and the vertical line, which is a fill line in the Y-axis direction, are filled points in both the X and Y axes. , its value is 7 or 6, and the black dots in the binarized figure pattern are 7 (for example, Q and R), and the white dots in the binarized figure pattern are 6 (for example, S). There is. By the way, since the explanation so far has been for the case where there is no chip in the outer frame, there may be no points in Fig. 2D that are filled only in the X-axis direction or points that are filled only in the Y-axis direction. If there is a chip in the outer frame, the value of points filled only in the X-axis direction is 2, and the value of points filled only in the Y-axis direction is 4. These relationships are summarized in the table below.

[Table] The outermost shell contour tracking unit 2 performs scanning with the X-axis direction as the main scanning direction and the Y-axis direction as the sub-scanning direction, as shown in FIG. Using the well-known 8-connection-boundary tracing algorithm starting from S _O (points with a value of 2 or more),
Contour tracing of black point or original white point and contour point sequence
Extract L _O (named the outermost contour line). Note that during contour tracking, the original white points and black points are distinguished by monitoring the values of the tracking points, and the original white points and total contour points are counted. The original white point number ratio calculation unit 3 calculates the original white point number ratio by dividing the original white point number by the total number of contour points. The termination determination unit 4 compares the original white point ratio and a predetermined threshold value and determines whether or not to terminate the process based on the magnitude relationship between the two.The setting of the predetermined threshold value will be explained using FIG. . For example, if the binarized pattern is as shown in FIG. 4A, the filling pattern will be as shown in FIG. 4B. Here, the white square marked "□" is a former white dot, and the "■" mark is a black dot. If the outermost shell contour is extracted from this filled pattern Fig. 4B, Fig. 4C is obtained, and since the total number of contour points = 72 and the original white point number = 0, the original white point number ratio is 0, and the outermost shell contour The erasing section 5 erases the outermost shell contour line just extracted. For erasing, for example, rewriting the value of the outermost contour point to 0, inverting the sign of the value, or arbitrarily selecting an appropriate method so that it is treated as if it were a white point in the next contour tracing. It's okay to do that. Next, when the outermost shell contour is extracted again from the fill pattern after erasing the outermost shell contour,
Figure 4D is obtained, but since the original white point ratio = 0/68, it is still 0, and after erasing the outermost shell contour that has just been extracted, the next outermost shell contour in Figure 4E is extracted. The total number of contour points in Figure 4 E is 64, and the original number of white points is 11.
Therefore, the original white point number ratio = 11/64 ≒ 0.17, and the final condition is not satisfied, so when the next outermost shell contour line F in Figure 4 is extracted, the total number of contour points is 60, and the original white point number is is 56, so the original white score ratio = 56/60≒
It becomes 0.93. If we set the predetermined threshold value for termination judgment to 0.95, then since the final termination condition is not satisfied, if we further extract the outermost contour line G in Figure 4, the total number of contour points = 52 and the number of original white points = 52. , the original white point ratio = 52/52 = 1, and it is determined that the game is over. At this point, the unerased black point (which naturally does not include the original white point since it is a black point) is H in Figure 4, so
This means that only the internal character part of FIG. 4A is extracted as a binarized pattern. In the above explanation, the predetermined threshold value for determination of termination is set to 0.95, but this value is not necessarily optimal and should be set taking into consideration the pattern to be recognized. For example, in the case of a flowchart, there is generally a gap between the symbol part and the internal character part, so the original white point ratio will almost certainly reach 1, so it is okay to set the predetermined threshold to a value of about 0.95 to 1. . However, in the case of seal stamp patterns, there are some patterns in which the outer seal impression frame and the character part are separated, as shown in Figure 5A, but this is not the case in all patterns, and the outer seal impression frame and the character part are separated. If there is a pattern that is in contact with the pattern, in that case, the original white point number ratio will never reach 1, and it is not appropriate to set the predetermined threshold value too high. According to experiments conducted by the inventors, the predetermined threshold value was set at approximately 0.7 to 0.75 for 414 types of seal imprint patterns obtained from 55 types of seals (including both seals in which the outer frame and the character part were in contact and those in which the character part was separated). The results show that it is appropriate to set it to .
Here, the pattern resolution is 8 dots/mm. As explained above, in the first embodiment, the outermost contour of a fill pattern is extracted, and the outermost contour is erased until the original white point number ratio of the contour exceeds a predetermined threshold. Since it extracts only the character part, it has the following advantages. In other words, when tracing the outline of a pattern that is not filled in, it is possible to trace the outline of a pattern with chips in the outer frame or patterns where the outer frame and internal characters are in contact, as shown in Figures 5A and 5B. In this case, the contour line becomes L _A or L _B , and it is not possible to extract only the outermost shell, but the inner part is also traced. However, since the contour of the filled pattern is traced, this does not occur. Only the outermost shell can be extracted. In addition, even in the case of a pattern where there is no chip in the outer frame and the outer frame and the inner characters do not contact each other, as shown in Fig. 4A, as the outermost shell outline is erased, the remaining pattern becomes substantially In other words, this is the same as a chip in the outer frame, and the same drawbacks as above occur. Furthermore, by introducing the concept of counting the original white points of the filled pattern, there is an advantage that effective information for determining the end of the pattern can be obtained. In the first embodiment, the figure recognition device erases the outermost shell contour line and extracts only the internal character part. Instead of rewriting the value of each contour point of the determined outermost shell contour to 0, for example, leave the absolute value as is and just change the sign from positive to negative, and then the final determination is made. By extracting only the points having a negative value at the point in time, it is possible to create a figure recognition device that separates and extracts only the outer frame. Further, in the first embodiment, an apparatus using a binary pattern has been described, but an apparatus using a multi-value pattern is also possible. In other words, when the multi-value pattern has N values, for example, the constant value added as a point value when filling in the X-axis direction was set to 2 in the first embodiment, but if it is set to a value of N+1 or more, the original multi-value pattern is Since it is possible to perform filling while preserving the value information, it is sufficient to create filling patterns in both the X and Y axes directions while taking such consideration. Furthermore, in the first embodiment, only the separation and extraction of the outer frame part and the inner character part was explained, but the thickness of the outer frame can be changed depending on how many outermost shell lines are erased before determining the end. It can also be recognized. In addition, in the first embodiment, in the fill pattern creation section, for the white point inside the outer frame (that is, the white point that becomes the original white point after filling), (1) fill only in the X-axis direction, and (2) fill in the Y-axis direction. Fill only, (3)
We have explained the case where the values are different depending on the three filling situations in both the X and Y axis directions, but it is possible to use a white point that is filled in at least one direction of the X axis or Y axis without making such a distinction. Even if the value of (that is, nothing but the original white point) is set to a common value, the same effect as in the first embodiment can be obtained. In other words, the fill pattern creation section converts all points within the figure existence permissible area into black points (black points of the original pattern),
They are classified into three types: original white points (white points filled in at least one direction, that is, white points inside the outer frame), and white points (white points that are not filled in at all, that is, white points outside the outer frame). However, it may be useful to set different values like 2 and 4 so that you can see if the outer frame is missing.
In contour tracking, for example, if the point values are 7, 4, 4,
If there is a succession of original white points for filling in the Y direction and a succession of original white points for filling in the X direction between two originally black points like 4, 4, 2, 2, 2, 2, 7, then
You can see the chipping of the outer frame. (Effects of the Invention) The present invention creates a fill pattern for a figure pattern, sequentially extracts its outermost contour, and compares the original white point number ratio of the contour with a predetermined threshold value. The outer frame of a graphic pattern that is composed of an outer frame and an inner character part, such as a seal or a flowchart, is determined by determining whether it belongs to the outer frame part or the inner character part. Since it is easy to separate and extract parts and internal character parts, it can be used in various graphic recognition devices.

[Brief explanation of drawings]

FIG. 1 is a block diagram of one embodiment of the present invention, and FIG.
Figures A to D are explanatory diagrams of a fill pattern creation unit in an embodiment of the present invention, Figure 3 is an explanatory diagram of an outermost shell contour tracking unit in an embodiment of the present invention, and Figures 4 A to F are diagrams of the invention FIGS. 5A and 5B, which are explanatory diagrams of the processing steps in one embodiment, are examples of patterns in which the outer frame is chipped or the outer frame and internal characters are in contact with each other. 1... Filling pattern creation section, 2... Outermost shell contour tracing section, 3... Original white point ratio calculation section, 4... End determination section, 5... Outermost shell contour erasing section.

Claims (1)

[Claims] 1. A figure recognition method for separately identifying an outer frame and an inner figure pattern,
The original pattern is filled in by adding a predetermined number to the outer frame and white points and black points inside the outer frame of the original pattern in at least one direction in the axial direction, and the outermost shell contour of the filled pattern is traced and extracted. ,
The original white point number ratio, which is the ratio of the number of original white point cells that are filled white points to the total number of cells of the extracted outermost shell contour, is calculated, and if the original white point number ratio is smaller than a predetermined value, In some cases, the deletion, value replacement, or sign replacement of the outermost cell of the outermost shell contour is repeated until the original white point number ratio reaches the predetermined value, and the remaining pattern is an internal figure pattern. A figure recognition method characterized by identifying. 2. The graphic recognition method for a seal imprint pattern according to claim 1, wherein the predetermined value is in the range of 0.70 to 0.75.