KR20230139449A - A recognizing method of assembled stamp using image feature analysis - Google Patents

Abstract

According to the present invention, provided is a method for identifying an assembled stamp using image element analysis, which is executed by a computer system including a display unit, a control unit, an input unit, and a storage unit. The method comprises: a stamp image preparing step (ST-110); a character outline detection step (ST-120) of detecting an outline of a prepared stamp image (100); a character unit integration step (ST-130) of performing integration by each character; a character area detection step (ST-140) of obtaining minimum square information by each character; a character alignment step (ST-150) of correcting an angle of a character; and a DB data comparison step (ST-160) of calculating aligned character information with DB data. In the DB data comparison step (ST-160), a filling rate of character information on the stamp image aligned in the character alignment step (ST-150) and a plurality of character information included in the DB data is calculated to determine the character information on the DB data with the highest filling rate. The DB data stores each character information together with classification information, and each character information included in the DB data is provided by executing the steps from the stamp image preparing step (ST-110) to the character alignment step (ST-150). Therefore, it can be accurately determined whether an assembled stamp is sealed.

Description

{A recognizing method of assembled stamp using image feature analysis}

The present invention relates to a method for identifying prints, and more specifically, to a method for identifying assembled prints using image element analysis that can accurately identify assembled prints for use.

A seal is a person's name or official name engraved on the face of the seal, and a seal is a trace of a stamp made by putting seal on the seal and pressing it. Seals have been used since ancient times, and are currently widely used as a means of identifying oneself in Eastern countries such as Korea, Japan, and China.

In Korea, seals are essential for important legal acts such as transfer of ownership of major assets such as real estate or creation of mortgages, with over 240 types of administrative office work requiring submission of seals and dozens of types of private transactions. As seals are used for important transactions and identity verification functions, crimes of counterfeiting them are also increasing day by day. In addition, the frequency of use of assembled seals is increasing, but there is a problem that there is no accurate standard for determining whether a seal is made by assembled seals.

Republic of Korea Publication No. 10-2011-0104724 Open Patent Publication Republic of Korea Publication No. 10-2011-0117543 Open Patent Publication

The present invention was proposed to solve the problems of the prior art as described above, and aims to provide a method for identifying an imprint using image element analysis that can accurately determine whether or not it is an assembled imprint stamped by an assembled seal. The purpose.

For the above purpose, the present invention is implemented in a computer system including a display unit, a control unit, an input unit, and a storage unit; A print image preparation step, a character outline detection step in which the outline of the prepared print image is detected, a character unit integration step in which each character is integrated, a character-specific area detection step in which minimum square information for each character is acquired, and the angle of the character is corrected. It includes a character sorting step and a comparison step with DB data for calculating the sorted character information with DB data;

In the step of comparing with the DB data, the character information of the image image sorted in the character sorting step and the fill rate of the plurality of character information included in the DB data are calculated, and the character information of the DB data with the highest fill rate is determined;

In the DB data, each character information is stored along with font information, and each character information included in the DB data is identified by assembled impressions using image element analysis, which are prepared by executing steps from the impression image preparation step to the character alignment step. Provides a method.

In the above, the character outline detection step includes a binarization step in which the image prepared in the print image preparation step is binarized, an outline detection step in which the binarized image information is operated with a differential operator to detect the outline, and an outline classification step in which the outline is classified. and; In the outline classification step, the two outlines with the largest horizontal and vertical lengths among the plurality of outlines detected in the outline detection step are classified as outlines.

In the above, after the character outline detection step, it further includes a noise removal step in which noise outlines among the human print outlines are removed; In the noise removal step, the horizontal and vertical lengths of the plurality of outlines detected in the outline detection step are derived, and the horizontal ratio of the print outline to the horizontal length of the print (contour horizontal length ratio) and the print outline to the vertical length of the print are derived. The vertical length ratio (contour vertical length ratio) is calculated, and if either the horizontal length ratio of the outline or the vertical length ratio of the outline is less than a set value, it is classified as noise and is not included in the contour information.

In the above, the set value is characterized as 2%.

In the above, in the character unit integration step, a plurality of human print outlines are clustered for each character and classified into character units.

In the above, in the character unit integration step, the center position of the pixel set forming each print outline, excluding the outline, is calculated and the plurality of print outlines are classified into character units.

In the above, the area detection step for each character includes a convex set polygon pixel information deriving step in which convex set polygon information formed by the print outline forming each character obtained in the character unit integration step is derived, and the convex set polygon pixel information deriving step is derived. It is characterized by including a minimum square information derivation step in which minimum square information surrounding the convex set polygon is derived.

In the above, in the character alignment step, the slope difference between a randomly selected component of the pixels forming the print outline of the character and the remaining pixel components is calculated, and if the slope difference is less than a set value, it is judged to be a continuous pixel component, and the remaining pixel components forming the character are judged as continuous. It includes a step of deriving the number of continuity components in which the number of continuity pixel components is counted and stored;

The step of deriving the number of continuity components is repeatedly performed for the remaining components that form the print outline of the character, and histogram information for each component is prepared; It is characterized by rotating the component with the most continuous pixel components so that it is horizontal.

In the above, in the step of comparing with the DB data, pixels forming the boundary of the character area are derived using a morphology operation; A weight smaller than 1 is assigned to pixels forming the boundary to calculate the fill rate.

In the above, an erosion operation and an expansion operation are performed on the pixels forming the character area of the prepared impression aligned through the character alignment step, so that the area where the pixel value changes is derived as a pixel forming the border of the character area;

By performing an erosion operation and an expansion operation on the pixels forming the character area of the print included in the DB data, the area where the pixel value changes is derived as a pixel forming the border of the character area;

The pixels forming the boundary are given a weight value less than 1 to calculate the fill rate.

According to the method of identifying a seal using image element analysis according to the present invention, it is possible to provide accurate and rapid information as to whether a seal is made by assembled and used seal.

1 is a flowchart of the execution of a human image identification method using image element analysis according to the present invention;
Figure 2 is a flow chart illustrating the character outline detection step of Figure 1;
Figure 3 is an example diagram of pixels and differential operators shown to explain the contour detection step;
Figure 4 is a human print image illustrating an exemplary outline derived by executing the human print identification method using image element analysis according to the present invention;
Figure 5 is an image showing an example to explain the noise removal step,
Figure 6 is a print image illustratively shown to illustrate the character unit integration step of the print identification method using image element analysis according to the present invention;
Figure 7 is a partial image of a human impression shown as an example to explain the character-specific area detection step of the human impression identification method using image element analysis according to the present invention;
Figure 8 is a diagram to explain the area detection step for each character that constitutes the human image identification method using image element analysis according to the present invention;
Figure 9 is an example of a human print image in which the human print identification method using image element analysis according to the present invention is executed and the minimum square information for each character is displayed;
Figure 10 is a partial image of a human print shown to explain the character alignment step of the human print identification method using image element analysis according to the present invention;
Figure 11 is a histogram graph shown to explain the character sorting step of the human image identification method using image element analysis according to the present invention.

All technical and scientific terms used in the description of the present invention, unless otherwise defined, have meanings commonly understood by those skilled in the art to which the present disclosure pertains. All terms used in this disclosure are selected for the purpose of more clearly explaining this disclosure and are not selected to limit the scope of rights according to this disclosure.

Expressions such as “comprising,” “comprising,” “having,” etc. used in the description of the present invention imply the possibility of including other embodiments, unless otherwise stated in the phrase or sentence containing the expression. It should be understood as open-ended terms.

The singular expressions used in the description of the present invention may include plural meanings unless otherwise specified, and this also applies to the singular expressions recited in the claims.

Expressions such as “first” and “second” used in the description of the present invention are used to distinguish a plurality of components from each other, and do not limit the order or importance of the components.

In the description of the invention, when a component is referred to as being “connected” or “coupled” to another component, it means that the component can be directly connected to or coupled to the other component, or as a new component. It should be understood as something that can be connected or combined through components.

Hereinafter, with reference to the attached drawings, the method for identifying assembled prints using image element analysis of the present invention will be described in detail.

The method of identifying assembled prints using image element analysis of the present invention is executed as a program in a computer system including a storage display unit, a control unit, an input unit, and a storage unit.

DB data is stored in the storage unit. The DB data is stored by matching character information and classification information for a plurality of human images. Classification information includes assembly stamp manufacturer, font, and character.

Character information about a plurality of human print images included in the DB data is integrated for each character through a print image preparation step (ST-110) and a character outline detection step (ST-120) in which the outline of the prepared print image 100 is detected. Prepared and stored through a character unit integration step (ST-130), a character-specific area detection step (ST-140) in which the minimum square information for each character is obtained, and a character alignment step (ST-150) in which the angle of the character is corrected. do.

The information of the print image compared with the DB data includes the print image preparation step (ST-110), the character outline detection step (ST-120) in which the outline of the prepared print image 100 is detected, and the character unit integration integrated for each character. The aligned characters are prepared through a step (ST-130), an area detection step for each character (ST-140) in which the minimum square information for each character is obtained, and a character alignment step (ST-150) in which the angle of the character is corrected. Information is calculated with DB data through a comparison step (ST-160).

The text information included in the DB data and the text information of the print image to be compared with the DB data are calculated and prepared in the same way.

In the DB data and comparison step (ST-160), the fill rate of the character information of the impression image sorted in the character sorting step (ST-150) and the plurality of character information included in the DB data are calculated to determine the DB data with the highest fill rate. It is judged to correspond to text information.

In the print image preparation step (ST-110), the print image stamped on paper is scanned or photographed to prepare a print image. The print image is stored in the storage unit.

The imprint image 100 has an imprint area 120 and an outline area 110 as areas where the ink is stamped, and has a background area 120 where the ink is not imprinted.

The character outline detection step (ST-120) includes a binarization step (ST-121) in which the print image 100 prepared in the print image preparation step (ST-110) is binarized, and the binarized image information is operated with a differential operator to create an outline. It includes an outline detection step (ST-123) where the outline is detected, and an outline classification step (ST-127) where the outline area is classified.

The set of pixels forming each outline detected in the outline detection step (ST-123) is expressed as the equation below.

And the horizontal and vertical positions of Pn are xn and yn, respectively.

In Figure 3, a differential operator for contour detection is exemplarily shown, and a part of the print image is exemplarily shown as a grid on the leftmost side. In Figure 3, the shaded pixels can be the print area or outline area, and the remaining pixels are the background area. The binarized image is multiplied and summed by the differential operator shown as an example on the right side of FIG. 3 to calculate the value of each pixel and detect the outline. For reference, the pixels that are multiplied by the differential operator and whose summed pixel value is not '0' become the pixels that form the outline. In Figure 3, the Sobel differential operator (x-direction, y-direction) is shown, but it is not limited to this, and various known differential operators can be calculated to detect the contour.

To explain the operation with the x-direction differential operator as an example, the x-direction ⑤ pixel value is ①×(-1)+②×0+③×1+④×(-2)+⑤×0+⑥×1+ It becomes ⑦×(-1)+⑧×0+⑨×1.

As shown in FIG. 4, the outline detected in the outline detection step (ST-123) includes a print outline 131, which is the outline of the print area, and outer outlines 111 and 112, which are the outlines of the outline area.

In the outline classification step (ST-127), the two outlines with the largest horizontal and vertical lengths among the plurality of outlines detected in the outline detection step (ST-123) are classified as outline lines 111 and 112. In Figure 4, lymax is the maximum value of the print vertical, lymin is the minimum value of the print vertical, lxmax is the maximum value of the print width, and lxmin is the minimum value of the print width. Specifically, the coordinate value of the pixel with the largest vertical coordinate value among the pixels forming the print area (print area and outline area) becomes the maximum print vertical value, and the vertical coordinate value among the pixels forming the print area (print area and outline area) becomes the maximum vertical coordinate value. The coordinate value of this smallest pixel becomes the minimum vertical engraving value. In this way, the maximum value of the printed price and the minimum value of the quoted price are derived.

The lymax is the vertical maximum value of the outer outline 111, lymin is the vertical minimum value of the outer outline 111, lxmax is the horizontal maximum value of the outer outline 111, lxmin is the outer outline ( 111) is the horizontal minimum value. From these values, the horizontal length (lxmax-lxmin) and the vertical length (lymax-lymin) of the outer outline 111 are calculated. Similarly, for the inner outline 112, the horizontal maximum value, horizontal minimum value, vertical maximum value, and vertical minimum value are derived, and the horizontal length and vertical length are calculated.

As described above, the horizontal maximum value, horizontal minimum value, vertical maximum value, and vertical minimum value are derived for each of the print outlines 131 forming the print area, and the horizontal and vertical lengths are calculated. The calculated horizontal and vertical lengths are compared, and the two contour lines with the largest horizontal and vertical lengths are classified as the outer contour lines (111, 112). In other words, the horizontal and vertical lengths of all contours are calculated and compared, and the two contours with the largest horizontal and vertical lengths are classified as outer contours.

The character outline detection step (ST-120) further includes a noise removal step (ST-125) in which noise outlines among the outlines are removed. In the noise removal step (ST-125), the horizontal and vertical lengths of the plurality of outlines detected in the outline detection step (ST-123) are calculated, and the horizontal length ratio of the outline to the horizontal length of the impression (lxmax-lxmin) ( The horizontal length ratio of the outline) and the vertical length ratio of the outline (the vertical length ratio of the outline) to the vertical length of the impression (lymax-lymin) are calculated, and if either the horizontal length ratio of the outline or the vertical length ratio of the outline is less than the set value, there is noise. It is classified and processed so that it is not included in the outline information.

It was confirmed that it is appropriate to set the above setting value to 2%. When the setting value was set larger than 2%, there was a problem in that even though the outline forming the print area was not included in the outline information, when the setting value was set smaller than 2%, there was a problem in that noise was not completely removed.

Figure 5 is shown to exemplarily explain the noise removal step (ST-125). In Figure 5, lymax is the vertical maximum value of the print (the vertical coordinate of the pixel with the vertical maximum value among the pixels forming the print), lymin is the vertical minimum value of the print (the vertical coordinate of the pixel with the minimum vertical value among the pixels making the print), and lxmax is the horizontal maximum value of the print (the horizontal coordinate of the pixel with the maximum horizontal value among the pixels forming the print), and lxmin is the horizontal minimum value of the print (the horizontal coordinate of the pixel with the minimum horizontal value among the pixels making the print).

The horizontal length (xlength) of the seal is calculated as lxmax-lxmin, and the vertical length (ylength) is calculated as lymax-lymin.

In Figure 5, Pymax is the vertical maximum value of one of the print areas, Pymin is the vertical minimum value, Pxmax is the horizontal maximum value, and Pxmin is the horizontal minimum value. The horizontal length (Ixlength) of the corresponding printing area is calculated as Pxmax-Pxmin, and the vertical length (Iylength) is calculated as Pymax-Pymin. The vertical length ratio of the outline of the impression is calculated as Iylength/ylength×100, and the horizontal length ratio is calculated as Ixlength/xlength×100. The vertical and horizontal ratios are calculated for all print areas, and if either the vertical or horizontal ratio is less than a set value, it is judged to be noise such as ink blur.

In the character unit integration step (ST-130), the plurality of print areas 130 are clustered and classified into character units.

In the character unit integration step (ST-130), for each print outline 131, the average coordinate value (xm, ym) of the pixels forming the print outline 131 is calculated. That is, the average value of the coordinates of each impression outline 131 is calculated.

In FIG. 6, Cymax is the maximum vertical coordinate value of the pixels forming the print areas (maximum vertical value of the print area), and Cymin is the minimum vertical coordinate value of the pixels forming the print areas (minimum vertical value of the print area). If the distance between the vertical average values (ym) of each impression outline 131 is Cymax-Cymin/3, it is judged to be included in the character unit forming one letter. Since the calculation is performed as above and the print areas are classified into print areas forming character units, the clustering process is simple and a plurality of print outlines 131 can be classified into each character unit with one operation. If the vertical distance between the centers of neighboring print outlines 131 is less than the reference value, they are judged to be print outlines forming the same character. More precisely, if it is less than 80% of the standard value, it is classified and stored as a print outline forming the same character.

In the character-specific area detection step (ST-140), the convex set polygon multi-pixel information derivation step in which the convex set polygon information formed by the impression outline 131 forming each character obtained in the character unit integration step (ST-130) is derived. and a minimum square information derivation step in which minimum square information surrounding the convex set polygon 133 derived in the convex polygon multi-pixel information derivation step is derived.

In the convex set polygon multi-pixel information deriving step, the convex set polygon 133 information shown as an example in FIG. 7 is derived by the known Convex Hull method. In the minimum square information derivation step, as shown in FIG. 8, a minimum area square is derived for the convex set polygon 133 of each character derived in the convex set polygon pixel information derivation step. In the minimum square information derivation step, the minimum area square is derived by rotating the square surrounding the convex set polygon 133 by a small angle using the known Rotating Caliper method.

Figure 9 is a human image that has progressed to the step of detecting the area for each character (ST-140), and reference numeral 135 is the minimum square of each character.

In the character alignment step (ST-150), the slope difference between the components of randomly selected pixels (marked as "A" in FIG. 10) and the remaining pixel components among the pixels forming the print outline of the character is calculated, and the slope difference is less than or equal to the set value. It includes a step of deriving the number of continuity components in which the number of continuity pixel components for the remaining part of the character is counted and stored by determining the back surface continuity pixel components. Among the pixels forming the print outline, pixels in some sections are selected, the remaining pixels are also selected in the same number as the pixels in the selected section, and the slope for the pixels in the selected section is calculated. If the calculated slope is less than the set value (5°), it is judged to be a continuous pixel component and its number is counted (reference numeral 131-2 in Fig. 10), and if it is greater than the set value, it is judged to be a discontinuous pixel component (reference numeral 131-3 in Fig. 10). ). In this way, pixels are selected for all sections forming the print outline, and calculations are made for the remaining sections. In FIG. 10, the part indicated by “B” indicates a pixel area selected in addition to “A”, and reference numeral 131-2 indicated in yellow indicates a pixel component that is continuous with the part “B”. And reference numeral 131-3 shows the “A” section and discontinuous pixel components.

Figure 11 shows histogram information for each component prepared by repeatedly performing the step of deriving the number of continuity components for the remaining components forming the print outline of the character in the same manner as above. The coordinates of the pixels that make up the character are rotated and aligned so that the component with the most continuous pixel components is in the horizontal direction. In Figure 11, the component with the second largest number of continuous pixel components is in the vertical direction.

DB data is prepared by executing the above-described method on a plurality of stamped seal images with confirmed classification information (manufacturer, font, character). In DB data, classification information and character information are matched and stored.

The character information of the human image to be identified is compared with the character information of the human image included in the DB data, and it is determined which character information in the DB data the human image to be identified corresponds to.

In the DB data and comparison step (ST-160), the fill rate of the character information of the impression image sorted in the character sorting step (ST-150) and the plurality of character information included in the DB data are calculated to determine the DB data with the highest fill rate. It is judged based on text information.

In the step of comparing with the DB data (ST-160), pixels forming the boundary of the character area are derived using a morphology operation; A weight less than 1 is assigned to pixels forming the border to calculate the fill factor.

The known erosion and dilation operations are performed on the pixels that form the character area of the prepared impression aligned through the character alignment step (ST-150), so that the area where the pixel value changes is the boundary of the character area. It is derived from pixels. By performing an erosion operation and an expansion operation on the pixels forming the character area of the print included in the DB data, the area where the pixel value changes is derived as a pixel forming the border of the text area, and a weight value less than 1 is applied to the pixel forming the border. (e.g., 0.5) is assigned to calculate the filling rate.

The filling rate is calculated as the ratio of the number of pixels that overlap with the pixels that make up the character area of the print to be identified to the number of pixels that make up the character area of the prints included in the DB data. Among the pixels that form the character area of the impressions included in the DB data, the pixels that form the border are multiplied by the weight value when counting the number, and the pixels that form the border among the overlapping pixels are also multiplied by the weight value when counting the number. do.

Information on the filling rate is described in Korean Patent Publication No. 10-2011-0104724, and detailed description thereof will be omitted.

In the same way as above, it is possible to check which manufacturer manufactured the assembled seal.

100: Printed image 110: Outline area
120: background area 130: print area

Claims (10)

Runs on a computer system including a display unit, control unit, input unit, and storage unit;
A print image preparation step (ST-110), a character outline detection step (ST-120) in which the outline of the prepared print image 100 is detected, a character-by-character integration step (ST-130), and a character-by-character integration step (ST-130). A character-specific area detection step (ST-140) in which minimum square information is acquired, a character alignment step (ST-150) in which the angle of the character is corrected, and a comparison step with DB data in which the sorted character information is calculated with DB data (ST-140) ST-160);
In the DB data and comparison step (ST-160), the fill rate of the character information of the impression image sorted in the character sorting step (ST-150) and the plurality of character information included in the DB data are calculated to determine the DB data with the highest fill rate. It is judged based on text information;
In the DB data, each character information is stored along with classification information, and each character information included in the DB data undergoes steps from the impression image preparation step (ST-110) to the character sorting step (ST-150). A method for identifying assembled prints using image element analysis, characterized in that provided. The method of claim 1, wherein the character outline detection step (ST-120) includes a binarization step (ST-121) in which the print image 100 prepared in the print image preparation step (ST-110) is binarized, and the binarized image information is It includes a contour detection step (ST-123) in which the outline is detected by calculating with a differential operator, and an outline classification step (ST-127) in which the outline area is classified;
In the outline classification step (ST-127), the two outlines with the largest horizontal and vertical lengths among the plurality of outlines detected in the outline detection step (ST-123) are classified as outline lines 111 and 112. Assembly imprint identification method using image element analysis. The method of claim 2, wherein the character outline detection step (ST-120) further includes a noise removal step (ST-125) in which noise outlines are removed from among the outlines; In the noise removal step (ST-125), the horizontal and vertical lengths of the plurality of outlines detected in the outline detection step (ST-123) are derived, and the ratio of the horizontal length of the outline to the horizontal length of the print (contour horizontal length ratio) The ratio of the vertical length of the outline to the vertical length of the impression (contour vertical length ratio) is calculated, and if either the horizontal length ratio of the outline or the vertical length ratio of the outline is less than the set value, it is classified as noise and is not included in the outline information. A method for identifying assembled prints using characteristic image element analysis. The method of identifying assembled prints using image element analysis according to claim 3, wherein the set value is 2%. The method of identifying assembled prints using image element analysis according to claim 3, wherein in the character unit integration step (ST-130), a plurality of outlines are clustered for each character and classified into character units. The method of claim 5, wherein in the character unit integration step (ST-130), the center position of the pixel set forming each print outline (131) excluding the outer outline is calculated and the plurality of print outlines (131) are classified into character units. A method for identifying assembled prints using image element analysis. The method of claim 1 or 6, wherein in the character-specific area detection step (ST-140), the convex set polygon information formed by the impression outline 131 of each character obtained in the character unit integration step (ST-130) is Characterized by comprising a step of deriving convex set polygon multi-pixel information, and a step of deriving minimum square information surrounding the convex set polygon 133 derived in the step of deriving the convex set polygon multi-pixel information. A method for identifying assembled prints using image element analysis. The method of claim 7, wherein in the character alignment step (ST-150), the slope difference between a randomly selected component of the pixels forming the print outline of the character and the remaining pixel components is calculated, and if the slope difference is less than a set value, it is determined to be a continuous pixel component. It includes a step of deriving the number of continuity components in which the number of continuity pixel components for the remaining portion of the character is counted and stored in a method that can be used;
The step of deriving the number of continuity components is repeatedly performed for the remaining components that form the print outline of the character, and histogram information for each component is prepared; A method for identifying assembled prints using image element analysis, characterized in that the component with the most continuous pixel components is rotated so that it is horizontal. The method of claim 8, wherein in the step of comparing with the DB data (ST-160), pixels forming the boundary of the character area are derived using a morphology operation; A method for identifying assembled prints using image element analysis, wherein the filling rate is calculated by assigning a weight less than 1 to the bordering pixels. The method of claim 9, wherein the area where the pixel value changes by performing an erosion operation and an expansion operation on the pixels forming the character area of the prepared impression aligned through the character alignment step (ST-150) is a pixel forming a boundary of the character area. It is derived as;
By performing an erosion operation and an expansion operation on the pixels forming the character area of the print included in the DB data, the area where the pixel value changes is derived as a pixel forming the border of the character area;
A method for identifying assembled prints using image element analysis, wherein the filling rate is calculated by assigning a weight value less than 1 to the pixels forming the boundary.

Images