JPH10105715A - Generation method of template for pattern matching - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the capacity of a storage device for template which is used for pattern matching to calculate the position of a feature part of a light- sectioned picture. SOLUTION: A light cut picture of a master work is scanned at a prescribed pitch, and positions of picture points matching with individual scanning lines of the light-sectioned picture are detected, and stroke data representing the positions of these picture points is generated, and stroke data in a prescribed range having a feature point as the center is stored. At the time of measurement of the work, picture data for a template TP representing the shape of the feature part is generated from stored stroke data by graphic processing. At the time of the graphic processing, expansion processing (B) and smoothing processing (C) are performed after linearization processing (A), where lines connecting individual points of stroke data are plotted, to generate picture data of a luminance distribution approximating a luminance distribution (D) of the light-sectioned picture.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for creating a template used in a pattern matching process for measuring the position of a characteristic portion such as a bent portion, a step portion, and an edge portion of a work by a light cutting method.

[0002]

2. Description of the Related Art In a light cutting method, a work is irradiated with slit light, and the slit light applied to the work is irradiated by an image pickup device arranged in such a manner that an optical axis is obliquely oblique to the light surface of the slit light. In this method, a light section image of a workpiece to be drawn is captured, and the workpiece is measured from the light section image of the workpiece appearing on an imaging screen. If the workpiece has a characteristic portion such as a bent portion, a step portion, or an edge portion, the light-cut image of the work also has a shape having a characteristic portion corresponding to the bent portion, and the characteristic of the work is determined from the position of the characteristic portion of the light-cut image. The position of the part can be measured.

[0003] Conventionally, in order to determine the position of the characteristic portion of the light-section image, as shown in Japanese Patent Application Laid-Open No. 5-288513, the imaging screen is scanned at a minute pitch, and each of the light-section images of the workpiece is scanned. The position of an image point that matches the scanning line is detected as the area centroid of the luminance distribution on each scanning line, and point sequence data representing the position of these image points is created. The position of the characteristic part of the image is determined.

[0004]

In the above-mentioned conventional method, it takes a long time to perform measurement because the imaging screen is scanned at a small pitch over the entire area. When performing online measurement, the tact time of a line is adjusted in accordance with the measurement time. Sometimes it has to be long.

As a method of solving such a problem, it is conceivable to determine the position of a characteristic portion of a light-cut image of a workpiece by pattern matching. Pattern matching is
This method uses a template that represents the shape of a feature to determine the position of the shape that matches the template using the normalized correlation method. it can.

When measuring the assembly accuracy of an automobile body, it is necessary to measure the body at a number of measurement points, and since the shape of the characteristic portion to be measured differs at each measurement point, pattern matching is performed as described above. To determine the position of the feature, it is necessary to prepare a large number of templates. The template is stored as image data representing the shape of the characteristic portion. When a large number of templates are prepared, a large storage capacity is required, and the size of the storage device is increased and the cost is increased.

[0007] In view of the above, it is an object of the present invention to provide a method for creating a template that requires a small storage capacity.

[0008]

Means for Solving the Problems In order to solve the above problems,
The present invention relates to a method for creating a template used for determining the position of a characteristic portion of a light-cut image of a work appearing on an imaging screen by pattern matching when a light-cut image of the work is drawn by slit light applied to the work. And
Irradiating the master work with slit light to capture a light cut image of the master work, and scanning the imaging screen at a predetermined pitch to detect the position of image points that match each scanning line of the light cut image of the master work Generating point sequence data representing the positions of these image points; storing a predetermined range of point sequence data centered on the characteristic portion; and performing graphic processing on the stored point sequence data by graphic processing. Creating image data for a template representing the shape of the template.

According to the present invention, since image data for a template is created from point sequence data at the time of pattern matching, only the point sequence data needs to be stored. Therefore, the storage capacity is much smaller than that for storing image data, and even when a large number of templates need to be prepared, a small storage device is sufficient and the cost can be reduced.

Incidentally, since the light-section image of the work has a certain width, it is desirable to perform a dilation process and a smoothing process at the time of graphic processing to create a template that approximates the actual light-section image.

[0011]

FIG. 1 shows a work A such as an automobile body.
1 shows an outline of an optical measuring device used for measurement of the object A. The device comprises a slit light source 1 composed of a slit laser or the like that irradiates a workpiece A with slit light, an imager 2 composed of a CCD camera, and an imager 2. And an image processing device 3 for inputting the image signal.

The slit light source 1 and the image pickup device 2 are mounted on a measuring head (not shown) attached to the operating end of a moving mechanism such as a robot, and the measuring head is sequentially moved to a position facing each of a plurality of measuring parts of the work A. Then, each measurement site is measured.
The slit light source 1 and the image pickup device 2 are mounted on the measuring head in such a positional relationship that the optical axis of the image pickup device 2 obliquely intersects the optical surface SP of the slit light at a predetermined angle θ (for example, 45 °).

Hereinafter, measurement of a cross section of a work will be described. The cross section measurement is performed to measure the position of a characteristic portion such as a bent portion, an edge portion, or a step portion of the work A. At the time of measurement, a slit light is radiated in a state where the slit light source 1 is directly opposed to the work A, and a light cut image S of the work A drawn by the slit light is captured by the image pickup device 2. Here, assuming that a coordinate axis of the imaging screen orthogonal to the optical plane SP of the slit light and parallel to a plane including the optical axis of the imaging device 2 is an X axis, and a coordinate axis orthogonal to the X axis is a Y axis, as shown in FIG. At a measurement site having a simple S-shaped cross section, an S-shaped light section image S bent in the X-axis direction corresponding to the cross-sectional shape of the work A appears on the imaging screen as shown in FIG. Part is, for example, a mountain-folded bent part Sa
And measure its position. It should be noted that a disturbance image R due to the reflection image r of the light-section image s that appears on the work surface bent in a valley-fold shape also appears on the imaging screen.

Since the bent portion Sa has a radius, it is difficult to uniquely specify the position. Therefore, the equation of the first image line LS1 approximating the upper image portion in the Y-axis direction that is continuous with the bent portion Sa, and the second image line approximating the lower image portion in the Y-axis direction that is continuous with the bent portion Sa LS2 equation is obtained, and the position of the bent portion Sa is measured based on the coordinates of the intersection Q of the two image lines LS1 and LS2.

Here, the equation of each of the image lines LS1 and LS2 is obtained by setting a plurality of scanning lines in the image processing device 3 so as to intersect each of the image portions and matching each scanning line of the light-section image. The coordinates of the image points to be detected are detected by regression processing from these image points.
In order to accurately determine the equation of S2, it is necessary to improve the setting accuracy of each scanning line and the accuracy of detecting image points in each scanning line.

Therefore, in the present embodiment, in order to improve the setting accuracy of each scanning line, pattern matching is performed on the imaging screen using a template representing the shape of the bent portion Sa, which is a characteristic portion of the light-section image S. Bending section Sa
Is determined, and each scanning line can be accurately set in a required image portion based on this position.

The template is created based on data at the time of teaching for performing measurement on the master work. Hereinafter, the measurement at the time of teaching will be described in detail.

In the teaching measurement, first, the master work is irradiated with slit light, and a light cut image s of the master work is obtained.
Is imaged. Next, as shown in FIG. 3A, a large number of scanning lines L parallel to the X axis are set at a predetermined pitch in the Y axis direction on the imaging screen, and each scanning of the light-cut image S is determined from the luminance distribution on each scanning line. The X coordinate of the image point PS matching the line is detected. Hereinafter, 480 pixels (pixels) in the Y-axis direction of the imaging screen
Are described, and a case where 95 scanning lines L are set at a pitch of 5 pixels in the Y-axis direction will be described as an example.

The luminance distribution on the scanning line Ln in FIG. 3A is as shown in FIG. 4A, and when this luminance distribution is smoothed, a luminance distribution graph as shown in FIG. 4B is obtained. . In this brightness distribution graph, a left peak derived from the light section image S and a right peak derived from the disturbance image R appear.

Next, the peak degree P of the luminance change at each point on the scanning line is obtained, and the distribution of the peak degree P on the scanning line is shown in FIG.
A peak degree distribution graph as shown in FIG. X
As shown in FIG. 5A, the peak degree P of a point on the scanning line having the coordinate n is represented by a luminance point on the luminance distribution graph at the point.
n, the luminance points on the luminance distribution graph at both end points of the measurement range W of a predetermined width set on the scanning line with the point as the center are bn and cn, and the height of an with respect to the connection line connecting bn and cn is It is obtained as a value to represent. In this case, the height of an with respect to the midpoint of the connection (= ｛2an− (bn
+ Cn)｝ / 2) may be set as the peak degree P, and the length of a perpendicular line descending from an in the connection may be set as the peak degree P. Here, in the luminance distribution of FIG. 5A, X = n1
Has a positive value, and the point P at X = n2 has a negative value.

Since the peaks derived from the disturbance image R appearing in the luminance distribution graph are relatively wide, the peak degree P is small, while the peaks derived from the light-section image S are relatively narrow, so that the peaks are small. The degree P increases. If the width of the measurement range W is set to an appropriate value (for example, 10 pixels), the peaks having a height equal to or more than a predetermined value among the peaks of the peak degree distribution graph are derived from the light-section image S. Can be determined. However, if the width of the disturbance image R is narrow, the peak degree P may become a predetermined value or more. Here, the disturbance image R due to the re-reflection of the slit light appears at a position shifted to the imaging direction side, that is, to the right side of the light-section image S on the imaging screen due to the imaging direction. For this reason, half of the maximum value in the peak degree distribution graph is set to the predetermined value, peaks having a height lower than the predetermined value are excluded from the processing target, and the leftmost peak of the remaining peaks is optically cut. It is desirable to select as the one derived from the image S. In addition, the peak degree P
Therefore, it is desirable to determine the peak degree P from a smoothed luminance distribution graph as shown in FIG.

The position of the peak of the peak of the peak degree distribution graph selected as described above can be set as the position of the image point that matches the scanning line of the light-section image S, but the position of the peak is measured. Therefore, it is conceivable that the position of the area centroid of the peak of the luminance distribution graph corresponding to the selected peak of the peak degree distribution graph is used as the position of the image point. In this case, it is desirable to calculate the area center of gravity excluding both skirts of the mountain part affected by the background illumination. However, how to define the measurement range of the area center of gravity becomes a problem.

Here, the peak degree distribution curve once becomes a negative value at the rising portions of both skirts of the peak portion of the luminance distribution graph, and becomes a positive value when it rises to some extent. Then, a zero point P0 at which the peak degree becomes zero appears on both sides of the peak of the peak degree distribution curve, and a section between the zero points P0 is a part which is not affected by the background illumination. Therefore, as shown in FIG. 5B, if the area centroid G ′ in the section between the zero point P0 of the luminance distribution graph is obtained and the position is set as the position of the image point, the influence of the background illumination is eliminated and the image point is removed. Position can be accurately detected.

However, the inclination of the peak of the brightness distribution graph on the imaging direction side (right side) tends to be gentler than the inclination on the opposite side, so that the area center of gravity G 'of the peak is the position of the peak of the peak. From the camera to the imaging direction. On the other hand, the peaks of the peak degree distribution graph have substantially the same inclination on both sides even if the inclinations of the peaks of the luminance distribution graph are different on both sides. Therefore,
The area center of gravity G of the peak in the section between the zero points P0 of the peak degree distribution graph does not shift to the left from the position of the peak of the peak.
Therefore, in order to improve the detection accuracy of the image points, the area centroid G of the section between the zero points P0 of the peak degree distribution graph is obtained,
It is desirable that the position of the center of gravity G be the X coordinate of the image point PS.

As described above, when the positions of the image points PS corresponding to the respective scanning lines of the light-section image are detected, these image points P
Point sequence data representing the position of S is created. FIG. 3B shows a graph of the point sequence data. Next, the peak degree of the change in the position of the image point PS in the X-axis direction of each scanning line is obtained. The peak degree of the scanning line having the Y-axis coordinate m is defined as am at the image point of the scanning line, bm at the image point of the scanning line at both ends of the measurement range of a predetermined width set around the scanning line, and
As cm, it is calculated as a value representing the separation distance of am with respect to the connection line connecting bm and cm. In this case, the distance between the midpoint of the connection and am may be used as the peak degree, and the length of a perpendicular line descending from am on the connection may be used as the peak degree.

Next, the image point where the peak degree becomes maximum, FIG.
In (B), an image point am1 of a scanning line whose Y-axis coordinate is m1,
The image point having the minimum peak value (negative value), and the image point am2 of the scanning line whose m-axis coordinate is m2 in FIG. 3B, are extracted as feature points. When a plurality of image points having the maximum peak degree and the minimum peak degree are continuously arranged, an intermediate image point is extracted as a feature point. Further, an image point at the end of the light-section image, in FIG. 3B, an image point am of the scanning line L1
Image points am3 of the scanning lines 0 and L95 are also extracted as feature points. If the image points of two adjacent scanning lines are separated by a predetermined value (for example, 10 pixels) or more in the X-axis direction, it can be determined that the light-section image is discontinuous there. The image point is extracted as a feature point as an image point at the end of the light-section image.

Next, a desired image point is selected by an external operation from the image points extracted as the feature points. As in the present embodiment, when the position of the light section image S is measured with the bent portion Sa of the mountain cut as a characteristic portion, the image point am1
Select When the image points are selected in this manner, the point sequence data representing the positions of the image points in a predetermined range in the Y-axis direction around the selected image points is displayed together with the data indicating the relevance of these image points together with the data for template creation. Stored as data. The image points of the two upper and lower feature points that are closest to the selected image point are also stored as points defining the end of the image part for which the equations of the image lines LS1 and LS2 are obtained. For example, when the image point am1 is selected, the position data of the image point am0 and the image point am2 is stored as the position of the end of the image part for which the image line is to be calculated.

FIG. 3C shows point sequence data and relevance data stored as template creation data. In this example, No. 10 is the data of the selected image point am1, and data of a total of 21 image points, 10 in each of the upper and lower image points, is stored. The relevance data indicates that “1” is the upper end point, “2” is the continuous point, and “3” is the lower end point.

At the time of measuring the cross section of the work, the template creation data corresponding to the measurement site of the work is read out, and the characteristic portion to be measured, in the above example, the mountain bent portion Sa
The image data for the template representing the shape of is created by graphic processing. At the time of graphic processing, first, as shown in FIG. 6A, linearization processing for connecting continuous points between the upper end point and the lower end point with a line based on the point sequence data and the relevancy data is performed. A dilation process is performed to make the luminance of eight pixels around each pixel equal to the luminance of each pixel, and the line is thickened as shown in FIG.
Finally, the brightness of each pixel is calculated by comparing each pixel with its surroundings.
A smoothing process is performed to set the average value of the luminance of the individual pixels, and template image data having a luminance distribution as shown in FIG. 6C is created. The luminance distribution of the light-section image S is shown in FIG.
The luminance distribution in FIG. 6C is very similar to this. FIG. 7A is a view visually showing the state of execution of pattern matching using the template TP with respect to the imaging screen obtained by capturing the light-section image S of the work A. position,
That is, the position corresponding to the image point am1 is determined by pattern matching using the normalized correlation method or the like. In this embodiment, in order to reduce the processing time of the pattern matching, the pattern matching is performed by reducing the imaging screen and the template TP at the same ratio (for example, 1/4).

At the step portion of the work, the light-section image S is divided as shown in FIG. 8A, and the image points am (n) and am (n + 1) of the upper and lower scanning lines at the division point in the X-axis direction. Image points am (n) and am (n + 1) are extracted as feature points representing edges of the image at a distance of a predetermined value or more. When these image points am (n) and am (n + 1) are captured in the template creation data, the relevance data of the image point am (n) is “3” (lower end point), and the image point am (n +
The relevance data of 1) is “1” (upper end point). In this case, if the image data of the template is created such that the image point am (n) is the lower end point and the image point am (n + 1) is the upper end point as shown in FIG. The degree of approximation decreases. Therefore, in this case, the image of the template is changed to the image point a as shown in FIG.
It is created by extending 走 査 of the scanning pitch below m (n) and above image point am (n + 1), respectively.

The position of the characteristic portion of the light cut image S of the work to be measured by pattern matching, that is, the point Pam corresponding to the image point am1 that matches the bent portion Sa of the mountain fold.
Once the position of the image point is determined, the positions of the two image points am0 and am2 stored as the positions of the ends of the image part for which the image line is to be calculated are detected at the reference position of the point Pam1 (detected by the teaching measurement). The position of the points Pam0 and Pam2 corresponding to the image points am0 and am2 is calculated by correcting the deviation from the position of the image point am1).

The point Pam1 is shifted to the left in the X-axis direction from the reference position when the work A is displaced in the direction approaching the measuring head.
The shift to the right in the axial direction is when the work A is displaced in the direction away from the measuring head. When the work A is shifted to the left in the X-axis direction due to the difference in perspective from the image pickup device 2, the light-cut image is slightly generated. When S is enlarged and shifted to the right in the X-axis direction, the light-section image S is reduced. Therefore, the image point am
In addition to moving 0 and am2 in parallel in both X and Y axes from the reference position of point Pam1 in both X and Y axes, enlargement and reduction of the light-cut image S due to displacement in X axis direction. It is desirable to calculate the positions of the points Pam0 and Pam2 in consideration of the above. When the positions of the points Pam0 and Pam2 are calculated so as to be outside the screen, the points are replaced with the intersections of the straight line connecting the points Pam1 and the outer edge of the screen.

As described above, the points Pam1, Pam0, Pam
Next, as shown in FIG. 7B, the connection between the points Pam1 and Pam0 and the points Pam1 and Pam1 are determined.
The position of each of the equally dividing points that divides the connection with m2 into a plurality of equal parts, for example, six equal parts, is determined, and a scanning line L having a predetermined length parallel to the X axis and having these respective equally dividing points as a middle point is set as a point Pam1. Five points are respectively set between the point Pam0 and the point Pam1 and the point Pam2.

Next, the coordinates of an image point PS that matches each scanning line L of the light-section image S is detected from the luminance distribution on each scanning line L. Here, the detection of the image point PS is performed in the same manner as the detection of the image point PS during the teaching measurement. That is, FIG.
As shown in (B), a peak degree distribution graph is created from the luminance distribution graph, and among the peaks of the peak degree distribution graph, the zero point P0 of the leftmost peak located at a height equal to or higher than a predetermined value. The area center of gravity G in the section between them is determined, and the position of the center of gravity G is set as the position of the image point PS.

Next, as shown in FIG. 7 (C), the equation of the first image line LS1 approximating the upper image portion which is continuous with the mountain-bent bent portion Sa of the light-section image S is expressed by the points Pam1 and Pam1.
The equation of the second image line LS2 that is obtained by regression processing from the coordinates of the five image points PS between m0 and m0 and that approximates the lower image portion that is continuous with the fold-bent portion Sa of the light-section image S , Are obtained by regression processing from the coordinates of the five image points PS between the points Pam1 and Pam2, and the coordinates of the intersection Q of the first and second image points LS1 and LS2 are calculated.

When the length of the connection between the characteristic portion to be measured and the end of the image portion for which the image line is to be calculated, for example, the connection between the points Pam1 and Pam2 is less than a predetermined value (for example, 100 pixels). When this connection is divided into six equal parts, FIG.
As shown in (A), the scanning lines set at the uppermost and lowermost equal points intersect with the rounded and curved valleys of the light-section image S, and the image line is correctly calculated. become unable. Therefore, when the connection length is equal to or less than the predetermined value, a scanning line is set at each equally dividing point that divides a plurality of intermediate portions excluding the vicinity of the characteristic portion and the end portion. For example, as shown in FIG. 9B, the connection is divided into eight equal parts, and the scanning lines L are set at five intermediate points other than the uppermost and lowermost equal points.

Further, the image line is calculated as a straight line or a circle so that the regression processing for obtaining the equations of the image lines LS1 and LS2 can be performed in a short time. Is determined in advance at the time of teaching. The determination method is as shown in FIG.
First, at STP1, it is determined whether or not teaching is performed. At the time of teaching, the process proceeds to STP2, and whether or not the length of a connection between the characteristic portion to be measured and the end of the image portion for which the image line is to be calculated is greater than or equal to a predetermined value. Is determined. When the length of the connection is short, when the image line is calculated as a circle, there is a variation in measurement every time.
The variation rapidly increases when the connection length is shorter than 100 pixels. Therefore, when the predetermined value is set to 100 pixels and the connection length is shorter than the predetermined value, the image line is calculated as a regression line (a line that minimizes the total amount of deviation from the line of each image point) in STP3. ,
At STP4, it is determined whether or not the calculation is successful. If the calculation is not successful, error processing is performed at STP5.

If the connection length is longer than a predetermined value, the STP
In step 6, the image line is calculated as a regression circle (a circle that minimizes the amount of deviation of each image point from the circle), and it is determined in STP7 whether or not the calculation was successful. If the process is successful, the process proceeds to STP9, and it is determined whether the calculated radius of the regression circle is equal to or larger than a predetermined value (for example, 5000 pixels). If the radius is large, the process proceeds to STP3 to calculate the image line as a regression line. If the radius is smaller than the predetermined value, the radius is stored as teaching data in STP10.

When measuring the cross section of the work, the process proceeds from STP1 to STP11, and it is determined whether or not it is possible to calculate the regression circle as the teaching radius. When the light-section image deviates greatly from the reference position in the X-axis direction,
The ratio of enlargement and reduction of the light-section image also increases, and it is not possible to calculate the regression circle with the radius of teaching. In this case, the process proceeds to STP2, and the same processing as during teaching is performed. Otherwise, the process proceeds to STP12, where it is determined whether or not the teaching data specifies that the image line is calculated as a straight line. If so, the image line is obtained as a regression line in STP3. S if not a straight line
Proceeding to TP13, the image line is calculated as a regression circle having the taught radius, and whether or not the calculation is successful is determined in STP14. If the calculation is not successful, error processing is performed in STP15.

According to the present embodiment, the scanning line L can be accurately set in a required positional relationship in the image portion where the image line is to be calculated, and the image point P which matches the scanning line L of the light-section image S
The coordinates of S can also be accurately detected by using the peak distribution, and the calculation accuracy of the image line, and therefore, the accuracy of the cross-section measurement of the work is improved. Furthermore, since only a limited portion of the captured image is scanned, the measurement time can be shortened, and online measurement on a line with a short tact time can be sufficiently dealt with.

Further, by simply selecting an image point corresponding to a characteristic portion to be measured from image points extracted as characteristic points during teaching, point sequence data in a predetermined range centered on the image point is created. Therefore, a template representing the shape of the characteristic portion can be easily and accurately created, and the accuracy of pattern matching is improved.

Further, since the image data of the template is created by graphic processing from the point sequence data stored as described above, the number of data to be stored is much smaller than when the image data itself is stored. I can do it. Therefore, a small storage device is sufficient for a work having a large number of measurement parts such as an automobile body.

[Brief description of the drawings]

FIG. 1 is a perspective view showing an outline of an optical measuring device.

FIG. 2 is a diagram showing an imaging screen imaged in the state shown in FIG. 1;

FIGS. 3A and 3B are diagrams showing (A) a setting of a scanning line on an imaging screen at the time of teaching measurement, (B) a diagram showing extraction of point sequence data and a feature point of an optical cutting image of a master work, and (C) a template creation. Diagram showing the data of

4A is a graph showing a luminance distribution on a scanning line of an imaging screen, FIG. 4B is a graph showing a luminance distribution after a smoothing process,
(C) Graph showing peak degree distribution

FIG. 5A is a diagram illustrating a method of calculating a peak degree, and FIG. 5B is a diagram illustrating a method of calculating an image point.

FIGS. 6A, 6B, and 6C are diagrams illustrating a method of creating a template, and FIG. 6D is a diagram illustrating a luminance distribution of a light-section image.

FIG. 7A is a diagram showing a pattern matching situation;
(B) A diagram showing the setting of scanning lines for obtaining image lines,
(C) Diagram showing image lines

FIGS. 8A, 8B, and 8C show a method of creating a template in an image dividing unit. FIGS.

9A and 9B are diagrams showing a method of setting a scanning line in a short section.

FIG. 10 is a flowchart showing image line calculation processing during teaching and work measurement.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Slit light source 2 Imager 3 Image processing device S Light cutting image L Scan line PS Image point TP Template

Claims (2)

[Claims] 1. Creation of a template for use in pattern-matching to determine the position of a characteristic portion of a light-cut image of a work appearing on an image screen when a light-cut image of the work drawn by slit light applied to the work is picked up. A method of irradiating a slit light on a master work to capture an optical cut image of the master work, and scanning an imaging screen at a predetermined pitch to match an image corresponding to each scanning line of the optical cut image of the master work. A step of detecting point positions and creating point sequence data representing the positions of these image points; a step of storing point sequence data in a predetermined range centering on the feature portion; and a graphic from the stored point sequence data. Generating image data for a template representing the shape of the characteristic portion by processing. Forming method. 2. The method according to claim 1, wherein an expansion process and a smoothing process are performed in the graphic process.